Word-Formation. Volume 2 Word-Formation: An International Handbook of the Languages of Europe 9783110246278, 9783110246261

Citation preview

Word-Formation HSK 40.2

Handbücher zur Sprach- und Kommunikationswissenschaft Handbooks of Linguistics and Communication Science Manuels de linguistique et des sciences de communication Mitbegründet von Gerold Ungeheuer Mitherausgegeben (1985–2001) von Hugo Steger

Herausgegeben von / Edited by / Edités par Herbert Ernst Wiegand

Band 40.2

De Gruyter Mouton

Word-Formation An International Handbook of the Languages of Europe Volume 2 Edited by Peter O. Müller Ingeborg Ohnheiser Susan Olsen Franz Rainer

De Gruyter Mouton

ISBN 978-3-11-024626-1 e-ISBN (PDF) 978-3-11-024627-8 e-ISBN (EPUB) 978-3-11-039468-9 ISSN 1861-5090 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2015 Walter de Gruyter GmbH, Berlin/Boston Typesetting: Meta Systems Publishing & Printservices GmbH, Wustermark Printing and binding: Hubert & Co. GmbH & Co. KG, Göttingen Cover design: Martin Zech, Bremen 앝 Printed on acid-free paper 앪 Printed in Germany www.degruyter.com

Contents

Volume 2 IV. Rules and restrictions in word-formation I: General aspects 45. 46. 47. 48.

Rules, patterns and schemata in word-formation · Heike Baeskow Word-formation and analogy · Sabine Arndt-Lappe . . . . . . . . . Productivity · Livio Gaeta and Davide Ricca . . . . . . . . . . . . Restrictions in word-formation · Livio Gaeta . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

803 822 842 859

. .

876 894 918

. . . .

932 945 957 972

. . . . . . . . .

984 1002 1020

. . . .

. . . .

1035 1056 1099 1128

63. Noun-noun compounds · Christina L. Gagné and Thomas L. Spalding . . . . 64. Gender marking · Ursula Doleschal . . . . . . . . . . . . . . . . . . . . . . . . 65. Singulatives · Paolo Acquaviva . . . . . . . . . . . . . . . . . . . . . . . . . .

1143 1159 1171

V. Rules and restrictions in word-formation II: Special cases 49. 50. 51. 52.

Argument-structural restrictions on word-formation patterns · Holden Härtl Phonological restrictions on English word-formation · Renate Raffelsiefen Morphological restrictions on English word-formation · Lothar Peter . . . Semantic restrictions on word-formation: the English suffix -ee · Heike Baeskow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53. Dissimilatory phenomena in French word-formation · Marc Plénat . . . . . 54. Closing suffixes · Stela Manova . . . . . . . . . . . . . . . . . . . . . . . . . 55. Closing suffix patterns in Russian · Dmitri Sitchinava . . . . . . . . . . . .

VI. Semantics and pragmatics in word-formation I: General aspects 56. 57. 58. 59.

Motivation, compositionality, idiomatization · Daniela Marzo . . . . . . Word-formation and folk etymology · Sascha Michel . . . . . . . . . . . Categories of word-formation · Volkmar Lehmann . . . . . . . . . . . . Schemata and semantic roles in word-formation · Hanspeter Ortner and Lorelies Ortner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60. Word-formation and argument structure · Manfred Bierwisch . . . . . . 61. Word-formation and metonymy · Manfred Bierwisch . . . . . . . . . . . 62. The pragmatics of word-formation · Lavinia Merlini Barbaresi . . . . .

. . . .

VII. Semantics and pragmatics in word-formation II: Special cases

vi

Contents 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80.

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

1183 1195 1209 1230 1241 1253 1268 1285 1304 1316 1327 1339 1351 1360

. . . .

. . . .

. . . .

1373 1390 1406 1424

. . . . . .

1467 1482

. . . . . .

1492 1515

. . . . . .

1528 1545

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

v vii

81. 82. 83. 84. 85. 86. 87. 88. 89.

Collectives · Wiltrud Mihatsch . . . . . . . . . . . . . . . . . . . . . . . . Action nouns · Maria Koptjevskaja-Tamm . . . . . . . . . . . . . . . . . Action nouns in Romance · Livio Gaeta . . . . . . . . . . . . . . . . . . Verbal nouns in Celtic · Paul Russell . . . . . . . . . . . . . . . . . . . . Nominalization in Hungarian · Tibor Laczkó . . . . . . . . . . . . . . . . Result nouns · Chiara Melloni . . . . . . . . . . . . . . . . . . . . . . . . Quality nouns · Franz Rainer . . . . . . . . . . . . . . . . . . . . . . . . . Status nouns · Hans Christian Luschützky . . . . . . . . . . . . . . . . . Agent and instrument nouns · Franz Rainer . . . . . . . . . . . . . . . . Patient nouns · Susanne Mühleisen . . . . . . . . . . . . . . . . . . . . . Place nouns · Bogdan Szymanek . . . . . . . . . . . . . . . . . . . . . . Intensification · Franz Rainer . . . . . . . . . . . . . . . . . . . . . . . . . Negation · Marisa Montero Curiel . . . . . . . . . . . . . . . . . . . . . . Negation in the Slavic and Germanic languages · Jozef Pavlovič . . . . Spatial and temporal relations in German word-formation · Ludwig M. Eichinger . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adverbial categories · Davide Ricca . . . . . . . . . . . . . . . . . . . . Denominal verbs · Andrew McIntyre . . . . . . . . . . . . . . . . . . . . Valency-changing word-formation · Dieter Wunderlich . . . . . . . . . . Word-formation and lexical aspect: deverbal verbs in Italian · Nicola Grandi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Word-formation and aspect in Samoyedic · Beáta Wagner-Nagy . . . . Verbal prefixation in Slavic: a minimalist approach · Petr Biskup and Gerhild Zybatow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Denumeral categories · Bernhard Fradin . . . . . . . . . . . . . . . . . . The semantics and pragmatics of Romance evaluative suffixes · Martin Hummel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Morphopragmatics in Slavic · Alicja Nagórko . . . . . . . . . . . . . . .

Volume 1

I.

Word-formation as a linguistic discipline

1. The scope of word-formation research · Hans-Jörg Schmid . . . . . . . . . 2. Word-formation research from its beginnings to the 19 th century · Barbara Kaltz and Odile Leclercq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Word-formation in historical-comparative grammar · Thomas Lindner . . . 4. Word-formation in structuralism · Wolfgang Motsch . . . . . . . . . . . . . 5. Word-formation in inhaltbezogene Grammatik · Johannes Erben . . . . . . 6. Word-formation in onomasiology · Joachim Grzega . . . . . . . . . . . . . 7. Word-formation in generative grammar · Rochelle Lieber . . . . . . . . . . 8. Word-formation in categorial grammar · Ulrich Wandruszka . . . . . . . . .

.

1

. . . . . . .

22 38 52 66 79 94 112

Contents 9. 10. 11. 12. 13.

vii

Word-formation in natural morphology · Hans Christian Luschützky . . Word-formation in cognitive grammar · John R. Taylor . . . . . . . . . Word-formation in optimality theory · Renate Raffelsiefen . . . . . . . . Word-formation in construction grammar · Geert Booij . . . . . . . . . Word-formation in psycholinguistics and neurocognitive research · Gary Libben . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

123 145 158 188

. . .

203

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

218 235 301 322 340 352 364 386 413 434 450 467 485 500

. . . . . . . . . . . . . . . .

517 524 537 551 568 582 594 611 627 660 673 688 707 727 742 757

.

780

II. Units and processes in word-formation I: General aspects 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27.

The delimitation of derivation and inflection · Pavol Štekauer . . . Units of word-formation · Joachim Mugdan . . . . . . . . . . . . . Derivation · Andrew Spencer . . . . . . . . . . . . . . . . . . . . . . Conversion · Salvador Valera . . . . . . . . . . . . . . . . . . . . . . Backformation · Pavol Štekauer . . . . . . . . . . . . . . . . . . . . Clipping · Anja Steinhauer . . . . . . . . . . . . . . . . . . . . . . . Composition · Susan Olsen . . . . . . . . . . . . . . . . . . . . . . . Blending · Bernhard Fradin . . . . . . . . . . . . . . . . . . . . . . . Incorporation · Jason D. Haugen . . . . . . . . . . . . . . . . . . . . Particle-verb formation · Andrew McIntyre . . . . . . . . . . . . . Multi-word expressions · Matthias Hüning and Barbara Schlücker Reduplication · Thomas Schwaiger . . . . . . . . . . . . . . . . . . Word-creation · Elke Ronneberger-Sibold . . . . . . . . . . . . . . Allomorphy · Wolfgang U. Dressler . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

III. Units and processes in word-formation II: Special cases 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44.

Affective palatalization in Basque · José Ignacio Hualde . . . . . . . . . . . Parasynthesis in Romance · David Serrano-Dolader . . . . . . . . . . . . . . Affix pleonasm · Francesco Gardani . . . . . . . . . . . . . . . . . . . . . . Interfixes in Romance · Michel Roché . . . . . . . . . . . . . . . . . . . . . Linking elements in Germanic · Nanna Fuhrhop and Sebastian Kürschner Synthetic compounds in German · Martin Neef . . . . . . . . . . . . . . . . Verbal pseudo-compounds in German · Christian Fortmann . . . . . . . . . Particle verbs in Germanic · Nicole Dehé . . . . . . . . . . . . . . . . . . . Particle verbs in Romance · Claudio Iacobini . . . . . . . . . . . . . . . . . Particle verbs in Hungarian · Mária Ladányi . . . . . . . . . . . . . . . . . . Noun-noun compounds in French · Pierre J. L. Arnaud . . . . . . . . . . . Verb-noun compounds in Romance · Davide Ricca . . . . . . . . . . . . . . Co-compounds · Bernhard Wälchli . . . . . . . . . . . . . . . . . . . . . . . Multi-word units in French · Salah Mejri . . . . . . . . . . . . . . . . . . . . Multi-word expressions and univerbation in Slavic · Olga Martincová . . . Compounds and multi-word expressions in Slavic · Ingeborg Ohnheiser . . Paradigmatically determined allomorphy: the “participial stem” from Latin to Italian · Anna M. Thornton . . . . . . . . . . . . . . . . . . . . . . . . . .

viii

Contents

Volume 3 VIII. Foreign word-formation, language planning and purism I: General aspects 90. Types of foreign word-formation · Wieland Eins 91. Word-formation in Neo-Latin · Thomas Lindner and Franz Rainer 92. Foreign word-formation, language planning and purism · Wolfgang Pöckl

IX. Foreign word-formation, language planning and purism II: Special cases 93. 94. 95. 96. 97. 98. 99. 100. 101.

Foreign word-formation in German · Peter O. Müller Foreign word-formation in English · Klaus Dietz Foreign word-formation in Italian · Claudio Iacobini Foreign word-formation in Polish · Krystyna Waszakowa Word-formation and purism in German · Mechthild Habermann Word-formation and purism in French · Petra Braselmann Word-formation and purism in Croatian · Branko Tošović Word-formation and language planning in Estonian · Virve Raag Individual initiatives and concepts for expanding the lexicon in Russian · Wolfgang Eismann

X. Historical word-formation I: General aspects 102. Mechanisms and motives of change in word-formation · Franz Rainer 103. Change in productivity · Carmen Scherer

XI. Historical word-formation II: Special cases 104. Grammaticalization in German word-formation · Mechthild Habermann 105. The grammaticalization of prepositions in French word-formation · Dany Amiot 106. The Romance adverbs in -mente: a case study in grammaticalization · Ulrich Detges 107. Grammaticalization in Slavic word-formation · Krystyna Kleszczowa 108. The origin of suffixes in Romance · David Pharies

XII. Historical word-formation III: Language sketches 109. 110. 111. 112. 113.

Historical word-formation in German · Peter O. Müller Historical word-formation in English · Klaus Dietz From Latin to Romance · Éva Buchi and Jean-Paul Chauveau From Latin to Romanian · Marina Rădulescu Sala From Old French to Modern French · Franz Rainer and Claude Buridant

Contents 114. 115. 116. 117. 118. 119.

ix

From Old Irish to Modern Irish · David Stifter Historical word-formation in Slavic · Swetlana Mengel From Ancient Greek to Modern Greek · Io Manolessou and Angela Ralli The history of word-formation in Uralic · Johanna Laakso From Old Hungarian to Modern Hungarian · Tamás Forgács Historical word-formation in Turkish · Claus Schönig

XIII. Word-formation in language acquisition and aphasia 120. Word-formation in first language acquisition · Hilke Elsen and Karin Schlipphak 121. Word-formation in second language acquisition · Cornelia Tschichold and Pius ten Hacken 122. Word-formation in aphasia · Carlo Semenza and Sara Mondini

XIV. Word-formation and language use 123. 124. 125. 126. 127. 128. 129. 130.

Word-formation Word-formation Word-formation Word-formation Word-formation Word-formation Word-formation Word-formation

and and and and and and and and

text · Anja Seiffert brand names · Elke Ronneberger-Sibold planned languages · Klaus Schubert sign languages · Ronnie B. Wilbur technical languages · Ivana Bozděchová literature · Peter Handler orthography · Hannelore Poethe visuality · Lorelies Ortner

XV. Tools in word-formation research 131. Dictionaries · Renate Belentschikow 132. Corpora · Ulrich Heid 133. Internet · Georgette Dal and Fiammetta Namer

Volume 4 XVI. Word-formation in the individual European languages Indo-European Germanic 134. German · Irmhild Barz 135. English · Ingo Plag

x

Contents 136. 137. 138. 139. 140. 141. 142. 143.

Dutch · Geert Booij Frisian · Jarich F. Hoekstra Yiddish · Simon Neuberg Faroese · Hjalmar P. Petersen Danish · Hans Götzsche Norwegian · John Ole Askedal Swedish · Kristina Kotcheva Icelandic · Þorsteinn G. Indriðason

Romance 144. 145. 146. 147. 148. 149. 150. 151.

Portuguese · Bernhard Pöll Spanish · Franz Rainer Catalan · Maria Teresa Cabré Castellví French · Franck Floricic Ladin · Heidi Siller-Runggaldier Sardinian · Immacolata Pinto Italian · Franz Rainer Romanian · Maria Grossmann

Celtic 152. Breton · Elmar Ternes 153. Welsh · Paul Russell 154. Irish · Brian Ó Curnáin

Slavic 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168.

Upper Sorbian · Anja Pohontsch Polish · Alicja Nagórko Kashubian · Edward Breza Czech · Ivana Bozděchová Slovak · Martina Ivanová and Martin Ološtiak Ukrainian · Ievgeniia Karpilovska Belarusian · Alâksandr Lukašanec Russian · Igor’ S. Uluhanov Slovene · Irena Stramljič Breznik Croatian · Mario Grčević Serbian · Božo Ćorić Bosnian · Branko Tošović Bulgarian · Cvetanka Avramova and Julia Baltova Macedonian · Lidija Arizankovska

Map of languages

Contents

Volume 5 Baltic 169. Lithuanian · Bonifacas Stundžia 170. Latvian · Agnė Navickaitė-Klišauskienė

Albanian 171. Albanian · Monica Genesin and Joachim Matzinger

Greek 172. Greek · Angela Ralli

Indo-Iranian 173. Ossetic · David Erschler 174. Tat · Gilles Authier

Uralic 175. 176. 177. 178. 179. 180. 181.

Nenets · Beáta Wagner-Nagy Finnish · Kaarina Pitkänen-Heikkilä Estonian · Krista Kerge Permic · László Fejes Mari · Timothy Riese Mordvinic · Sándor Maticsák Hungarian · Ferenc Kiefer

Basque 182. Basque · Xabier Artiagoitia, José Ignacio Hualde and Jon Ortiz de Urbina

Semitic 183. Maltese · Joseph Brincat and Manwel Mifsud

Turkic 184. Turkish · Jens Wilkens 185. Bashkir · Gulnara Iskandarova 186. Tatar · László Károly

xi

xii

Contents 187. 188. 189. 190.

Crimean Tatar · Henryk Jankowski Gagauz · Astrid Menz Karaim · Éva Á. Csató Chuvash · Galina N. Semenova and Alena M. Ivanova

Mongolic 191. Kalmyk · Danara Suseeva

Northwest Caucasian 192. Abkhaz · Viacheslav A. Chirikba 193. Adyghe · Yury Lander 194. Kabardian · Ranko Matasović

Northeast Caucasian 195. 196. 197. 198. 199. 200. 201. 202. 203. 204. 205. 206. 207.

Rutul · Mikhail Alekseyev Budugh · Gilles Authier and Adigözel Haciyev Udi · Wolfgang Schulze Aghul · Timur Maisak and Dmitry Ganenkov Archi · Marina Chumakina Khinalug · Wolfgang Schulze Lak · Wolfgang Schulze Dargwa · Nina Sumbatova Bezhta · Madzhid Khalilov and Zaira Khalilova Botlikh · Mikhail Alekseyev Akhwakh · Denis Creissels Avar · Madzhid Khalilov and Zaira Khalilova Khwarshi · Zaira Khalilova

Subject index Map of languages

IV. Rules and restrictions in word-formation I: General aspects 45. Rules, patterns and schemata in word-formation 1. 2. 3. 4. 5.

Introduction The pre-generative era Generative approaches Schemata in word-formation (an outline) References

Abstract The observation that word-formation is systematic to a relatively high degree has inspired generations of linguists to identify patterns, rules and schemata and to describe regularities as well as idiosyncrasies from different perspectives. This article presents selected approaches, which have successfully contributed to the scientific attractiveness of word-formation, and takes account of morpho-syntactic, semantic, phonological and cognitive considerations.

1. Introduction Since the early seventies of the last century, word-formation has opened up a highly complex and continually expanding field of research, which owes its attractiveness to the fact that the combination of morphemes is not restricted to morpho-syntactic considerations, but also involves aspects of other linguistic components, especially of phonology and semantics. Since the interaction of these components is most striking in derivation, this article will concentrate on the rules, patterns and schemata involved in wordformation processes of this type. For instance, from a morpho-syntactic point of view, the derivative writer consists of the verbal base write and the nominalizing suffix -er, which constitutes the head of the complex word because it determines its categorial properties. Semantically, the deverbal derivative writer is an agent noun because it refers to the class of individuals actively or habitually involved in a writing event. Of course, the properties depicted here rather pre-theoretically are not individual properties of writer, but apply to a whole range of derivatives, as examples like driver, reader, singer, runner, producer and many more suggest. Moreover, sets of data like these, which are easily retrievable from reverse dictionaries (e.g., Lehnert 1971, Muthmann 2002 or the OneLook Reverse Dictionary provided by the internet), reveal the selectional preferences of a suffix. Thus, although the highly productive suffix -er is relatively flexible with respect to its input (cf. potter, stranger, upper ‘anti-depressant pill’, fiver ‘five pound note’, penny-a-liner ‘journalist’), it prototypically combines with verbal bases.

804

IV. Rules and restrictions in word-formation I: General aspects The interaction between derivation and phonology becomes evident in pairs like divine [dɪ'vɑɪn] : divinity [dɪ'vɪnəti], serene [sə'ri:n] : serenity [sə'renəti] or profound [prə'faʋnd] : profundity [prə'fʌndəti]. Unlike the functionally similar native suffix -ness, which preserves the sound pattern of its base, (e.g., serene : sereneness), the Latinate suffix -ity often reduces a long vowel or diphthong contained in the final syllable of a polysyllabic adjectival base. Since it is the antepenultimate syllable of an -ity derivative which is affected by vowel reduction, this rule is referred to as trisyllabic shortening. Of course, prefixation is an “interactive process” as well. In English, prefixes do not normally change the categorial properties of their bases because they occur in the nonhead position, but they add semantic information content. For example, the prefix un-, which typically combines with adjectives, has a negativizing function. As observed by Jespersen 1974 [1942]: 466) and confirmed by Zimmer (1964: 36−37), un- shows a relatively strong preference for adjectival bases denoting positive qualities. There are derivatives like unwise, unhealthy, unclean or unhandsome, but forms like *undumb, *unill, *undirty or *unugly are unacceptable. According to the redundancy restriction formulated by Lieber (2004: 161), “[a]ffixes do not add semantic content that is already available within a base word (simplex or derived)”. As far as un- is concerned, there are only a few exceptions to the redundancy restriction, including forms like uncorrupt, unguilty, unselfish or unvulgar (Zimmer 1964: 36; Lieber 2004: 159). The redundancy restriction also predicts the compatibility of un- with semantically neutral bases, which according to Zimmer is attested in a number of cases (unbindable, uncountable, unseen, etc.). Although derivation − unlike syntax − involves a rather high degree of idiosyncrasy, which has to be listed in the lexicon or a comparable storage place, the preliminary examples provided so far illustrate that morpho-syntactic, semantic and phonological regularities, which account for well-formedness, rarely affect individual words, but operate over sets of derivatives. The observation that there are regularities in word-formation by derivation has inspired generations of linguists to identify patterns and schemata and to formulate rules which serve to generate new complex words on the one hand and to analyse existing ones on the other. However, since word-formation − like language in general − is a dynamic system, these rules do not have absolute character, but express strong tendencies which may be overridden by exceptions.

2. The pre-generative era Before representative rules, patterns and schemata and the way they are treated in modern linguistics are dealt with in more detail, some pre-generative approaches which have had a considerable influence on subsequent morphological research will be considered briefly. Most valuable insights, which are ascribed to modern word-formation theories nowadays, were already formulated by Hermann Paul (1846−1921). In his lecture “Ueber die Aufgaben der Wortbildungslehre” [On the tasks of word-formation theory], whose content is provided by Henne and Kilian (1998: 171−192), he convincingly argues that inflection and word-formation should be treated as different domains of morphology.

45. Rules, patterns and schemata in word-formation This view arose from the observation that inflection is relatively systematic and functionally determined by the syntax, whereas word-formation is only partially regular. Paul does not deny an analogy between both domains: there are inflectional categories such as the genitive or the subjunctive on the one hand and word-formation categories like nomina agentis or nomina actionis on the other. However, word-formation involves aspects of meaning, which are neither predicted by the syntax nor reducible to dictionary entries. Moreover, in word-formation, the relation between form and function is often obscured by lexicalization, as in the case of German Schöpfer ‘Creator’, whose relation to schöpfen ‘to create’ is only etymologically traceable. By emphasizing the necessity to identify the morphological and functional aspects which favour or restrict the productivity of a suffix, Hermann Paul anticipates the mechanisms of generative word-formation. Paul’s differentiation between form and function or meaning is mirrored by Ferdinand de Saussure’s (2005 [1916]: 98−102) view of the linguistic sign. According to Saussure (1857−1913), whose name is inextricably linked with European structuralism, a linguistic sign is associated with a concept (signifié) and a sound sequence (signifiant), which are related to each other in an arbitrary way. For instance, the fact that the concept HORSE is referred to as horse /hɔ:s/ in English, cheval /ʃəval/ in French, Pferd /pfe:rt/ in German or caballo /kaßaʎo/ in Spanish is entirely arbitrary or unmotivated. As far as word-formation is concerned, Saussure explicitly distinguishes between semantic and grammatical functions of derivational suffixes. For example, the suffix -tēr of the Greek noun zeuk-tēr ‘one who yokes’ denotes an agent or initiator (l’agent, l’auteur de l’action). On the other hand, the suffix -nū as displayed by zéug-nū(mi) is a marker of the present tense and hence fulfils a grammatical function. Importantly, however, complex signs are never entirely motivated a) because their parts are inherently arbitrary and b) because their semantics is not strictly compositional (e.g., pain + ful), but results from the interaction of the meanings of their parts (e.g., pain × ful). This part-whole relation plays a key role in Saussure’s approach. The view that the linguistic sign or signal consists of sound and meaning is preserved by the American structuralist Leonard Bloomfield (1887−1949). In particular, sound and meaning are involved in the definition of the morpheme: “A linguistic form which bears no partial phonetic-semantic resemblance to any other form, is a simple form or morpheme” (Bloomfield 1933: 161). The meaning of a morpheme is referred to as a sememe, i.e. a feature of the practical world, and sememes are arbitrarily linked to phonemes. However, the description of meaning is considered to be beyond the scope of linguistics, which according to Bloomfield is an independent science. In this respect, American structuralism differs from European structuralism, where meaning was part of the linguistic description. As far as word-formation is concerned, a remarkable consequence of the purely structural Bloomfieldian approach is that primary words like hammer, rudder, spider, bitter, linger or under, which bear a partial resemblance to secondary words like danc-er, lead-er or rid-er, are supposed to consist of two immediate constituents, namely of a root (/hɛm-/, /rod-/, /spajd-/, etc.) and an affix-like constituent -er (1933: 240−241). The constituent which resembles the derivational suffix -er is referred to as a primary affix because it occurs in primary words, i.e. in words which contain no free forms among their immediate constituents. Further examples of primary affixes are -ow (e.g., furrow, yellow, borrow), -ock (e.g., hummock, mattock, hassock) or de- (e.g., deceive,

805

806

IV. Rules and restrictions in word-formation I: General aspects deduce, detain). Primary affixes which convey semantic information are typically found in American Indian languages, whose analysis was the main concern of American structuralism. The descriptive approach to language initiated in structuralism is maintained in the comprehensive works on English word-formation presented by Koziol (1971 [1937]), Jespersen (1974 [1942]) and Marchand (1969). These works include large sets of data, which are annotated with diachronic and synchronic information. Koziol lists his examples in chronological order, beginning with Old English. Jespersen’s work on morphology, which is part of his Modern English Grammar in seven volumes, is based on historical principles, but focuses on present-day English and thus on “living, i.e. productive formations” (1974: 4), including their phonological behaviour. A milestone in the literature on English word-formation is Marchand’s synchronic-diachronic approach first published in 1960. Since this book provides a wealth of data, it offers a solid, corpus-like basis for morphological analyses. As far as the theoretical framework is concerned, Marchand’s (1969) work ranges between traditional and generative approaches. On the one hand, it is in the tradition of Koziol and Jespersen in that it includes etymological aspects. On the other hand, it is already partly influenced by Lees (1960), who derived nominalizations from underlying sentences via transformations. Thus, Marchand refers to complex words as “grammatical syntagmas” (1969: 2) and assumes that a denominal agent noun like potter derives from an underlying sentence He makes pots (1969: 276). However, since the complex machinery of early transformational grammar (i.e. Chomsky 1957) is not applied by Marchand, his work is closer to the traditional than to the generative school of thought. Although concrete mechanisms of word-formation are not developed by Koziol, Jespersen and Marchand, their works constitute reference books of great value, which are consulted in international research up to the present, as for instance in Bauer, Lieber and Plag (2013).

3. Generative approaches The idea that nominalizations are derived from underlying sentences via transformations as suggested by Lees (1960) and accepted, e.g., by Marchand (1969) and Lakoff (1970) was criticized by Chomsky in his influential article “Remarks on nominalization” (1970). Although it was Chomsky himself who developed generative transformational grammar, he claims that the idiosyncrasies involved in word-formation should be dealt with in the lexicon. In the seventies of the last century, this position gave rise to the lexicalist hypothesis, according to which the lexicon functions not only to store the idiosyncratic properties of lexical items, but also to accommodate word-formation processes. Unlike the authors of traditional works, who confined themselves to descriptive, etymologically oriented analyses of complex words, the representatives of generative approaches have always attempted to reveal morpho-syntactic, semantic and phonological generalizations over patterns of word-formation, some of which will be presented in the following sections.

45. Rules, patterns and schemata in word-formation

3.1. Derivational mechanisms 3.1.1. Word-formation rules, adjustment rules and word-formation patterns A major achievement of generative word-formation theories is the explicit formalization of the observation anchored already in traditional approaches that the combination of derivational affixes with bases is not arbitrary, but somehow systematic. Morris Halle was the first to introduce word-formation rules (WFRs) as a formal device which tells us “how the morphemes are to be arranged in sequence to form actual words” (1973: 4) and which − together with a list of morphemes − provides the potential words of a language. The actual words are separated from potential words by means of a filter and find their way into the dictionary. Potential words (e.g., *derival, *confusal, *arrivation) are marked as [−lexical insertion] by the filter and thus prevented from entering the dictionary. The filter also accounts for idiosyncratic semantic and phonological information associated with lexical items. It ensures that only well-formed actual words are listed in the dictionary. Halle’s pioneering model of word-formation, which describes regular patterns as well as idiosyncrasies without making use of syntactic transformations, was elaborated by Jackendoff (1975), Aronoff (1976) and later on by Bochner (1993). However, contrary to the morpheme-based approach presented by Halle, these authors prefer a full-entry theory, i.e. a theory of derivation in which WFRs take actual words from the dictionary as their input instead of morphemes. Thus, they avoid the problem of assigning lexical properties to hapax legomena such as butch (in butcher) or non-native roots like mit (in permit, submit, transmit, etc.). According to Aronoff (1976: 49), both the input and the output of a WFR are members of major lexical categories. The following WFR combines the suffix -er with actual words of the category “verb”. More precisely, the verb serving as an input may be either transitive (e.g., teacher) or intransitive (e.g., sleeper). The output of the WFR belongs to the category “noun” and receives an agentive interpretation, which is “a function of the meaning of the base” (Aronoff 1976: 50). (1)

[X]V [±transitive]

→

[[X]V

#er]N

[±transitive]

‘one who Vs habitually, professionally, …’ As indicated already by Halle (1973: 16), a high proportion of complex words from the dictionary are stored already in the speaker’s permanent memory, so that strictly speaking, a WFR is activated only when he/she encounters or forms new words. Aronoff (1976: 31) explicitly accounts for this observation by claiming that WFRs function not only to generate new words, but also constitute redundancy rules, i.e. rules which help to analyze existing words. For example, the rule in (1) reveals that baker consists of the verb bake and the nominalizing suffix -er and thus has the structure bakeV#erN. The structurally similar form butcher is analysable only as butch#erN. Since *butch does not exist in English, this string − unlike bakeV − cannot be assigned a categorial label by the relevant WFR. A problem with the word-based approach is that the output of a WFR is not always well-formed and may require adjustment before the rules of phonology apply. For instance, the rule which combines verbs like nominate, evacuate, vaccinate, etc. with the

807

808

IV. Rules and restrictions in word-formation I: General aspects suffix -ee generates the sequences *nominatee, *evacuatee or *vaccinatee, which do not exist in this shape. Aronoff (1976: 88−98) solves this problem by introducing a truncation rule, i.e. an adjustment rule which deletes the verb-forming suffix -ate to yield the well-formed sequences nominee, evacuee or vaccinee. Another type of adjustment rule is the allomorphy rule, which serves to adjust the form of a morpheme to its phonological environment. For example, this type of rule determines the distribution of the phonologically conditioned variants +Ation, +ion, and +tion associated with the suffix -ion. Aronoff’s comprehensive work has set a standard for generative word-formation. By separating the lexical component from the syntax, Aronoff − like Halle (1973) and Jackendoff (1975) − explicitly accounts for the fact that words are more than syntactic building blocks and that the speakers of a language are well aware of their internal structure and their interaction with other linguistic components. The full-entry theory proposed by Jackendoff (1975), in which idiosyncratic information is measured by relating derivatives to their bases via redundancy rules (e.g., decide ↔ decision), inspired Motsch (2004) to describe word-formation processes on the basis of word-formation patterns. A complex word consists of a semantic pattern (“semantisches Muster”) and a phonological form. In the lexicon, these components are paired with syntactic and morphological information respectively. The potential to form neologisms is inherent to the semanticosyntactic patterns. As shown in (2), such a pattern is conceived of as a predicate-argument structure containing a categorial label, e.g., (N), and an argument position (x). In the course of a word-formation process, the categorial label is replaced by a concrete lexical item, e.g., HUND (‘DOG’). The semantic representation of the newly generated word hündisch ‘dog-like’, which is predicated of the referent of (x), is specified in the form of the paraphrase ‘a referent x has prominent properties of N (e.g., of dogs)’. (2)

SEMANTISCHES MUSTER: [WIE (N)] (x) ‘ein Bezugswort x hat prominente Eigenschaften von N’ SEMANTISCHE REPRÄSENTATION VON WORTBILDUNGEN: hündisch [WIE (HUND)] (x) ‘ein Bezugswort x hat prominente Eigenschaften von Hunden’

Word-formation patterns allow for a direct mapping of meaning structures onto sound structures and thus render mediating syntactic rules redundant.

3.1.2. Subcategorization frames The introduction of a lexicon as an autonomous component of the grammar removed from the syntax the burden of handling word-formation processes with all their idiosyncrasies. However, the generative theories presented so far indicate that the shift of wordformation from the syntax to the lexicon coincided with the introduction of quite a few morphological devices. Halle’s model requires a filter, and the output of Aronoff’s WFRs may be forced to undergo truncation or allomorphy rules.

45. Rules, patterns and schemata in word-formation A less costly and more elegant description of derivational processes is achieved by Lieber’s (1981) morpheme-based approach, which provides lexical entries for the idiosyncratic properties of free and bound morphemes. The lexical entry of a morpheme is composed of its phonological and semantic representation, categorial information and diacritical features (e.g., [+Latinate]). Significantly, the entries of affixes display subcategorization frames which specify the category of their input and output, e.g., (3)

-ee: ]V ___ ]N

Subcategorization frames have their origin in generative transformational grammar (e.g., Chomsky 1965), where they revealed the number and category of the complements required by verbs. A considerable advantage of Lieber’s model is that it dispenses not only with WFRs, but also with adjustment rules. Since affixes have access to bound morphemes, truncation is no longer required. Thus, -ee directly selects the bound root nomin- of the verb nominate and inserts it into its subcategorization frame. The abolishment of truncation is advantageous especially for languages whose verbs have infinitival endings (e.g., German fahren → Fahr-er, French danser → dans-eur, Spanish pensar → pensa-miento). Allomorphy rules are no longer required either because according to Lieber (1981: 141), all allomorphy in English is confined to stems. Analogously to inflectional morphology, she postulates lexical classes for ordered pairs of Latinate stem allomorphs, which are defined over morpho-lexical rules (e.g., Xduce ~ Xduct, Xscribe ~ Xscript, Xmit ~ Xmis etc). In the lexical entry of -ion (and -ive), it is stated that these suffixes combine with the marked member of each pair, i.e. with the one which does not occur independently (e.g., reduct-ion, inscript-ion, permiss-ive). Given this mechanism, the shapes of the suffixes remain invariable. Although Lieber’s approach is morphemebased, it does not imply that morphemes must be meaningful. A morpheme-based, lexicalist approach in the sense of Lieber (1981) is also favoured by Selkirk (1982), Olsen (1986), Dalton-Puffer (1996) and Baeskow (2002). In Selkirk’s model, the notion of subcategorization is extended to include the selectional behaviour of bound roots like -ceive (as in deceive, receive, conceive, etc.) or moll- (as in mollify), whose combination with other morphemes is as obligatory as the addition of an affix to an appropriate base.

3.2. Semantic restrictions in generative word-formation The constraints imposed upon derivational processes by lexicalists affect not only the subcategorial behaviour of affixes, but also the interpretation of their output. In generative word-formation, semantic aspects of derivation are usually expressed in terms of argument structure, thematic relations and mechanisms which co-ordinate the word-internal assignment of thematic information. Although it is widely accepted today that argument structure is associated not only with verbs, but also with nouns, adjectives and prepositions (cf., e.g., Williams 1981; Higginbotham 1985; Rauh 1988; Zwarts 1992; Lieber 2004), the focus of attention has for a long time been on deverbal derivatives in research on word-formation.

809

810

IV. Rules and restrictions in word-formation I: General aspects

3.2.1. Absorption and inheritance In lexicalist approaches (e.g., Olsen 1986: 78; Randall 1988: 143−145) it is assumed that the much discussed suffix -er absorbs the theta-role of the external argument of its verbal base. As a result, this role is no longer assignable outside the derivative. Thus, a phrase like *the builder of the ship by John is ill-formed because the agent role of build is absorbed by -er and simultaneously assigned to the prepositional phrase by John, which leads to a violation of the theta-criterion postulated by Chomsky (1993: 36). According to this restriction on theta-role assignment, “[e]ach argument bears one and only one theta-role, and each theta-role is assigned to one and only one argument”. Formally, the process of absorption may be represented as follows: (4)

a. [[build ] -er ] b. [[build ] -er ]

Apart from the referential argument , which indicates that build denotes an event, (4a) specifies the theta-roles AGENT and THEME, which the verb assigns to its external and internal argument respectively. This representation is referred to as the verb’s thetagrid. Since -er derivatives denote sets of entities in the world (i.e. human beings or objects), this suffix is associated with the referential argument introduced by Williams (1981: 86). In the course of the derivation, the AGENT role of to build is absorbed by -er and hence added to the suffix’s theta-grid, as shown in (4b). The process of absorption (first introduced by Jaeggli 1986 for passive constructions) may be considered a kind of morphological licensing because it ensures the interpretation of derivatives like builder, writer, singer, etc. as agent nouns. Since -er specifically (though of course not exclusively) refers to the base verb’s external argument, Burzio (1986) claims that unaccusative verbs like fall, die or arrive, which only have an internal argument in their argument structure, are excluded from being selected by this suffix. The THEME role of to build is passed on to the derivative builder by inheritance (e.g., Selkirk 1982: 33; Olsen 1986: 78−88; Rappaport Hovav and Levin 1992: 130−131). In contrast to an absorbed role, an inherited role is available for the assignment to a modifier (e.g., shipbuilder) or a syntactic phrase (e.g., a builder of ships). This process, however, is semantically restricted. As noted by Rappaport Hovav and Levin, English instrument nouns do not inherit the internal argument of their base verb, with the result that there are no corresponding phrases introduced by of which convey an instrumental reading. For example, a wiper of windshields does not refer to an instrument. Only an agentive reading would be possible here, but this is blocked by extralinguistic considerations. If wiper is extended to windshield wiper or a wiper for windshields, the noun windshield does not realize the internal argument of to wipe because wiper did not inherit this argument from its base verb and hence cannot assign the THEME role. Thus, the instrumental role for windshield has to be reconstructed via inference. The word-internal assignment of the THEME role is also blocked in the case of instrument nouns like tin opener, hairdryer, vegetable steamer, etc. (but assigned via the inference process just mentioned). Moreover, there is a correlation between inheritance and the

45. Rules, patterns and schemata in word-formation event interpretation of -er nominals. For example, the agentive reading of a phrase like a grinder of imported coffees presupposes that an event of grinding has actually occurred. On the other hand, instrument nouns like grinder or coffee grinder emphasize the purpose of a machine rather than an event in which it is used. Thus, instrument nouns (or non-event nominals) are incompatible with adverbs of frequency such as frequent or constant (e.g., I know that Dan is a frequent waxer of parquet floors vs. *I know that this mop is a frequent floor waxer). As observed by Olsen (1986: 82−83) for German, inheritance is not generally optional. If the derivative generated by -er constitutes a relational noun, which requires complementation, its THEME role has to be saturated word-internally. Examples are *Hemmer, *Vertilger or *Treter, which are well-formed only in synthetic compounds like Appetithemmer ‘appetite suppressant’, Unkrautvertilger ‘herbicide’ or Balltreter ‘football player, kicker’. Although absorption and inheritance successfully restrict the semantic output of regular derivational processes, a problem with these mechanisms is that there is not always a one-to-one relation between a suffix and a particular argument position of the base. For instance, established derivatives like roaster, fryer, broiler or formations like wilter, dyer or fader collected by Ryder (1999) from everyday contexts violate Burzio’s generalization because the suffix -er makes reference to the internal argument of its unaccusative verbal bases. The same problem is observable for the suffix -ee, which typically, but not necessarily absorbs the role of the base verb’s first internal argument, i.e. of the argument which syntactically occupies the position of the direct object. This behaviour is responsible for the passive character of many -ee derivatives (e.g., employee, trainee, visitee). Nevertheless, there are nouns like addressee, experimentee, amputee, escapee or attendee, which cannot be linked to this particular argument position. Moreover, as far as denominal derivatives like potter, Marxist or festschriftee are concerned, there is no thematic argument position for the suffix to correspond with because the nominal bases only have a referential argument in their argument structure. Authors of more recent generative approaches have attempted to solve this matching problem. Barker (1998) postulates an individual theta-role for the suffix -ee, which episodically links the sentient, non-volitional referent of an -ee derivative to the event denoted by or associated with the base. This approach will be dealt with in more detail in article 52 on semantic restrictions on word-formation: the English suffix -ee. In Lieber’s (2004, 2006) lexical semantic approach, the input and output of word-formation processes is defined over sets of well-motivated semantic features. An outline of this approach will be presented in the following section. Like Barker’s proposal, Lieber’s treatment of -ee derivatives is also part of article 52.

3.2.2. The anatomy of lexical items: semantic skeleton and semantic body In lexicalist approaches, lexical entries specify the ensemble of orthographic, phonological, morpho-syntactic and semantic information associated with individual morphological building blocks (words or morphemes). The most delicate matter is the representation of meaning. On the one hand, the referents of many lexical items have highly specific extra-linguistic properties which require complex definitions. Scientific or technical

811

812

IV. Rules and restrictions in word-formation I: General aspects terms are a case in point. Even the definition of the apparently simple item dog goes beyond the listing of properties like ‘has four legs’, ‘has fur’ and ‘barks’ if biological and evolutionary facts are taken into consideration. On the other hand, the profundity of world knowledge differs considerably among the speakers of a language. As pointed out by Moravcsik (1981), a child’s understanding of objects and concepts is definitely incomplete, and so is the knowledge of the layperson in comparison to the knowledge of the expert. However, despite the discrepancy between the infinite complexity of facts, causal relations and events in the world and the relative narrowness of human knowledge, people are able to communicate more or less successfully. Thus, the mental lexicon in the generative sense should not be expected to deal scientifically with the referents of lexical items. Instead, it makes more sense to distinguish between semantic knowledge and world knowledge, the latter of which is also referred to as conceptual or encyclopedic knowledge. This distinction is at the heart of Lieber’s (2004, 2006) lexical semantics model of word-formation. In this model, lexical items are assumed to consist of two components: the semantic skeleton and the semantic body. The skeleton provides elementary lexical information such as an item’s category membership and argument structure. The body consists of substantial encyclopedic information which differs from one speaker to another − as pointed out above. The focus of attention is on the semantic skeleton. Significantly, skeletal properties are defined over cross-categorial semantic features like [+material] (man, chair, book), [−material] (peace, time, love), [+dynamic] (sing, write, teach), [−dynamic] (know, own; tall, intelligent) or [dynamic], the latter of which is a privative nominal feature, i.e. it is present only in nouns which imply activity, like author, mother or effort. These features instantiate the ontological categories SUBSTANCES/THINGS/ESSENCES and SITUATIONS that structure the extra-linguistic reality in a very abstract but universal way. Another feature, namely [+IEPS] “Inferable Eventual Position or State”, adds a PATH component of meaning to the skeleton of verbs denoting a change of position (e.g., descend, fall, go) or change of state (e.g., evaporate, forget, grow). A sample of semantic skeletons, which include placeholders for argument positions, is given in (5). (5)

a. b. c. d. e.

chair mother employ grow happy

[+material, ([ ])] [+material, dynamic ([ ])] [+dynamic ([ ], [ ])] [+dynamic, +IEPS ([ ]) [−dynamic, +scalar ([ ])]

Affixes, like simplex lexical items, are defined by means of semantic features as well. In the following lexical entry postulated by Lieber (2004: 62), the suffix -er is specified for the features [+material] and [dynamic]. The brackets are placeholders for the referential argument : (6)

-er Syntactic subcategorization: attaches to V, N Skeleton: [+material, dynamic, ([ ], )]

An essential ingredient of Lieber’s lexical semantics is the principle of co-indexation (2004: 61), which accounts for the interpretation of derivatives independently of the category of their input.

45. Rules, patterns and schemata in word-formation (7)

Principle of co-indexation In a configuration in which semantic skeletons are composed, co-index the highest nonhead argument with the highest (preferably unindexed) head argument. Indexing must be consistent with semantic conditions on the head argument, if any.

Applied to suffixation, this principle states that the highest argument of the base is coindexed with the highest argument of the head-forming suffix provided that these arguments are semantically compatible. The resulting derivative absorbs whatever thematic interpretation the base argument has. In the case of a deverbal -er derivative like driver, the external argument of drive is co-indexed with the referential argument of the suffix, and the derivative assumes an agentive interpretation. (8)

driver [+material, dynamic ([i ], [+dynamic]([i ], [ ])])] -er drive

If the base of an -er derivative is an unaccusative verb (e.g., sink), the principle of coindexation links the verb’s internal argument to the referential argument of the suffix. Since unaccusative verbs lack an external argument, the internal argument is the highest one. (9)

sinker [+material, dynamic ([i ], [+dynamic, +IEPS ([i ])])] -er sink

As suggested by (8) and (9), -er does not impose a semantic restriction on the non-head argument it is co-indexed with. In the case of -ee, the situation is different (cf. article 52 on the English suffix -ee). Subcategorization frames of a morpho-syntactic design gradually lose their significance in Lieber’s works on lexical semantics. The reason is that selection on the basis of syntactic category (c-selection) is replaced by semantic selection (s-selection), according to which affixes are sensitive to the semantic categories of their potential bases (cf. Lieber 2006: 266−267).

3.3. From affix ordering to lexical phonology The compatibility of affixes with bases is determined not only by the individual selectional behaviour of affixes and semantic mechanisms, but also by more general constraints, which have an impact on the structure of the lexicon.

3.3.1. Affix classes As first observed by Siegel (1974), there are several aspects which suggest that the English affix inventory falls into two classes − class I affixes (10) and class II affixes (11):

813

814

IV. Rules and restrictions in word-formation I: General aspects (10) in+, con+, de+, sub+, dis+; +ity, +ion, +ate, +al, +or, etc. (11) un#, be#, non#, fore#; #ness, #hood, #less, #ful, #en, etc. The distinction between these affix classes was basically a phonological one. The morpheme-boundary “+” and the word-boundary “#” were introduced by Chomsky and Halle (1968) in order to restrict the application of phonological rules. Since these rules operate across morpheme-boundaries, but not across word-boundaries, only class I affixes are able to change the phonological properties of their bases. Moreover, class I affixes themselves may be subject to phonological processes. The negative prefix in- is a case in point (e.g., inactive, impossible, irrelevant, illegal). On the other hand, class II affixes are phonologically neutral. From a morphological point of view, class I affixes combine with words (e.g., parent-al, modern-ity, de-limit) and with bound roots (e.g., astr-al, vivac-ity, de-duce), whereas class II affixes predominantly occur in the context of words (e.g., happi-ness, child-hood, un-kind). As a result, the former are nearer to the root than the latter in words which contain affixes of both classes (e.g., credul-ous-ness, romantic-ism). Moreover, while derivatives formed by class II affixes are semantically transparent, the semantics of derivatives displaying class I affixes is not always predictable. For example, as observed by Riddle (1985), derivatives ending in -ness (ethnicness, Africanness, pinkness, etc.) ascribe an embodied trait to the individuals they are predicted of, whereas -ity derivatives mainly denote abstract entities, i.e. names of concepts, situations, and of characteristics in the generic sense (ethnicity, hyperactivity, sectility). Sporadically, however, -ity derivatives also refer to concrete entities, e.g., curiosity in the sense of ‘an object of interest; any object valued as curious, rare, or strange’, cavity ‘a hollow place; a void or empty space within a solid body’ or oddity ‘an odd or peculiar person’, ‘an odd, peculiar, or grotesque thing’ (Oxford English Dictionary). The “concrete entity” reading is idiosyncratic and has to be specified in the lexicon. The most influential observation formulated by Siegel (1974) is that class I affixation precedes class II affixation in English. This ordering hypothesis, which was extended by Allen (1978) to include compounding, served as a basis for Kiparsky’s (1982) lexical phonology, in which a close interaction of morphological and phonological processes is assumed. The lexicon consists of three levels (or “strata”): Level 1 accommodates class I affixation (divine → divin+ity) as well as irregular inflection (keep ~ kept, index ~ indices). At level 2, class II affixation and compounding take place (sahib → sahib#hood; house#boat). Level 3 is reserved for regular inflection (dance ~ danced; book ~ books). Significantly, the output of the morphological processes taking place at a particular level is subject to a set of phonological rules of the same level. Since the rules of lexical phonology (e.g., trisyllabic shortening, velar softening) interact with morphological processes at each level of the lexicon, they are intrinsically cyclic and thus differ from rules of postlexical phonology (e.g., coalescence, insertion of intrusive -r). According to the bracket erasure convention (BEC), the internal structure of words generated at a particular level is invisible to subsequent levels. A hallmark of lexical phonology is the blocking effect. For instance, both the agent noun assist+ant and the irregular plural form oxen are generated at level 1, where they are lexically specified as [assistant]N [Agent] and [oxen]N [Plural] respectively. The BEC and the lexical information that assistant is an agent noun prevent the verb assist from being combined with the productive agent-noun forming suffix -er at level 2 (*assister). Likewise, the formation

45. Rules, patterns and schemata in word-formation of *oxes is blocked at level 3 because the plural form oxen was idiosyncratically formed at level 1. The non-application of a general rule, which is due to the application of a specific rule on an earlier level, is referred to as the “elsewhere condition”.

3.3.2. Problems with lexical phonology and alternative models Although lexical phonology was quite influential especially in the eighties of the last century and beyond (cf. Rubach 1984; Mohanan 1986; Booij 1997; McMahon 2000), it has been subject to severe criticism. As shown convincingly by Giegerich (1999), one major problem is that there are quite a few English suffixes which display dual class membership because their morphological or phonological behaviour enables them to select bases at level 1 and level 2. Idiosyncrasy resulting in dual class membership is particularly striking in the following contexts identified by Giegerich: − an affix which predominantly selects free morphemes sporadically occurs in the context of bound morphemes (e.g., gorm-less, wist-ful, scrib-er, astrolog-er, un-couth) − the phonological neutrality of a suffix is overridden (e.g., compárable vs. cómparable, wild [waɪld] → wilderness [wɪldənəs]) − an affix assigned to class I gains productivity and semantic transparency in certain jargons (e.g., -ant in chemo-technical/medical terms like depressant, digestant, propellant, coolant or in legal agent nouns such as defendant, consultant, complainant, claimant) − an affix which qualifies for class II membership precedes a class I affix (e.g., govern#ment+al, un#grammatical+ity). Obviously, it is problematic to generalize over the morphological, phonological and semantic properties of affixes and to aim at a neat division of the affix inventory, which eventually determines the structure of the lexicon. Since there is too much idiosyncrasy involved, the lexical properties of affixes do not constitute a reliable criterion for level ordering. Nevertheless, it cannot be denied that lexical phonology provided valuable insights into the complexity of morphological structure and the proceedings at the morphology-phonology interface. Moreover, numerous alternatives have been proposed which preserve the basic insights of this theory. Apart from revealing the shortcomings of affix-driven stratification, Giegerich (1999) develops an alternative which is intended to overcome the difficulties described above. On the basis of the observation formulated by Selkirk (1982: 98−99) that English morphology distinguishes between the categories “root” and “word”, Giegerich proposes a theory of base-driven stratification, according to which the levels (or strata) of the lexicon are no longer determined by affixes, but by their potential input. Significantly, the problem of dual membership, which blurred the distinction between class I and class II affixes in the earlier models of lexical phonology, does not arise in this approach. While the number and nature of lexical levels is fixed in a language (the German lexicon, for instance, consists of three strata which are determined by the categories “root”, “stem” and “word”, cf. Wiese 1996: 129), affixes may combine with their bases on more than one stratum. Influential works which entirely dispense with level ordering have been presented by Plag, who discovered a significant correlation between selectional restrictions and stack-

815

816

IV. Rules and restrictions in word-formation I: General aspects ing. For example, Plag (1999: 67−69) shows that the incompatibility of nominalizing suffixes like -age, -al, -ance, -ment or -y with verbs ending in -ize, -ify or -ate (e.g., *magnify-ation, *verbalize-al, *concentrate-ment) is not due to a selectional restriction which these suffixes impose on their bases, but follows from a base-driven selectional restriction: Complex bases ending in -ize, -ify or -ate typically select the suffix allomorphs -ation (verbalization), -cation (identification) and -ion (concentration) of the morpheme -ation and thus block the attachment of other nominalizers. An attractive phonological alternative to level ordering is offered by Raffelsiefen (1999), who claims that suffixes differ with respect to whether or not they form a phonological word (pword) with their base, where the pword is defined as the domain for syllabification. The idea behind this assumption is that only suffixes which trigger phonological rules, i.e. the former class I suffixes, are integrated into the pword of the base they operate on, e.g., (medícinal)ω. On the other hand, phonologically neutral suffixes, i.e. the former class II suffixes, appear outside the pword of their base, e.g., ((áccurate)ω ness)ω. This approach accounts for the observation that so-called stem-affixes are generally closer to their bases than word-affixes. Referring to an observation formulated already by Booij (1985), Raffelsiefen assumes that in English the suffixes which fuse with their bases usually begin with vowels, whereas consonant-initial suffixes are phonologically neutral. This is what she refers to as the “law of initials”. However, since there are some vowel-initial suffixes which are not integrated into the pword of their bases − examples are provided by ((vínegar)ω ish)ω or ((ínjur)ω able)ω − Raffelsiefen concludes that “[n]ot all, but only vowel-initial suffixes can induce stress shift” (1999: 229). The numerous attempts which have been made to preserve and reformulate the achievements of lexical phonology reveal the impact which this model has on contemporary morphology. The fruitful discussions triggered by this theory, which was doomed to extinction, will most probably serve as an input to further research.

4. Schemata in word-formation (an outline) In cognitive approaches to word-formation (e.g., Ungerer 2002; Ryder 1999; Panther and Thornburg 2002), grammatical information associated with the input and output of derivational processes is of secondary relevance. In these models, the compatibility of affixes with bases is primarily described by means of schemata, i.e. mental representations of the knowledge which human beings share about objects and events in the world. Schemata, which were first introduced by Sir Frederic Bartlett (1997 [1932]), can be activated any time and help to process new information, impressions and situations on the basis of former experience. In cognitive grammar, each linguistic building block is conceived of as a bipolar symbolic unit, consisting of a phonological and a semantic structure. Since cognitive grammar does not differentiate between linguistic and extralinguistic information, the knowledge associated with an entity is a continuum ranging from salient to marginal properties. Importantly, abstract units function as schemata for specific units. Complex words such as the -er nominalization pencil sharpener discussed by Langacker (1988: 19−20) form constructional schemata in which highly abstract symbolic units like [THING/X], [PROCESS/Y] and [ER/-er] are instantiated by specific symbolic units.

45. Rules, patterns and schemata in word-formation The shift of emphasis from grammatical information to world-knowledge enables adherents of cognitive approaches to describe a wide range of regular and idiosyncratic word-formation processes. As indicated already in section 1, the input of the productive suffix -er is not restricted to verbs. Evidence comes from derivatives like potter, stranger, upper, fiver, empty nester, etc., which weaken the classical generative assumption that a suffix is subcategorized for bases belonging to a particular category. Moreover, although a high proportion of -er derivatives constitute agent nouns, a variety of non-agentive readings are available. Apart from denoting human beings (e.g., singer, potter, left-hander), nouns ending in -er may refer to animals (e.g., retriever, pointer, sitter), plants (creeper, (late) bloomer, bedder), physical objects, especially instruments (e.g., toaster, receiver, tin-opener), articles of clothing (e.g., slipper, jumper, rompers), locations (e.g., sleeper, diner, kneeler) and events (e.g., no-brainer, thriller, laugher). Even a single -er derivative like birthdayer may be highly polysemous. According to Ryder (1999: 284), this form is associated with the meaning components ‘person having the birthday, person giving the party, person attending the party, present given’ or ‘birthday cake’. As far as their ambiguity is concerned, -er derivatives are comparable to noun-noun compounds, which Ryder (1994) describes on the basis of semantic information schemas. A noun-noun compound like garage man evokes multiple event schemas in which the referents of the nominal constituents participate. Thus, a garage man may be ‘a man who works in a professional garage, builds garages, hangs around in garages, is shaped like a garage’, etc. The most plausible interpretation is retrievable either from the context or from world knowledge. As far as derivatives are concerned, the interpretation is complicated by the fact that the referent of the head constituent, i.e. of the suffix -er, is indeterminate. Referent ambiguity is observable even for deverbal -er derivatives if they are presented context-independently. For example, the form smasher may me associated with a variety of meanings. The verbal base smash evokes an event schema which involves a number of participants. Prototypically, there is an agent who performs the activity of smashing (e.g., George) and a patient, i.e. an entity which is affected by the destructive activity (e.g., the rock). Optionally, the agent may use an instrument to smash the object (e.g., a sledge-hammer), and there may even be a benefactive for whom he does the smashing (e.g., George’s wife). Moreover, since events are located in space and time, the smashing event may have occurred in the backyard yesterday evening. In principle, the noun smasher could refer to the agent, the patient, the instrument, the benefactive or the place or time of the smashing event. However, according to Ryder (1999: 285), there are two conceptual factors which restrict the reference of such a derivative, namely salience and identifiability. Salience refers to the degree to which something is noticeable in comparison with its surroundings. As far as the smashing event is concerned, the agent, the patient and the instrument are more salient than the benefactive and the spatial or temporal location. Identifiability refers to the extent to which a participant is readily identifiable by mention of the event alone. Although the patient is one of the salient entities in the smashing event, it is unlikely to be identified as the referent of smasher because patients are identifiable only in very specific contexts (e.g., roaster and broiler are identifiable as patients of different cooking events). Likewise, smasher is unlikely to receive an instrument reading because the smashing event (unlike, for instance, the event of putting nails

817

818

IV. Rules and restrictions in word-formation I: General aspects into a wall or the event of cutting a piece of wood) is not associated with a particular tool. Thus, the agent is the participant most readily identifiable as the referent of smasher. Panther and Thornburg, two further representatives of cognitive word-formation, consider the suffix -er to be a polysemous symbolic unit with the central sense “a human Agent who performs an action or engages in an activity to the degree that doing so defines a primary occupation” (2002: 285), to which all other -er nominals are related. Given the central sense of the suffix, -er derivatives evoke a prototypical transitive scenario, which, like Ryder’s event schema, constitutes a knowledge representation. The prototypical transitive scenario has a setting (i.e. a place and a time in which an event takes place) and two distinct participants that are in an asymmetrical interaction. One participant is an intentionally acting human, whereas the other is directly affected/effected by the action. Significantly, the parameters of this multidimensional model are scalar, so that the scenario can be extended or reduced in various ways. Representative derivatives like teacher, baker, brewer, steel-worker, etc. range high on the transitivity scale because they fully correspond to the idealised cognitive model of human activities. The verbal bases serve as a reference point in that they allow mental access to other components of the respective scenario. This idea is the starting-point for Panther and Thornburg’s argumentation that the denotatum of a non-verbal base serves as a reference point from which the activity performed by the agent is accessed either metonymically (e.g., Wall Streeter ‘person professionally employed on Wall Street’) or via a combination of metonymy and metaphor (e.g., hoofer ‘professional (vaudeville/chorus) dancer’, upper ‘anti-depressant pill’). There are, however, a number of idiosyncrasies in derivation which are not accounted for by schemata alone. For instance, although Panther and Thornburg attempt to dispense with morpho-syntactic information, they correctly state (in a non-cognitive way) that the suffix -ist shows a preference for non-native bases (e.g., nihilist, analyst). Moreover, they observe that agent nouns ending in -ant or -ent are mainly derived from intransitive verbs (e.g., emigrant, convalescent), whereas instrument and patient nouns in -ant/-ent are restricted to transitive verbs (e.g., defoliant, ingestant). Observations like these suggest that word-formation is not entirely reducible to conceptual information − although this kind of information certainly plays an important role. The rules, patterns and schemata presented in this article, which are achievements from decades of morphological research, rather suggest that word-formation processes are determined by the interaction of morpho-syntactic, phonological and conceptual-semantic information.

Acknowledgement I would like to thank the Deutsche Forschungsgemeinschaft (DFG), whose support of my project on word-formation enabled me to write this article.

5. References Allen, Margaret R. 1978 Morphological investigations. Ph.D. dissertation, University of Connecticut. Aronoff, Mark 1976 Word Formation in Generative Grammar. Cambridge, MA: MIT Press.

45. Rules, patterns and schemata in word-formation Baeskow, Heike 2002 Abgeleitete Personenbezeichnungen im Deutschen und Englischen. Kontrastive Wortbildungsanalysen im Rahmen des Minimalistischen Programms und unter Berücksichtigung sprachhistorischer Aspekte. Berlin/New York: de Gruyter. Barker, Chris 1998 Episodic -ee in English: A thematic role constraint on new word formation. Language 74(4): 695−727. Bartlett, Frederic Charles 1997 [1932] Remembering. A Study in Experimental and Social Psychology. Cambridge: Cambridge University Press. Bauer, Laurie, Rochelle Lieber and Ingo Plag 2013 The Oxford Reference Guide to English Morphology. Oxford: Oxford University Press. Bloomfield, Leonard 1933 Language. New York: Hold, Rinehart and Winston. Bochner, Harry 1993 Simplicity in Generative Morphology. Berlin/New York: Mouton de Gruyter. Booij, Geert 1985 Coordination reduction in complex words: A case for prosodic phonology. In: Harry van der Hulst and Norval Smith (eds.), Advances in Nonlinear Phonology, 143−159. Dordrecht: Foris. Booij, Geert 1997 Allomorphy and the autonomy of morphology. Folia Linguistica 31(1−2): 25−56. Burzio, Luigi 1986 Italian Syntax. A Government-Binding Approach. Dordrecht: Reidel. Chomsky, Noam 1957 Syntactic Structures. The Hague: Mouton. Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, Noam 1970 Remarks on nominalization. In: Roderick A. Jacobs and Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar, 184−221. Waltham, MA: Blaisdell. Chomsky, Noam 1993 Lectures on Government and Binding. The Pisa Lectures. Reprint 7th ed. 1st ed. 1981. Berlin/New York: Mouton de Gruyter. Chomsky, Noam and Morris Halle 1968 The Sound Pattern of English. New York: Harper and Row. Dalton-Puffer, Christiane 1996 The French Influence on Middle English Morphology. A Corpus-based Study of Derivation. Berlin/New York: Mouton de Gruyter. Giegerich, Heinz 1999 Lexical Strata in English. Morphological Causes, Phonological Effects. Cambridge: Cambridge University Press. Halle, Morris 1973 Prolegomena to a theory of word-formation. Linguistic Inquiry 4: 3−36. Henne, Helmut and Jörg Kilian (eds.) 1998 Hermann Paul. Sprachtheorie, Sprachgeschichte, Philologie. Reden, Abhandlungen und Biographie. Tübingen: Niemeyer. Higginbotham, James 1985 On semantics. Linguistic Inquiry 16: 547−593. Jackendoff, Ray 1975 Morphological and semantic regularities in the lexicon. Language 51: 639−671.

819

820

IV. Rules and restrictions in word-formation I: General aspects Jaeggli, Oswaldo 1986 Passive. Linguistic Inquiry 17: 587−622. Jespersen, Otto 1974 [1942] A Modern English Grammar on Historical Principles. Part VI: Morphology. London: Allen & Unwin. Kiparsky, Paul 1982 From cyclic phonology to lexical phonology. In: Harry van der Hulst and Norval Smith (eds.), The Structure of Phonological Representations (Part I), 131−175. Dordrecht: Foris. Koziol, Herbert 1971 [1937] Handbuch der englischen Wortbildungslehre. Heidelberg: Winter. Lakoff, George 1970 Irregularity in Syntax. New York: Holt, Rinehart and Winston. Langacker, Ronald W. 1988 An overview of cognitive grammar. In: Brygida Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics, 3−47. Amsterdam/Philadelphia: Benjamins. Lees, Robert B. 1960 The Grammar of English Nominalizations. The Hague: Mouton. Lehnert, Martin 1971 Rückläufiges Wörterbuch der englischen Gegenwartssprache. Leipzig: VEB Verlag Enzyklopädie. Lieber, Rochelle 1981 On the organization of the lexicon. Bloomington, IN: Indiana University Linguistics Club. Lieber, Rochelle 2004 Morphology and Lexical Semantics. Cambridge: Cambridge University Press. Lieber, Rochelle 2006 The category of roots and the roots of categories: What we learn from selection in derivation. Morphology 16: 247−272. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. McMahon, April 2000 Lexical Phonology and the History of English. Cambridge: Cambridge University Press. Mohanan, Karuvannur P. 1986 The Theory of Lexical Phonology. Dordrecht: Reidel. Moravcsik, Julius M. 1981 How do words get their meanings? Journal of Philosophy 78: 5−24. Motsch, Wolfgang 2004 Deutsche Wortbildung in Grundzügen. 2nd ed. 1st ed. 1999. Berlin/New York: de Gruyter. Muthmann, Gustav 2002 Reverse English Dictionary. Based on Phonological and Morphological Principles. Berlin/New York: Mouton de Gruyter. Olsen, Susan 1986 Wortbildung im Deutschen. Stuttgart: Kröner. OneLook Reverse Dictionary online http://www.onelook.com/reverse-dictionary.shtml [last access 24 Aug 2010]. Oxford English Dictionary online http://dictionary.oed.com [last access 24 Aug 2010]. Panther, Klaus-Uwe and Linda Thornburg 2002 The roles of metaphor and metonymy in English -er nominals. In: René Dirven and Ralf Pörings (eds.), Metaphor and Metonymy in Comparison and Contrast, 279−319. Berlin/ New York: Mouton de Gruyter.

45. Rules, patterns and schemata in word-formation Plag, Ingo 1999 Morphological Productivity. Structural Constraints in English Derivation. Berlin/New York: Mouton de Gruyter. Raffelsiefen, Renate 1999 Phonological constraints on English word formation. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1998, 225−287. Dordrecht: Kluwer. Randall, Janet 1988 Inheritance. In: Wendy Wilkins (ed.), Syntax and Semantics 21. Thematic Relations, 129−146. San Diego: Academic Press. Rappaport Hovav, Malka and Beth Levin 1992 -Er nominals: Implications for the theory of argument structure. In: Tim Stowell and Eric Wehrli (eds.), Syntax and Semantics 26. Syntax and the Lexicon, 127−153. San Diego: Academic Press. Rauh, Gisa 1988 Tiefenkasus, thematische Relationen, Thetarollen. Die Entwicklung einer Theorie von semantischen Relationen. Tübingen: Narr. Riddle, Elizabeth M. 1985 A historical perspective on the productivity of the suffixes -ness and -ity. In: Jacek Fisiak (ed.), Historical Semantics Historical Word-Formation, 435−461. Berlin: Mouton. Rubach, Jerzy 1984 Cyclic and Lexical Phonology. Dordrecht: Foris. Ryder, Mary Ellen 1994 Ordered Chaos. The Interpretation of English Noun-Noun Compounds. Berkeley: University of California Press. Ryder, Mary Ellen 1999 Bankers and blue-chippers: An account of -er formation in present-day English. English Language and Linguistics 3(2): 269−297. Saussure, Ferdinand de 2005 [1916] Cours de linguistique générale. Paris: Payot. Selkirk, Elizabeth 1982 The Syntax of Words. Cambridge, MA: MIT Press. Siegel, Dorothy 1974 Topics in English morphology. Ph.D. dissertation, MIT. Ungerer, Friedrich 2002 The conceptual function of derivational word-formation in English. Anglia 120(4): 534− 567. Wiese, Richard 1996 Phonology of German. Oxford: Clarendon Press. Williams, Edwin 1981 Argument structure and morphology. Linguistic Review 1: 81−114. Zimmer, Karl 1964 Affixal Negation in English and Other Languages. An Investigation of Restricted Productivity. Supplement to Word 20, Monograph 5. New York: Clowes. Zwarts, Jost 1992 X'-Syntax − X'-Semantics. On the Interpretation of Functional and Lexical Heads. Utrecht: OTS Dissertation Series.

Heike Baeskow, Wuppertal (Germany)

821

822

IV. Rules and restrictions in word-formation I: General aspects

46. Word-formation and analogy 1. 2. 3. 4. 5. 6. 7.

Introduction Definition and terminology Senses of analogy in word-formation theory Complementary analogy Analogy in analogy-based approaches to word-formation Summary and conclusion References

Abstract The article discusses the much-debated status of analogy in contemporary theories of synchronic word-formation. It provides an overview of the key assumptions made in pertinent theoretical camps as well as of the major phenomena that have featured prominently as evidence in the debate. Theories can broadly be classified into those which assume that analogy is active only as a complementary mechanism, and those which assume that analogy is the central mechanism of productive word-formation. Among the latter, we can distinguish between general theoretical and computational analogical models. Based on a detailed definition and discussion of the analogical equation, different usages of analogy in the literature are shown to be closely tied to theory-dependent conceptualisations of productivity, predictability, and (ir-)regularity.

1. Introduction The term analogy is used in many different senses and in many different contexts within morphological theory. One subdiscipline with which analogy is frequently associated is diachronic morphology, where, ever since the Neogrammarian revolution, analogical change has been seen as a central mechanism of morphological change (cf., e.g., Hock 2003 for a summary). Developments which have come to be associated with analogy are, especially, analogical extension and levelling. In synchronic morphological theory, the type of analogy that is the subject of discussion is proportional analogy, i.e. a heuristic mechanism in which a new complex word is formed on the basis of a perceived similarity with existing base-derivative pairs (cf. section 2 below for a more detailed definition). The central question that has been debated in the contemporary literature is whether analogy is an active mechanism in synchronic morphology, and, if it is, what its relation is to other mechanisms in synchronic morphology, such as rules or constraints or schemas. Much of this debate, which has its origins far back in the history of morphological research (cf., e.g., Becker 1990; Anttila 2003 for a summary), has taken place in inflectional morphology. In this debate, the term analogy has come to be used in different senses, which are often heavily dependent on the underlying theory. In particular, senses of analogy are often tied to particular assumptions about central theoretical notions such as regularity, productivity, variability, and the nature of lexical representations.

46. Word-formation and analogy Thus, we find approaches that claim that analogy is the basis of any rule-based, productive behaviour in morphology (cf., e.g., Blevins and Blevins 2009b for an overview), whereas at the same time we find, especially generative, approaches that appeal to analogy exactly in those cases in which linguistic behaviour is not rule-governed, but exceptional, unproductive, unpredictable, or irregular (cf., e.g., Prasada and Pinker 1993; Pinker and Prince 1994). Also within word-formation theory, analogy has come to be used as a term opposite to the concept of the linguistic rule (cf. Bauer 1983, 2001). In usage-based and constructionist approaches, by contrast, it is argued that analogy forms the underlying principle of exemplar-based reasoning or the beginnings of low-level schematisation (Booij 2010: 88–93). Crucially, the implication in this latter group of approaches is that analogies in word-formation are regularly based on subsymbolic aspects of lexical representations. However, this assumption is not inherent in the definition of analogy per se, as we will see. This article is concerned with concepts of analogy in synchronic word-formation. The focus is on providing an overview of the different notions of analogy as they are used in different theories of word-formation, and of the different phenomena that have featured prominently in references to analogy in the word-formation literature. Reference to parallel developments in theories of inflection will be made occasionally, where necessary. The structure of the article is as follows: We will begin with a general definition of proportional analogy as a heuristic device (section 2). Section 3 will then provide an overview of the status of analogy in different theories of word-formation. We will see how different theories operationalise different aspects of the structure of proportional analogy in different ways, resulting in radically different views about the regularity, productivity, and predictability of analogical formations. Based on this overview, theories will be grouped broadly into a) theories that consider analogy to be an irregular or exceptional process and b) those that consider analogy to be the basic process underlying word-formation. Sections 4 and 5 will then be devoted to these two classes of theories, respectively, discussing pertinent word-formation phenomena for which analogy has been invoked. The discussion in section 5 will specifically focus on computational analogical models, which will be shown to provide interesting solutions to some of the criticism that has traditionally been mounted against analogy-based models, but will also be shown to be limited in terms of the range of processes covered in such approaches to date. The article ends with a conclusion (section 6).

2. Definition and terminology Analogy as used in the word-formation literature is usually described in terms of a proportion (proportional analogy), as in (1). (1)

a:b=c:x

In this equation, ‘x’ is the new form, i.e. a morphologically complex word that is about to be coined. ‘a’, ‘b’, and ‘c’ are forms that already exist in the lexicon. ‘a’ and ‘c’ are (potential) base forms, whereas ‘b’ is an existing complex form. What happens in an analogical formation, then, is that the relationship between ‘a’ and ‘b’ is used as a model example for the formation of ‘x’.

823

824

IV. Rules and restrictions in word-formation I: General aspects As an example, consider the English compound chairperson, which we may plausibly assume to have been formed on the basis of analogy with the existing compound chairman. In this case, we can fill the variables in the equation above as in (2). (2)

chair : chairman = chair : chairperson

Interestingly, there is no established terminology in the morphological literature to refer to most parts of the analogical equation. The only established term seems to be analogue, which is usually used to refer to the complex form on which a newly coined word is modelled (‘b’ in (1), chairman in (2)). In order to facilitate further discussion, however, it is useful to have labels to refer to the other parts of the equation as well. These labels are presented in Figure 46.1, again using chairperson as an example.

Fig. 46.1: Elements of the analogical equation

In accordance with most of the literature, we will use the term analogue to refer to the complex word that serves as a model for the coining of a new complex word. The new word that is about to be coined will be referred to as the new word. Finally, there are two bases involved in an analogy, for which we will use the terms base of the analogue and base of the new word, respectively. As is clear from the example, a key role in the process of analogy is played by the (perceived) similarity between the elements of the equation. Figure 46.2 provides an overview of (and labels for) the similarity relations we find within the analogical equation.

Fig. 46.2: Similarity relations in the analogical equation

46. Word-formation and analogy In the example chairperson the base of the analogue and the base of the new word are identical (both chair). Hence, it is easy to see that they may be perceived as being similar to each other by the hypothetical speaker(s) who coined chairperson. The second aspect where similarity plays a role is the relation between the base of the analogue (chair) and the analogue (chairman), which must be perceived as being similar to the relation between the base of the new word (chair) and the new word (chairperson). In chairperson, the relation between the analogue and its base is a morphological relation, pertaining to both the form and the meaning of the two lexemes. The form of the base, chair, appears as the first constituent of the compound chairman. Semantically, we could broadly say that the relation between the analogue and its base is that between a role (chair) and the occupant of that role, who is human and male (chairman). Like the relation between chair and chairman, the relation between chair and chairperson (i.e. that between the new form and its base) also pertains to both the form and the meaning of the elements involved. Again, chair appears as the first constituent of a compound. Semantically, the relation between the new word and its base is that between the role (chair) and the occupant of that role, who is human but, crucially, not necessarily male. Our example already indicates that similarity is not only a key determinant of a morphological analogy, but that it is also one of the key problems in defining and explaining analogies, and, on a theoretical level, a key challenge for any morphological theory that is based on analogies. The reason is that the basis for the computation of similarity is not part of the equation. Whereas the word-formation literature generally agrees that in cases of analogy similarity must be given both in terms of form and in terms of meaning, there is almost no restriction on precisely which formal and semantic properties can make an analogue and a potentially corresponding new word similar. In addition, there is no agreement about how analogical similarity relations (those schematised in Fig. 46.2) map onto morphological complexity relations. For example, when discussing chairperson as a product of analogical reasoning on the basis of chairman, we assumed that the relation between chair and chairman was the base for the analogy. At the same time, however, it is also clear that chairman is related to man as much as it is related to chair, and that the relation between man and chairman is similar to the relation between person and chairperson. Thus, Fig. 46.3 below is another plausible representation of an analogical relation between chairman and chairperson.

Fig. 46.3: chairman − chairperson − an alternative

825

826

IV. Rules and restrictions in word-formation I: General aspects What this tells us is that it is not always clear what exactly the base of an analogy is. There is, from a theoretical point of view, no restriction on which of the multiple similarity relations that exist between words in the mental lexicon may form the basis of an analogical formation. In sum, we see that analogy as a heuristic formalism does not say much about many of the issues that morphological theory needs to be explicit about. Specifically, it does not say anything about a) which features (formal, semantic, syntactic, etc.) establish similarity relations on which analogies may be based, and b) which of the existing similarity relations may or will form the basis of a new analogical formation. Furthermore, as we will see later in this article, it does not say anything about c) how many lexemes are involved in an analogy. This explains, in part, the great diversity of usages of the term analogy in the literature, to which we will now turn.

3. Senses of analogy in word-formation theory The focus in this section will be on showing the scope of senses in which analogy has been used in the literature on synchronic word-formation. Rather than attempting to be exhaustive, the discussion will be restricted to a sample of representative theories. Analogy is often discussed in the context of the theoretical divide between wordbased, paradigmatic, and syntagmatic approaches to morphology (cf., e.g., Becker 1990, 1993a and references therein for discussion). It is clear that approaches to word-formation which attribute a systematic role to analogy are all word-based approaches. However, not every word-based approach assumes that analogy is an active mechanism in productive word-formation. This is true in spite of claims often made in the literature that the mechanisms underlying word-based formalisms can be described in terms of a proportional analogy. Word-based approaches are divided in terms of whether they consider rules or analogy to be the central mechanism in productive word-formation. In rule-based approaches of this type, analogy is often invoked to explain irregular, or unproductive behaviour. In analogy-based approaches, analogy is invoked to explain regular, productive behaviour. The term paradigmatic approaches to morphology is difficult to apply here, because it is used in different senses in the literature. Whereas it is assumed to be synonymous with ‘word-based’ approaches by some authors, others use it rather in the sense that has been labelled ‘analogy-based’ above. In what follows we will discuss senses of analogy in word-based theories. One type of approach that takes an extreme position with respect to the rule-analogy divide is comprised of, mostly generative, paradigmatic frameworks which make a radical distinction between analogical formations and regular processes of word-formation. Regular processes are the product of an abstract formalism that operates independently from individual lexemes, on symbolic features that are shared by pertinent lexemes (cf., e.g., Aronoff 1976). Analogy, in this view, is always local in the sense that it affects only few and very specific lexical items. Productivity is rule-application, with the consequence that a low degree of variation is predicted for morphological rules. Unlike regular and productive word-formation, then, processes of analogy are unpredictable and unproductive. A clear expression of this view, which is found frequently in the generative

46. Word-formation and analogy literature, is found, for example, in Bauer’s (1983) textbook on English word-formation (but cf. also Bauer 2001: 75–97, where this view has been relativised considerably): If instances of word-formation arise by analogy then there is in principle no regularity involved, and each new word is produced without reference to generalizations provided by sets of other words with similar bases or the same affixes: a single existing word can provide a pattern, but there is no generalization. [...] If it is true that there are in principle no generalizations, then a generative account of word-formation is at best a convenient fiction and at worst an irrelevancy. (Bauer 1983: 294)

At the other end of the divide we find approaches that assume that analogy with existing lexemes is used regularly and productively in the formation of new words. In what follows I will refer to such approaches as “analogical” approaches. Conceptually, they are rooted mainly in two traditions: One is the Neogrammarian and the American descriptivist tradition (cf. esp. Becker 1990 for details and discussion), the other comprises functionalist, exemplar-based, and usage-based theories of grammar, i.e. theories that are grounded in the assumption that analogy constitutes a central cognitive mechanism that is active in human cognition in general, and in language in particular (cf. esp. Bybee 2001, 2010; Gahl and Yu 2006; Blevins and Blevins 2009a). Work in both types of theories has traditionally focussed on inflection and on diachronic language development. Recent times, however, have seen a growing number of publications devoted to word-formation phenomena. For example, Becker (1990, 1993a, 1993b) proposes for both inflection and derivation that all morphological operations are analogical, in the sense that they describe relations between existing words, on the basis of which speakers productively coin new words by means of proportional analogy (e.g., Becker 1990: 187). His proposal builds on and extends those of especially van Marle (1985, 1990), who also claims a synchronic relevance for analogical formation in word-formation, but distinguishes between (rulebased) productivity and (analogy-based) creativity, the former producing regularity, and the latter allowing for some degree of unpredictability. One obvious characteristic of approaches that do not embrace the distinction between analogical and rule-based word-formation is that they consider variability and gradience to be a key property of morphological operations. This variability has often been identified with unpredictability, and the failure of many analogical approaches to be predictive constitutes one of the key points of criticism against these approaches (cf. above and Bauer 2001: 75–97 for a summary of pertinent arguments − and counterarguments). However, there is also a growing body of literature springing from mainly quantitative work in morphology that challenges the view that analogical models must necessarily be non-predictive (cf., e.g., Baayen 2003; Hay and Baayen 2005). The main argument is that variability in general and different degrees of variability in particular can be predicted in a probabilistic approach to linguistic categorisation. One major class of such probabilistic approaches explicitly draws on analogy as the fundamental underlying principle of morphology and has produced a growing number of studies that model word-formation phenomena with the help of computational implementations of analogical models of grammar (specifically: AM(L), Skousen 1989, 1992; Skousen, Lonsdale and Parkinson 2002; TiMBL, Daelemans et al. 1999 ff.). What these models have in common is that they apply similarity-based, analogical reasoning, creating new forms on the basis of the similarity of the base of the new form with existing forms in the

827

828

IV. Rules and restrictions in word-formation I: General aspects lexicon. Analogies are therefore very rarely local; the idea advanced in many rule-based approaches that analogies must be based on a single lexical item is conceived to be only one of many possibilities. Much more frequently analogues in such approaches are sets of words in the lexicon. Behaviour that is described as rule-based in other approaches emerges exactly in situations in which analogues for a given new form comprise a large set. Details of such models will be discussed in section 5.2. Whereas the said analogical models fundamentally differ from rule-based models in the way in which they view variability, however, not all of them differ from rule-based models in terms of the way in which they conceptualise lexical representations. Thus, in some analogical approaches it is assumed that similarity relations between words are established on the basis of symbolic features (e.g., Becker 1990: 63–71). In some exemplar-based models, by contrast, it is assumed that the exemplars that serve as potential analogues also comprise information that is more detailed and specific than the abstract features traditionally associated with rule-based grammatical models (cf., e.g., work on compounding, esp. Krott, Baayen and Schreuder 2001, 2002; Plag, Kunter and Lappe 2007; Arndt-Lappe 2011). An intermediate position between views that consider analogy to be exceptional and those that consider it to be the basis of regular word-formation is found in constructionist theories. The interesting question here is how analogy is related to schemas or schematisation, which are considered to be the central mechanism in word-formation. A clear view on this is found in Booij’s recent proposal (Booij 2010). Here it is claimed that schemas and subschemas may operate on symbolic features, and that the crucial difference between analogical formations and schema-based formations lies in their making reference to different degrees of abstraction. Analogy in this model is defined as strictly local analogy, which is complementary to schemas and may constitute an initial stage of the development of a schema (cf. esp. Booij 2010: 88–93 and section 4.1 for examples). Thus, Booij’s constructionist approach is different from both analogical and non-analogical approaches outlined above. There are differences between what happens in productive word-formation and what happens in an analogical formation in his sense, but this difference is a gradual difference, and not, as in the generative tradition, a difference that concerns the fundamental nature of the system. In the remaining two sections of this article I will provide an overview of the type of word-formation phenomena for which analogy has been invoked in the literature. Section 4 will deal with analogical formation in approaches which attribute a complementary role to analogy, i.e. rule- or constraint- or schema-based approaches in the sense outlined above. Section 5 will be devoted to phenomena discussed in analogical approaches, both non-computational and computational.

4. Complementary analogy We find pertinent appeals to analogy mainly in three domains. The first comprises relatively local analogies, explaining the emergence of small-scale patterns (section 4.1). Such small-scale patterns are often seen as the precursors to morphological processes proper. The second domain comprises apparent cases of reanalysis (section 4.2). The third domain is affix selection (section 4.3).

46. Word-formation and analogy

829

4.1. Local analogies An analogy is usually considered to be local if a) the analogue is restricted to one particular lexeme, b) a very high degree of similarity is involved, and c) the productivity of the process is very limited, in the extreme case producing only one new word. Needless to say, ‘locality’ of an analogy is a gradual notion. Traditionally, many such cases are of the type chairman : chairperson discussed in section 2 above, where both formally and semantically there is complete identity of one of the bases involved. A representative recent analysis is Booij’s (2010) constructionist approach, where a distinction is made between “analogical word-formation in the strict sense” on the one hand and constructional schemas on the other hand. Booij’s examples of analogical word-formation are given in (3). They are all from Dutch. (3)

Examples of analogical word-formation (glosses are taken from the original) new word paniek-haas lit. ‘panic-hare, panicky person’ vader-taal lit. ‘father-language, native language of father’ muis-vaardig lit. ‘mouse-able, with mouse-handling skills’ oud-komer lit. ‘old-comer, immigrant who arrived a long time ago’

analogue angst-haas lit. ‘fear-hare, terrified person’ moeder-taal lit. ‘mother language, native language’ hand-vaardig lit. ‘hand-able, with manual skills’ nieuw-komer lit. ‘new-comer, recent immigrant’ (Booij 2010: 89)

For the distinction between analogical word-formation and constructional schemas, Booij considers it to be crucial that [f]or these words [i.e. those in (3), S. A.-L.] we can indeed point to one particular existing compound as the model for the formation of the new compound, and the meaning of this new compound is not retrievable without knowing the (idiomatic) meaning of the model compound. (Booij 2010: 90)

Even in approaches using the term analogy in such a restrictive sense, it is often noted that analogies of the type exemplified in (3) can give rise to new word-formation patterns (cf. Booij 2010: 90–91, Szymanek 2005 for examples and pertinent references). The claim in many theories is, however, that once a new pattern has been created, its description in terms of analogy is no longer appropriate. There are two conceptual problems involved for an analogical description of the new pattern. One is that, once a pattern has arisen, it is impossible to trace the analogy back to one single analogue lexeme. Another problem is that many of the examples that are usually quoted in the literature involve reanalysis of either the analogue or the base of the analogue as morphologically complex. The problem of reanalysis will be discussed in section 4.2 below. For the problem of analogue selection, reconsider our example chairperson from section 1, where we have assumed that its analogue is chairman. However, there also exists chairwoman, which, like chairman, predates chairperson (the Oxford English Dictionary, henceforth: OED, records 1699 as the date of the earliest

830

IV. Rules and restrictions in word-formation I: General aspects attestation of chairwoman and 1971 as the corresponding date for chairperson). It is, therefore, unclear whether chairperson was modelled on chairman or on chairwoman. Indeed, the definition of the OED as ‘a chairman or chairwoman’ (OED, s. v.) suggests that it may have been modelled on both. On a theoretical level, the example shows that the assumption that the distinction between “analogy in the strict sense” and a pattern is not without problems. Thus, if we take Booij’s approach, we would assume the existence of a schema to account for the triplet chairman − chairwoman − chairperson. Apart from the apparent stipulation that bases of analogy must be single lexemes, it is, however, unclear why chairperson should be attributed a different status in the word-formation system than cases like paniek-haas (cf. (3) above). An alternative approach would assume that analogues may be sets of lexemes, which is, for example, inherent in many traditional, analogy-based accounts of cases of reanalysis, to which we now turn.

4.2. Reanalysis based on analogy The reason why analogy is often invoked in reanalysis cases lies in the fact that reanalysis obviously happens on the basis of similar lexemes that are stored in the lexicon. There are two pertinent classes of processes: certain cases of backformation and affix secretion (Marchand 1969: 210–214). Note that for both classes also rule-based accounts have been proposed. Examples of backformation are given in (4). (4)

Cases of backformation new word burgle

base of new word analogue/base burglar write/writer, sing/singer, etc. (Marchand 1969: 391)

televise

television

act/action, revise/revision, etc. (Marchand 1969: 395)

self-destruct

self-destruction

cases of noun-verb pairs where -ion is added to the verbal base (Bauer 2001: 83)

Cases like to burgle (derived from: burglar) are commonly analysed as involving reanalysis of the base word as morphologically complex, on the basis of analogy with existing pairs of lexemes. In this case, the bases of the analogue and the new word share a form, -er, but this form does not have the same meaning in the two bases. Note, however, that still there is a semantic similarity between the two bases: In the case of burglar and its analogues, for example, they all denote agents. One major theoretical issue in approaches to backformation cases is whether backformation is a diachronic process (cf. Becker 1993a for discussion and a review). Examples like those in (4) have been used convincingly to demonstrate the synchronic relevance of the process. The basis of this argument is semantic complexity. Whereas, for example, to burgle means ‘to act the burglar’, the noun burglar does not mean ‘one who burgles’ (Marchand 1963, as discussed in Becker 1993a: 4–8).

46. Word-formation and analogy

831

Another case of apparent backformation which has been analysed by appealing to analogy is the case of bracketing paradoxes. A well-known representative is Spencer’s analysis of certain English person-denoting adjective-noun phrases (Spencer 1988). Pertinent data are given in (5). (5)

Bracketing paradoxes new phrase transformational grammarian atomic scientist moral philosopher

analogue transformational grammar atomic science moral philosophy

The phrases in (5) are part of a group of phrases that, for some morphological theories, form bracketing paradoxes because it is unclear to which base the person-noun forming suffix (in the examples in (5): -ian, -ist, -er) is attached. Thus, the morphological base for -ian suffixation is the noun grammar, while the semantic base is the phrase transformational grammar. According to Spencer, formations such as those in (5), which are clearly productive in English, pose a challenge to rule-based morphological theories because the derived person noun cannot be convincingly related to their bases via a syntagmatic morphological rule (which would, for example, involve suffixation in grammarian and suffix substitution in philosopher). He therefore argues that the relation between bases and derivatives is an analogical relation, pertaining between lexicalised phrases in the mental lexicon. Unlike in the “traditional” backformation cases discussed further above, then, Spencer’s claim is that analogical processes may be productive. Another group of reanalysis cases where analogy is often assumed to play a role involves cases where new morphological patterns emerge (affix secretion, in Marchand’s terminology, cf. Szymanek 2005: 431, 435–436 for English). Similarly to the backformation cases, in these cases the semantics of the base for the analogical formation arises through reanalysis of the analogue-base relation as morphologically complex. Unlike the cases in (4) and (5), however, analogue forms are not morphologically complex, at least not before the advent of secretion. Some of the pertinent cases have also been described as blends. Examples are given in (6). (6)

cases of affix secretion new word candyteria

analogue cafeteria

meaning ‘shop, store, or establishment selling food x’ (Marchand 1969: 211)

Monicagate

Watergate

‘political scandal involving x’ (Szymanek 2005: 436)

There is a formal overlap between the new word and the analogue (-teria, -gate). This form, however, is not a unit of meaning in the analogue base, but becomes a unit of meaning in the new word, i.e. the moment it is extended to other words. This is precisely the situation for which Booij (2010: 88–93) argues that schematisation takes place, which in his view replaces analogy as the underlying mechanism.

832

IV. Rules and restrictions in word-formation I: General aspects

4.3. Affix selection on the basis of analogy A third group of processes where analogy has been invoked in the literature is affix selection. Pertinent examples are given in (7). (7)

Irregular affix selection new word a. orienteer

analogue volunteer (Bauer 1983: 290)

b. womanity

humanity (Baeskow 2012: 9–10)

For examples like those in (7), analogy is often invoked to explain affix selection. The underlying assumption is that this selection is irregular or unproductive. Thus, in both cases in (7), the affix selected to derive an agent noun (-eer) and an abstract nominalisation (-ity) is allegedly unproductive for the bases orient and woman, respectively. Bauer (1983: 285–291) assumes that -eer is generally unproductive in Modern English, and appeals to analogy to provide an explanation for the form orienteer, which is an apparent counterexample. The basis for the analogy here is phonological similarity between volunt- and orient, i.e. the base of the analogue and the base of the new word. The form womanity (7b) is an apparent counterexample to the generalisation that English -ity attaches to Latinate bases (cf., e.g., Baeskow 2012 and Arndt-Lappe 2014) for discussion and further counterexamples). Like in the case of (7a), phonological and perhaps also semantic similarity between the bases of analogue and new word, woman and human, plays a large part in motivating the analogy. Unlike in (7a), the two bases are also semantically similar. Apart from affix selection, also other types of selection between grammatical alternatives have traditionally been explained with the help of analogy. A case in point is stress in English nominal compounds, where two types of stress are available: left stress and right stress. Whereas left stress has traditionally been assumed to be the default pattern, cases of right stress have often been explained to be the product of analogy. Oft-cited textbook examples are compounds denoting street names. The examples in (8) are taken from Plag (2003). Stress is marked by an acute accent on the pertinent vowel. (8)

Stress in English noun-noun compounds Óxford Street Fóurth Street

Madison Ávenue Fifth Ávenue (Plag 2003: 139)

The analogical effects exemplified in (8) differ in two important ways from the affix selection cases in (7). Thus, in (8) analogue bases are sets of words, not isolated words (i.e. all compounds whose second constituent is street or avenue, respectively). Also, within their domain, analogies in (8) are productive. In sum, we have seen in this section that appeals to analogy in rule-, constraint-, or schema-based theories by no means form a homogeneous group. What they have in common, is that analogy is conceived to be relatively local, usually affecting lexical

46. Word-formation and analogy

833

items which are highly similar both phonologically and semantically. However, approaches differ in terms of whether they define elements involved in the analogy as single lexemes or sets, and in terms of which similarity relation exactly they view as being crucial to trigger an analogical formation. Also, they vastly differ in terms of whether they view analogy as a (potentially) productive process. In the next section we turn to a group of theoretical approaches which not only assumes that analogy may be productive in morphological grammar, but that it in fact forms its underlying principle.

5. Analogy in analogy-based approaches to word-formation Two groups of phenomena feature prominently in analogical approaches to word-formation: productive replacive formations, and cases of variability that affect the formal properties of outputs of word-formation. The former has been the object of discussion in much of the general theoretical literature, whereas the latter has been in the focus of the literature working with computational analogical models.

5.1. Productive replacive formations The term replacive formation refers to a pattern where new words are coined from existing complex words via affix replacement (Becker 1993a: 9–12). The phenomena discussed in the literature are in part the same as the phenomena discussed in sections 4.1 and 4.2. The issue under debate between analogical and non-analogical approaches here is, however, the productivity of these phenomena. Whereas in much of the generative literature it is claimed that analogical patterns of the type discussed in section 4 are not productive, it is claimed in the analogical literature that they are productive. In addition, it is claimed that the distinction between analogy and rule-based behaviour cannot be upheld on formal grounds. An explicit discussion of the theoretical implications that the existence of productive replacive formation has in terms of an analogical approach to word-formation is found in Thomas Becker’s work (Becker 1990, 1993a, 1993b). The key argument has two parts: a) There are productive replacive word-formation patterns (Becker 1993a, 1993b), b) they are not different from patterns which have traditionally been described as rulegoverned. An example of a productive replacive formation pattern is the pattern producing inand ex- prefixed words in German (discussed in Becker 1993b: 194). Examples are given in (9). (9)

German pairs of in- and ex-prefixed words Immatrikulation Inkardination Internat

‘immatriculation’ Exmatrikulation ‘incardination’ Exkardination ‘boarding school’ Externat

‘exmatriculation’ ‘excardination’ ‘a school that accepts day students’ (Becker 1993b: 194)

834

IV. Rules and restrictions in word-formation I: General aspects In- and ex-prefixation is replacive in the sense that one prefixed word is derived from the other prefixed word. In all examples in (9), the ex- derivative has been coined on the basis of the in- derivative, but there are also examples in which the reverse is the case. The pattern constitutes evidence in favour of a paradigmatic approach to wordformation because, as Becker convincingly shows, many pertinent cases cannot be described in terms of a concatenation of in- or ex- and a base. For example, ternat is attested only in the pair Internat and Externat in German, which makes it difficult to analyse it as a base for prefixation. The second part of Becker’s argument, i.e. that replacive formations are not different from other, allegedly rule-governed patterns, is more difficult. In Becker’s analogical approach, both types of pattern are described in terms of a rule format that Becker shows to be formally equivalent to a proportional analogy, especially since the format crucially employs traditional symbolic representations of lexemes. For Becker, the only difference between word-formation patterns that have been described as rule-governed and those that have been described as analogical in the generative literature lies in their different degrees of productivity. However, like many strands of traditional generative theory, which claim exactly the opposite, also Becker’s theory is lacking a testable means to predict which patterns will be more productive and which ones will be less productive.

5.2. Cases of formal variability Testability of degrees of variability is one of the key issues that has been addressed in simulation studies employing computational analogical models. The discussion here will focus on work based on the two analogical algorithms which are most widely used to model word-formation phenomena: the Tilburg Memory Based Learner (TiMBL, Daelemans et al. 1999 ff.; cf. Daelemans and van den Bosch 2005) and Skousen’s analogical model of language (AM, Skousen and Stanford 2007; cf. Skousen 1989, 1992; Skousen, Lonsdale and Parkinson 2002). Another algorithm that has been used in much work on inflection is the generalized context model (Nosofsky 1986, 1990). Furthermore, there is also work investigating the role of analogical factors using statistical modelling, without the implementation of a formal analogical model (cf., e.g., Plag 2006, 2010 on English compound stress). An obvious question is what the exact nature of the analogical theory is that is implemented by algorithms like TiMBL and AM. We will address this issue after we have discussed relevant studies. Like many analogical models, the key focus in the initial stages of pertinent morphological research was on problems of inflection (cf. esp. Daelemans and van den Bosch 2005; Skousen, Lonsdale and Parkinson 2002 for a summary of central issues). Issues of word-formation addressed in the literature always concern cases in which outputs of word-formation exhibit some sort of semi-regular variability. They are “semi-regular” in the sense that deterministic rule-based models fail to predict the attested variability. In contrast to work on inflection where there is a wealth of literature exploring the predictive power of computational analogical models (cf. relevant references in Skousen, Lonsdale and Parkinson 2002, the AM bibliography at http://humanities.byu.edu/am/ambiblio.html, Daelemans and van den Bosch 2005, and, in particular, work on the English

46. Word-formation and analogy past tense, e.g., in Skousen 1989, Eddington 2000, Keuleers 2008), pertinent research on word-formation phenomena is still in its infancy. Existing research has focussed mainly on two word-formation phenomena: compounding and allomorphy in derivation. Pertinent studies are most often based on corpus data and, in some cases, on data in which novel complex words have been generated by experimental subjects. Simulation studies devoted to variability in compounding have investigated linking morphemes in Dutch and German (esp. Krott, Baayen and Schreuder 2001, 2002, Krott et al. 2007; cf. Krott 2009 for a summary) and stress assignment in English noun-noun compounds (esp. Plag, Kunter and Lappe 2007; Arndt-Lappe 2011, cf. also Plag 2006, 2010). Both TiMBL and AM were used as algorithms. The three options that are available as linking morphemes in Dutch are: -s-, -en-, and -0̸- (‘zero, no linking morpheme’). Examples are given in (10). For easier reading but contrary to orthographic conventions, the relevant morphological components in the Dutch words are separated by spaces. (10) Variability in Dutch linking morphemes (from Krott, Schreuder and Baayen 2002: 55 f.) thee-bus papier-handel plaatje s boek tabak s rook krent en brood boek en kast

‘teabox’ ‘paper trade’ ‘picture book’ ‘tobacco smoke’ ‘currant bread’ ‘book case’

In English noun-noun compounds, two options are available: stress on the first constituent (“left stress”) or stress on the second constituent (“right stress”, sometimes also referred to as “level stress”). Examples are given in (11). Stress is marked by an acute accent. (11) Variability in stress assignment in English noun-noun compounds ópera glasses wátch-maker clássroom Óxford Street

steel brídge morning páper silk tíe Madison Ávenue

In spite of the fact that they are concerned with different phenomena, the two groups of simulation studies show surprising agreement in terms of their findings. Thus, predictive power of the computational analogical models employed was greater than that of traditional rule-based models. Furthermore, the most important determinant of the variation was the constituent family. This means that linking morphemes or stress assignment of a given novel compound can be predicted best on the basis of the pertinent behaviour of existing compounds that share either the first or the second constituent with the novel compound. In addition, it was found in most studies that, apart from constituent family, also semantic factors, if included in the simulation, served to enhance predictive power. Another group of phenomena that has been studied is affix selection and allomorphy in derivation. Pertinent studies include, for example, work on diminutives (Daelemans,

835

836

IV. Rules and restrictions in word-formation I: General aspects Berck and Gillis 1997 on Dutch; Eddington 2002, 2004 on Spanish), negative prefixation (Chapman and Skousen 2005, with a diachronic perspective), nominalisation (Eddington 2006 on Spanish, Arndt-Lappe 2014 on English) and comparative formation (Elzinga 2006) in English. A representative study is Eddington’s study of diminutives (Eddington 2002, 2004). Examples are given in (12). (12) Variable allomorph selection in Spanish diminutives minut-ito gallet-ita vidri-ecito yerb-ecita pastor-cito joven-cita normal-ito naric-ita pec-ecito flor-ecita lej-itos Luqu-itas patron-cita

← ← ← ← ← ← ← ← ← ← ← ← ←

minuto galleta vidrio yerba pastor joven normal nariz pez flor lejos Lucas patrona

‘minute’ ‘cookie’ ‘glass’ ‘grass’ ‘shepherd’ ‘young girl’ ‘normal’ ‘nose’ ‘fish’ ‘flower’ ‘far away’ ‘Luke’ ‘patron saint’ (Eddington 2002: 402)

Variability affects at least two dimensions: the form of the diminutive suffix (mainly -ito/a, -cito/a, -ecito/a), and the form of the stem allomorph (truncated, not truncated). In Eddington’s simulation experiment AM is able to predict correctly some 92 % of the data. In addition, it is shown that the variability predicted by AM is plausibly similar to the variability that exists in real life. Thus, uncertainty in the model’s predictions occurs exactly where uncertainty in allomorph selection in real life manifests itself, for example, by the existence of doublets. Eddington’s study of Spanish diminutive allomorphy is also representative of this type of analogical approach in terms of the features that were given to the algorithm as its information source. These typically involve mainly the phonological (i.e. segmental and prosodic) shape, but also the relevant grammatical categories of the base words in the dataset. In Eddington’s study, the latter comprised gender information. In general, the studies that have been introduced here are representative of work employing computational analogical algorithms to model variability in word-formation. An obvious difference to much previous work in analogy is that analogical algorithms are predictive mechanisms. In what follows we will briefly address the question of how this is achieved. Major differences between AM and TiMBL will be mentioned, but will not be in the focus of the discussion. Like in all approaches discussed in this article, analogical word-formation is assumed to be the product of a perceived formal and semantic similarity between a form that is about to be coined and its analogue (cf. section 2 above). Unlike in other approaches, however, the scenario that the analogue is only a single form or a small set of forms that is maximally similar to the new form is only one of several potential scenarios. Instead, analogies are based on those exemplars in the lexicon that are informative with

46. Word-formation and analogy respect to the given task. This group of exemplars is often called the “analogical set” of a new form (esp. in the AM-based literature) or the “nearest neighbour set” (in the TiMBL-based literature). Often, exemplars in the analogical set will differ in terms of their similarity with the given item. Classification of a new form will therefore always incorporate an effect of (type-) frequency because all members of the analogical set will influence classification. Thus, one element that makes algorithms like AM and TiMBL predictive is the fact that, unlike other analogical approaches, they have a principled way of determining which exemplars in the lexicon will serve as analogues, i.e. will be part of the analogical set. The second element that makes the models predictive is that they have a principled method at their disposal to determine which types of similarity are relevant for a given classification. The basis of all computation of similarity is formed by those elements of lexical representations which are provided by the researcher as a set of coded features for each exemplar in the database. The nature of these features is, in principle, a matter of choice, and it is still an unresolved question, which types of features lead to the best predictive power of a model. Existing studies of compounding have successfully used features encoding aspects of the compositional semantics of the compound, as well as features encoding the particular identity of the compound constituents. Existing studies of derivational allomorphy have typically used syllable-based phonological representations of base words (e.g., phonemes of onset, nucleus, coda of the ultima, penult, etc.) as well as, in some cases, grammatical information such as gender, word class, etc. Crucially, the question of how abstract or symbolic representational features are is still a matter of debate. The problem of determining which exemplars end up in the analogical set for a given new form is resolved in different ways by TiMBL and AM. What they have in common, though, is that the analogical set / nearest-neighbour set comprises those items which are similar to the new form in terms of exactly those features that are most useful for the given task. To do this, most varieties of TiMBL weigh the coded features of all exemplars in their lexicon in terms of how informative they are with respect to the given task, and treat items that share more informative features as more similar to a given new form than items that share less informative features. This means, then, that the importance of a given feature for the computation of similarity is the same for the whole lexicon. This is different in AM, where the decision of which coded features are relevant for a given exemplar to end up in the analogical set is made for each new form on an individual basis. For each new form the algorithm determines which combinations of features shared or not shared with that new form behave in a homogeneous way with respect to the given task (cf., e.g., Skousen 2002a, 2002b for a discussion of “homogeneity”). A crucial property of computational analogical models that is particularly relevant for grammatical theory in general and morphological theory in particular is that they have been claimed to be able to account for both types of effect: one that has traditionally been described as local analogy and one that has traditionally been described as rulegoverned behaviour. This point has explicitly been made mainly for inflection (cf. esp. Derwing and Skousen 1994; Eddington 2000; Keuleers 2008), but is also often alluded to in work on derivation and compounding. Recall from section 4 above that in nonanalogical, rule-based approaches it has often been claimed that the distinguishing feature between regular and analogical processes is productivity. In an analogical model,

837

838

IV. Rules and restrictions in word-formation I: General aspects however, there exists only a gradual distinction between local analogies and less local analogies. A local analogy arises if only exemplars which share many features with the new form are incorporated in the analogical set. Typically, then, analogical sets will be very small, and members of these sets will be highly similar in terms of both their phonological structure and their meaning. By contrast, behaviour that looks like rulegoverned behaviour in the traditional sense will arise if the analogical set is large, with exemplars in that set sharing fewer features. Thus, members of the analogical set will be less similar to each other both phonologically and semantically.

6. Summary and conclusion This article has presented an overview of approaches to analogy in word-formation theory. It has become clear that we have to distinguish analogy as a heuristic device from analogy as a construct in word-formation theory. The former is a mechanism that is very open and, in principle, underspecified in terms of many issues that need to be addressed in morphological theory. This openness of analogy is reduced in the specific usages of analogy as a construct in word-formation theory. These usages are closely tied to the theories’ basic assumptions about productivity, regularity, and variability. Strikingly, analogy has often been used in a narrow sense to denote local analogies, where both formally and semantically a very high degree of similarity is involved. The implication of this usage is that it is assumed that less local generalisations are non-analogical, because they involve higher degrees of abstraction (in terms of representations, rules, constraints, or schemas). This view is not shared by analogical approaches, where local analogies are considered to be only a special case of analogy. Here emphasis is put on the generality of analogy as a mechanism, with the implication that predictability and regularity of morphological operations is gradient. Constraints on analogy are often seen as a consequence of the nature of lexical representations and usage-based factors. In computational analogical theories, we furthermore observe that analogy is conceptualised as a predictive mechanism, where predictability emerges from the fact that analogues are not selected by chance, but by algorithms that have a principled, information-theoretic method at their disposal to distinguish informative and non-informative features and select sets of analogues accordingly.

7. References AM research group (ed.) AM bibliography. http://humanities.byu.edu/am/am-biblio.html [last access 20 Oct 2014]. Anttila, Raimo 2003 Analogy: The warp and woof of cognition. In: Brian D. Joseph and Richard D. Janda (eds.), The Handbook of Historical Linguistics, 425−440. Malden, MA: Blackwell. Arndt-Lappe, Sabine 2011 Towards an exemplar-based model of stress in English noun–noun compounds. Jounal of Linguistics 47(11): 549–585.

46. Word-formation and analogy Arndt-Lappe, Sabine 2014 Analogy in suffix rivalry − the case of -ity and -ness. English Language and Linguistics 18(3): 497–547. Aronoff, Mark 1976 Word Formation in Generative Grammar. Cambridge, MA: MIT Press. Baayen, Harald R. 2003 Probabilistic approaches to morphology. In: Rens Bod, Jennifer Hay and Stefanie Jannedy (eds.), Probabilistic Linguistics, 229−287. Cambridge, MA: MIT Press. Baeskow, Heike 2012 -Ness and -ity: Phonological exponents of n or meaningful nominalizers of different adjectival domains? Journal of English Linguistics 40(1): 6–40. Bauer, Laurie 1983 English Word-Formation. Cambridge: Cambridge University Press. Bauer, Laurie 2001 Morphological Productivity. Cambridge: Cambridge University Press. Becker, Thomas 1990 Analogie und morphologische Theorie. München: Fink. Becker, Thomas 1993a Back-formation, cross-formation and ‘bracketing paradoxes’ in paradigmatic morphology. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1993, 1−25. Dordrecht: Kluwer. Becker, Thomas 1993b Morphologische Ersetzungsbildungen im Deutschen. Zeitschrift für Sprachwissenschaft 12(2): 185−217. Blevins, James P. and Juliette Blevins (eds.) 2009a Analogy in Grammar. Oxford: Oxford University Press. Blevins, James P. and Juliette Blevins 2009b Introduction: Analogy in grammar. In: James P. Blevins and Juliette Blevins (eds.), Analogy in Grammar, 1−12. Oxford: Oxford University Press. Booij, Geert 2010 Construction Morphology. Oxford: Oxford University Press. Bybee, Joan 2001 Phonology and Language Use. Cambridge: Cambridge University Press. Bybee, Joan 2010 Language, Usage and Cognition. Cambridge: Cambridge University Press. Chapman, Don and Royal Skousen 2005 Analogical modeling and morphological change: The case of the adjectival negative prefix in English. English Language and Linguistics 9(2): 333−357. Daelemans, Walter, Peter Berck and Steven Gillis 1997 Data mining as a method for linguistic analysis: Dutch diminutives. Folia Linguistica 31: 57−75. Daelemans, Walter and Antal van den Bosch 2005 Memory-Based Language Processing. Cambridge: Cambridge University Press. Daelemans, Walter, Jakub Zavrel, Ko van der Sloot and Antal van den Bosch 1999 ff. TiMBL: Tilburg Memory Based Learner. Available from http://ilk.uvt.nl/timbl/. Derwing, Bruce I. and Royal Skousen 1994 Productivity and the English past tense: Testing Skousen’s analogical model. In: Susan D. Lima, Roberta Corrigan and Gregory K. Iverson (eds.), The Reality of Linguistic Rules, 193−218. Amsterdam/Philadelphia: Benjamins. Eddington, David 2000 Analogy and the dual-route model of morphology. Lingua 110: 281−298.

839

840

IV. Rules and restrictions in word-formation I: General aspects Eddington, David 2002 Spanish diminutive formation without rules or constraints. Linguistics 40(2): 395−419. Eddington, David 2004 Spanish Phonology and Morphology. Experimental and Quantitative Perspectives. Amsterdam/Philadelphia: Benjamins. Eddington, David 2006 Look Ma, no rules: Applying Skousen’s analogical approach to Spanish nominals in -ión. In: Grace Wiebe, Gary Libben, Tom Priestly, Ron Smyth and H. S. Wang (eds.), Phonology, Morphology, and the Empirical Imperative. Papers in Honour of Bruce L. Derwing, 371–407. Taipei: Crane. Elzinga, Dirk 2006 English adjective comparison and analogy. Lingua 116(6): 757−770. Gahl, Susanne and Alan C. L. Yu (eds.) 2006 Special Issue on Exemplar-based Models in Linguistics. The Linguistic Review (23). Berlin: Mouton de Gruyter. Hay, Jennifer and Harald R. Baayen 2005 Shifting paradigms: Gradient structure in morphology. Trends in Cognitive Science 9: 342−348. Hock, Hans H. 2003 Analogical change. In: Brian D. Joseph and Richard D. Janda (eds.), The Handbook of Historical Linguistics, 441−460. Malden, MA: Blackwell. Keuleers, Emmanuel 2008 Memory-Based Learning of Inflectional Morphology. Ph.D. dissertation, University of Antwerp. Krott, Andrea 2009 The role of analogy for compound words. In: James P. Blevins and Juliette Blevins (eds.), Analogy in Grammar, 118−136. Oxford: Oxford University Press. Krott, Andrea, Harald R. Baayen and Rob Schreuder 2001 Analogy in morphology: Modeling the choice of linking morphemes in Dutch. Linguistics 39: 51−93. Krott, Andrea, Rob Schreuder and Harald R. Baayen 2002 Analogical hierarchy: Exemplar-based modeling of linkers in Dutch noun-noun compounds. In: Royal Skousen, Deryle Lonsdale and Dilworth B. Parkinson (eds.), Analogical Modeling, 181−206. Amsterdam/Philadelphia: Benjamins. Krott, Andrea, Rob Schreuder, Harald R. Baayen and Wolfgang U. Dressler 2007 Analogical effects on linking elements in German compounds. Language and Cognitive Processes 22: 25−57. Marchand, Hans 1963 On content as a criterion of derivational relationship with backderived words. Indogermanische Forschungen 68: 170−175. Marchand, Hans 1969 Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. Nosofsky, Robert M. 1986 Attention, similarity, and the identification-categorization relationship. Journal of Experimental Psychology: General 115: 39−57. Nosofsky, Robert M. 1990 Relations between exemplar similarity and likelihood models of classification. Journal of Mathematical Psychology 34: 393−418. OED = Oxford English Dictionary. http://www.oed.com [last access 30 Sept 2011].

46. Word-formation and analogy

841

Pinker, Steven and Alan Prince 1994 Regular and irregular morphology and the status of psychological rules in grammar. In: Susan D. Lima, Roberta Corrigan and Gregory K. Iverson (eds.), The Reality of Linguistic Rules, 321−351. Amsterdam/Philadelphia: Benjamins. Plag, Ingo 2003 Word-Formation in English. Cambridge: Cambridge University Press. Plag, Ingo 2006 The variability of compound stress in English: Structural, semantic, and analogical factors. English Language and Linguistics 10(1): 143−172. Plag, Ingo 2010 Compound stress assignment by analogy: The constituent family bias. Zeitschrift für Sprachwissenschaft 29(2): 243−282. Plag, Ingo, Gero Kunter and Sabine Lappe 2007 Testing hypotheses about compound stress assignment in English: A corpus-based investigation. Corpus Linguistics and Linguistic Theory 3(2): 199−233. Prasada, Sandeep and Stephen Pinker 1993 Generalization of regular and irregular morphological patterns. Language and Cognitive Processes 8: 1−56. Skousen, Royal 1989 Analogical Modeling of Language. Dordrecht: Kluwer. Skousen, Royal 1992 Analogy and Structure. Dordrecht: Kluwer. Skousen, Royal 2002a An overview of analogical modeling. In: Royal Skousen, Deryle Lonsdale and Dilworth B. Parkinson (eds.), Analogical Modeling, 11−26. Amsterdam/Philadelphia: Benjamins. Skousen, Royal 2002b Issues in analogical modeling. In: Royal Skousen, Deryle Lonsdale and Dilworth B. Parkinson (eds.), Analogical Modeling, 27−48. Amsterdam/Philadelphia: Benjamins. Skousen, Royal, Deryle Lonsdale and Dilworth B. Parkinson (eds.) 2002 Analogical Modeling. Amsterdam/Philadelphia: Benjamins. Skousen, Royal and Thereon Stanford 2007 AM: Parallel. Available from http://humanities.byu.edu/am/. Spencer, Andrew 1988 Bracketing paradoxes in the English lexicon. Language 64(4): 663−682. Szymanek, Bogdan 2005 The latest trends in English word-formation. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 429−448. Dordrecht: Springer. van Marle, Jaap 1985 On the Paradigmatic Dimension of Morphological Creativity. Dordrecht: Foris. van Marle, Jaap 1990 Rule-creating creativity − analogy as a synchronic morphological process. In: Wolfgang U. Dressler, Hans C. Luschützky, Oskar E. Pfeiffer and John R. Rennison (eds.), Contemporary Morphology, 267−273. Dordrecht: Foris.

Sabine Arndt-Lappe, Düsseldorf (Germany)

842

IV. Rules and restrictions in word-formation I: General aspects

47. Productivity 1. 2. 3. 4. 5.

Introduction Quantitative approaches Qualitative approaches Productivity of inflectional classes References

Abstract The article deals with the many facets of the concept of productivity in word-formation, focusing on the one hand on the recent elaboration of various statistical quantitative measures made possible by the availability of large textual corpora. On the other hand, the article will provide a brief account of the qualitative impact on productivity given by the different kinds of system-related and cognitively rooted restrictions on wordformation. The distinct, but related, notion of productivity of inflectional classes is also discussed.

1. Introduction Productivity is a central issue in linguistics which has been traditionally of relevance for morphology but has recently gained attention in other domains of language as well (for instance, in syntax, cf. Barðdal 2008). In a nutshell, productivity in word-formation can be conceived of as the capacity of natural languages to form an in principle uncountable number of new words with the help of morphological means in an unintentional way, as in Schultink’s (1961) classical definition. Several objections have been raised against this definition. First, it is not clear what “an in principle uncountable number” really means, because the potential of creating new words is normally restricted by the size of the input domain selected. Thus, if a certain word-formation rule (henceforth, WFR) takes as an input only, let’s say, kinship terms, then its productivity can be quite high because it can freely form derivatives in this domain, but the number of words formed is perfectly countable and in fact corresponds to the whole set of kinship terms of the language. The latter is usually limited to at most a couple dozens of words. As we will see below, this problem leads us to the distinction between qualitative aspects of productivity focusing on the selectional properties of the WFRs and quantitative aspects focusing on the concrete number of lexical units involved. It is not entirely clear what the relation should be between these two faces of the coin. Second, it is not clear at all what a new word definitely is. This depends very much on a number of idiosyncratic factors relating among others to the personal education of the speakers, to his/her familiarity with specific speech registers or experiential (ontological) domains, etc. Thus, what is new for one speaker can sound perfectly familiar to another. As we will see, there are several proposals for approximating the intuitive idea of ‘new formation’.

47. Productivity Third, “unintentional” refers to the fact that the productivity of a WFR cannot be attested by nonce formations expressly created by the speakers to reach particular stylistic effects, as in the well-known Heideggerian expression: Wahrnis des Wesens der Wahrheit ‘true-ness of the essence of truth’ (cf. Fleischer 1975: 71), in which Wahrnis is Heidegger’s genuine creation but cannot seriously be taken to testify to the productivity of the suffix -nis because the philosopher is well known for his attitude towards coining affected (and we may add: abstruse) expressions. As commented upon by Bauer (2001: 56), “it would not be expected that such words would be adopted by the speech community, since they would run counter to the rules accepted by the speech community”. On the other hand, it is not at all clear whether a creation, even Heidegger’s Wahrnis, has to be considered as the result of an intentional, conscious process. At any rate, even if Heidegger had created Wahrnis unconsciously, the term could not signal the productivity of the suffix -nis in German. Conversely, it is not necessarily true that intentional, conscious creations could not be instantiations of productive WFRs (cf. Bauer 2001: 66– 68). The latter is normally the case with terminologies, which are the result of the intentional agreement of a special speech community on a certain way of denominating referents familiar to the community. Finally, Schultink’s definition does not reflect the scalar character of productivity, whereby WFRs are often available to different degrees to form new words, because it characterizes productivity as an absolute property (cf. Bauer 2005: 330). To cope with the multifaceted nature of productivity, Rainer (1987) identifies at least six different meanings associated with this term in the domain of word-formation: (1)

a. b. c. d. e. f.

the number of words formed with a certain WFR; the number of new words coined with a certain WFR in a given time span; the possibility of coining new words with a certain WFR; the probability of coining new words with a certain WFR; the number of possible (or generatable by rule) words formed with a certain WFR; the relation between occurring and possible words formed with a certain WFR.

The six meanings highlight different properties of productivity, and in particular: (1a) simply equates productivity with the vocabulary size of the words formed with a certain WFR; (1b) introduces the variable of time thus rendering productivity basically a diachronic notion; (1c) and (1d) focus on the speakers’ competence by referring respectively to their capacity to use a certain WFR and to the probability that a certain WFR is used; (1e) attempts to quantify the concept of possible word and finally (1f) relates it to the mere quantitative count provided by (1a). These meanings cross-cut the main distinction between a qualitative and a quantitative understanding of productivity. In fact, the qualitative approach basically sees productivity as inversely relating to the number of selectional restrictions of a WFR (cf. Booij 1977): with their increase we generally observe the decrease of the productivity, namely the possibility of using the WFR, which relates qualitative productivity essentially with the meaning (1c) above. Clearly, the idea of a possible word is strictly related to such an approach. Given the indirect relation between qualitative properties and quantitative factors, this relation is at least difficult to spell out on a precise basis.

843

844

IV. Rules and restrictions in word-formation I: General aspects The other meanings besides (1c) focus on a quantitative understanding of productivity, and in fact the discussion has centered on how to elaborate precise methods to measure the quantitative impact of the WFRs. In what follows, we will first discuss in section 2 the quantitative side of the issue, and we will especially focus on recent attempts to find reliable methods to measure productivity in large text corpora. Subsequently, in section 3 we will devote our attention to the qualitative perspective, trying to assess those qualitative factors which also influence the quantitative instantiation of WFRs. Finally, in section 4, we will briefly discuss the inflectional side of productivity, to the extent that it is relevant for word-formation as well.

2. Quantitative approaches The earlier attempts to deal quantitively with the notion of productivity have mainly relied upon the use of large dictionaries (including frequency dictionaries) and, less commonly, on text samples to substantiate intuitions about the degree of productivity of individual WFRs, especially in English (cf. Neuhaus 1973; Plag 1999: 115; Anshen and Aronoff 1999; Bauer 2001: 157), but also in Italian (e.g., Iacobini and Thornton 1992; Thornton 1998), Dutch (Al and Booij 1981), German (Wellmann 1975). However, reflection and research on a quantitative, measurable correlate to the notion of productivity in word-formation has undergone a dramatic development starting from the early nineties of last century under the decisive impulse of Baayen’s work (e.g., Baayen 1992, 1993; Baayen and Renouf 1996; more recent references are found in the following discussion of the article). The key strategy introduced by these works, made possible by the availability of very large, computer-analyzable corpora, mainly relies on the role of hapax legomena, i.e. words occurring only once in such corpora (typically sized in tens or hundreds of millions of tokens), as will be seen below. The basic concept underlying corpus-based approaches to productivity is the notion of vocabulary growth curve of a given affix or morphological procedure, namely the plotting of the number V(N) of different types of derivatives formed with that affix as a function of the total number N of the tokens of the same affix occurring in the corpus. This assumes of course that the figures for the function V(N), as those for N itself, are computed progressively as long as the corpus is being processed. To give an idea of the shape of the vocabulary growth curves, four instances of V(N) are reported in Fig. 47.1, taken from Gaeta and Ricca (2006: 58): they refer to the Italian suffixes -mente, forming adverbs, and -mento, -(t)ura and -nza, forming action nouns, progressively sampled from three years of the Italian newspaper La Stampa. If an affix is even minimally productive, new types will be encountered as long as the sampling proceeds: mathematically V(N) is a non-decreasing monotonic function. However, for every affix the slope of the curve V(N) will progressively decrease, since it will become more and more probable that new tokens of the affix will be occurrences of already attested types. The curves in Fig. 47.1 have been taken from a procedure of real progressive sampling of a newspaper corpus. Fortunately, at present several statistic models and packagings are available (although the mathematics behind them is far from obvious, cf. Baayen 2001; Evert 2004; Evert and Baroni 2007) which provide reliable fittings for the growth

47. Productivity

Fig. 47.1: Vocabulary growth curve V(N) for four Italian derivational suffixes (from Gaeta and Ricca 2006: 58)

curves, at least for interpolations: extrapolating the curves beyond the real sample size raises much more difficult and debated reliability issues, which cannot be mentioned here. These fitting algorithms need as input just data obtainable from the full corpus sampling, which makes things operationally much easier. From Fig. 47.1, it is evident that the curves V(N) for the four suffixes increase at different rates. Whereas the curve of the suffix -nza immediately reaches almost the whole number of possible types and then approximates a horizontal line, for the other suffixes the curve is clearly still increasing, although with different slopes, at the end of the sampling procedure. The behaviour of the curve for -nza illustrates the case of a qualitatively unproductive suffix: once all (or nearly all) the well-established items derived with the suffix occur in a (sufficiently large) corpus, there is no expectation of finding instances of new types. All three other curves display the behaviour expected for qualitatively productive wordforming procedures, although they differ according to several measurable parameters. The simplest parameter to measure and understand is the type frequency V in the full corpus, which can be seen as the size of the morphological category (Baayen 2009: 901). A high value of V means that many items have been derived with that affix and are currently in use in the language community. A corpus-based value of V may be compared with the type frequency resulting from lexicographical counts, and usually is much more indicative of the real “currency” of the given affix than the latter, assuming we are dealing with a sufficiently large and balanced corpus which can reasonably approximate real language behaviour. This is because a corpus will not include many obsolete words

845

846

IV. Rules and restrictions in word-formation I: General aspects that dictionaries record plenty of, will not overstate the impact of special language terminologies, and, for productive procedures, will contain many new formations too “regular” and unnoticeable to become quickly registered in dictionaries. A large-scale comparison between type frequency data taken from corpora and dictionaries is provided for Italian in Gaeta and Ricca (2003). However, V is an indication of how much success a morphological procedure has enjoyed in a given language, and in this regard it roughly corresponds to Corbin’s (1987: 177) notion of rentabilité ‘profitability’ (“la possibilité de s’appliquer à un grand nombre de bases” [the possibility of applying to a large number of bases]), but does not tell the whole truth about how easily the given procedure is currently available to form new items, which is intuitively the very core of the notion of productivity. An affix which has been very fashionable in the past may keep a trace of its past relevance by still displaying a relatively high V, even if speakers do not use it anymore to form new words. That’s why Baayen (2009: 901) suggests the label “realized productivity” for the notion measured by V. As a matter of fact, considering the bulk of Italian derivation taken into account in Gaeta and Ricca (2003: 83–84), this effect is much less significant than expected. The (nearly) unproductive suffixes with highest V in the Italian corpus examined are the denominal agentive -aio (giornale → giornalaio ‘newspaper seller’) and the deverbal potential adjectival -evole (piegare → pieghevole ‘folding’), and they are ranked rather low in the list. Clearly, the distortion effect would be much more important if dictionary-based type frequency were considered, for the reasons mentioned above. Indeed, most productive formation rules tend to produce such a high amount of hapax legomena and very low-frequency items registered in a corpus (but not in a dictionary!), that the well-established items which survive the marginalization of their word-formation procedure cannot really reverse the picture. This leads to the consideration of the central role of the hapax legomena in measuring productivity. Mathematically, the hapax number h after N occurrences of an affix have been processed is directly connected to the slope of the curve V(N), which can be demonstrated (Baayen 1992: 115) to be equal to the ratio h/N. Measuring the slope of the growth curve at a given point N means evaluating the speed at which new types of a certain affix emerge in the sample: the steeper the slope, the higher the production rate of new types. For this reason, comparing the slopes of the growth curves of two affixes for a given value of N is a way of comparing the contribution of the two affixes to the growth rate of the vocabulary in a corpus, i.e. it is a way to rank their productivities. As stressed by Gaeta and Ricca (2006), who applied the procedure to a significant portion of productive Italian derivational morphology, the slope comparison at equal values of N allows one to plausibly rank the productivity of affixes with different token frequencies, whose overall growth curves in a fixed corpus differ substantially in length (for instance, the nearly synonymous -tura and -mento in Fig. 47.1). Gaeta and Ricca define their procedure as “variable corpus approach”, because, in order to compare at equal token number two affixes of different token frequency, one has to process subcorpora of different size. Baayen terms this facet of productivity “expanding productivity”, contrasting it with the “realized productivity” measured by the sheer V value, and confirms that the former measure produces rankings which are “reasonable reflections of linguists’ overall intuitions about degrees of productivity” (Baayen 2009: 905). However, he states that the same results can be obtained by a still simpler measure, namely by comparing directly

47. Productivity the absolute values of h in the whole corpus for any two (or more) affixes. On the basis of his suggestion, Gaeta and Ricca (2006: 73–74) verified that there is indeed a high correlation between the two measures; however, they also showed that this is not always the case. In particular, comparing the two Italian deverbal suffixes -tore for male/generic agent and -trice for female agent, the simple number of hapaxes h strongly favours the more frequent suffix -tore, which has twice as many hapaxes as -trice, while the two display about the same value for Gaeta and Ricca’s “variable corpus” measure. Being essentially linked with hapax legomena, the expanding productivity may be related to Corbin’s (1987: 177) notion of disponibilité ‘availability’, although the latter (“la possibilité de construire des dérivés non attestés” [the possibility of forming unattested derivatives]) can also be interpreted as a purely yes/no notion, as done for instance by Bauer (2001: 205). A third, quite different measure originally suggested by Baayen is the h/N value calculated for each affix at the full corpus value, which means the slope of the growth curves in Fig. 47.1 at their endpoints. Given that for any growth curve, the slope steadily decreases for increasing N, this measure would get lower results for the very frequent affixes (e.g., for -mente and -mento compared with -tura in Fig. 47.1). The resulting rankings are quite perplexing from the point of view of linguists’ intuitions, as shown by Gaeta and Ricca (2006: 72). For instance, the adverbial suffix -mente ‘-ly’ is unanimously considered extremely productive. Its generality locates it at the border between derivation and inflection, and for some even beyond it, analogously to its English counterpart (see, e.g., Haspelmath 1996: 49–50). But according to this measure, -mente would be ranked below some clearly derivational suffixes, and particularly below the relatively infrequent female agent suffix -trice, which in turn would result as over five times more productive than its male agent counterpart -tore. Similar considerations are entertained by Plag (1999: 113) about the data for English -ly in Baayen and Renouf (1996). However, Baayen (2009) gives an interpretation of the measure above: the ratio h/N at full corpus should essentially measure how far a given procedure is from saturating its domain of application. This is reflected by his proposed label of “potential productivity”. From this point of view, there is a further way to evaluate the grade of saturation. An estimation of the maximal size S (in types) of the domain of a given morphological procedure, for an ideally infinite corpus, can be provided by statistical models, such as those proposed in Evert (2004). Then the ratio I = S/V(N) would measure the inverse of the saturation ratio, which Baayen (2009: 907) claims to show a high positive correlation with his potential productivity. The narrow connection between the number of hapax legomena in a corpus and the productivity measures has often been criticized with the objection that hapaxes in a corpus are not necessarily new formations. On the one hand, this is undoubtedly true, but becomes progressively less relevant as corpus size increases: for corpora of many tenths of millions tokens, most hapaxes indeed turn out to be un-established words (cf. Baayen and Renouf 1996 on a 80 million-word Times corpus). On the other hand, as Baayen (2009: 906) puts it, “hapax legomena are not a goal in itself, they only function as a tool for a statistical estimation method aimed at gauging the rate of expansion of morphological categories”. For hapaxes to be a reliable tool, however, it is necessary that corpus data are carefully and time-consumingly checked by manual inspection: a fully automatic listing of items associated with a given ending in a corpus would indeed produce huge distortions (cf. Evert and Lüdeling 2001). Suffice it to think that most

847

848

IV. Rules and restrictions in word-formation I: General aspects misprints would turn out as hapaxes, thus heavily polluting any quantitative measure based on them. A more serious operational problem met with any quantitative approach is establishing what exactly counts as a type of a given affix. For instance, to what extent should opaque derivations be included? More intriguingly, as pointed out by Bauer (2001: 151), should underdevelopment be counted as a different type of -ment (considering it as derivation from the verb underdevelop), or should its occurrences be merged with those of development into a single type? The latter instance raises another problem. Should underdevelopment count as a type of under-, of -ment, or of both? These issues have been tackled at length in Gaeta and Ricca (2006), to which we refer. The most interesting is perhaps the third one, namely how to deal coherently with multiply affixed words, like conventionalize or reprintable. The standard choice in these cases (defended explicitly in Plag 1999: 29) has always been to select as affix tokens only those words in which the affix is attached last (i.e. -ize and -able in the examples given above, called “outer cycles” in Gaeta and Ricca 2006). The main reason for this choice – which is also often operationally easier – is that it keeps the tokens of the different affixes as independent subsets within the corpus, thus allowing for statistical testing (Baayen 2009: 903–904). Linguistically, however, it would be preferable to count both re- and -able in reprintable as occurences of the respective affixes. Fortunately, Gaeta and Ricca (2006: 79–83) have shown that in most cases their “variable corpus” productivity measure is practically insensitive to the inclusion of the inner derivational cycles, even when these amount to a great number of tokens. In the latter case, however, the same does not hold for Baayen’s potential productivity. The brief discussion above should have made clear that there is not a unique quantitative concept of productivity in word-formation. Different facets of this complex phenomenon may be reflected quantitatively by different statistical measures. However, there can be little doubt that statistical work on large corpora has contributed decisively to a deeper understanding of the notion of productivity and to the disentanglement of its diverse components. Moreover, quantitative methods make it possible to rank and compare effectively with each other different morphological procedures competing in the same domain (and perhaps also in different domains) by means of reproducible and falsifiable data. Finally, statistical corpus-based methods, dealing with large amounts of real language production, have the great merit of stressing that the coinage of new words is deeply rooted in the external communicative needs of the speech community, and therefore can hardly be treated as a prevailingly system-based phenomenon, contrary to what is suggested in approaches like the one in Dressler and Ladányi (2000). This opens the wide issue of comparing the morphological productivity of the same procedures across genres and registers, not to speak of the spoken-written distinction. In principle, a careful construction and comparative investigation of corpora contrasting along genre/ register dimensions may be the ideal empirical tool to reach sound and reliable conclusions in this domain, although much is still to be done (cf. Plag, Dalton-Puffer and Baayen 1999). Another aspect in which much research work remains to be done is the systematic cross-linguistic collection and comparison of quantitative data on productivity, especially in the light of the increasing availability of large electronic corpora in many languages other than English. First results in this direction are available at least for Dutch (Baayen 1994; Baayen and Neijt 1997), German (Lüdeling and Evert 2005; from a diachronic

47. Productivity

849

perspective, see Scherer 2005 and article 103), Italian (Gaeta and Ricca 2003; diachronically, see Štichauer 2009), and French (cf. Dal 2003 and the papers contained there).

3. Qualitative approaches Following Aronoff (1976), a number of investigations have focused on the idea that productivity should be strictly related to the concept of possible word. From this perspective, productivity depends on the possibility offered by the grammar of a WFR to apply to a certain set of bases. Accordingly, Booij (1977) suggests to inversely relate the productivity of WFRs to the number of restrictions pertaining to a certain WFR when applied to a set of bases. These restrictions constrain the theoretically unlimited application of a WFR, and therefore have an impact on its productivity. However, the number of restrictions pertaining to a WFR cannot be directly reflected in quantitative terms, as argued by Booij, because their quantitative impact does not relate in linear terms with the selectional properties displayed by the WFRs. Thus, it is not safe to claim that a WFR displaying more restrictions is also quantitatively less productive than a WFR displaying a smaller number of restrictions. The restrictions can come from every level of linguistic analysis, from phonology to syntax, including semantic and pragmatic aspects (cf. article 48 on restrictions in wordformation). To mention only a couple of cases with an immediate impact on the quantitative dimension, it has been observed that the interaction of prefixation and suffixation creates interesting telescoping effects, which limit or enhance the productivity of WFRs when they are considered in connection with each other. Thus, in Italian deverbal potential adjectives formed with a negative prefix like invendibile ‘unsalable’ display an apparent distortion of the quantitative measure of their expanding productivity as discussed in section 2. Notice that these adjectives display a clear derivational sequence, because the prefix cannot take verbs as an input: (2)

a. vendere ‘to sell’ b. vendere

→ →

vendibile *invendere

→ →

invendibile invendibile

If the inner derivational cycles are taken into account, the calculation of the expanding productivity of the suffix -bile varies considerably (cf. Gaeta and Ricca 2003, 2006 for details). In fact, the inclusion of the inner cycles (thus for instance the prefixed [in[vendi-bile]]) strongly lowers the expanding productivity of -bile with regard to the value obtained including the outer cycles only. Far from undermining the reliability of the whole procedure, Gaeta and Ricca argue that this is due to the fact that many derivatives formed with -bile are current only if co-occurring with a further prefixal derivation. This is shown by cases like introvabile ‘untraceable’, instancabile ‘tireless’, imperturbabile ‘imperturbable’, etc., whose unprefixed bases (??trovabile, ??stancabile, ??perturbabile) are at most marginal due to pragmatic reasons. Therefore, the exclusion of the inner cycles has the effect of introducing many “spurious” hapaxes not flanked by their much commoner prefixed counterparts, which considerably overestimates the productivity value of -bile. Thus, although the question of the inclusion of the inner/outer derivational cycles can be generally kept under control by the procedure described in the previous

850

IV. Rules and restrictions in word-formation I: General aspects section, in some cases the interaction of the qualitative properties of prefixation and suffixation can lead to apparently distorted results. A second example has to do with the stratification of the vocabulary which is often neglected or marginalized in the theoretical debate. For instance, Japanese has borrowed a significant part of its lexicon from Chinese. This includes especially VN compounds (3a), which are at odds with the native NV compounds (3b) (cf. Kageyama 1982): (3)

a. satu-zin ‘killing a man’ kill-man hoo-bei ‘visiting the USA’ visit-US

b. yama-nobori ‘mountain climbing’ mountain-climb booru-nage ‘ball throwing’ ball-throw

The pattern in (3a) has become productive in Japanese, but it is still restricted only to items of Chinese origin. Thus, a VN compound containing a lexical unit not belonging to the Chinese stratum is unacceptable: *hoo-Amerika ‘visiting America’. Therefore, to judge the quantitative relevance of these compounds, one should preliminarily know the size of the Chinese stratum in the Japanese lexicon. Furthermore, the comparison with the productivity of the native compounds is not obvious because it crucially depends on the mechanism of borrowing, which is in a way extra-grammatical. Similar problems are found in many other languages as well, most notoriously in the case of the English suffixes -ive, -ity, -ous and -al (as in dialectal, parental, etc.) which are mostly restricted to Latinate bases; unfortunately, they are scarcely considered in quantitative works on productivity. From a more general perspective, a number of universal restrictions have been suggested which allegedly limit the productivity of WFRs besides the aforementioned ones concerning individual, language-specific cases. Accordingly, one can distinguish (i) general claims relating to the general format of the WFRs, (ii) specific claims relating to particular properties of the grammar, (iii) predictions resulting from the particular format of the grammar, and (iv) general properties of the lexical items depending on their accessibility at the cognitive level. These restrictions are of a very different nature, going from those reflecting a certain view of WFRs or of grammar, i.e. the way in which our language competence is structured, to those relating to more general properties of our cognitive endowment when it performs linguistic operations. In the following a quick survey of the different theoretical options will be offered, which also reflects the historical trend from more competence-oriented approaches typical of the early discussions on the productivity of WFRs to the more performance-oriented views which characterize recent research carried out with the help of large electronic corpora (cf. Bauer 2005 and Rainer 2005 for more detailed recent surveys). With regard to general claims relating to the format of the WFRs, they clearly have an impact on their potential application. For instance, Aronoff’s (1976) unitary base hypothesis (UBH) severely constrains the input base to a single word category. Given the very different size of, for instance, the nominal with regard to the verbal lexical class normally present in a language, it is straightforward to expect that, at least in purely absolute numbers, denominal WFRs will be more productive than deverbal WFRs. In spite of its coarseness, this estimation should not be forgotten when WFRs based on nouns and verbs are compared to each other. On the other hand, it has been suggested

47. Productivity that the UBH might be a reflex of the semantics of the WFRs (cf. Plag 1998: 237): if it is true, this casts some doubts on the relevance of the UBH as an independent universal constraint. As for specific claims relating to the particular properties of the grammar, several universal restrictions have been suggested like the adjacency condition (Siegel 1977), the atom condition (Williams 1981), etc. As commented upon by Rainer (2005: 336): “Most of these constraints were flawed from the beginning by an insufficient empirical underpinning, and when the conditions-on-rules approach went out of fashion in generative grammar, they met the quiet death they deserved.” Other predictions result from the particular format of the grammar assumed by a certain framework. In this perspective, it may for instance be relevant which stance the researcher takes on the question of whether restrictions are generally affix- or basedriven. The former approach has been by far the most common in lexicalist frameworks, but Plag (1999: 67–76, 2002) suggests on the contrary a generalized base-driven approach, giving by way of illustration the treatment of the complex domain of English deverbal abstracts. Accordingly, the selective correlation between the verbs formed with the suffix -ize and the nominalizing suffix -ation is taken to be driven by -ize, which at the same time accounts for the restriction on deverbal abstracts displaying the sequence *-ize-ment. This approach is more economic than the former because it does not require one to explicitly state the restriction on *-ize-ment among the selectional properties of -ment. On the other hand, it crucially excludes doublets of derivatives from the same base, which appears to be contradicted by the Italian verbalizing suffix -eggia- selecting -mento (corteggiare ‘to court’ → corteggiamento), -tura (tinteggiare ‘to paint’ → tinteggiatura) and -ìo (lampeggiare ‘to blink’ → lampeggìo, see Gaeta 2005 for a detailed discussion of the question based on Italian). Finally, a number of restrictions have been claimed to be due to general properties of the lexical items insofar as they are more or less easily parsable or accessible at the cognitive level. In this regard, Hay (2000, 2002) has suggested that the whole question of selectional restrictions be conceived from a perspective dubbed by Plag (2002) “complexity-based ordering”. Accordingly, the prediction is made that the more easily parsable affixes should be normally less restricted than (and accordingly should occur after) the less easily parsable ones. In other words, words containing less easily parsable affixes are more likely to be stored in our mental lexicon and therefore directly accessed as units. Factors influencing the ease of parsability are frequency, and especially relative frequency, namely the frequency of a derivative with respect to its base, and phonotactics. The latter is more difficult to operationalize, but generally has to do with the occurrence of less frequent sound clusters resulting from the combination of two morphemes. They are a better cue for detecting and parsing a morphological boundary than more frequent sound clusters which often occur inside morphemes. Taking into consideration these parameters allows Hay to construct a hierarchy which expresses the combinability potential of the individual affixes with regard to the base: affixes scoring higher in terms of relative frequency and phonotactics are normally placed closer to the base. This shows that the selectional properties of the affixes can in principle be explained only with the help of these performance-oriented factors. However, Hay and Plag (2004: 590) observe that this approach is in fact unable to exclude all impossible combinations and conclude that “both selectional restrictions and processing constraints are instrumental in determining suffix ordering”.

851

852

IV. Rules and restrictions in word-formation I: General aspects A further example of the impact of cognitive aspects relating to our storage memory is given by blocking, which for instance causes the relatively marginal status of °stealer with respect to thief. As shown by Rainer (1988), the main factor influencing synonymy blocking due to a well-established lexical item is the frequency of the latter. As it decreases, doublets of synonymous words become acceptable. For instance, the German suffix -heit forms quality nouns from monosyllabic adjectives, but it is blocked when an adjective forms a quality noun with another unproductive suffix: reich ‘rich’ → Reichtum/*Reichheit. However, the strength of the blocking effect decreases in correlation with the frequency of the established quality noun: herb ‘acerb’ → Herbe/Herbheit, because Herbe is comparatively uncommon. It is precisely the relevance of frequency to blocking that shows that blocking is a qualitatively different phenomenon from the grammar-related restrictions discussed above, since it is not a constraint on formation proper, but rather on the lexicalization/ entrenchment chances of a given word. While a word like *dier (from die) may be considered ill-formed due to a semantic restriction on the base, a word like °stealer is not ill-formed, it is simply unlikely to get established in the mental lexicon due to the existence of the well-entrenched synonym thief.

4. Productivity of inflectional classes A different notion of productivity which is perhaps only tangentially relevant to wordformation concerns the productivity of inflectional classes. However, the topic is worth mentioning in the context of this handbook, because the inflectional classes are obviously a partition of a major word class, and consequently their (in)accessibility to new entries has long-range consequences on the general shape of the lexicon in a given language. Thus, for instance, if it can be shown that a certain WFR is sensitive to a constraint on the inflectional class to which its input base belongs, the productivity of the inflectional class has a great impact on the productivity of the WFR. In Classical Latin, the two adverbializing suffixes -ē and -iter took as an input adjectives belonging to the first and to the second inflectional class (cf. respectively lentus/-a/-um ‘slow’ → lentē vs. fortis/-e ‘strong’ → fortiter). Given the different distribution of the two classes, this has a consequence on the productivity of the two adverbializing suffixes. Probably the qualitative approach has been dominant in dealing with this facet of productivity (especially due to the massive work by Wolfgang U. Dressler and his school in this domain); however, there are also quantitative considerations which can corroborate some relevant points of the issue (see Gaeta 2007 discussed below). Dressler’s approach to productivity in inflectional classes, although qualitative, is not really two-valued, but rather based on a gradual scale of criteria, which test the accessibility of a new item to the given inflectional class (or single microclass, according to Dressler’s terminology, cf. Dressler 2003). Hierarchically, the strictest criteria concern the integration of loanwords. Following Wurzel (1984), Dressler further distinguishes the integration of those loanwords that do not originally fit in the class they enter (according to phonological and/or structural properties) from those which already have fitting properties. For instance, the integration of English verbs into Italian (to dribble → dribblare, to set → settare; similarly for Spanish) requires also the insertion of the

47. Productivity thematic vowel -a- to comply with the system adequacy requirement for Italian verbs (Dressler 2003: 37). The integration of unfitting loanwords is seen by Dressler as marking the highest level of productivity, which can obviously get lost in diachrony. As an illustration, in older Italian the masculine nominal class identified by sg. -o/pl. -i usually received loanwords also ending in a consonant (e.g., stoccafisso/-i ‘stockfish’ < Old Dutch stocvisch, 15th century, tallero/-i < Ger. Thaler (an old currency), 16th century) but nowadays these items are put in the invariable class (killer → It. killer/*killero, Ger. Blitz ‘flash’ → It. blitz/*blitzo ‘swoop, police raid’). According to Dressler’s approach, this means that the -o/-i class is now only intermediately productive and no longer fully productive. Notice that even a fitting loanword is not necessarily integrated into an existing inflectional class. A clear case is given by the contrast between feminine and masculine loanwords ending in -a in Italian: feminines like dacia ‘dacha’, sauna, geisha vs. masculine lama (both the Tibetan monk and the animal), koala, tanga ‘thong underwear’. Italian has two inflectional classes whose singular ends in -a, namely the nouns with pl. -e (all feminines, e.g., casa/e ‘house(s)’), and those with plural -i (all masculines, e.g., poeta/i ‘poet’, barring two exceptions). Feminine loanwords in -a enter the -a/-e class without exceptions, while masculine loanwords never enter the -a/-i class, but remain invariable. Like the case of the -o/-i class seen above, the impossibility of integrating into the -a/-e class the unfitting feminine loanwords like la jeep, pl. le jeep (never *la jeeppa/ *le jeeppe) makes this class only strongly productive, not fully so. On the other hand, the masculine -a/-i class is labelled as plainly unproductive, because it answers negatively also to a weaker test posited by Dressler: the integration of indigenous non-derivational lexemes (i.e. those arising from “extramorphological” procedures like clippings or acronyms, or from conversions). The partially positive response to the latter criterion (the conversions from verb infinitives like sapere ‘to know’ → il sapere ‘the knowledge’, pl. i saperi ‘the knowledge fields’) together with a negative response to the stronger first two, locates a further Italian noun class, the -e/-i of f. fronte ‘forehead’ and m. ponte ‘bridge’, in an intermediate position between the -a-/e and the -a/-i classes and thus defines it as “slightly productive”. The Italian -a/-i noun class only responds positively to a last criterion: new items steadily enrich it via the very productive suffix -ista ‘-ist’. However, in Dressler’s model this just testifies to the stability, not the productivity of the class, because in this case the productivity is seen as a property of the derivational suffix only (Dressler 2003: 43). Similarly, in Polish the neuter microclass in -e is considered as unproductive, despite the existence of a productive word-formation process, abstract nouns in -anie. The same is said of German neuter nouns forming the plural by umlaut + -er (e.g., Land ‘country’ → Länder), although the derivational suffix -tum (pl. -tümer) is productive. By the same token, instances of class shift (like Ger. der Mops ‘the pug’, older plural die Mops-e > today’s Umlaut plural die Möps-e) may at most distinguish between stable and recessive classes (Dressler 2003: 43) – and not necessarily so, as shown by the example above, where the two classes involved are said to display the same level of productivity. Dressler’s approach provides the field with a comprehensive framework which has been thoroughly applied to a number of languages, e.g., Italian (Dressler and Thornton 1996; Gardani 2013), Polish (Dressler, Dziubalska-Kołaczyk and Fabiszak 1997), Russian (Dressler and Gagarina 1999), French (Kilani-Schoch and Dressler 2005) and re-

853

854

IV. Rules and restrictions in word-formation I: General aspects cently also in a wider diachronic perspective from Latin to Old Italian (Gardani 2013). However, many open points remain, for instance: (i)

There is still limited support for the claim that the above criteria can be ordered hierarchically according to an implicational scale, which is crucial for the crosslinguistic reliability of pseudo-quantitative evaluations like ‘strongly > weakly > slightly productive’. (ii) The qualitative approach, although being discrete in principle, cannot conceal the existence of grey areas between the grades. For example, in Italian, a minority of -o masculines are apparently integrated in rare cases (i/dei tornado has ten times the occurrences of i/dei tornadi on the web), and conversely a minority of -e nouns usually do (i/dei droni from il drone ‘military crewless airplane’ is largely dominant, despite the fact that the borrowing is quite recent); in general what impact the variable time may have on the integration of lexical material is not discussed, especially when a language also has inflectional classes exploiting zero marking (see below). (iii) It is unclear whether in a given language the invariable class has to be considered as a mere default reservoir of non-integrated items, or rather as a full inflectional class on its own. For instance, Russian has basically no native indeclinable nouns, so an indeclinable pal’to ‘coat’ (< Fr. paletot) has to be considered a non-integrated loanword, which does not initiate a productive class; but the same item paltò in Italian goes into an open class of the native vocabulary (enrichable via the productive suffix -ità ‘-ity’), and in doing so it fully reflects system adequacy given its phonetic shape (all Italian words ending in a final stressed vowel are invariable). Invariable nouns are rightly considered a productive class in Italian (Dressler and Thornton 1996: 7); however, it is a bit paradoxical that by the criteria discussed above they turn out to be the only fully productive class in the language. A different approach to the productivity of inflectional classes is taken by Gaeta (2007), who for the first time applies the quantitative, corpus-based methodology sketched in section 2 systematically to the domain of inflection. Some of his results concern inflection proper, and appear to strongly support the validity of the “variable corpus” method developed in Gaeta and Ricca (2006). In particular, Gaeta shows that the “variable corpus” productivity values for different endings varying along a dimension of contextual inflection (in the sense of Booij 1996) are exactly the same, as should be expected linguistically: the growth curves for the types of 3rd sg. imperfect -ava and 3rd pl. imperfect -avano can be superposed, although the tokens of the former are three times more numerous, and consequently its growth curve is much longer. No similar result is obtained by applying the other measures of productivity described in section 2. Variation in the inherent inflection component (e.g., 3rd sg. imperfect -ava vs. 3rd conditional sg. -erebbe) has an impact on the productivity values (see Gaeta 2007 for its evaluation) but interestingly the different values are all ranked above those of the most productive derivational affixes, another linguistically meaningful result not matched by the other measures. The main relevance of Gaeta (2007) for the topic under discussion here, however, concerns the sharp contrast which emerges in quantitative productivity data among the productive verb class of -are verbs, and the other two conjugations (-ere and -ire verbs). Dressler and Thornton (1991) merge the latter into a single essentially unproductive

47. Productivity macroclass, a theoretical and descriptive proposal shared also outside the framework of natural morphology (cf., e.g., Vincent 1988: 294; Maiden 1992: 309). Contrary to -are verbs, in Gaeta’s work the productivity values for the inflections of -ere and -ire verbs are very low, in the range of those of the scarcely productive derivational suffixes. This result validates quantitatively the qualitative distinction between productive and unproductive inflectional classes, and also the “polarization” effect in productivity when inflection is concerned, suggested, e.g., by Aronoff and Anshen (1998: 246–247).

5. References Al, Bernard and Geert Booij 1981 De produktiviteit van woordvormings-regels: Enige kwantitatieve verkenningen op het gebied van de nomina actionis. Forum der Letteren 22: 26–38. Anshen, Frank and Mark Aronoff 1999 Using dictionaries to study the mental lexicon. Brain and Language 68: 16–26. Aronoff, Mark 1976 Word-Formation in Generative Grammar. Cambridge, MA: MIT Press. Aronoff, Mark and Frank Anshen 1998 Morphology and the lexicon: Lexicalization and productivity. In: Andrew Spencer and Arnold M. Zwicky (eds.), The Handbook of Morphology, 237–247. Oxford: Blackwell. Baayen, R. Harald 1992 Quantitative aspects of morphological productivity. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1991, 109–149. Dordrecht: Kluwer. Baayen, R. Harald 1993 On frequency, transparency and productivity. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1992, 181–208. Dordrecht: Kluwer. Baayen, R. Harald 1994 Productivity in language production. Language and Cognitive Processes 9: 447–469. Baayen, R. Harald 2001 Word-Frequency Distributions. Dordrecht: Kluwer. Baayen, R. Harald 2009 Corpus linguistics in morphology: Morphological productivity. In: Anke Lüdeling and Merja Kytö (eds.), Corpus Linguistics. An International Handbook. Vol. 2, 899–919. Berlin/New York: Mouton de Gruyter. Baayen, R. Harald and Anneke Neijt 1997 Productivity in context: A case study of a Dutch suffix. Linguistics 35: 565–587. Baayen, R. Harald and Antoinette Renouf 1996 Chronicling the Times: Productive lexical innovations in an English newspaper. Language 72: 69–96. Barðdal, Jóhanna 2008 Productivity. Evidence from Case and Argument Structure in Icelandic. Amsterdam/ Philadelphia: Benjamins. Bauer, Laurie 2001 Morphological Productivity. Cambridge: Cambridge University Press. Bauer, Laurie 2005 Productivity: Theories. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 315–334. Dordrecht: Springer. Booij, Geert 1977 Dutch Morphology. Lisse: De Ridder.

855

856

IV. Rules and restrictions in word-formation I: General aspects Booij, Geert 1996 Inherent versus contextual inflection and the split morphology hypothesis. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1996, 1–16. Dordrecht: Kluwer. Corbin, Danielle 1987 Morphologie dérivationelle et structuration du lexique. Tübingen: Niemeyer. Dal, Georgette 2003 Productivité morphologique: Définitions et notions connexes. In: Georgette Dal (ed.), La productivité morphologique en questions et en expérimentations. Special issue of Langue Française 140: 3–23. Dressler, Wolfgang U. 2003 Degrees of grammatical productivity in inflectional morphology. In: Mark Aronoff and Livio Gaeta (eds.), Morphological Productivity. Special issue of Italian Journal of Linguistics 15(1): 31–62. Dressler, Wolfgang U., Katarzyna Dziubalska-Kołaczyk and Małgorzata Fabiszak 1997 Polish inflection classes within Natural Morphology. Bulletin de la Société Polonaise de Linguistique 53: 95–119. Dressler, Wolfgang U. and Natalia Gagarina 1999 Basic questions in establishing the verb classes of Contemporary Russian. In: Lazar Fleishman, Mikhail Gasparov, Tatiana Nikolaeva, Alexander Ospovat, Vladimir Toporov, Alekseĭ Vigasin, Ronald Vroon and Andrej Zaliznjak (eds.), Essays in Poetics, Literary History and Linguistics. Presented to Viacheslav V. Ivanov on the Occasion of his Seventieth Birthday, 754–760. Moscow: OGI. Dressler, Wolfgang U. and Mária Ladányi 2000 Productivity in word-formation (WF): A morphological approach. Acta Linguistica Hungarica 47: 103–144. Dressler, Wolfgang U. and Anna M. Thornton 1991 Doppie basi e binarismo nella morfologia italiana. Rivista di Linguistica 3(1): 3–22. Dressler, Wolfgang U. and Anna M. Thornton 1996 Italian nominal inflection. Wiener Linguistische Gazette 55–57: 1–24. Evert, Stefan 2004 A simple LNRE model for random character sequences. In: Gérald Purnelle, Cédrick Fairon and Anne Dister (eds.), Le Poids des Mots. Proceedings of the 7th International Conference on Textual Data Statistical Analysis, 411–422. Louvain-la-Neuve: UCL. Evert, Stefan and Marco Baroni 2007 ZipfR: Word frequency distributions in R. In: Annie Zaenen and Antal van den Bosch (eds.), Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics. Posters and Demonstrations Session, 904–911. East Stroudsburg, PA: ACL. Evert, Stefan and Anke Lüdeling 2001 Measuring morphological productivity: Is automatic preprocessing sufficient? In: Paul Rayson, Andrew Wilson, Tony McEnery, Andrew Hardie and Shereen Khoja (eds.), Proceedings of the Corpus Linguistics 2001 Conference. Special issue of UCREL Technical Paper 13: 167–175. Lancaster: Lancaster University. Fleischer, Wolfgang 1975 Wortbildung der deutschen Gegenwartssprache. 4th ed. Tübingen: Niemeyer. Gaeta, Livio 2005 Combinazioni di suffissi in italiano. In: Maria Grossmann and Anna M. Thornton (eds.), La formazione delle parole. Atti del XXXVII Congresso Internazionale di Studi della Società di Linguistica Italiana, 229–247. Roma: Bulzoni. Gaeta, Livio 2007 On the double nature of productivity in inflectional morphology. Morphology 17: 181– 205.

47. Productivity Gaeta, Livio and Davide Ricca 2003 Frequency and productivity in Italian derivation: A comparison between corpus-based and lexicographical data. In: Mark Aronoff and Livio Gaeta (eds.), Morphological Productivity. Special issue of Italian Journal of Linguistics 15(1): 63–98. Gaeta, Livio and Davide Ricca 2006 Productivity in Italian word-formation: A variable-corpus approach. Linguistics 44(1): 57–89. Gardani, Francesco 2013 Dynamics of Morphological Productivity. The Evolution of Noun Classes from Latin to Italian. Leiden: Brill. Haspelmath, Martin 1996 Word-class-changing inflection and morphological theory. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1995, 43–66. Dordrecht: Kluwer. Hay, Jennifer 2000 Causes and consequences of word structure. Ph.D. dissertation, Northwestern University. Hay, Jennifer 2002 From speech perception to morphology: Affix-ordering revisited. Language 78: 527– 555. Hay, Jennifer and Ingo Plag 2004 What constrains possible suffix combinations? On the interaction of grammatical and processing restrictions in derivational morphology. Natural Language and Linguistic Theory 22: 565–596. Iacobini, Claudio and Anna M. Thornton 1992 Tendenze nella formazione delle parole nell’italiano del ventesimo secolo. In: Bruno Moretti, Dario Petrini and Sandro Bianconi (eds.), Linee di tendenza dell’italiano contemporaneo. Atti del XXV Congresso della Società di Linguistica Italiana, 25–55. Roma: Bulzoni. Kageyama, Taro 1982 Word-formation in Japanese. Lingua 57(2–4): 215–258. Kilani-Schoch, Marianne and Wolfgang U. Dressler 2005 Morphologie naturelle et flexion du verbe français. Tübingen: Narr. Lüdeling, Anke and Stefan Evert 2005 The emergence of productive non-medical -itis. Corpus evidence and qualitative analysis. In: Stephan Kepser and Marga Reis (eds.), Linguistic Evidence. Empirical, Theoretical, and Computational Perspectives, 351–370. Berlin/New York: Mouton de Gruyter. Maiden, Martin 1992 Irregularity as a determinant of morphological change. Journal of Linguistics 28: 285– 312. Neuhaus, Heinz Joachim 1973 Zur Theorie der Produktivität von Wortbildungssystemen. In: Abraham P. ten Cate and Peter Jordens (eds.), Linguistische Perspektiven. Referate des VII. Linguistischen Kolloquiums, 305–317. Tübingen: Niemeyer. Plag, Ingo 1998 The polysemy of -ize derivatives: On the role of semantics in word-formation. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1997, 219–242. Dordrecht: Kluwer. Plag, Ingo 1999 Morphological Productivity. Structural Constraints in English Derivation. Berlin/New York: Mouton de Gruyter. Plag, Ingo 2002 The role of selectional restrictions, phonotactics and parsing in constraining suffix ordering in English. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 2001, 285–394. Dordrecht: Kluwer.

857

858

IV. Rules and restrictions in word-formation I: General aspects Plag, Ingo, Christiane Dalton-Puffer and R. Harald Baayen 1999 Morphological productivity across speech and writing. English Language and Linguistics 3: 209–228. Rainer, Franz 1987 Produktivitätsbegriffe in der Wortbildungslehre. In: Wolf Dietrich, Hans-Martin Gauger and Horst Geckeler (eds.), Grammatik und Wortbildung romanischer Sprachen, 187– 202. Tübingen: Narr. Rainer, Franz 1988 Towards a theory of blocking: The case of Italian and German quality nouns. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1988, 155–185. Dordrecht: Foris. Rainer, Franz 2005 Constraints on productivity. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 335–352. Dordrecht: Springer. Scherer, Carmen 2005 Wortbildungswandel und Produktivität. Eine empirische Studie zur nominalen -er-Derivation im Deutschen. Tübingen: Niemeyer. Schultink, Henk 1961 Produktiviteit als morfologisch fenomeen. Forum der Letteren 2: 110–125. Siegel, Dorothy 1977 The adjacency condition and the theory of morphology. In: Mark J. Stein (ed.), Proceedings of the Eighth Annual Meeting of the North East Linguistic Society, 189–197. Amherst, MA: North East Linguistic Society. Štichauer, Pavel 2009 La produttività morfologica in diacronia. I suffissi -mento, -zione e -gione in italiano antico dal Duecento al Cinquecento. Praha: Karolinum. Thornton, Anna M. 1998 Quali suffissi nel “Vocabolario di Base”? In: Federico Albano Leoni, Daniele Gambarara, Stefano Gensini, Franco Lo Piparo and Raffaele Simone (eds.), Ai limiti del linguaggio, 385–397. Roma/Bari: Laterza. Vincent, Nigel 1988 Italian. In: Martin Harris and Nigel Vincent (eds.), The Romance Languages, 279–313. London: Routledge. Wellmann, Hans 1975 Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. 2. Hauptteil: Das Substantiv. Düsseldorf: Schwann. Williams, Edwin 1981 On the notions ‘lexically related’ and ‘head of a word’. Linguistic Inquiry 12: 245–274. Wurzel, Wolfgang U. 1984 Flexionsmorphologie und Natürlichkeit. Berlin: Akademie Verlag.

Livio Gaeta and Davide Ricca, Turin (Italy)

48. Restrictions in word-formation

48. Restrictions in word-formation 1. 2. 3. 4.

Introduction Constraints on word-formation rules Domain-specific restrictions References

Abstract The main factors taken to be responsible for constraining or restricting the application of word-formation rules are surveyed. On the one hand, constraints of a general nature will be discussed which may be due to several distinct reasons ranging from our concrete cognitive abilities to different views of approaching word-formation from a theoretical point of view. On the other, a typology of more specific restrictions will be provided which result from the interaction of the different levels of the linguistic analysis.

1. Introduction Word-formation rules (WFRs) are typically subject to a number of general constraints or more specific restrictions conditioning or limiting their productivity, the latter intended in a broad sense as the possibility of applying to lexical bases serving as an input (see article 47 on productivity for a survey). Rainer (2005a: 335) observes that the question of restrictions only arises for productive WFRs, for which the application domain has to be defined intensionally, i.e. by indicating one or more features that any potential base must or should possess as well as additional factors from outside the pattern itself that may be relevant. For unproductive rules the domain is generally described extensionally by enumerating the set of bases to which the rule applies. However, this does not exclude that intensionally defined features may also synthetically summarize the properties shared by the enumerated bases, especially when the latter are quite numerous. What is more, productivity is likely to be a gradient notion to the effect that in some cases a very low degree of productivity approximates unproductivity. The question of the restrictions on WFRs has been the object of wide investigations ever since, and recent surveys can be found in Bauer (2005) and Rainer (2000, 2005a). As a matter of fact, a big part of the research carried out in word-formation focuses on the restrictions displayed by WFRs. They can be approached by adopting two different, although interwoven, perspectives: theory-driven restrictions or constraints of a general nature, and specific restrictions empirically resulting from the analysis of individual language-specific patterns. The latter also come from the interaction of morphology with the other components of the language. In what follows, a survey of the different aspects of the question will be offered starting (i) with different views of looking at constraints and restrictions and subsequently (ii) developing a typology of restrictions resulting from the interaction of the different levels of the linguistic analysis. In this regard, it must be observed that constraints are

859

860

IV. Rules and restrictions in word-formation I: General aspects usually considered to be those absolute limitations on WFRs which are of a rather general nature while restrictions have a more narrow scope (cf. Rainer 2005a). On the other hand, in the wake of optimality theory constraints can also be taken to be violable, much more limited in scope and hierarchically ordered (cf. Bauer 2001: 126). However, the literature is not always consistent with these distinctions; in this article I will follow Rainer’s distinction and generally speak of constraints with regard to general limits on WFRs which are independent of the particular linguistic level considered while restrictions are held to be of a more reduced scope.

2. Constraints on word-formation rules Generally, there are two possible ways of looking at the question of constraints on WFRs: the first view adopts a top-down perspective, according to which there are constraints due to the format of the grammar and more generally of the language faculty; this view is accordingly competence-oriented. The opposite view is performance-oriented and treats the constraints as resulting in a bottom-up fashion from the way in which our language faculty concretely treats lexical items when they are processed by our cognitive equipment. From the interplay of these two opposite views four possible families of constraints can be identified, which also reflect the historical trend from more competence-oriented approaches typical of the early models of word-formation to the more performance-oriented views characterizing more recent research supported by the use of large electronic corpora.

2.1. Constraints relating to the format of the word-formation rules A first type of constraints directly depends on how WFRs are generally conceived. In this regard, a question which has been discussed at length concerns the input of WFRs, whether they select as a base a possible or an actual word or rather an abstract morpheme. In a nutshell, while nothing seems to hinge a priori on whether a word is possible or actual, i.e. stabilized or entrenched in our mental lexicon, the question of word- or morpheme-based WFRs is much thornier. As to the first point, it can be easily shown that possible but unattested words can constitute the input of WFRs, as in decaffeinate which presupposes the unattested °caffeinate. On the other hand, WFRs may be sensitive to the actual status of the base, as in the cases of paradigmatic word-formation pointed out by Rainer (1993: 29) in which a complex word is formed on the basis of another complex word as in the German compound Volkszählung ‘population census, lit. population count’ → Volkszähler ‘person carrying out the census’, which can only be interpreted with regard to the idiosyncratic meaning of the base. Thus, a conceivable form ??Volksberechner is odd because no base ??Volksberechnung ‘population count’ occurs. Furthermore, bases that are stabilized in the lexicon and give rise to instances of paradigmatic word-formation may also be larger than one word, as in baroque flute → baroque flutist, while ??wooden flutist is odd because wooden flute is not stabilized in the lexicon (cf. Spencer 1988). This latter example calls into play another family of constraints which aim at limiting the access of syntactic patterns and rules below the word level

48. Restrictions in word-formation and go under various names such as “lexical integrity principle”, “no phrase constraint”, etc. (see Gaeta 2006 for a recent survey). As to the second point, Aronoff (1976) launched the slogan of a word-based wordformation intending that the input of WFRs cannot consist of (bound) morphemes but rather must consist of full lexemes (possibly deprived of their inflectional endings). This is motivated by word pairs like aggression/aggressive which are not to be derived from a bound stem *aggress- but rather form a series of derivatives Xion/Xive, in which the adjective is formed on the basis of the action noun. Substantive evidence in support of this analysis is provided among other things by those cases in which an available verb stem cannot be the base of the adjective which is rather formed from the action noun: induction → inductive in spite of induce (cf. Aronoff 1976: 28–30). However, although words intended as lexemes are undoubtedly the prototypical input of WFRs, the restriction against bound stems cannot be universal as is shown on the one hand by bound stems like log- occurring in logic, logistics, etc. On the other, for languages of the polysynthetic or strongly fusional (including the introflecting) type, the concept of lexeme may be much more difficult to define. For instance, in Montagnais, an Algic language spoken in Canada, a strictly morpheme-based approach has been defended (cf. Drapeau 1980). However, morphological templates characterizing non-concatenative processes might be more easily treated in a word-based fashion rather than in a morpheme-based framework which can only accommodate a linear concatenation of morphemes intended as atomic units. For instance, in Hindi/Urdu the anticausative verb form is claimed to be straightforwardly obtained by shortening the root vowel: [XV1V1]V ‘A causes B to happen’ ↔ [XV1]V ‘B happens’, as in maar-/mar- ‘to kill/die’ (cf. Haspelmath 2002: 49). This question has gained more relevance in recent times after Aronoff’s (1994) “morphomic” turn especially from the perspective of a realizational approach to inflectional morphology as suggested by Stump (2001: 2). This view has repercussions for WFRs that manifest themselves in a general tendency towards the maximization of base allomorphy with respect to affix allomorphy (cf. Loporcaro 2012 for a discussion with regard to inflectional morphology). The base allomorphy is accordingly dealt with in terms of different “morphomes”, i.e. concrete formats of a certain lexeme, selected by the affix. In this way, a bias arises towards favoring as input to WFRs existing morphomes, while abstract, underspecified morphemes like stems which increase affix allomorphy are taken to be costly. In brief, the problem is how to deal with cases which allow different interpretations going back to different input bases. This also implies a different format for the affix. For instance, there are at least two different allomorphs for the Italian suffix forming agent or instrument nouns found in stampare ‘to print’ → stampatore ‘printer’ or udire ‘to hear’ → uditore ‘hearer’ in contrast with aggredire ‘to attack’ → aggressore ‘mugger’, distribuire ‘to distribute’ → distributore ‘distributor’, etc. (see article 44 on paradigmatically determined allomorphy: the “participial stem” from Latin to Italian). The latter derivatives select as base the Latinate perfect participle which does not match the actual past participle (cf. *aggresso vs. aggredito, *distributo vs. distribuito). On the other hand, the former may be either analyzed as formed by -tore plus the so-called verbal stem (formed in turn by the root plus the thematic vowel: stampa-tore), or by -ore plus the stem of the past participle (cf. stampat-ore). The latter analysis presents the advantage of minimizing the suffix allomorphy at the expense of the base allomorphy for which two distinct morphomes have to be assumed. In addition, it also accounts for

861

862

IV. Rules and restrictions in word-formation I: General aspects derivatives like diffondere ‘to diffuse’ → diffusore ‘diffuser’ based on the Italian past participle diffuso. However, this choice leaves unexplained on the one hand those cases which require as a base the verbal stem instead of the past participle as scoprire ‘to discover’ → scopritore ‘discoverer’ (cf. the past participle scoperto), and on the other new formations which are based neither on the Latin nor on the Italian participle but are rather formed from the parallel action nouns suffixed with -ione like estorcere ‘to extort’ → estorsore ‘blackmailer’ / estorsione ‘extortion’ (cf. the Latin and Italian past participles extortus/estorto, see Rainer 2001), possibly under the influence of Neo-Latin patterns (see article 91 on word-formation in Neo-Latin). A second type of constraints on the format of WFRs focuses on their possible input or output, and maintains that any WFR should be limited to one single word category in input (unitary base hypothesis, UBH, cf. Aronoff 1976: 47) or in output (unitary output hypothesis, UOH, cf. Scalise 1984: 137). As for the UBH, it has been pointed out that practically any combination of features such as [±N], [±V] and the like has been suggested, which means that “by choosing the appropriate feature system the UBH can be immunized against refutation” (Plag 1999: 48). Furthermore, the process of base selection is most likely guided by rule-specific semantic principles (cf. Plag 1998: 237) rather than by purely abstract features. Finally, WFRs are often sensitive to well-defined lexical sub-domains on the basis of a unitary meaning of the process (cf. Plank 1981: 43–65, Rainer 2005b). At any rate, in many cases the decision of considering two derivatives from different bases (like for instance the denominal fashion-able and the deverbal accept-able) as related to the same or to a different WFR depends on whether we look with favor at the occurrence of affixal homonymy or whether we rather prefer the assumption of rules of semantic extension such as those discussed in article 74 on agent and instrument nouns. While the UBH has been discussed at length, the UOH has received much less discussion and has been, to a large extent, taken for granted. In principle, the UOH opens two different perspectives depending on whether the formal or the semantic aspect of the WFR is in focus. From the formal perspective, the UOH is strictly connected to the degree of allomorphy one is willing to tolerate before considering two affixes as distinct. For instance, in the Italian case discussed above one might be tempted to postulate two different WFRs, a first one selecting a suffix -tore and a second one selecting -ore. In virtue of their identical meaning, however, this choice is likely to be inadequate. On the other hand, nobody would postulate a unitary WFR for two utterly different affixes sharing the same meaning such as -ant in inhabit → inhabitant and -er in sleep → sleeper. In other words, suppletion is generally admitted for lexical bases (e.g., the French pair eau ‘water’ / hydr-ique ‘hydric’, cf. Schwarze 1970) but much less so for derivational affixes. Notice that this does not hold for inflectional rules (e.g., the Hungarian second person singular suffix of the indefinite present takes the form -ol after sibilants or affricates and -(a)sz elsewhere, cf. Carstairs 1988: 70), probably because of the stronger paradigmatic force displayed by inflection vs. that of word-formation. Scalise (1984) suggests that the UOH might be valid only for the formal aspect of WFRs, not for their semantic aspect. However, that the question is much more complex is shown by the Italian suffix -ino which gives rise to several different sorts of derivatives: mare ‘sea’ → mar-ino ‘marine’, tavolo ‘table’ → tavol-ino ‘table-DIM’, bocca ‘mouth’ → bocch-ino ‘mouthpiece’ and stagno ‘tin’ → stagn-ino ‘tinker’. The first case can be explained away as an instance of affixal homonymy because the output word

48. Restrictions in word-formation category is clearly different (an adjective vs. a noun), although this criterion is not uncontroversial as the objections raised against the UBH above also apply here. The other three examples are more complicated, because the purely diminutive value found in tavolino can also be traced back in instrument nouns like bocchino that denote little objects, and even the agent nouns like stagnino generally refer to humble, in a way “little” professions. Even worse, these cases are paralleled by deverbal instrument and agent nouns like, respectively, cancellare ‘to erase’ → cancellino ‘eraser’ and spazzare ‘to sweep’ → spazz-ino ‘street sweeper’, which display exactly the same properties. It is not easy to decide whether all of this should be assigned to the same or to different WFRs. Similar to what we observed above for the UBH, the decision depends on the plausibility of assuming rules of semantic extension; as an alternative, one might also think of a relationship in terms of family resemblance of a Wittgensteinian kind among the several nominal types that can in any case be kept apart from the adjectival homonym (see article 59 on schemata and semantic roles in word-formation). Finally, a further constraint generally assumed is the open-class base hypothesis which requires that only major lexical classes can be input of WFRs, namely nouns, verbs, adjectives and adverbs. This excludes, for instance, adpositions and pronouns from being involved in WFRs, which forces an analysis of certain patterns like German hinauf ‘thereon’, darunter ‘there:below’, etc. as resulting from a process different from word-formation proper. Moreover, it does not lead us to expect to find cases like Spanish le ‘her’ → leísmo ‘use of the form le for direct objects’, in which a pronoun serves as the input of a WFR. However, one can conclude that the major word classes represent the most common or prototypical input of WFRs, although this restriction probably has to be related to the main function of vocabulary enrichment typical of word-formation. In this light, words belonging to the minor word classes usually display grammatical meaning which is only in restricted cases salient enough to be used in word-formation as shown by the Spanish example mentioned above or by numerals (see article 87 on denumeral categories).

2.2. Constraints relating to general properties of the grammar In the light of its general value, the last constraint might also be treated in this section, which discusses constraints depending on general properties usually held to shape our language faculty. One such property is expressed by the compositionality or Frege’s principle, because the German logician Gottlob Frege is generally credited for its first modern formulation (but see Klos 2011): it requires that the meaning of a complex word resulting from a WFR be a function of the meaning of the rule and the base. Against Frege’s principle, clear cases of analogical formations have been mentioned which require a holistic reference to another complex word like the German compound Doktormutter ‘female thesis supervisor’ with regard to its male counterpart Doktorvater. In general, a holistic approach has to be assumed when affix substitution occurs like the Italian verb svitare ‘to unscrew’ which can only be interpreted by making reference to a previous avvitare ‘to screw’ (cf. *vitare), or with instances of bracketing paradoxes like multiconfessional which is formally derived by prefixation [[multi]confessional] but semantically requires the analysis [[multiconfession]al]. At any rate, this anisomorphism

863

864

IV. Rules and restrictions in word-formation I: General aspects between form and meaning can be treated in terms of the paradigmatic relations mentioned above and is not substantially different from a strictly compositional approach on condition that the lexical status of the pattern is duly taken into account. In other terms, the compositionality and the holistic approach simply reflect the two different routes followed by the speakers when they access complex words, namely decomposition or full lexical access (cf. Baayen and Schreuder 2003 for a recent survey). Further constraints focus on the limits imposed on WFRs which result from the interaction with other components of the language. In particular, general trends favoring haplology have been observed for many languages, which block the application of a WFR if a phonological string is replicated by the addition of an affix. For instance, the Italian deanthroponymic suffix -iano is normally blocked when the base ends with the same string: Gadda → gaddiano, but Flaiano → ??flaianiano. In spite of the apparently universal character of this tendency towards the avoidance of cacophonic repetitions, formulating a general rule is not an easy task (see section 3.6). A second more debated case concerns recursion which is normally widely present in syntax, but much less so in word-formation. Recursion seems to be generally possible in compounding (although languages may differ as to its extent) but much more restricted in affixation. In contrast to syntactic recursion, recursion in word-formation is strongly limited by two aspects: on the one hand, WFRs often are property-changing, which prevents their immediate reapplication to the output. On the other, the systematic reapplication of WFRs leads to long chains of morphemes which may present problems from the viewpoint of their processing, especially when property-changing affixes occur as, for instance, in organizationalization. When the last two factors do not intervene, recursion can be generally observed as in the case of the Italian evaluatives casa ‘house’ → cas-etta ‘small house’ → cas-ett-ina ‘small small house’. Finally, while the reapplication of two identical suffixes seems quite rare (again with the remarkable exception of evaluative suffixes as in the colloquial Spanish examples ahora ‘now’ → ahor-it-ita ‘right now’, amigo ‘friend’ → amig-az-azo ‘close friend’, cf. Rainer 1993: 108), prefixes are in general more liberal, probably because they are mostly not property-changing, as in re-rewrite and the like. Finally, general constraints can also come from the interaction with factors external to the language faculty but of high relevance with regard to the function of lexical enrichment generally assigned to word-formation. A first constraint has to do with the demand of new words which is of greatest importance for those WFRs which are more connected with the naming function rather than with other functions carried out by WFRs like the mere transcategorization. This is clearly the case with WFRs forming agent or instrument nouns which presuppose the existence of a certain profession or device. This fact contributes to a large extent to shape our lexicon as the result of our cultural historical development and to motivate the varied degree of acceptability of certain formations which synchronically lack a referent like Spanish arzobispa lit. ‘archbishop (fem.)’, calienta-ojos lit. ‘eyes-warmer’, etc. (cf. Rainer 1993: 113). The possible unacceptability of these formations is likely to be guided not by grammatical – i.e. competence-oriented – principles, but rather by performance-oriented conditions, also connected with our world knowledge (see section 3.6 and article 53 on dissimilatory phenomena in French word-formation). Similar observations also hold for a constraint such as neophobia which has been invoked to account for the low acceptability of new formations simply because they are

48. Restrictions in word-formation unusual words (cf. Gyurko 1971 on Spanish). This is especially the case with neologisms which are launched in creative writing (intending on the one hand literary works and on the other products which involve the conscious manipulation of language like advertisements). What appears more acceptable in certain contexts allowing for more creativity may be rejected in contexts requiring a stricter subscription to a shared norm.

2.3. Constraints relating to the particular format of the grammar While the former constraints can be considered to be theory-independent and therefore universal, the constraints discussed in this section are strictly related to a certain format attributed to the grammar. For instance, in the late seventies a number of locality conditions were formulated which aimed to restrict the number of features visible to a certain WFR in a given domain like Siegel’s (1977) adjacency principle, Williams’ (1981) atom condition, or Kiparsky’s (1982) bracket erasure convention (cf. Plag 1999: 45–46 for a brief survey). As repeatedly emphasized in the literature, these constraints were flawed by serious problems, due among others to an insufficient empirical basis underlying their formulation. At any rate, when the interest in generative grammar sailed towards theoretical shores different from conditions on rules, these constraints were simply abandoned. A similar problem concerns the binary branching hypothesis (cf. Scalise 1984: 146–151), which – far from being universal – is systematically falsified by coordinative compounds like German-French-English (corporation) and therefore best to be viewed as consequence of the semantics of determinative compounds rather than as the result of a formal constraint on the grammar format (cf. Barri 1977). Stratal conditions on WFRs deserve a partially different discussion. The latter were originally formulated to account for the well-known fact that the WFRs may be sensitive to certain sets of lexical items (lexical strata, cf. Saciuk 1969) characterized for instance in etymological terms (e.g., “of Latin origin”). This idea was further expanded by assuming at least two different and serially ordered derivational strata or levels to which the affixes belong (cf. Siegel 1979). Accordingly, the properties shared by different groups of affixes result from the specific level assigned to them and need not be specified for the individual WFRs. Against the stratal view it has been generally objected that it is largely impossible to account for the severe restrictions on the combinability of the affixes especially when they belong to the same level (cf. Fabb 1988). Furthermore, in several cases affixes have to be assigned to more than one stratum in order to account for their selective and allomorphic properties. This weakens the stratal approach considerably. On the other hand, there are surely languages in which the lexicon is sharply compartmentalized into separate strata (see article 183 on Maltese), displaying robustly different properties. For instance, in German the native WFRs do not generally produce prosodic changes on the bases, while the non-native WFRs are largely characterized by stress shifts: Wíssenschaft ‘science’ → Wíssenschaftler ‘scientist’ vs. Térror ‘terror’ → Terroríst ‘terrorist’. An orthogonal question relates to the source of the selective properties, whether they must be sought in the WFRs, i.e. in the affixes, as generally assumed by those who support a stratal approach, or in the bases, as maintained by Plag (1999: 67–76) who defends a generalized base-driven approach. Accordingly, the selective correlation be-

865

866

IV. Rules and restrictions in word-formation I: General aspects tween the German non-native noun-forming suffix -ität and the non-native adjectival bases as in banal ‘banal’ → Banalität ‘banality’ is taken to be driven by the latter. This approach is more economic than the former because it does not require us to assume a complex mechanism of rule-by-rule blocking to account for the oddness of the conceivable form *Banalheit. Furthermore, it also accounts for cases in which the non-native suffix is selected by native bases like schwul ‘gay, queer’ → Schwulität ‘embarrassing situation’, which violate a rigid stratal view (see section 2.4). However, doublets of derivatives from the same base should in principle be excluded, but exceptions of this kind are not uncommon, as shown by cases like absurd ‘absurd’ → Absurdität/Absurdheit ‘absurdity’, naiv ‘ingenuous’ → Naivität/Naivheit ‘ingenuity’, etc.

2.4. Constraints depending on lexical accessibility The problems relating to the lexical strata and their ordering have been approached recently from a completely different perspective, which has been termed by Plag complexity-based ordering. This refers to the general properties displayed by the lexical items when they are processed by our cognitive capacities. In particular, Hay (2000, 2002) suggests that more easily parsable affixes should be normally less restricted than less easily parsable ones and accordingly should occur more externally. On the other hand, words containing less easily parsable affixes are more likely to be directly accessed as units entrenched in our mental lexicon. The ease of parsability, or its counterpart lexical entrenchment, are influenced by factors like frequency, especially relative frequency, i.e. the frequency of a derivative with respect to its base, and phonotactics. The latter refers to the occurrence of less frequent sound clusters resulting from the combination of two morphemes which are a better cue for detecting and parsing a morphological boundary than sound clusters occurring frequently inside morphemes. Relying on these parameters, Hay provides a hierarchy expressing the combinability force of the individual affixes with regard to the base: affixes scoring higher in terms of relative frequency and phonotactics are likely to be placed closer to the base. In spite of the attractiveness of this entirely performance-oriented approach, it is empirically insufficient because of the fact that specific selectional restrictions are also required to account for the number of impossible combinations normally observed (cf. Hay and Plag 2004). A second family of constraints relating to lexical accessibility goes under the broad label of blocking, although substantially different concepts are to be understood here. First of all, one must distinguish between homonymy and synonymy blocking: the first type has been suggested to account for the non-occurrence of denominal verbs like spring → *to spring, fall → *to fall parallel to summer → to summer, winter → to winter because of the mere presence in our mental lexicon of the corresponding homophonous verbs. However, the non-occurrence – at least in British English, see Bauer (1983: 97) – of *to autumn in the absence of any homonymous verb casts doubt on the reliability of this explanation. More in general, “this approach fails to expound why language tolerates innumerable ambiguities, but should avoid this particular one” (Plag 1999: 50). Much more relevance has been attributed to the second type of synonymy blocking. Two cases have to be distinguished: word or (perhaps slightly emphatically) Paul’s blocking, in which the occurrence of one synonymic lexeme in our mental lexicon is

48. Restrictions in word-formation made responsible for the non-occurrence of a possible derivative as in the classic example of thief blocking the formation of ??stealer. In this case, which is a true instance of lexical blocking as already envisaged by Hermann Paul (1896: 704), the accessibility of the established word is of crucial relevance: as argued by Rainer (1988: 163), the blocking force of the established word is a direct function of its frequency and of the productivity of the intervening WFR. On the other hand, the blocked word is not really illformed, but a potentially usable word – and indeed often attested – provided that for some reason a speaker fails to retrieve the established word responsible for the blocking and/or is in search of a particular meaning effect as in scene-stealer (cf. Rainer 2012). Notice that potential words like ??stealer are different from possible words like °caffeinate seen in section 2.1 also because they are usually inert to further derivation as shown by the impossibility of ??stealerless with regard to pairs like leader → leaderless, teacher → teacherless, etc., while °caffeination is a possible word exactly like its base. In this regard, Rainer (2005a: 337) formulates a possible-base constraint according to which bases of WFRs must be possible words, while merely potential words are excluded. The second case is more complicated and can be referred to as rule or Pāṇini’s blocking, because the non-occurrence of a derivative is accounted for by the fact that a synonymous pattern takes precedence provided that both patterns are productive. This reminds us very closely of the so-called Pāṇini’s, or elsewhere principle whereby the application of a more specific rule blocks that of a more general one, as already envisaged by the Indian grammarian Pāṇini (cf. Kiparsky 1973; for a different view see Giegerich 2001). Rainer (1988) suggests to account in these terms for the lexical domain of the German quality nouns formed on the basis of end-stressed adjectives. The latter select different suffixes depending on a set of features restricting in a cumulative way their scope of application. Thus, the suffix -heit normally combines with end-stressed adjectives: gewiss ‘sure’ → Gewissheit ‘sureness’, ordinär ‘vulgar, common’ → Ordinärheit ‘vulgarity’, etc., unless they display a learned flavor; in this case they select -ität: binär ‘binary’ → Binarität ‘binarity’/??Binärheit, cf. also ??Ordinarität, only possible with a mathematical meaning: ‘the property of being a common event’. Finally, if a learned, endstressed adjective ends with the bound stem -phil, it selects the suffix -ie: xenophil ‘xenophile’ → Xenophilie ‘xenophilia’/??Xenophilität/??Xenophilheit. Notice that the simple occurrence of an ending /-fil/ does not trigger the application of -ie and the superordinate preference for -ität applies in the light of the learned flavor: monofil ‘unifilar’ → Monofilität ‘unifilarity’/??Monofilie/??Monofilheit. Although they rely on a similar synonymic mechanism, Paul’s and Pāṇini’s blocking are two completely different phenomena because the former refers to the degree of entrenchment of a word in our mental lexicon which is measurable in frequency terms, while the latter is due to the selective specificity of two rules applying to (portions of) the same set of lexical bases. In fact, Rainer (2005a: 337) observes that Pāṇini’s blocking may also “apply even when no actual blocking word formed according to the rival pattern exists”. Furthermore, while a word like ??Xenophilheit can be said to be illformed because of the conditions on the selected base, ??stealer is not ill-formed stricto sensu, as discussed above. At any rate, frequency may also play a role in the case of Pāṇini’s blocking as shown by the occurrence of doublets of derivatives from the same end-stressed adjectives if the latter “have become part of a more colloquial register” (Rainer 2005: 338): debil ‘stupid’ → Debilheit ‘stupidity’, beside established Debilität, skurril ‘droll’ → Skurrilheit ‘drollery’, beside established Skurrilität, etc. (see the pairs

867

868

IV. Rules and restrictions in word-formation I: General aspects Absurdität/Absurdheit, etc. mentioned in section 2.3). Clearly, the property of becoming part of a more colloquial register is also connected with an increase of frequency, which has the effect of relaxing the strict condition on learnedness. On the other hand, Rainer (1988: 172) has suggested that frequency may interfere in cases of affix rivalry systematically blocking a derivative when a clearly more frequent competitor occurs. This is allegedly the case of the Italian deadjectival nouns formed with the two highly productive suffixes -ismo and -ità, whereby frequent quality nouns selecting -ismo (by dropping the ending -ico of the base) are said to block the possible formation of -ità derivatives: cinico ‘cynic’ → cinismo ‘cynism’/??cinicità, patriottico ‘patriotic’ → patriottismo ‘patriotism’/??patriotticità, etc. Although Rainer maintains that this should be interpreted as a case of Paul’s blocking, the high productivity of the two WFRs might be regarded as a clue that indeed an intertwining of the two types of blocking is going on here, because the frequency of the individual derivatives cannot be kept totally distinct from the availability of the two synonymic WFRs expressed by productivity. The latter is in fact related to frequency (see article 47 on productivity). In other words, a productive WFR can be blocked by the intervention of another productive WFR forming more frequent derivatives.

3. Domain-specific restrictions Let us now turn to specific restrictions relating to the different levels of linguistic analysis which have been pointed out in the literature. The discussion will be in some cases brief because several issues have been already touched upon in the foregoing sections.

3.1. Phonological restrictions Besides the constraint on haplology mentioned above in section 2.2, there are generally three types of restrictions of a phonological nature. First, there may be selectional restrictions of a positive or a negative value relating to the segmental make-up of the base. For instance, a certain stem ending or the occurrence of certain segments within a stem may favor or hinder the combination with a certain suffix: respectively, the suffix -eer preferably selects bases ending with a dental voiceless obstruent: musketeer, profiteer, etc. (cf. Rainer 2005a: 344), while the Dutch noun-forming suffix -te as in koelte ‘coolness’ cannot be added to adjectives ending in a vowel (cf. Bauer 2001: 129). More complex and much discussed especially from the viewpoint of an autosegmental approach to phonology is the case of the Latin suffix -ālis in capitālis ‘capital’, nāvālis ‘naval’, etc., which takes the form -āris if the base contains a lateral: lūnāris/*lūnālis ‘lunar’ (cf. Cser 2010). Notice that prefixes are generally held to be far less sensitive to base-driven phonological restrictions (cf. Rainer 2000: 881). Second, the selectional restrictions may relate to the prosodic shape of the base; in particular word stress may play a role guiding, for instance, the positive selection of the suffix -al with regard to verbs stressed on the final syllable: arrival, rebuttal, etc. On the other hand, this restriction might also be seen as due to the preference for Latinate prefix-root verbs, which all happen to have final stress (cf. Malicka-Kleparska 1992:

48. Restrictions in word-formation 437). Word stress is relevant for the derivation of circumfixal abstract nouns in German, insofar as only bases displaying initial stress are possible: klatschen ‘to clap’ → Geklatsche ‘clapping’ but applaudieren ‘to clap’ → *Geapplaudiere, etc. Third, the length of the base computed in syllables may be relevant, as in the suffix -C1oj ‘-ish’ found in the Mayan language Tz’utujil spoken in Guatemala, which only selects monosyllabic adjectives rax-roj ‘greenish’, q’eq-q’oj ‘blackish’ (cf. Bauer 2001: 129). The stress position and the syllable number may also form a joint restriction as in the case of the suffix -eer mentioned above which preferably selects bisyllabic trochaic bases: cameleer vs. *giraffeer, profiteer vs. *gaineer, racketeer vs. *fraudeer, etc. (cf. Rainer 2005a: 344).

3.2. Morphological restrictions We have already seen some examples of morphological restrictions above when stratal constraints were discussed. In general, three types of morphology-driven restrictions can be determined. First, the base can belong to a class which is morphologically welldefined by means of stratal constraints or some other morphological feature like gender as in the case of the Hebrew sarcastic diminutive of the form CCaCCaC which can only be formed from masculine nouns: zakan ‘beard’ → zkankan ‘little beard’ (cf. Bauer 2001: 130). In this regard, the reference to “etymological” information such as “of foreign origin” and the like mentioned in section 2.3 above might also be labeled as morphological (or lexical, possibly) because “most speakers do not have in their mental lexicons information about the sources of the words they use” (Bauer 2001: 130). Rather, the latter “are perceived as belonging to various synchronic classes” which “mimic etymological provenance (because that is their origin), but the mental listing involves assigning them to classes which are as random as (perhaps more random than) gender classes” (Bauer 2001: 131). In fact, we have also seen above that the etymological categorization often “leaks”, insofar as words of a wrong etymological type are included. Notice incidentally that reference to some information about the base, including when the latter already contains an affix, comes into conflict with those approaches which are typically represented by Anderson’s (1992) “a-morphous morphology”, because they assume that a process of bracketing erasure cancels any morphological information contained in the base, which is therefore inaccessible to further WFRs. This view is too radical, as shown by the highly productive selectional solidarity of -ize and -ation which does not hold when the ending has no morphological status: to realize → realization, but to surmise → *surmisation, etc. (cf. Rainer 2005a: 345). Second, the base may have to show a particular morphological structure. In this regard, examples are found in which an affix only applies to complex bases, as the Punjabi prefix gair- ‘un-’, which only selects derived bases, e.g., deriving gairsarkaarii ‘non-governmental’ from sarkaarii ‘governmental’, itself derived from sarkaar ‘government’ (cf. Bauer 2001: 131). The third possible case of morphological restrictions relates to the presence or the absence of a particular affix in the base, as in the case of the Dutch female suffix -ster which requires the presence of the suffix -aard ‘-er’ in the base: wandelaarster ‘female hiker’ (cf. Bauer 2001: 131), or, conversely, the German suffix -heit which can be combined with compounds (e.g., Schreib-faul-heit ‘the quality of being a bad correspondent;

869

870

IV. Rules and restrictions in word-formation I: General aspects lit. write-lazy-ness’), prefixed adjectives (un-gleich ‘un-equal’ → Ungleichheit ‘inequality’) or circumfixed past participles (ge-schloss-en ‘closed’ → Geschlossenheit ‘closure’), but does not generally apply to already suffixed bases as shown by freund-lich ‘friend-ly’ → *Freundlichheit vs. Freundlichkeit ‘friendliness’, ein-sam ‘lonely’ → *Einsamheit vs. Einsamkeit ‘loneliness’, in which the allomorph -keit has to be selected (cf. Aronoff and Fuhrhop 2002: 459), although sparse exceptions like blei-ern ‘lead-en’ → Bleiernheit ‘leaden-ness’ are attested. A positive correlation can give rise to the phenomenon of potentiation when the productivity of an affix is reinforced by the productivity of the affix in the base (cf. Williams 1981: 250). On the other hand, a negative correlation has been referred to in terms of closing morphemes, namely morphemes that “‘close’ the construction to other morphemes” (Nida 1949: 85, cf. Marle 1985: 234–238 for Dutch, and Aronoff and Fuhrhop 2002 for German). The closing property is considered an idiosyncratic feature of the individual affixes which has the effect of pre-empting the application domain of another affix as, for instance, in the case of the Bulgarian suffix -ski forming denominal adjectives like pisatel ‘writer’ → pisatelski ‘writer’s’ which cannot be further derived although there are recent Russian borrowings like russk-ost ‘Russian-like style’, svet-sk-ost ‘worldly-minded style’, etc. (see article 54 on closing suffixes for a detailed discussion).

3.3. Syntactic restrictions Although at first sight one might expect to observe a number of clear-cut restrictions resulting from the interaction of morphology and syntax, in practice this turns out to “be more illusory than real” (Bauer 2001: 133). The often mentioned importance of the syntactic category of the base as a milestone for the WFRs has been overestimated, as pointed by several authors (see for instance Plank 1981: 43–45), while Plag (1998: 237) even dismisses it as “a by-product of the semantics of the process”. More generally, allegedly syntactic restrictions can always be reinterpreted as morphological (or possibly lexical) in nature because of the way in which “a word is used depends to some extent upon the class it belongs to” and therefore “it might seem preferable to merge these two” (Bauer 2001: 133). However, in a more loose parlance one may treat under the label of syntactic restrictions those instances which refer to abstract properties of the bases which have an immediate effect on their syntactic behavior. One such case is represented by examples in which the argument structure of the verbal base is involved as suggested by the so-called “constructional” approaches to argument structure according to which “meaning resides in the syntactic context” (cf. Levin and Rappaport Hovav 2005: 18). For instance, it has been repeatedly claimed that the suffix -able normally combines only with transitive verbs: visitable vs. *goable, observable vs. *lookable, etc. On the other hand, depending on the theory, transitivity has also be seen as a semantic, not a syntactic feature (cf. Rainer 2005a: 348). Similarly, in Apalai, a Cariban language spoken in Brazil, two different suffixes are used to form agent nouns, -ne with transitive and -kety with intransitive verbs (cf. Bauer 2001: 133): parata wo-ne ‘rubber cutt-er’ and wa-kety ‘danc-er’. Finally, a particularly tricky example is provided by the Australian language Diyari in which the attributive suffix -kaɲʧi is used on “the set of common nouns which take the inchoative verbalizer and appear in the ergative case when used predicatively” (Austin 1981: 39) as in ŋudu ‘power’ → ŋudukaɲʧi ‘powerful one’.

48. Restrictions in word-formation

3.4. Lexical restrictions Since the role of the lexicon is ubiquitous in word-formation, it is difficult to identify genuinely lexical restrictions. One might conceive of two different sources for lexical restrictions. First, considering that unproductive WFRs normally give rise to shorter or longer lists of words in our mental lexicon, these lists have been generally assumed to form the lexical restrictions of the WFRs. Particular blatant are those cases in which the domain of a WFR is restricted to one or two single entries, as in bishopric, the only English word testifying of a suffix -ric, or laughter and slaughter which testify of the suffix -ter. Similarly, in Punjabi the nominalizing suffix -aapaa is found only in the noun kuʈaapaa ‘beating’ from the verb kuʈʈ ‘to beat’, and in Abkhaz the intensifier -samsal appears only in the adjective àyk°ac°’a-samsal ‘very black’ (Bauer 2001: 135). Second, lexical restrictions may relate to class properties of the bases which have to do with their status within our mental lexicon. One example is given by the stratal conditions repeatedly discussed above, which can also be treated as lexical in nature if one thinks that they refer to the architecture of our mental lexicon rather than to formspecific properties of the words. A similar conclusion can also be reached if a “projectionist” approach to argument structure is adopted, which maintains that the latter results from the projection onto syntax of lexically specified information contained in the verb (cf. Levin and Rappaport Hovav 2005: 18).

3.5. Semantic restrictions Similarly to the lexicon, also the role of semantics is ubiquitous because any WFR displays a meaning side which selects portions of the lexicon on the basis of their content. In this light, since a projectionist approach to argument structure may also be interpreted as involving a semantic restriction on the possible input, this would subsume under semantics all the examples discussed above. In general, semantic restrictions are invoked when highly specific meaning aspects of the base domain are required in order to delimit the input of a WFR. For instance, the Italian suffix -eto combines only with plant or fruit names and forms nouns referring to the corresponding grove: canna ‘reed’ → canneto ‘grove of reeds’, arancia ‘orange’ → aranceto ‘orange grove’, etc. Similarly, in the Australian language Mangarayi the ethnic suffix -ŋuŋuŋ combines only with bases referring to a place name or a language: Guwiɲilen-ŋuŋuŋ ‘Queenslander’ (cf. Bauer 2001: 134). A certain debate has been kindled by the question of the boundary between word meaning and world knowledge insofar as this is relevant for WFRs. For instance, the reversative prefix un- can only be applied to verbal bases displaying a reversible meaning: unfold, unscrew vs. *unswim, *unkill. While for the unacceptability of unswim a true semantic restriction may be invoked because the atelic process of swimming cannot give rise to any reversative interpretation, the unacceptability of unkill might be due to our encyclopedic knowledge which tells us that death is an irreversible state. The latter condition, however, might not hold in other possible worlds: for instance in the jargon of video-game aficionados unkill is a possible verb consisting in bringing back to life a character.

871

872

IV. Rules and restrictions in word-formation I: General aspects

3.6. Pragmatic restrictions As briefly hinted at in the previous section, a lively debate has focused on the possible distinction of the word meaning from our encyclopedic knowledge which is necessary in order to correctly understand “the nature of the real-world referent of the word” (Bauer 2001: 135) when it is used in a certain context. The latter perspective can also open the door for investigating the role of pragmatically-oriented restrictions on WFRs. For instance, the Dyirbal suffix -ginay meaning ‘covered with, full of’ is normally restricted to bases denoting something dirty or unpleasant as in gunaginay ‘covered with faeces’. Similarly, in Kusaie, an Austronesian language spoken in Micronesia, the inchoative suffix -yak combines typically with names of insects (with the meaning ‘to become infested with’) or diseases (with the meaning ‘to be badly affected by’). In Kannada, the adverbializing suffix -va:ra is generally restricted to a bureaucratic language, as in ko:mu-va:ra ‘community-wise’ (Bauer 2001: 135 and further references there). This reminds us of the German suffix -ität seen in section 2.4 above, which is sensitive to the stylistic register in which the base is employed. A full-fledged system of restrictions relating to stylistic-sociolinguistic features is provided by the Javanese “cromification” rules (cf. Becker 1990: 20–23). Finally, restrictions of an “aesthetic” nature have been invoked for the speakers’ rejection of certain words which are theoretically well-formed: for instance, Guilbert (1975: 191) discusses an aesthetic reaction against very long words in French as the reason preventing the formation of the adverb *oppositionellement ‘oppositionally’ from its base oppositionel ‘oppositional’. In this vein, an aesthetic reason might be made responsible for the haplological blocking of *sillily in English. However, in the absence of solid investigations these observations have an impressionistic flavor.

4. References Anderson, Stephen R. 1992 A-Morphous Morphology. Cambridge: Cambridge University Press. Aronoff, Mark 1976 Word-Formation in Generative Grammar. Cambridge, MA: MIT Press. Aronoff, Mark 1994 Morphology by Itself. Cambridge, MA: MIT Press. Aronoff, Mark and Nanna Fuhrhop 2002 Restricting suffix combinations in German and English: Closing suffixes and the monosuffix constraint. Natural Language and Linguistic Theory 20: 451–490. Austin, Peter 1981 A Grammar of Diyari, South Australia. Cambridge: Cambridge University Press. Baayen, Harald and Robert Schreuder 2003 Morphological Structure in Language Processing. Berlin/New York: Mouton de Gruyter. Barri, Nimrod 1977 Giving up word formation in structural linguistics. Folia Linguistica 11: 13–37. Bauer, Laurie 1983 English Word-formation. Cambridge: Cambridge University Press. Bauer, Laurie 2001 Morphological Productivity. Cambridge: Cambridge University Press.

48. Restrictions in word-formation Bauer, Laurie 2005 Productivity: Theories. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 315–334. Dordrecht: Springer. Becker, Thomas 1990 Analogie und morphologische Theorie. München: Fink. Carstairs, Andrew 1988 Some implications of phonologically conditioned suppletion. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1988, 67–94. Dordrecht: Kluwer. Cser, András 2010 The -alis/-aris allomorphy revisited. In: Franz Rainer, Wolfgang U. Dressler, Dieter Kastovsky and Hans Christian Luschützky (eds.), Variation and Change in Morphology. Selected Papers from the 13th International Morphology Meeting, Vienna, February 2008, 33–51. Amsterdam/Philadelphia: Benjamins. Drapeau, Lynn 1980 Le rôle des racines en morphologie dérivationnelle. Recherches Linguistiques à Montréal 14: 299–326. Fabb, Nigel 1988 English suffixation is constrained only by selectional restrictions. Natural Language and Linguistic Theory 6: 527–539. Gaeta, Livio 2006 The lexical integrity principle as a constructional strategy. Lingue e Linguaggio 5(1): 1– 16. Giegerich, Heinz J. 2001 Synonymy blocking and the elsewhere condition: Lexical morphology and the speaker. Transactions of the Philological Society 99: 65–98. Guilbert, Louis 1975 La créativité linguistique. Paris: Larousse. Gyurko, Lanin A. 1971 Affixal negation in Spanish. Romance Philology 25(2): 225–239. Haspelmath, Martin 2002 Understanding Morphology. London: Arnold. Hay, Jennifer 2000 Causes and consequences of word structure. Ph.D. dissertation, Northwestern University. Hay, Jennifer 2002 From speech perception to morphology: Affix-ordering revisited. Language 78: 527– 555. Hay, Jennifer and Ingo Plag 2004 What constrains possible suffix combinations? On the interaction of grammatical and processing restrictions in derivational morphology. Natural Language and Linguistic Theory 22: 565–596. Kiparsky, Paul 1973 “Elsewhere” in phonology. In: Stephen R. Anderson and Paul Kiparsky (eds.), A Festschrift for Morris Halle, 93–106. New York: Holt, Rinehart and Winston. Kiparsky, Paul 1982 Lexical morphology and phonology. In: In-Seok Yang (ed.), Linguistics in the Morning Calm, 3–91. Seoul: Hanshin. Klos, Verena 2011 Komposition und Kompositionalität. Möglichkeiten und Grenzen der semantischen Dekodierung von Substantivkomposita. Berlin/New York: de Gruyter. Levin, Beth and Malka Rappaport Hovav 2005 Argument Realization. Cambridge: Cambridge University Press.

873

874

IV. Rules and restrictions in word-formation I: General aspects Loporcaro, Michele 2012 Stems, endings and inflectional classes in Logudorese verb morphology. Lingue e Linguaggio 11(1): 5–34. Malicka-Kleparska, Anna 1992 Against phonological conditioning of WFRs. In: Jacek Fisiak and Stanislaw Puppel (eds.), Phonological Investigations, 423–442. Amsterdam/Philadelphia: Benjamins. Marle, Jaap van 1985 On the Paradigmatic Dimension of Morphological Productivity. Dordrecht: Foris. Nida, Eugene A. 1949 Morphology. The descriptive analysis of words. Ann Arbor: University of Michigan Press. Paul, Hermann 1896 Über die Aufgaben der Wortbildungslehre. Sitzungsberichte der philosophisch-philologischen und der historischen Classe der königlichen bayerischen Akademie der Wissenschaften zu München, Jahrgang 1896, 696–713 [Reprinted in: Leonhard Lipka and Hartmut Günther (eds.), Wortbildung, 17–35. Darmstadt: Wissenschaftliche Buchgesellschaft, 1981]. Plag, Ingo 1998 The polysemy of -ize derivatives: On the role of semantics in word-formation. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1997, 219–242. Dordrecht: Kluwer. Plag, Ingo 1999 Morphological Productivity. Structural Constraints in English Derivation. Berlin/New York: Mouton de Gruyter. Plank, Frans 1981 Morphologische (Ir-)Regularitäten. Tübingen: Narr. Rainer, Franz 1988 Towards a theory of blocking: The case of Italian and German quality nouns. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1988, 155–185. Dordrecht: Foris. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rainer, Franz 2000 Produktivitätsbeschränkungen. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 877–885. Berlin/New York: Mouton de Gruyter. Rainer, Franz 2001 Compositionality and paradigmatically determined allomorphy in Italian word-formation. In: Chris Schaner-Wolles, John Rennison and Friedrich Neubarth (eds.), Naturally! Linguistic studies in honour of Wolfgang Ulrich Dressler presented on the occasion of his 60th birthday, 383–392. Torino: Rosenberg & Sellier. Rainer, Franz 2005a Constraints on productivity. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 335–352. Dordrecht: Springer. Rainer, Franz 2005b Semantic change in word formation. Linguistics 43(2): 414–441. Rainer, Franz 2012 Morphological metaphysics: Virtual, potential, and actual words. Word Structure 5: 165– 182. Saciuk, Bohdan 1969 The stratal division of the lexicon. Papers in Linguistics 1: 464–532.

48. Restrictions in word-formation

875

Scalise, Sergio 1984 Generative Morphology. Dordrecht: Foris. Schwarze, Christian 1970 Suppletion und Alternanz im Französischen. Linguistische Berichte 6: 21–34. Siegel, Dorothy 1977 The adjacency condition and the theory of morphology. In: Mark J. Stein (ed.), Proceedings of the Eighth Annual Meeting of the North East Linguistic Society, 189–197. Amherst, MA: North East Linguistic Society. Siegel, Dorothy 1979 Topics in English Morphology. New York: Garland. Spencer, Andrew 1988 Bracketing paradoxes and the English lexicon. Language 64: 663–682. Stump, Gregory T. 2001 Inflectional Morphology. A Theory of Paradigm Structure. Cambridge: Cambridge University Press. Williams, Edwin 1981 Argument structure and morphology. The Linguistic Review 1: 81–114.

Livio Gaeta, Turin (Italy)

V. Rules and restrictions in word-formation II: Special cases 49. Argument-structural restrictions on word-formation patterns 1. 2. 3. 4. 5.

Introduction Word-formation and the syntax-morphology interface Structural principles Conditions and operations References

Abstract The implementation of argument-structural effects on word-formation is a vital aspect in modeling the lexical system and the interface between morphology and syntax. The current article provides an overview of theoretical perspectives in the field and presents analyses of structural principles holding in the domain. A number of test cases relating to fundamental operations, e.g., in compounding and nominalization are discussed, as well as specific conditions restricting the formation of morphologically complex words.

1. Introduction The relation between argument structure (AS) and word-formation patterns is a central topic in the theoretical description of the structural operations available in language. In particular, a correspondence between full sentences and certain types of nominalizations (cf. John described the city and John’s description of the city), where each of the predicates’ arguments link systematically to specific structural positions, has long been assumed in the literature (e.g., Lees 1960; Levi 1978; Marchand 1969). For example, Marchand’s (1969) classification of compound nouns is based upon the syntactic function of the compound’s head constituent, so that beer drinker classifies as subject-type nominalization and eating apple as object-type. The parallels between nominalizations and sentences are also evident when aspectual properties of a verbal predicate are inherited by a nominal (cf. giving vs. gift), which, at the same time, have been argued to determine the argument realization qualities of the head noun, cf. The frequent expression *(of one’s feelings) is desirable, in which the event reading of expression forces the object argument to be realized overtly, cf. Grimshaw (1990: 50). The examples illustrate that a deeper understanding of AS regularities in processes of word-formation can also give us a broader insight into the characteristics of the interfaces between the different structure-building components of grammar. Specifically, an investigation can help us find an answer to the intensely debated assumption of an autonomous

49. Argument-structural restrictions on word-formation patterns morphological, word-formation component, which is attached to the lexical system and as such isolated from syntax. The various perspectives on this matter, as will be shown in the next section, can differ radically in their assumptions about the general architecture of grammar and the locus of word-formation, as well as in their theoretical presuppositions (for outlines see, among others, Carstairs-McCarthy 2010; Meyer 1993; Olsen 1989). This is also reflected in the terminological conventions used in the literature, e.g., when the labels of “external argument” from a syntactic angle and “agentive role” from a lexical-semantic perspective are used to denote the same thing, i.e. a “subject” nominal of some kind. Hence, discussing word-formation regularities in a theory-neutral fashion is rather difficult.

2. Word-formation and the syntax-morphology interface According to the classical lexicalist-morphological stance, word-formation is part of an autonomous component of grammar, i.e. the lexical system, which organizes the formation of novel lexemes and can, as such, be seen as the basis of lexical productivity. The history of the debate about the appropriateness of this perspective leads us back to Chomsky’s seminal “Remarks on Nominalization” (Chomsky 1970), in which he localizes nominalizations and word-formation in general as part of the lexicon and thus deprives the lexicon of regular syntactic structure building mechanisms, see article 45 on rules, patterns and schemata in word-formation, Bauer (1983: 75 ff.), Roeper (2005). Initially, word-formation was considered for the most part idiosyncratic, and it was only later that such a lexicalist approach to word-formation was bolstered with systemic lexical and AS rules in their own right as have been developed, for example, by Di Sciullo and Williams (1987), Jackendoff (1975), Lieber (2004), Williams (1981a). Marchand (1969) can be considered a precursor of lexicalism, cf. Kastovsky (2005). To consolidate the assumption that morphological rules are different from syntactic transformations (cf. Scalise and Guevara 2005: 150), often the principle of lexical integrity is employed (cf. Anderson 1992). The principle states that syntactic operations cannot access word-internal structures and thus explains, for instance, the ungrammaticality of “stranded” noun-noun compounds as in *Morphology, she would never give a ____ lecture (see Spencer 2005: 78). However, apparent counter-examples as they are related, e.g., to the bracketing paradox (evident in phrases like transformational grammarian, where the adjective forms a constituent with a subpart of the head noun, i.e. grammar, cf. Booij 2009a for discussion) can be utilized to promote the exact opposite, nonlexicalist position, in which the internal structure of complex words is indeed open to syntactic operations. According to such an integrative view, products of word-formation are generated by the same recursive mechanisms as syntactic phrases, with the implication that syntactic operations like movement or binding apply at word level as well. In this manner, for example, Lieber investigates cases of sublexical binding as in Max’s argument was pointiless, but Pete’s did have onei, which displays pronominal binding below the level of X0 through reference between one and the sublexical noun point in pointless (cf. Lieber 1992: 130). The origins of the syntactic approach can be traced back to transformationalist accounts of nominalization as we find them in Lakoff (1970). Several theoretical variants

877

878

V. Rules and restrictions in word-formation II: Special cases of the integrative view of word-formation have been implemented in quite different grammar models since then, among them distributed morphology (cf. Harley 2008; Lieber 2006) and also construction grammar (cf. Booij 2009a, 2009b; Schlücker and Plag 2011, Borer 2003 and article 12 on construction grammar). A position mediating between the syntactic and the lexicalist stance is taken by Borer (1991), who promotes a parallel architecture. Here, internal word-structure is subject to a separate morphological rule system whose output, however, is visible to syntax in the derivation of the structural environment as well as the subcategorization features of complex words.

3. Structural principles In order to capture the argument-structural characteristics of complex expressions in a principled manner, proponents of the different theories sketched above have formulated a number of rules relating to issues like the following: How is the AS of a verbal stem transferred to a derived form? What linking regularities underlie the linear and thematic organization of an output form? And what types of modifiers can a complex noun host? Certain answers to these questions might entail, for instance, that a phrasal modifier cannot occur within a synthetic compound: *apple on a stick taster, cf. Roeper (1988). Lieber (1992: 59 f.) explains this behavior on syntactic grounds when she argues that a phrase, i.e. a maximal projection like apple on a stick, is case-licensed in the complement position to the right of the head only and, therefore, cannot be moved leftward.

3.1. Principles of argument projection A central research question in the word-formation domain under discussion concerns the process by which AS features are projected up from lexical entries to produce complex word structures and, thus, grasp the intuition that the AS of a compound verb like panfry is a function of the AS of its head. Lieber (1983) conceives of this in terms of a feature percolation mechanism, which transfers the morpho-syntactic features (including the AS features) to the first non-branching node dominating that morpheme, see (ibid.: 252) and, for critical discussion, Lieber (1992: 86 ff.). Specific AS realizations are then derived from her argument linking principle see Lieber (1983: 258). It dictates that if a verbal head appears as sister to a (potential) internal argument that is the logical object, this argument slot will be linked (i.e. satisfied), thus bringing about the configuration of synthetic compounds like beer drinker as [[beerN drinkV] -erN]. In the case of a semantic argument of the head, e.g, the instrument hand in hand-weave, the verb’s AS features percolate to the compound verb, which then satisfies its internal role outside the compound, as in hand-weave the cloth (cf. Spencer 1991: 331 f. for critical discussion). One problem with this analysis is that in the derivation of deverbal synthetic compounds like beer drinker a verbal element would be involved, which, however, is not a possible expression: *John likes to beer-drink, cf. Carstairs-McCarthy (2010: 26 ff.) for discussion. Hence, in Lieber (2004), the theoretical focus shifts to lexical-conceptual aspects of synthetic compounding when the author formulates her principle of co-indexation. This maintains that the head’s highest argument, in our case the referential argument of

49. Argument-structural restrictions on word-formation patterns -er, and the non-head’s (drink-) highest argument, are co-indexed, which renders an agentive interpretation of drinker with the internal argument role still active (for the details see article 45 on rules, patterns and schemata in word-formation, Lieber 2004: 83, 2005: 382 f.).

3.2. Thematic regularities A significant number of scholars take into account thematic criteria in their description of the AS regularities in word-formation. For example, Baker (1998: 190) refers to Chomsky’s (1981) theta criterion to rule out cases like *a truck-driver of 14-wheelers, where the PATIENT role of drive is realized twice, which violates the criterion and, at the same time, illustrates that it governs not only phrasal syntax but the construction of compound structures as well. Also, again from a syntactic perspective, Lieber (1992: 61) exploits Baker’s (1988) uniformity of theta assignment hypothesis to motivate the deepstructural identity of phrases and compounds of the type quencher of thirst and thirst quencher, respectively. Grimshaw (1990: 14) refers to the specific semantic content of thematic roles when she formulates her prominence theory. According to this approach, for example, a GOAL argument is more prominent than a THEME argument and a non-head of a compound must realize the least prominent argument. This is illustrated by the ungrammaticality of *child-giving of gifts in which child denotes a GOAL. Consequently, gift-giving to children, which has the THEME argument inside the compound, is grammatical. Note, however, that Selkirk (1982: 37) considers an equivalent example like *toy-giving to children unacceptable (see Härtl 2001: 82 f. for further discussion). Another aspect Grimshaw examines in this context is the syntactic type of a noun’s argument: Sentential complements of a deverbal nominal are always optional, cf. The announcement (that an investigation has been initiated) was inaccurate, even if the underlying verb takes an obligatory complement, i.e. an object NP, cf. *They announced (see Grimshaw 1990: 74). The author concludes that nouns do not directly theta-mark sentential complements; an assumption which is also supported by the unavailability of sentential complements to -er nominals, cf. *the observer that water boils at a certain pressure (see Grimshaw 1990: 101 ff.). As a final matter, the theta-assigning behavior of affixes shall be mentioned here. Lieber (1992: 57) assumes that affixes like de- and en- as in defuzz and encase assign a theme and a location role, respectively, to their base nouns. In contrast, a suffix like -ize does not assign a role to its base but rather assigns a theme role to a word-external NP, cf. modernize the monarchy, and Lieber concludes that only verbalizing prefixes can assign theta-roles word-internally. Later, Lieber (1998) revises this position in reference to examples like apologize or texturize, in which the nominal base seems indeed to be assigned a theme role, which leads the author to favor a lexical-semantic analysis over a purely syntactic approach.

3.3. Linearization regularities The question of whether and how affixes assign thematic roles hinges on whether an affix figures as head or not. Williams (1981b: 248) formulated a righthand head rule for

879

880

V. Rules and restrictions in word-formation II: Special cases English, which defines the right-hand member of a complex word as the head of that word. Hence, for example, the suffix -ion in construction functions as the head. The rigidity of this (parameterized) rule is called into question by apparently left-headed complex verbs containing prefixes like en-, which seem to determine the syntactic category of an output form, cf. entomb, [[en-V tombN]V]; see Lieber (1992: 31), Selkirk (1982), Williams (1981a: 249 f.) and, for discussion article 23 on particle-verb formation. Addressing this problem, Olsen (1992: 12), following Wunderlich (1987), argues for German prefixed forms like [[Ge-N spöttV]N] (‘mockery’) or [[ver-V armA]-enV] (‘to impoverish’) that they do not contradict the righthand head rule. On diachronic grounds, Olsen characterizes cases like these as instances of a conversion, which triggers a categorical change of the head, with the assumption that it is the right-hand element, i.e. spott- and arm, respectively, which functions as the head of the complex word. To guarantee a match between morpho-syntactic and morpho-phonological configurations applying to affixes and heads, Ackema and Neeleman (2004: 140) assume a linear correspondence principle, which controls the linear organization of complex words, cf. also Spencer (2005: 91). From a transformational standpoint, Roeper and Siegel (1978) assume a first sister principle, which states that verbal compounds always incorporate the first sister of the underlying verb, thus excluding ungrammatical forms like *quickly-smoker, in which quickly does not figure as first sister, cf. John smokes cigarettes quickly. Bauer (1983: 180 f.) argues that the first sister principle is empirically incorrect because it does not predict examples of verbal compounds like evening smoker, in which an adverbial occurs as non-head. Bauer’s more general proposal implies that any noun can be used in the formation of synthetic compounds containing a transitive verb (for discussion see also Lieber 1983: 282 f.; Spencer 1991: 326 f.). A refined ordering principle, which is related to the first sister principle, was formulated by Selkirk (1982: 37). Her first order projection condition states that all internal arguments need to be realized “within the first order projection of Xi”, thus excluding cases like *pizza restaurant eating, where the internal argument pizza of the verb eat is realized outside the first projection of the compound’s head, cf. Olsen (2000: 907 f.), Spencer (1991: 328 f.).

4. Conditions and operations Word-formation operations that are associated with the AS of lexical elements are restricted by mechanisms of quite different provenance. AS can be affected in many ways when a complex word is produced and, thus, we find operations in which AS features are simply passed on to some output form (describe sth. → the description of sth.), but also operations of AS reduction (tell sb. sth. → retell sth.) and AS extension (grow → outgrow sth.), cf. Bauer (1983: 177 ff.). Williams (1981a) was the first to formalize AS operations in terms of an externalization and internalization of arguments. He assumes, for example, the rule in (1) for suffixation with -able, which implies two stages: (i) the promotion of a new external argument (cf. Williams 1981a, Spencer 1991: 192 f.): (1)

read (AGENT,

THEME)

→ readable (AGENT,

THEME)

49. Argument-structural restrictions on word-formation patterns Rules like this enable us to capture meaning relations between sentences like John read the book and The book is readable in structural terms. Structural configurations are central as well for the interpretation of complex expressions. For example, the compound noun soldier brother is interpreted as denoting a brother of a soldier due to the fact that the relational noun brother contains an argument slot to be obligatorily filled, cf. The brother ??(of Max) smokes. In contrast, the interpretation of computer brother, because of the inanimate non-head noun, can only be deduced by referring to conceptual knowledge and, thus, be possibly understood as the brother who is a computer expert (cf. Meyer 1993: 104 ff., Štekauer 2005: 28 ff.). Along with the mere presence of an argument slot, it is also the thematic content of the slot, which governs the interpretation of complex words. From a processing perspective, Gagné and Shoben (1997) have developed a thematic relation model based on the assumption that thematic information associated with a noun is a key factor in the interpretation of noun-noun compounds. For instance, the noun mountain in mountain cabin, has a locative role as its primary thematic function (as part of its qualia structure, see Pustejovsky 1995) and, thus, tends to be interpreted as a cabin on a mountain. Word-formation operations are also sensitive to the number of arguments. This is evident in compounding where a restriction holds that no compound can be formed from a verb that has two obligatory arguments, cf. the example *the book-putting on the table, which can be explained under reference to Selkirk’s (1982) first order projection condition, see section 3.3 and Baker (1998: 191 ff.) for details. Further, Di Sciullo (2005) formulates a restriction which holds that as soon as an argument position is satisfied within a compound it is no longer accessible to any compound-external NP as *bikeride a scooter illustrates (cf. Di Sciullo 2005: 27). In this context, cases of apparent double argument saturations are challenging as in Personenbeschreibung der Täter ‘person description of the culprits’, where the predicate’s THEME role is associated with two nominals expressions, i.e. Person and Täter, and where the distinction between a synthetic and root compound is blurred, cf. Solstad (2010). Moreover, Randall (2010) observes that the grammatical difference between argument and adjunct affects compound formation. In passive compounds, for example, the lefthand element must be an adjunct, cf. hand-sewn clothes vs. *away-given clothes, and the externalized argument must be internal to the verb, machine-washed fabrics vs. *hoarseshouted throat (cf. Randall 2010: 210). Further, only (resultative) arguments but not adjuncts can occur as right-hand member in a passive compound: watered-flat tulips vs. *picked-late grapes (ibid.: 148 f.). Note, however, that Randall’s restriction is possibly subject to parameterization, as the availability of corresponding German examples indicates: der heisergeschrieene Hals ‘the hoarse-shouted throat’, die weggegebene Kleidung ‘the away-given clothes’.

4.1. Prefixation and suffixation The connection between AS and the various operations of prefixation and particle verb formation are well described in the literature, cf., among others, Booij (1992), Dehé et al. (2002), Günther (1987), McIntyre (2003), Olsen (1997a, 1998), Stiebels (1996), Wunderlich (1987) as well as articles 23 and 35 on particle-verb formation and particle

881

882

V. Rules and restrictions in word-formation II: Special cases verbs in Germanic. For example, in Germanic languages like German and Dutch, the prefix be- attached to an intransitive verb like gehen ‘to walk’ introduces an internal argument, cf. Sie begehen die Insel ‘they walk the island’, which Booij (1992) considers the outcome of a rule applying at the level of lexical-conceptual structure. A similar modification is the locative alternation, which is morphologically marked in German and Dutch but not in English, cf. Rappaport and Levin (1988), Olsen (1994) for an analysis: (2)

a. Er pflanzte Blumen auf das Beet. b. Er bepflanzte das Beet mit Blumen.

(3)

a. He planted flowers in the bed. b. He planted the bed with flowers.

Likewise, the prefixes ver- and über- in German affect AS in that the output form is always a transitive verb while the input’s AS can be intransitive, cf. schreiten ‘step’ → etwas überschreiten ‘to step over sth.; lit. to over-step sth.’. In contrast, particles like ab- or aus- do not introduce a new argument slot, cf. fahren ‘to drive’ → abfahren ‘to depart’, schlafen ‘to sleep’ → ausschlafen ‘to sleep in’. Particles like zu- add a dative argument, which is inserted to the lexical representation of the base via its goal argument P, (cf. the simplified representation in 4, see article 23 on particle-verb formation, Olsen 1997b: 317 and Wegener 1991): (4)

a. werfen ‘throw’ λP λy λx [THROW(x,y) and P(y)] b. zu ‘to’ λzDATIVE λy [BECOME(LOC(y,AT(z)))] c. zuwerfen ‘throw to; lit. toPART-throw’ λzDATIVE λy λx [THROW(x,y) and BECOME(LOC(y,AT(z)))]

The dative argument must be satisfied by an expression denoting an animate goal in German, see (5a). Inanimate entities can link with a corresponding (directional) prepositional phrase only, see (5b), cf. Olsen (1997b: 325), Witt (1998: 85 f.): (5)

a. den Ball dem Kind / *dem Korb zuwerfen the ball the childDATIVE / the basketDATIVE toPART-throw b. den Ball zu dem Korb werfen the ball to the basket throw

Similarly, particle verbs with the particle ein- ‘in’ do not accept animate goals linked with a PP, cf. ibid.: (6)

das Gebiss *in den Patienten / in den Mund / dem Patienten einlegen the denture into the patientACCUSATIVE / in the mouth / the patientDATIVE insert

In addition to such systematic derivational constraints, any theory of linking in prefixation must also allow for specific lexical differences between the derived forms. For example, the internal argument slot of the verb believe can be realized by an ACI, cf. I

49. Argument-structural restrictions on word-formation patterns believe him to be smart, whereas the prefixed form disbelieve cannot, *I disbelieve him to be smart, cf. Bauer (1983: 60), Carlson and Roeper (1981). Like prefixation, suffixation affects the AS of the input form. For instance, the suffix -ize attaches to nominal and adjectival bases and produces a verb with an internal argument, i.e. causative/transitive verbs like symbolize, modernize or inchoative/unaccusative verbs like oxidize, aerosolize. Despite their wide-ranging polysemy (cf. Lieber 2004: 77), Plag (1999: 137) assumes a unified lexico-conceptual representation for -ize verbs, which can realize both a transitive and an unaccusative verb form achieved through the optionality of the constant CAUSE. Note, though, that the implication of this assumption, i.e. the non-causativity of inchoative verbs, is subject to constant debate, cf. Bierwisch (2006), Chierchia (2004), Härtl (2013), Koontz-Garboden (2009), Levin and RappaportHovav (1995). Lieber (2004) proposes a unitary lexical template for -ize and -ify verbs as well but derives their individual differences from the semantic category of the base and specific co-indexation configurations holding between the arguments of the affix and the base, cf. Lieber (2004: 81 ff.). A more abstract perspective is taken by Williams (1981a), where an -ize derivation is achieved through the mechanisms of externalization and internalization of argument slots, cf. also Spencer (1991: 193) and section 4 above: (7)

modern (THEME) → modernize (AGENT,

THEME)

As we have seen, any theorizing about the link between word-formation and AS has to consider a wide range of linguistic phenomena, such as thematic role content, animacy, case, morpho-syntactic marking, etc., as well as structural configurations like transitivity or externalization. Another word-formation domain where the interplay of a broad variety of linguistic factors is particularly evident is that of nominalization, which we shall have a more detailed look at in the following section.

4.2. Nominalization The term nominalization covers a broad range of morpho-syntactic operations, which all produce a nominal of some kind. Thus, e.g., gerunds like criticizing, agent and instrument nouns (opener), deverbal nouns (description) in general as well as synthetic compounds (car driver) fall under this category, with the question being relevant here if and how they inherit the AS of the underlying verb. The perspectives on this issue vary radically: from the assumption that deverbal nouns do not contain any AS features or that they have their own AS to the classical view that the AS of the underlying verb is fully inherited by the derived form; for overviews see Alexiadou (2010), Spencer (1991: 324 ff.) and article 33 on synthetic compounds in German. According to the standard view, i.e. that the AS of the verb is copied over to the deverbal nominal, linking conditions control the verb’s internal argument, which is assigned structural accusative case in languages like German, to be realized as a structural genitive, cf. die Stadt beschreiben ‘to describe the city’ → die Beschreibung der Stadt ‘the description of the city’, cf. Olsen (1986). Such canonical linking postulations, however, are challenged by deviations where the internal argument links with a PP in a derived nominal, cf. die Feinde hassen ‘to hate the enemies’ → der Hass *der Feinde /

883

884

V. Rules and restrictions in word-formation II: Special cases auf die Feinde ‘the hatred of the enemies / towards the enemies’, cf. Lindauer (1995), Ehrich and Rapp (2000). This has led some researchers to conclude that derived nominals are equipped with their own AS, which is determined by semantic aspects like the eventstructural properties of the nominal, cf. Grimshaw (1990), or the affectedness of the lowest argument, cf. Ehrich and Rapp (2000) and section 4.2.3 for further details. On the other end of the theoretical spectrum we find approaches in which no verbal AS features are present in the grammatical representation of derived nominals. To substantiate this conception, in many cases the ontological differences between nouns and verbs are brought forward and, in particular, the optionality of the arguments of nouns, cf. Dowty (1998), Kayne (2008). Kaufmann (2002) argues that nouns do not exhibit a fixed array of linkers and considers the “arguments” of nouns to be semantic attributes instead, for which certain interpretative defaults apply. Likewise, Fanselow (1988) employs what he calls prominent meaning relations holding between the constituents of complex nouns, thus, making lexical-semantic argument positions redundant. According to Fanselow, this applies to derived nominals like Verfasser des Buches ‘composer of the book’ as well, for which a stereotypical relation like WRITE needs to be deduced thus explaining its parallels in meaning to non-derived nouns like Autor des Buches ‘author of the book’ (cf. Olsen 1992 for critical discussion). Problematic for such concept-based approaches are linking differences between deverbal nominals like Jill’s shock vs. Jill’s attempt. Here, parallel prominence relations link crosswise such that the genitive NP of a nominalized psych-predicate like shock links with an internal experiencer argument, whereas with attempt the genitive is linked with the external agent argument (cf. Bauer 1983: 77). This behavior can only be explained by dint of the predicates’ lexical-semantic properties, which have to be somehow active in the derivation.

4.2.1. Linking conditions on nominalization According to several theories, inter alia Grimshaw’s (1990) prominence theory, external arguments cannot be realized within synthetic compounds, cf. *gourmet-eating, *touristarriving, *child-sleeper. A similar restriction is implemented by Selkirk (1982), where the author employs her subject condition to allow only internal arguments to appear within a synthetic compound (cf. also Chomsky 1970). Borer (2003) doubts the validity of a general constraint against external arguments occurring within derived nominals, providing examples of -ion nouns, where a genitive NP is linked with an agent role, i.e. an apparent external argument, cf. the enemy’s destruction of the city. Also, Di Sciullo (1992) questions the rigidity of the constraint in reference to examples like expert-tested, in which the noun contained in the compound is associated with the external argument role of the base verb as well (ibid.: 66). Baker (1998) makes the same observation although with a different interpretation: According to Baker, the linking behavior of such adjectival constructions (i.e. expert-tested as in expert-tested guide) is expected under the subject condition because the agent role of a past participle form does not figure as an external argument but as an internal one. Rather, it is the theme (i.e. guide), which functions as the external argument of the adjectival predicate (ibid.: 191). Note, however, that AS based approaches, in general, are weakened by the noticeable degree of nonproductivity of the construction. While, for example, constructions like expert-tested

49. Argument-structural restrictions on word-formation patterns guide or chef-cooked dish may well be acceptable, a less stereotypical relation between the roles involved renders the expression odd, cf. ??grandmother-knitted sweater, ??professor-taught subject. Alternatively, what seems to play a role here is the conceptual salience of the property expressed with the adjective, which determines its interpretability and which makes its analysis as synthetic compound in the narrow sense redundant. Such a view is compatible with approaches which favor an analysis based on free interpretation, like Marantz’s (1997). These assume, along the lines of Grimshaw (1990) and the above restriction against external arguments, that agent-like genitives in phrases like the King’s separation of the family should rather be characterized as POSSESSORS, which happen to correspond to an agent interpretation based on conceptual knowledge, cf. Borer (2003) for critical discussion. The accessibility of such agent readings independent of AS is also evident in NPs like the German invasion, where the modifier German can receive both an agent interpretation as well as a theme interpretation, cf. Roeper and van Hout (1999). It is clear, however, that the adjective does not function as an argument, at least on the theme interpretation: as soon as an explicit agent is provided, the theme reading of German is no longer available, cf. *the German invasion by France (ibid.: 8).

4.2.2. Nominalization with -er and -ee The structural status of the arguments as external or internal is also relevant in -er nominalizations. A standard assumption comes from Levin and Rappaport (1988: 1068), who formulate a requirement for the bases of -er nominals that they contain an external argument, cf. appealer vs. *appearer, which is bound by the affix (cf. Di Sciullo 1992: 73). The specific thematic content of the role is not decisive, see Fleischer and Barz (1995: 151 ff.), Lieber (2004: 17) for lists of possible meanings of -er nominals. Furthermore, instrument interpretations are grammatical if this role can also be realized as subject of a corresponding proposition, cf. Levin and Rappaport (1988: 1071 f.), Rainer (2005: 348 f.): (8)

a. A metal gadget opened the can. → can openerINSTRUMENT b. *A silver fork ate the meat. → *meat eaterINSTRUMENT

It is commonly assumed that deverbal -er nominals (or a subset of them, see below), in some way, inherit the object arguments of the base, cf. baker of bread, giver of presents to children, which Lieber (2004: 61 f.) captures using her principle of co-indexation, see also section 3.1 above. Object arguments are not inherited in compound expressions containing a gerund, cf. *baking man of bread, *frying pan of meat, cf. Di Sciullo and Williams (1987). Besides, there are also several instances of -er nominals like villager and Londoner, which are not related to a verbal base, cf. Booij and Lieber (2004), Fleischer and Barz (1995: 154 f.) and, for diachronic aspects relevant in this context, Meibauer, Guttropf and Scherer (2004). Di Sciullo (1992) examines Italian verb-noun compounds like taglia-carte ‘paper cutter; lit. cut-paper’ and claims that the external argument of the verbal part is realized

885

886

V. Rules and restrictions in word-formation II: Special cases as pro (existent in Italian but not in English or German) inside the compound. According to Di Sciullo, this explains the unavailability of synthetic -ore ‘-er’ compounds in Italian, as this affix, too, binds an external argument role. As a result, the external argument would be satisfied twice in a synthetic -ore compound thus producing a theta criterion violation, cf. *tagliatore-carte (ibid.: 72). Note that Di Sciullo uses this argumentation to strengthen her reservations against the subject condition, which bans external arguments from being realized within compounds, see the previous section. A concept-based restriction on deverbal -er nominals (and synthetic compounds in general) is that they cannot contain cognate objects as non-head, cf. *tear crier, *dream dreamer, as they render the compound’s meaning tautological. Instead, a cognate object requires a taxonomic specification of the argument expression: false tears crier, nightmare dreamer. Similar observations have been made for unacceptable noun-noun compounds like *furniture chair or *animal horse, with the explanation that a modifier of a compound must always bring about an ontological specification of the head noun’s extension, cf. Meyer (1993: 102), Štekauer (2005: 11). Along with event structural factors, which we shall examine in the next section, it is also the optionality of the predicate’s internal argument, which determines the interpretation of deverbal -er nominals. Olsen (2000: 907) observes that, for example, tree devourer, due to the obligatory internal argument of devour, receives an interpretation of an entity that devours trees, whereas tree in tree eater, which contains a predicate with an omissible internal argument, is open for an interpretation as a locative modifier, i.e. an eater in trees. Nouns with the -ee suffix (present in English but not in Dutch and as a less productive equivalent -ling in German, as in Prüfling ‘test-ee’, Ankömmling ‘arriv-ee’) can be derivatives of transitive verbs, cf. employee, trainee. In these cases, the derived noun is related to the object argument of the predicate. But we also find subject-oriented -ee nouns, like escapee, attendee, and nouns that derive from genuinely intransitive verbs, like standee, again questioning conventional accounts based on AS inheritance (cf. Barker 1998, Spencer 2005 and article 52 on semantic restrictions on word-formation). The selectional characteristics of -ee have also led to several semantic treatments of the derivation, where semantic-conceptual features associated with volitionality and sentience are put in focus of the theoretical description, cf., e.g., Booij and Lieber (2004). Note that -er nominals with an of-complement cannot receive an instrument reading as only an agentive-eventive interpretation is possible with them: *openerINSTRUMENT of cans, *sharpenerINSTRUMENT of knives. This has led to the well-known assumption that only eventive -er nominals inherit the verbal AS and can hence realize an of-complement, whereas non-eventive ones cannot, cf. van Hout and Roeper (1998), Levin and Rappaport (1988). Thus, for instance, destroyer of the city denotes somebody who has actually destroyed something at some time, whereas a destroyer, i.e. a warship, may never destroy anything (ibid.: 1069). Olsen (1992: 23 f.), however, points to the influence of the determiner semantics in this context and discusses examples like closer of gates, which, although an -erAS nominal in Levin and Rappaport’s conception, receive a noneventive, generic interpretation, cf. also Alexiadou and Schäfer (2010) for a related aspectual analysis as well as McIntyre (2010) for discussion. Generic qualities are also reflected in compounds and in the well-described non-referentiality of the modifier of a compound (cf. Lawrenz 1996, Meibauer 2007), which, in turn, promotes the instrument reading of a synthetic -er compound like knife sharpener.

49. Argument-structural restrictions on word-formation patterns

4.2.3. Event structural conditions on nominalization Event structural properties have long been argued to determine the availability of AS in nominalizations. One of the standard assumptions can be traced back to Grimshaw (1990). She assumes that the presence of AS in a deverbal nominal depends on whether the nominal denotes a process, i.e. a complex event, or rather a non-eventive result of some event (ibid.: 49): (9)

a. The examinationPROCESS of the student was in the office at 12:00. b. The examRESULT (??of the student) was in the drawer.

Process nominals can be identified in time and space and can hence combine with temporal and spatial modifiers, cf. (9a), whereas result nominals can only be spatially identified, as (9b) illustrates. The underlying idea is that complements in NPs are not altogether optional; instead, only nominals lacking aspectual structure do not exhibit AS. A number of grammatical criteria have been isolated to substantiate the grammatical distinction displayed in (9), one of them being that a genitive NP in process nominals is linked with an agent role, whereas it is linked with a possessor role in result nominals, cf. the teacher’s examination of the student vs. the teacher’s exam, cf. (ibid.: 51) and Alexiadou (2001: 10 ff.), Alexiadou and Grimshaw (2008) for overviews of the differences between the two types. Criticism raised against Grimshaw’s original concept holds that, among other things, process nominals, too, do not necessarily require all their roles to be realized, as is illustrated in An unskilled instructor’s examination will take a long time, where the internal argument of examine is not realized, cf. Pustejovsky (1995: 257 f.). Problematic for the above distinction is also the significant number of deverbal nouns which realize their internal argument overtly but can still receive a result interpretation as in The written description of the painting is in the drawer, cf. Bierwisch (1989), (2009). As McIntyre (p.c.) notes, however, the problem dissolves under the assumption that the PP of the painting in this example does not link with the object argument of the verbal base but rather figures as an of-complementation to a relational noun on a par with non-deverbal nouns like replica, as in replica of the painting. A related assumption is implied in Grimshaw’s (1990) distinction between of-phrases functioning as arguments (“a-adjuncts” in her terminology) and those functioning as “modifiers”. For example, of the girl in picture of the girl containing the relational noun picture is described as a modifier by Grimshaw (1990: 144). Following this logic, the PP of the painting in descriptionRESULT of the painting figures as a modifier just as it figures as a modifier of the noun replica or picture. The distinction between of-modifier and of-argument is reflected in the separability of of-modifiers from their head, cf. The picture was of the girl vs. *The destruction was of the city, which Grimshaw attributes to the locality restriction of theta-assignment holding for arguments but not for modifiers. This, in turn, predicts that the above -ion noun with a result reading can be separated from a (non-argument) ofphrase, whereas the corresponding process nominal is predicted not to be detachable from the of-phrase. This is indeed supported by the following contrast: (10) a. The written descriptionRESULT was [of the painting]MODIFIER. b. *The frequent descriptionPROCESS was [of the painting]ARGUMENT.

887

888

V. Rules and restrictions in word-formation II: Special cases Alternative perspectives on the correlation between nominalization and AS realization put a stronger focus on the lexical-semantic qualities of the nouns involved. For example, Ehrich and Rapp (2000) consider verbal and nominal ASs to be completely independent of each other, each equipped with its own individual linking rules. Here, the linking properties of a deverbal noun, process nominal or not, are not derived from the underlying verb and, as the authors assume, it is the feature of affectedness, which determines the linking properties of arguments. The basic idea is that the interpretation of a postnominal genitive NP, in German, depends on whether the noun’s semantic representation contains a BECOME-operator: a postnominal genitive will always be interpreted, when present, as the lowest argument under BECOME, i.e. as an affected theme. This explains why postnominal genitives in NPs like Hinrichtung des Henkers ‘execution of the hangman’, which involve an affected object, can only be interpreted as theme, while nonaffecting predicates can realize any role in this position, cf. Entdeckung des SeefahrersAGENT/THEME ‘discovery of the sailor’, Verehrung der MädchenEXPERIENCER/THEME ‘adoration of the girls’, cf. (ibid.: 279 f.). The factor of affectedness has also been observed to have an impact on the preposing of object NPs, which are banned from a prenominal position in a deverbal nominal if they denote an unaffected object: *the fact’s knowledge vs. the city’s destruction, cf. Anderson (1977, 2007: 121 ff.). It has been argued that the affectedness constraint on preposed NPs is subject to parameterization as no restriction in terms of NP-internal fronting is active, for example, in Greek, cf. Alexiadou (2001: 94 ff.) for discussion. Event-structural conditions on AS linking can be found to be active elsewhere in word-formation. For instance, aspectual properties have also been described as a key factor determining the locative alternation (see section 4.1) and producing the meaning differences anchored in the alternating pairs, cf. Olsen (1994). This illustrates, all in all, that only a wide-ranging and interrelated view on the different components of the linguistic system and its interfaces will contribute to a full understanding of the lexical productivity in human language.

5. References Ackema, Peter and Ad Neeleman 2004 Beyond Morphology. Interface Conditions on Word Formation. Oxford: Oxford University Press. Alexiadou, Artemis 2001 Functional Structure in Nominals. Nominalization and Ergativity. Amsterdam/Philadelphia: Benjamins. Alexiadou, Artemis 2010 Nominalizations: A probe into the architecture of grammar. Language and Linguistics Compass 4: 496−511. Alexiadou, Artemis and Jane Grimshaw 2008 Verbs, nouns and affixation. In: Florian Schäfer (ed.), SinSpeC 1. Working Papers of the SFB 732, 1−16. Stuttgart: OPUS. Alexiadou, Artemis and Florian Schäfer 2010 On the syntax of episodical vs. dispositional -er nominals. In: Artemis Alexiadou and Monika Rathert (eds.), Nominalizations Across Languages and Frameworks, 9−38. Berlin: Mouton de Gruyter.

49. Argument-structural restrictions on word-formation patterns Anderson, Mona 1977 NP preposing in noun phrases. North-Eastern Linguistic Society 8: 12−21. Anderson, Mona 2007 Affectedness. In: Martin Everaert and Henk van Riemsdijk (eds.), The Blackwell Companion to Syntax, 121−141. Malden, MA: Blackwell. Anderson, Stephen R. 1992 A-morphous Morphology. Cambridge, MA: Cambridge University Press. Baker, Mark 1988 Incorporation. A Theory of Grammatical Function Changing. Chicago, IL: University of Chicago Press. Baker, Mark 1998 Comments on the paper by Sadock. In: Steven G. Lapointe, Diane K. Brentari and Patrick M. Farrell (eds.), Morphology and its Relation to Phonology and Syntax, 188− 212. Stanford, CA: CSLI. Barker, Chris 1998 Episodic -ee in English: A thematic role constraint on new word formation. Language 74: 695−727. Bauer, Laurie 1983 English Word-Formation. Cambridge, MA: Cambridge University Press. Bierwisch, Manfred 1989 Event nominalizations: Proposals and problems. In: Wolfgang Motsch (ed.), Wortstruktur und Satzstruktur, 1−73. Berlin: Akademie Verlag. Bierwisch, Manfred 2006 German reflexives as proper and improper arguments. In: Patrick Brandt and Eric Fuß (eds.), Form, Structure, and Grammar. A Festschrift Presented to Günther Grewendorf on Occasion of his 60th Birthday, 15−36. Berlin: Akademie Verlag. Bierwisch, Manfred 2009 Nominalization − lexical and syntactic aspects. In: Anastasia Giannakidou and Monika Rathert (eds.), Quantification, Definiteness, and Nominalization. Oxford: Oxford University Press. Booij, Geert 1992 Morphology, semantics, and argument structure. In: Iggy Roca (ed.), Thematic Structure. Its Role in Grammar, 27−50. Dordrecht: Foris. Booij, Geert 2009a Lexical integrity as a formal universal: A constructionist view. In: Sergio Scalise, Elisabetta Magni and Antonietta Bisetto (eds.), Universals of Language Today, 83−100. Berlin: Springer. Booij, Geert 2009b Phrasal names: A constructionist analysis. Word Structure 2: 219−240. Booij, Geert and Rochelle Lieber 2004 On the paradigmatic nature of affixal semantics in English and Dutch. Linguistics 42: 327−357. Borer, Hagit 1991 The causative-inchoative alternation: A case study in parallel morphology. The Linguistic Review 8(2−4): 119−158. Borer, Hagit 2003 Exo-skeletal vs. endo-skeletal explanations: Syntactic projections and the lexicon. In: John Moore and Maria Polinsky (eds.), The Nature of Explanation in Linguistic Theory, 31−67. Stanford, CA: CSLI. Carlson, Greg and Thomas Roeper 1981 Morphology and subcategorization: Case and the unmarked complex verb. In: Teun Hoekstra, Harry van der Hulst and Michael Moortgat (eds.), Lexical Grammar, 123− 164. Dordrecht: Foris.

889

890

V. Rules and restrictions in word-formation II: Special cases Carstairs-McCarthy, Andrew 2010 The Evolution of Morphology. Oxford: Oxford University Press. Chierchia, Gennaro 2004 A semantics for unaccusatives and its syntactic consequences. In: Artemis Alexiadou, Elena Anagnostopoulou and Martin Everaert (eds.), The Unaccusativity Puzzle, 22−59. Oxford: Oxford University Press. Chomsky, Noam 1970 Remarks on nominalization. In: Roderick A. Jacobs and Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar, 184−221. Waltham, MA: Blaisdell. Chomsky, Noam 1981 Lectures on Government and Binding. Dordrecht: Foris. Dehé, Nicole, Ray Jackendoff, Andrew McIntyre and Silke Urban (eds.) 2002 Verb Particle Explorations. Berlin: Mouton de Gruyter. Di Sciullo, Anna Maria 1992 Deverbal compounds and the external argument. In: Iggy M. Roca (ed.), Thematic Structure. Its Role in Grammar, 65−78. Dordrecht: Foris. Di Sciullo, Anna Maria 2005 Decomposing compounds. SKASE Journal of Theoretical Linguistics 2(3): 14−33. Di Sciullo, Anna Maria and Edwin Williams 1987 On the Definition of Word. Cambridge, MA: MIT Press. Dowty, David 1989 On the semantic content of the notion thematic role. In: Gennaro Chierchia, Barbara Partee and Raymond Turner (eds.), Properties, Types and Meanings. Semantic Issues, 69−129. Dordrecht: Kluwer. Ehrich, Veronika and Irene Rapp 2000 Sortale Bedeutung und Argumentstruktur: ung-Nominalisierungen im Deutschen. Zeitschrift für Sprachwissenschaft 19(2): 245−303. Fanselow, Gisbert 1988 Word syntax and semantic principles. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1988, 95−122. Dordrecht: Foris. Fleischer, Wolfgang and Irmhild Barz 1995 Wortbildung der deutschen Gegenwartssprache. 2nd ed. Tübingen: Niemeyer. Gagné, Christina L. and Edward J. Shoben 1997 The influence of thematic relations on the comprehension of non-predicating conceptual combinations. Journal of Experimental Psychology: Learning, Memory, and Cognition 23: 71−87. Grimshaw, Jane 1990 Argument Structure. Cambridge, MA: MIT Press. Günther, Hartmut 1987 Wortbildung, Syntax, be-Verben und das Lexikon. Beiträge zur Geschichte der deutschen Sprache und Literatur 109: 179−201. Harley, Heidi 2008 Compounding in distributed morphology. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 129−144. Oxford: Oxford University Press. Härtl, Holden 2001 CAUSE und CHANGE. Thematische Relationen und Ereignisstrukturen in Konzeptualisierung und Grammatikalisierung. Berlin: Akademie Verlag. Härtl, Holden 2013 Generic rescue: Argument alternations and the monotonicity condition. In: Patrick Brandt and Eric Fuß (eds.), Repairs. The Added Value of Being Wrong, 95−130. Berlin/ New York: Mouton de Gruyter.

49. Argument-structural restrictions on word-formation patterns Hout, Angeliek van and Thomas Roeper 1998 Events and aspectual structure in derivational morphology. MIT Working Papers in Linguistics 32: 175−220. Jackendoff, Ray 1975 Morphological and semantic regularities in the lexicon. Language 51(3): 639−671. Kastovsky, Dieter 2005 Hans Marchand and the Marchandeans. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 99−124. Dordrecht: Springer. Kaufmann, Ingrid 2002 Infinitivnominalisierungen von reflexiven Verben: Evidenz gegen Argumentstrukturvererbung? In: Claudia Maienborn (ed.), (A)Symmetrien − (A)Symmetries. Beiträge zu Ehren von Ewald Lang / Papers in Honor of Ewald Lang, 203−232. Tübingen: Stauffenburg. Kayne, Richard 2008 Antisymmetry and the lexicon. Linguistic Variation Yearbook 8: 1−32. Koontz-Garboden, Andrew 2009 Anticausativization. Natural Language and Linguistic Theory 27: 77−138. Lakoff, George 1970 Irregularity in Syntax. New York: Holt, Rinehart and Winston. Lawrenz, Birgit 1996 Der Zwischen-den-Mahlzeiten-Imbiß und der Herren-der-Welt-Größenwahn: Aspekte der Struktur und Bildungsweisen von Phrasenkomposita im Deutschen. Zeitschrift für Germanistische Linguistik 24: 1−15. Lees, Robert B. 1960 The Grammar of English Nominalizations. Bloomington, IN: Indiana University Press. Levi, Judith N. 1978 The Syntax and Semantics of Complex Nominals. New York: Academic Press. Levin, Beth and Malka Rappaport 1988 Non-event -er nominals: A probe into argument structure. Linguistics 26: 1067−1083. Levin, Beth and Malka Rappaport Hovav 1995 Unaccusativity. At the Syntax-Lexical Semantics Interface. Cambridge, MA: MIT Press. Lieber, Rochelle 1983 Argument linking and compounds in English. Linguistic Inquiry 14(2): 251−285. Lieber, Rochelle 1992 Deconstructing Morphology. Word Formation in Syntactic Theory. Chicago: University of Chicago Press. Lieber, Rochelle 1998 The suffix -ize in English: Implications for morphology. In: Steven G. Lapointe, Diane K. Brentari and Patrick M. Farrell (eds.), Morphology and its Relation to Phonology and Syntax, 12−33. Stanford, CA: CSLI. Lieber, Rochelle 2004 Morphology and Lexical Semantics. Cambridge: Cambridge University Press. Lieber, Rochelle 2005 Word formation processes in English. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 375−427. Dordrecht: Springer. Lieber, Rochelle 2006 The category of roots and the roots of categories. What we learn from selection in derivation. Morphology 16: 247−272. Lindauer, Thomas 1995 Genitivattribute. Eine morphosyntaktische Untersuchung zum DP/NP-System. Tübingen: Niemeyer.

891

892

V. Rules and restrictions in word-formation II: Special cases Marantz, Alec 1997 No escape from syntax. University of Pennsylvania Working Papers in Linguistics 4(2): 201−225. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. McIntyre, Andrew 2003 Preverbs, argument linking and verb semantics. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 2003, 119−144. Dordrecht: Kluwer. McIntyre, Andrew 2010 Agentive Nominals and Argument Structure. Ms. Université de Neuchâtel. Meibauer, Jörg 2007 How marginal are phrasal compounds? Generalized insertion, expressivity, and I/Qinteraction. Morphology 17: 233−259. Meibauer, Jörg, Anja Guttropf and Carmen Scherer 2004 Dynamic aspects of German -er-nominals: A probe into the interrelation of language change and language acquisition. Linguistics 42(1): 155−193. Meyer, Ralf 1993 Compound Comprehension in Isolation and in Context. The Contribution of Conceptual and Discourse Knowledge to the Comprehension of German Novel Noun-Noun Compounds. Tübingen: Niemeyer. Olsen, Susan 1986 Wortbildung im Deutschen. Stuttgart: Kröner. Olsen, Susan 1989 Zur Stellung der Wortbildung in der Grammatik. Drei Theorien der Affigierung. Papiere zur Linguistik 40(1): 3−22. Olsen, Susan 1992 Zur Grammatik des Wortes: Argumente zur Argumentvererbung. Linguistische Berichte 137: 3−32. Olsen, Susan 1994 Lokativalternation im Deutschen und Englischen. Zeitschrift für Sprachwissenschaft 13: 201−235. Olsen, Susan 1997a Zur Kategorie ‚Verbpartikel‘. Beiträge zur Geschichte der deutschen Sprache und Literatur 119(1): 1−32. Olsen, Susan 1997b Der Dativ bei Partikelverben. In: Christa Dürscheid, Monika Schwarz and Karl-Heinz Ramers (eds.), Festschrift für Heinz Vater, 307−328. Tübingen: Niemeyer. Olsen, Susan 1998 Semantische und konzeptuelle Aspekte der Partikelverbbildung mit ein-. Tübingen: Stauffenburg. Olsen, Susan 2000 Composition. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 897−916. Berlin/New York: de Gruyter. Plag, Ingo 1999 Morphological Productivity. Structural Constraints in English Derivation. Berlin/New York: Mouton de Gruyter. Pustejovsky, James 1995 The Generative Lexicon. Cambridge, MA: MIT Press.

49. Argument-structural restrictions on word-formation patterns Rainer, Franz 2005 Constraints on productivity. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 335−352. Dordrecht: Springer. Randall, Janet 2010 The Geometry of Argument Structure. Dortrecht: Springer. Rappaport, Malka and Beth Levin 1988 What to do with theta-roles. In: Wendy Wilkins (ed.), Thematic Relations, 7−36. New York: Academic Press. Roeper, Thomas 1988 Compound syntax and head movement. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1988, 187−228. Dordrecht: Foris. Roeper, Thomas 2005 Chomsky’s remarks and the transformationalist hypothesis. In: Rochelle Lieber and Pavol Štekauer (eds.), Handbook of English Word-Formation, 187−228. Dordrecht: Springer. Roeper, Thomas and Muffy Siegel 1978 Lexical transformation for verbal compounds. Linguistic Inquiry 9: 199−260. Roeper, Thomas and Angeliek van Hout 1999 The impact of nominalization on passive, -able and middle. Burzio’s generalization and feature-movement in the lexicon. In: Liina Pylkkänen, Angeliek van Hout and Heidi Harley (eds.), Papers from the UPenn/MIT Roundtable on the Lexicon, 185−211. Cambridge, MA: MIT Press. Online: http://people.umass.edu/roeper/online_papers/ NOM-PASSIVE%20MITWPL%20VAN%20HOUT00.pdf [last access 7 Jan 2015]. Schlücker, Barbara and Ingo Plag 2011 Compound or phrase? Analogy in naming. Lingua 121: 1539−1551. Scalise, Sergio and Emiliano Guevara 2005 The lexicalist approach to word-formation and the notion of the lexicon. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 147−187. Dordrecht: Springer. Selkirk, Elisabeth O. 1982 The Syntax of Words. Cambridge, MA: MIT Press. Solstad, Torgrim 2010 Postnominal genitives and prepositional phrases in German: A uniform analysis. In: Artemis Alexiadou and Monika Rathert (eds.), The Syntax of Nominalizations across Languages and Frameworks, 219−252. Berlin: Mouton de Gruyter. Spencer, Andrew 1991 Morphological Theory. Oxford: Blackwell. Spencer, Andrew 2005 Word-formation and syntax. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 73−97. Dordrecht: Springer. Štekauer, Pavol 2005 Meaning Predictability in Word Formation. Novel, Context-Free Naming Units. Amsterdam/ Philadelphia: Benjamins. Stiebels, Barbara 1996 Lexikalische Argumente und Adjunkte. Zum semantischen Beitrag von verbalen Präfixen und Partikeln. Berlin: Akademie Verlag. Wegener, Heide 1991 Der Dativ − ein struktureller Kasus? In: Gisbert Fanselow and Sascha W. Felix (eds.), Strukturen und Merkmale syntaktischer Kategorien, 70−103. Tübingen: Narr. Williams, Edwin S. 1981a Argument structure and morphology. The Linguistic Review 1: 81−114.

893

894

V. Rules and restrictions in word-formation II: Special cases Williams, Edwin S. 1981b On the notions ‘lexically related’ and ‘head of a word’. Linguistic Inquiry 12: 245−274. Witt, James 1998 Kompositionalität und Regularität im System der Partikelverben mit ein-. In: Susan Olsen (ed.), Semantische und konzeptuelle Aspekte der Partikelverbbildung mit ein-, 27−104. Tübingen: Stauffenburg. Wunderlich, Dieter 1987 An investigation of lexical composition: The case of German be-Verbs. Linguistics 25: 283− 331.

Holden Härtl, Kassel (Germany)

50. Phonological restrictions on English word-formation 1. 2. 3. 4. 5. 6. 7. 8.

Introduction Background Prosodic structure Cohering suffixation Prefixes vs. modifiers Implications for the mental lexicon Summary References

Abstract Word-formation rules differ from syntactic rules in that they, apart from obeying morphological and semantic constraints, can also be − and often are − restricted phonologically. The present article includes an overview of the relevant phenomena in English and discusses the consequences for the representation of words in the mental lexicon and for grammar.

1. Introduction Phonological restrictions on word-formation, as commonly understood, concern cases where affixation rules are sensitive to sound structure. The existence of such restrictions in English is not well established. The relevant sections in current English grammars with fairly comprehensive sections on word-formation, in particular Quirk, Greenbaum

50. Phonological restrictions on English word-formation and Leech (1972) and Huddleston and Pullum (2002), contain little, and often inaccurate relevant information. There is indeed not a single phonological restriction on affixation which is noted in both grammars; even the rather sparse information on phonological restrictions found in traditional handbooks (Jespersen 1942; Marchand 1969) goes unmentioned. This neglect is remarkable in view of the abundant evidence for phonological restrictions on English affixation. Before delving into this subject it is worth inspecting more marginal types of word-formation where phonological restrictions can hardly be overlooked. The formation of acronyms (e.g., DOS, NAFTA), as opposed to initialisms (e.g., GRE, BMX), appears to be conditioned by the organizability of the relevant segments into wellformed prosodic constituents. That is, if the phonemes associated with the sequence of letters can be grouped into unmarked syllables, in which a vocalic nucleus is flanked by single consonants, acronyms can be formed as in (1a, b). (Syllables are grouped into feet in English. Every phonological word includes minimally one foot. “ω” = phonological word, “Σ” = foot, “σ” = syllable, “S” = strong, “W” = weak, “O” = onset, “N” = nucleus, “C” = coda.) (1)

a.

b.

If the relevant sequences − due to the distribution of vowels and consonants − cannot be parsed into such syllables, a copulative compound of the names of the letters is formed instead. In such compounds the last member receives primary stress: (2)

(dʒi)ω (ɑɹ)ω (í)ω ‘GRE’ (bi)ω (ɛm)ω (ɛ´ks)ω ‘BMX’

A second example for conspicuous phonological restrictions concerns certain blends, in which the initial onset of one word is combined with the complete structure, except for the initial onset, of another word. In such cases the choice and order of the parts is restricted phonologically: the word whose initial onset is complex supplies that onset, the other word supplies the remaining structure, cf. (3). (3)

smoke + fog breakfast + lunch stalker + fan sneeze + fart brother + romance

smog brunch stan snart bromance

*foke *leakfast *falker *feeze *rother

895

896

V. Rules and restrictions in word-formation II: Special cases If neither word has a complex onset, blending appears to be inhibited, unless the onsets are followed by identical VC-sequences (e.g., Bill + Hillary → Billary, guess + estimate → guesstimate). In such cases, the order is still determined phonologically, such that both the properties of the onsets (e.g., minimal sonority) and foot structure play a role. The best-known cases of non-affixational word coinage are perhaps clippings (cf. Marchand 1969; Lappe 2007). The examples in (4) suggest a preference for reducing the word to the first syllable, unless the second syllable carries main stress. In the latter case, that syllable is included in the clipped form. (4)

a. fábulous rádical pícture Jàpanése síster géntleman

→ → → → → →

fab rad pic Jap sis gent

b. obítuary exécutive celébrity delícious legítimate ridículous

→ → → → → →

obít exéc celéb delísh legít ridíc

The patterns in (4) indicate a particular solution to the conflicting goals of shortening the word as much as possible while preserving as much of the salient sound structure of the base as possible. The conflict is solved by retaining the word-initial segmental material including the first stressed syllable. (The few counter-examples generally reflect a preference for the first syllable, regardless of stress, or a preference for the stressed syllable, regardless of its position, e.g., spagétti → spag, detéctive → tec.) There appears to be a preference for the clipped form to end in a consonant with relatively low sonority, with the result that certain postvocalic clusters are typically preserved (e.g., chìmpanzée → chimp, Yánkee → Yank, but cústomer → cus) while words with single postvocalic /r/, /l/, or /n/ tend to resist clipping altogether (rélevant → ??rel, itínerary → ??itín, facílity → ??facíl). Regardless of whether or not the above formations are recognized as word-formation proper, they do exhibit systematic properties that need to be described in a comprehensive grammar of English. Moreover the types of phonological constraints observed in these word coinages shed light on the sort of constraints to be expected in more ordinary affixation. Here, too, it is the prosodic organization of segments, which can be shown to play a central role. The sort of structure referred to is precisely the one seen in the above investigation of more marginal morphology: phonological word, foot, and syllable structure, as well as the sonority of phonemes.

2. Background Descriptions of word-formation in traditional handbooks are typically organized in form of a list of affixes, each individually detailed with the affix’s historical development and combinatory potential. For some affixes, the relevant entry also includes phonological stem properties as is illustrated in (5): (5)

Suffix -ive

restricted to: stems ending in -t, -s (Jespersen 1954: 454; Marchand 1969: 316)

50. Phonological restrictions on English word-formation -al (nominal)

stems with final stress (Jespersen 1954: 386)

-en (verbal)

stems ending in stop or fricative (Marchand 1969: 272)

The already fragmentary information given in the handbooks is largely ignored in modern grammars. One reason for neglecting the relevant rules may concern the distinction between the actual and the potential. For example, the restriction on -al-suffixation stated in (5) could be overlooked because of apparent counter-examples among English nouns ending in -al. Here it is significant, that none of the relevant exceptions originate from native -al-suffixation; the noun burial, for instance, goes back to the Old English noun buriels meaning ‘burying-place, tomb’. While the change in the spelling ( → ) along with the meaning change (‘burying-place’ → ‘act of burying’) indicate the likely historical reanalysis of the ending /əl/ as the suffix -al, the fact remains that no word has been coined in English by attaching -al to a verbal stem with non-final stress. To capture this generalization it is necessary to treat separately the conditions for “analysis”, including the conditions for affix recognition and the conditions for recognizing paradigmatic relatedness between words, and “synthesis”, which concerns the rules for forming new words based on known words as well as the ability to judge the grammaticality of non-existing words (cf. Aronoff 1976). Significantly, phonological restrictions on word-formation concern synthesis, that is, the potential for forming new words, not the analysis of given, that is inherited or borrowed, words. A second obstacle to the detection of phonological restrictions on word-formation can be traced to the indiscriminate use of the notions “stem” and “base”. The remarkable restriction on -ive-suffixation stated in (5) contrasts hence with the more fuzzy one offered in Huddleston and Pullum (2002), where the relevant domain is identified as “mostly Latinate verbs in /d/, /t/ or /s/” (2002: 1711). The relative weakness of the latter restriction results from the lack of distinction between “bases”, which concern paradigmatic relations between whole words such as the adjective evasive and the verb evade, as opposed to “stems”, which concern syntagmatic part-whole relations such as evas- and evasive. A stem is accordingly the part of a word which remains after an affix has been stripped off. Significantly, the sort of requirements illustrated in (5) concern stems, not bases, a point easily overlooked in English due to the frequent occurrence of so-called “free stems”, which are homophonous to bases (e.g., disrúpt-), as opposed to “bound stems”, which are not (e.g., suscépt- from suscépt-ible). The restriction relevant to -ive-suffixation in English concerns the availability of a stem, free or bound, which ends in -t or -s. If such a stem exists, -ive-suffixation may occur as is shown by the native coinages obséssive, sécretive, caréssive, púrposive, excúsive (not because of the verb excú/z/ ‘excuse’, but because of the noun excú/s/ ‘excuse’), even sticktóitive (‘stick to it + -ive’), where a quotation of a phrase is treated as a stem. If there is no such stem, native -ive-suffixation is ruled out, even for Latinate verbs in /d/ such as encode (e.g., *encódive, because the stem encód- ends in a phoneme other than /t/ or /s/, *encósive, *encótive because there are no stems *encós-, *encót-). The third and perhaps most important obstacle to the detection of phonological restrictions on word-formation lies in the common assumption that word-formation rules are properly described by specifying inputs, that is, the items that potentially undergo

897

898

V. Rules and restrictions in word-formation II: Special cases the rule, along with the relevant operations. From that perspective, the exclusion of certain items from the domain of a rule for strictly phonological reasons appears to have no function and seems whimsical at best. The crucial generalization, missed also in traditional work, concerns the observation that phonological constraints on affixation typically refer to only those aspects of stem structure which are part of the prosodic constituents encompassing the affixes in output forms. For instance, the targeting of the stem-final consonant in -ive-suffixation relates to that consonant’s role in the output, where it forms the onset of the syllable encompassing the suffix, cf. (6a). The observation that nominalizing -al attaches only to verbs with final stress relates to the fact that the relevant syllables together build a trochaic foot in the output, cf. (6b): (6)

b.

a.

From the perspective of output-oriented grammar models, in particular optimality theory, the restrictions observed in (5) indicate the interaction of independently motivated phonological markedness constraints such as foot binarity (a foot must consist of two syllables) and paradigm uniformity constraints (Prince and Smolensky 1993). The latter constraints are crucial for explaining the unacceptability of coinages such as *caróu/s/ive or *promísal. These derivations are ruled out because of the non-identity between the segments (caróu/s/ive − caróu/z/) or the stress patterns (promísal − prómise) in the derived forms compared to the respective base words. Such deviations violate a constraint, referred to as “uniform exponence” (Flemming 1995), which requires derivational paradigms to be associated with a single stem form. Once phonological restrictions are conceived of as expressions of markedness constraints on output forms, it becomes evident that besides positive requirements (trochee in (6)) there can also be negative ones, that is, the avoidance of word-formation for phonological reasons. For instance, there are numerous systematic gaps in English wordformation which indicate constraints against identical or near-identical segments within adjacent syllable constituents. Also many stressed suffixes will not attach to stems with final stress, apparently to avoid a stress clash. Some examples are listed in (7) (the symbol “#” marks the end of the stem): (7)

a.

-ish -ése -ity -al -ify´

sheepish, oldish Sudanese, Nepalese oddity, torpidity withdrawal, referral gasify, nullify

b. *[cor fric]# *[cor fric]# */t/# */l/# */f/#

*horse-ish, *fresh-ish *Bangladesh-ese, *Greece-ese *acute-ity, *concrete-ity *appeal-al, *result-al *shelf-ify, *stiff-ify

50. Phonological restrictions on English word-formation -ée -éer -étte -ésque

trainee, divorcee pistoleer, rocketeer kitchenette, wagonette Pinteresque, Kafkaesque

*/i/# *σ´#, */ɹ/# *σ´#, */t/# *σ´#, */s/#

899 *free-ee, *pity-ee *gún-eer, *revolver-eer *garáge-ette, *closet-ette *Camús-esque, *Wallace-esque

None of the restrictions in (7) are noted in Jespersen (1942) or Marchand (1969), apparently due to the fact that they are not easily expressed as selection (rather than avoidance). It is also significant that Aronoff, who did note the absence of stem-final coronal fricatives in -ish-suffixation and offered an explanation in terms of output constraints, went on to assert the existence of general constraints against coronal fricatives in adjacent syllables in English phonology, insisting that these constraints do not pertain to word-formation proper (Aronoff 1976: 82). There are many counter-examples to this assertion, including the words listed in (8a). Additionally, there are several affixation rules which appear to be insensitive to the constraint in question. The formations in (8b) show that -ship freely attaches despite the presence of a coronal fricative in the stemfinal syllable. (8)

a. hashish, shashlik, thesis, seize, zest, scissors, missis, sizzle, season, assassin b. musicianship, professorship, censorship, citizenship, lectureship, apprenticeship

Unlike Aronoff, Dressler (1977) relegates dissimilation to morphology. His constraint stated in (9), where # represents the boundary between a stem and a derivational suffix, accounts for ungrammatical derivations like *fishish, without affecting the well-formed words in (8): (9)

*XViCj#ViCjY

However, as a result of requiring identity of both vowel and consonant, the constraint in (8) is too specific to account for any of the ungrammatical examples in (7b). Generalizing the constraint to the effect that suffix-initial vowels may not be flanked by identical consonants would solve that problem, but falsely rule out the suffixations illustrated in (10): (10) -ic hierarchic, anarchic, monarchic, psychic -able bribable, (in)describable, (im)perturbable, clubbable -er bearer, explorer, admirer, caterer, murderer, sufferer It appears then that phonological restrictions in English word-formation concern outputoriented markedness constraints associated with individual affixes. Dressler’s (1977: 42) observation that, unlike sequences of the type in (9), those of the type XCiVj#CiVjY occur frequently indicates the insensitivity of consonant-initial suffixes to dissimilatory constraints. The wider generalization is that consonant-initial syllabic suffixes are generally rather insensitive to phonological structure in English. The causal connection between the sound shape of affixes and their potential sensitivity to phonological structure is the topic of the next section.

900

V. Rules and restrictions in word-formation II: Special cases

3. Prosodic structure In English derivational morphology, the occurrence of phonological restrictions appears to be linked to certain properties of affixes. Phonological sensitivity is typical for vowelinitial or vowelless suffixes, less so for consonant-initial suffixes which form a separate syllable (cf. Raffelsiefen 1999; Plag 2003). This generalization indicates the relevance of the prosodic organization of affixes to their potential sensitivity to sound structure. Significantly, vowel-initial or vowelless suffixes form a prosodic domain together with the stem, a property referred to as “coherence”. Other suffixes are non-cohering, in that they are prosodified separately (cf. Booij 1985 for a similar generalization in Dutch). The trees in (11) illustrate the basic connection between morphological and prosodic structure: prosodic word boundaries regularly coincide with stem boundaries as in (11b, c), unless a cohering suffix follows as in (11a) (“C” = clitic group). (11)

The terms “cohering” vs. “non-cohering” suffixes were introduced by Dixon (1977) to capture the morphophonological behavior of monosyllabic vs. disyllabic suffixes in Yidiny. The insight is that suffixes which are in some sense deficient cohere, in that they form a single domain for prosodic organization together with the stem, whereas nondeficient suffixes form a separate prosodic domain. In English, vowel-initial suffixes are deficient in that syllables formed from them lack an onset, whereas vowelless suffixes are deficient in that they lack a potential nucleus. Consequently both of these suffixes cohere, cf. (11a). Other suffixes are prosodically organized separately and are attached outside of the phonological word of the stem, cf. (11b, c). The importance of the division between cohering and non-cohering suffixes for the purpose of the present article lies in the causal connection between prosodic representation and potential phonological restrictions. It is plausible that affixes which form a prosodic domain together with the stem interact with that stem more extensively than do suffixes which are not part of that domain. Before further reviewing the relevant evidence from English, it is in order to discuss the strictly phonological motivation for the different prosodic groupings illustrated in (11). The fundamental division between cohering and non-cohering suffixes is motivated both on phonological and on phonetic grounds.

50. Phonological restrictions on English word-formation A rather striking generalization following from the notion of suffix coherence is that only vowel-initial suffixes can bear main stress in English. This is because vowel-initial suffixes, as a consequence of being integrated into the phonological word of the stem, are part of the domain for foot formation: in crispation the last two syllables form a binary foot, which regularly attracts main stress, cf. (11a). Some monosyllabic suffixes, including those in (12a), are lexically marked for carrying main stress in their respective domain. Suffixes with more than one syllable can also be lexically associated with secondary or no stress, cf. (12b): (12) a. /ɛ´sk/ ‘-esque’, /ɛ´t/ ‘-ette’, /í/ ‘-ee’, /íz/ ‘-ese’, /ɪ´ɹ/ ‘-eer’ b. /ɪzəm/ ‘-ism’, /ətɔ̀ɹi/ ‘-atory’, /ɪfài/ ‘-ify’, /ɪti/ ‘-ity’, /əbəl/ ‘-able’, /ənsi/ ̀ ‘-ancy’ c. /ɪv/ ‘-ive’, /íz/ ‘-ese’, /àiz/ ‘-ize’, /ɪʃ/ ‘-ish’, /ɪdʒ/ ‘-age’ /θ/ ‘-th’ By contrast, non-cohering suffixes are never lexically marked for stress. Unlike the bior trisyllabic cohering suffixes in (12b), non-cohering suffixes always form a single syllable, which is organized outside of the phonological word and consequently outside of the domain of foot formation. Single syllables are not grouped into feet and are therefore stressless, cf. (11b). The only exception is the suffix -hood shown in (11c), which forms a separate foot due to the necessary alignment of the phoneme /h/ with a left foot boundary in English. The presence of the foot is accordingly conditioned by segmental structure, not by lexical marking. A complete list of English non-cohering suffixes illustrated with native coinages is given in (13). English non-cohering suffixes are nominal or adjectival; none are verbal. (13) a. /fəl/ ‘-ful’ /dəm/ ‘-dom’ /səm/ ‘-some’ /mənt/ ‘-ment’ /mən/ ‘-man’ /uəɹd/ ‘-ward’ /iəɹ/ ‘-ior’, ‘-ure’ b. /lɪŋ/ ‘-ling’ /nɪk/ ‘-nik’ /ʃɪp/ ‘-ship’ /lɪs/ ~ /ləs/ ‘-less’ /nɪs/ ~ /nəs/ ‘-ness’ /lɪt/ ~ /lət/ ‘-let’ c. /li/ ‘-ly’ d. /hʊd/ ‘-hood’

purposeful, harmful kingdom, sheriffdom frolicsome, fearsome ailment, bereavemen craftsman, infantryman northward, downward failure, behavior youngling, nursling peacenik, fashionnik courtship, governorship goalless, effortless stubbornness, foulness leaflet, streamlet orderly, scholarly parenthood, priesthood

A closer examination of the phonological structure of the non-cohering suffixes reveals a remarkable degree of neutralization of phonemic contrast. Certain marked consonants such as voiced fricatives or dental fricatives are entirely absent, in contrast to the full range of fricatives found in cohering suffixes, cf. (12c). Apart from the presence of alveolar obstruents in marginal positions in -ment and -ward, the non-cohering suffixes generally have a simple onset-nucleus-coda structure as is illustrated in (14a). The pho-

901

902

V. Rules and restrictions in word-formation II: Special cases nemic representations of the suffixes -ward and -ior/-ure in (14b) is based on the assumption that high tense vowels can occur in syllable margin positions in English, where they are phonetically realized as glides [j] and [w]. The suffix -/iəɹ/ constitutes a borderline case in that its prosodic coherence is contextually conditioned: it systematically fuses with stems ending in an alveolar obstruent (e.g., (ɪɹéis)ω ‘erase’ − (ɪɹéiʃəɹ)ω ‘erasure’, (kəmpóuz)ω ‘compose’ − (kəmpóuʒəɹ)ω ‘composure’ − (feil)ω ‘fail’ − (féil)ωiəɹ ‘failure’). Such dual behavior is perhaps generally restricted to suffixes with an initial high front vowel. (14)

The most striking constraints on the distribution of phonemes in non-cohering suffixes, compared to cohering suffixes, concern the syllable nucleus. Here the default vowel conditioned by the stresslessness is schwa (cf. 13a), which in specific segmental contexts alternates with high vowels. The front high lax vowel [ɪ] is the regular allophone of schwa before the velar nasal, cf. (13b), (14c); in non-cohering suffixes that vowel appears before all coda obstruents, where it varies with schwa before alveolars, cf. (13b). Schwa is also banned from foot-initial syllables in English: here the back high lax vowel appears (/hʊd/ ‘-hood’, cf. (11c)), in accordance with the general affinity of back high vowels with stress in English (cf. the acronym SNAFU, with final stress, vs. MORI, with non-final stress). The occurrence of tense /i/ in /li/ -ly appears to be conditioned by the absence of a following consonant. Assuming that high tense vowels alone associate with both nucleus and coda as in (14d), the /i/ ensures syllable closure. The occurrence of vowels in non-cohering suffixes is then determined entirely by the consonantal structure, with no potential contrast. The constraints on the phonemic make-up of suffixes illustrated in (14) hold only for true suffixes. So-called semi-suffixes (Marchand 1969: 356−358), which correspond to free-standing words, form separate phonological words and exhibit the full range of phonemic contrasts (e.g., /laik/ -like, /uaiz/ -wise, /fʊl/ -ful). Apart from the phonemic structure of the suffixes themselves, there are additional phonological differences between cohering and non-cohering suffixation. The latter exhibit various juncture effects, that is, phoneme combinations and stress patterns resulting from the combination of stem and suffix, which occur neither within simplexes nor within cohering suffixations. These include “deviant” consonant clusters (e.g., nonhomorganic nasal-stop clusters (/ŋd/, /nf/), obstruent clusters which differ in voicing (/ds/, /gf/), fricative clusters (/sʃ/, /fs/), and irregular stress patterns (e.g., góvernership, which ends in three unstressed syllables). A particular characteristic of non-cohering suffixation is the frequent occurrence of overly complex rhymes preceding the suffix, including rhymes with two consonants (e.g., [ɑrm] in harmless), tense vowel plus consonant (e.g., [i:m] in teamster), or diphthong plus consonant (e.g., [eim] in aimless). Significantly, such juncture effects are generally stable for as long as a non-cohering suffix is recognized, regardless of whether or not there is a free stem, as is illustrated in (15) (cf. Raffelsiefen 2005: 241−242).

50. Phonological restrictions on English word-formation (15) (ɡɔɹm)ωləs (uɪst)ωfəl (ɔint)ωmənt (gæst)ωli

‘gormless’ ‘wistful’ ‘ointment’ ‘ghastly’

(cf. (cf. (cf. (cf.

903

†gome ‘attention’, ‘care’) †wistly ‘with close attention’) †oint ‘to smear with oil’) †gast ‘to frighten’)

Consonant-initial suffixes fuse regularly with the stem only when they are non-recurring and therefore cannot be learned and recognized, cf. (16a). Apart from such cases of systematic fusion, there are isolated cases of prosodic fusion in non-cohering suffixation, illustrated in (16b). These cases, like occasional fusion of compounds (e.g., cupboard, shepherd, necklace, nothing), are rather rare historical changes affecting individual words. They thereby differ sharply from the prosodic fusion of cohering suffixes, which is entirely systematic, conditioned by the phonological shape of the suffix (i.e. vowelinitial or vowelless). (16) a. (nɑ.lɪdʒ)ω (hɛi.tɹəd)ω b. (hæn.səm)ω (uɹ.ʃɪp)ω (bɪz.nəs)ω

‘knowledge’ ‘hatred’ ‘handsome’ ‘worship’ ‘business’

(cf. (cf. (cf. (cf. (cf.

know, -lɪdʒ) hate, -rəd) hand) worth) busy)

Additional evidence for the fundamental distinction between cohering and non-cohering suffixation pertains to phonetic realization. The narrow phonetic transcription of the word ransom in (17a), compared to the lexical phonemic representation, illustrates the processes of schwa reduction and stop epenthesis, which applies between a nasal and a homorganic fricative in foot-internal position. The fact that neither of these processes applies in winsome indicates accordingly the non-integration of the suffix -some into the foot of the stem (the phonetic transcriptions are adopted from the Cambridge English Pronouncing Dictionary (Jones 2011) and Muthmann 1999). Similarly, the duration of the stressed vowels in lawful or freeness, compared to that of the corresponding vowels in offal or Venus, respectively, indicate the non-integration of the suffix -ful into the foot of the stem, cf. (17b, c). This is because stressed vowels at the end of a phonological word are longer than word-internal stressed vowels. (17) a. b. c.

Phonemic: Phonetic: Phonemic: Phonetic: Phonemic: Phonetic:

Non-cohering suffixation (uɪn)ωsəm ‘winsome’ [wɪ´nsəm] (lɑ)ωfəl ‘lawful’ [lɑ:fəl] (fri)ωnəs ‘freeness’ [fri:nəs]

Simplex (ɹænsəm)ω ‘ransom’ ´ ntsəm] [ɹæ (ɑfəl)ω ‘offal’ [ɑfəl] (vinəs)ω ‘Venus’ [vi:nəs]

In general it holds that words which include non-cohering suffixes never rhyme perfectly with other words in careful speech, even when the relevant phoneme sequences are identical (cf. lawful and offal, freeness and Venus). This is because the separate prosodic organization of stem vs. suffix phonemes necessarily yields subtle differences in the phonetic realization compared to the joint organization of the entire phoneme sequence.

904

V. Rules and restrictions in word-formation II: Special cases By contrast, words with cohering suffixes can rhyme perfectly with simplexes. This is illustrated with the vowel-initial suffixes in (18a) and the vowelless suffixes in (18b): (18) a. [[fiud]STEM[əl]SUFFIX]WORD [[hɛl]STEM[ɪʃ]SUFFIX]WORD [[sɹpənt]STEM[ain]SUFFIX]WORD

b. [[tɹu]STEM[θ]SUFFIX]WORD [[ɡɹəu]STEM[θ]SUFFIX]WORD [[hai]STEM[t]SUFFIX]WORD

→ (fiudəl)ω ‘feudal’ → (hɛlɪʃ)ω ‘hellish’ → (sɹpəntain)ω ‘serpentine’ → (tɹuθ)ω ‘truth’ → (ɡɹəuθ)ω ‘growth’ → (hait)ω ‘height’

[nudəl]WORD [ɹɛlɪʃ]WORD [tɹpəntain]WORD [buθ]WORD [bəuθ]WORD [fait]WORD

→ (nudəl)ω ‘noodle’ → (ɹɛlɪʃ)ω ‘relish’ → (tɹpəntain)ω ‘turpentine’ → (buθ)ω ‘booth’ → (bəuθ)ω ‘both’ → (fait)ω ‘fight’

The assumption of a fundamental division between English affixes into prosodically cohering vs. non-cohering ones is then independently motivated by phonetic structure. Phonetic evidence for prosodic organization is particularly valuable because, unlike phonotactic rules, phonetic processes are free of lexical exceptions.

4. Cohering suffixation The importance of the division between cohering and non-cohering suffixes for English word-formation is reflected in the correlations listed in (19), which indicate the relatedness of phonological restrictions to other morphophonological phenomena in the language (cf. Raffelsiefen 1999, 2004, 2007; Plag 2003). (19) (i) (ii) (iii) (iv) (v) (vi)

only cohering suffixes can be sensitive to the phonological shape of the stem, resulting in gaps only cohering suffixes can trigger stem allomorphy only cohering suffixes may exhibit allomorphy only cohering suffixes can attach to bound stems (of other suffixations) cohering suffixes cannot follow non-cohering suffixes only cohering suffixes may “fuse” with other suffixes

The relatedness of the phenomena in (19) concerns the prominent role of phonological markedness constraints. These apply only within specific prosodic domains, which include cohering, but not non-cohering suffixes. The close relation between gaps and allomorphy can be illustrated by comparing the suffixes -eer and -ese. The suffix -eer, which has been somewhat productive in English in certain semantic fields, including military jargon, attaches only to trochaic stems as in (20a). (20) a. pístol+éer rócket+éer

→ (pìstoléer)ω → (ròcketéer)ω

b. Vietnám+ése → (Viètnamése)ω Sudán+ése → (Sùdanése)ω

50. Phonological restrictions on English word-formation wéapon+éer → (wèaponéer)ω platóon+éer → 0̸

Nepál+ése Saigón+ése

905 → (Nèpalése)ω → (Sàigonése)ω

The avoidance of stems with final stress such as platóon is indicative of an inherent conflict: attaching the suffix would either violate a phonological markedness constraint against adjacent stressed syllables (cf. *platòonéer), or would lead to non-uniform paradigms, where the stressed syllables in the stem of the derived form and in the base fail to correspond (cf. plàtoonéer − platóon), yielding multiple stem forms (cf. {platóon-, plátoon-}). To avoid violation of either of these constraints, the suffix is not attached. Instead there is a gap. The suffix -ese also does not tolerate a stress clash. The suffixes differ, in that the constraint on paradigm uniformity is violated in -ese-suffixation to ensure a phonologically well-formed alternating stress pattern, cf. (20b). Consequently, there are no gaps in -ese-suffixation. Instead there is stem allomorphy (e.g., {Vietnám-, Viétnam-}). The third possibility concerning the interaction of the markedness constraint against adjacent stressed syllables and the relevant paradigm uniformity constraint can be illustrated by -ee-suffixation: (21) a. ábsent+ée → (àbsentée)ω lícense+ée → (lìcensée)ω pátent+ée → (pàtentée)ω

b. seléct+ée → (selèctée)ω divórce+ée → (divòrcée)ω advíse+ée → (advìsée)ω

The suffix -ee attaches to any stems, including stems with final stress, without adjustments of the foot structure. As a result, -ee is like the suffix -eer in that there is no stemallomorphy, and like the suffix -ese, in that there are no stress-related gaps. The suffixes -eer, -ese, and -ee, all of which are lexically marked for carrying main stress, hence illustrate the basic possibility space observed in the interaction of phonological markedness constraints and paradigm uniformity constraints as well as the claim that such interactions are suffix-specific. Phonologically conditioned gaps and stem allomorphy are rarely found in non-cohering suffixation. One of the few examples is the adverbial suffix -ly, which does not attach to stems ending in /li/ (e.g., *sillily, *holily). The suffix -ful avoids stems ending in /f/ or /v/ (e.g., *loveful, *griefful, cf. Chapin 1970: 54). Avoidance of (near-)identical phonemes in junctures is not generally characteristic of non-cohering suffixation (cf. goalless, embalmment, stubbornness, fashionnik). There are no systematic cases of stemallomorphy in non-cohering suffixation other than “degemination”, often with schwa loss, in stems ending in /əl/ before adverbial -ly-suffixation (e.g., /bɹutəl/ ‘brutal’ − /bɹutəli/ ‘brutally’, /sʌtəl/ ‘subtle’ − /sʌt(ə)li/ ‘subtly’, but /kul/ − /kulli/ ‘coolly’). All other relevant examples concern historical prosodic fusion in individual words (e.g., the “stress shift” in (evidéntly)ω or (insíghtful)ω, compared to regular nonfused (élegant)ωly, (cónfident)ωly, (púrpose)ω ful, (wórship)ω ful). Allomorphy in cohering suffixes mentioned in (19iii) is illustrated with the homophonous suffixes in (22). While deverbal /əl/ does not attach to stems with final /l/ to avoid violation of the constraint against nuclei flanked by identical segments, the denominal suffix satisfies that constraint by violating suffix uniformity in a minimal manner: the phoneme /l/ is substituted by the other liquid, /ɹ/. Consequently, there are no gaps in the latter suffixation.

906

V. Rules and restrictions in word-formation II: Special cases (22) a. withdráw+al revérse+al renéw+al revéal+al fulfíll+al inhále+al

→ (withdráwal)ω → (revérsal)ω → (renéwal)ω → 0̸ → 0̸ → 0̸

b. séason+al bride+al coast+al prótocòl+ar enámel+ar corólla+ar

→ → → → → →

(séasonal)ω (brídal)ω (cóastal)ω (prótocòlar)ω (enámelar)ω (coróllar)ω

Additional examples of allomorphy in cohering suffixation are given in (23). The suffix /əɹi/ appears without the initial schwa in (23a) to avoid words ending in three unstressed syllables (cf. Marchand 1969: 285). To satisfy the same constraint, the suffix /ənsi/, which replaces the suffix /ənt/, appears without the final /i/ in (23b). (When attaching to verbs, the suffix is always /əns/, regardless of stress, e.g., gúidance, occúrrence, útterance, inhéritance.) (23)

a. bribe+ery → forge+ery → can+ery → rével+ery → húsband+ery→ mímic+ery →

(bríbery)ω (fórgery)ω (cánnery)ω (révelry)ω (húsbandry)ω (mímicry)ω

b. prégn-ant+ancy stríd-ent+ency delínqu-ent+ency ímman-ent+ence sáli-ent+ence hésit-ant+ance

→ → → → → →

(prégnancy)ω (strídency)ω (delínquency)ω (ímmanence)ω (sálience)ω (hésitance)ω

The selection and “replacement” of certain suffixes illustrated in (23b) is another property distinguishing cohering from non-cohering suffixes. Some of these relations, referred to as “correlative derivation” (Marchand 1969: 216), are asymmetrical and less regular (nutrit-ion licenses nutrit-ious, cf. (24a)), others are symmetrical and fully regular in that either suffixation licenses the word with the other suffix (e.g., femin-ism licenses feminist as femin-ist licenses femin-ism, cf. (24b)). Phonemic alternation in the relevant suffix pairs appears to be confined to coronals and vowels differing in stress. (24)

a. /ən/ → /əs/ /ənt/ → /ənsi/ /ət/ → /əsi/ b. /ɪst/ ↔ /ɪzəm/ /ɛit/ ↔ /ɛiʃən/ /ɪfai/ ↔ /ɪfɪkɛiʃən

nutrition − nutritious, caution − cautious, oblivion − oblivious blatant − blatancy, redundant − redundancy, lenient − lenience prophet − prophesy, pirate − piracy, secret − secrecy feminist − feminism, tourist − tourism, nudist − nudism frustrate − frustration, exacerbate − exacerbation purify − purification, falsify − falsification

The examples in (24) illustrate the ability of cohering suffixes to combine with bound stems. That ability is further illustrated by the native coinages in (25a), where the respective suffixed words which qualify as plausible sources of the relevant bound stems are listed in (25b). While it is true that there are also a few cases of bound stems in noncohering suffixation, listed in (25c), there is a crucial difference between the relevant coinages: according to the Oxford English Dictionary (2012) all of the coinages with non-cohering suffixes were based on a free stem, when the words were first coined. The relevant words, listed in (25d), became obsolete only after the suffixed words had come into existence. By contrast, native word-formation with cohering suffixes can in many

50. Phonological restrictions on English word-formation

907

cases be shown to involve bound stems (that is, stems found in words with other suffixes), from the start: (25)

a. Nat. coinage: (àmput-ée)ω (aggréss-ive)ω (biógraph-er)ω (cóal-ìze)ω (compétit-ive)ω (skélet-al)ω (sy´phil-òid)ω

b. Hist. base: (ámput-àte)ω (aggréss-or)ω (biógraph-y)ω (còal-ítion)ω (compétit-or)ω (skélet-on)ω (sy´phil-is)ω

c. Nat. coinage: (ruth)ωless (gorm)ωless (feck)ωless (grate)ωful (wist)ωful (dole)ωful (bale)ωful

d. Hist. base: † † † † † † †

ruthe ‘pity’ gome ‘attention’ feck ‘efficacy’ grate ‘agreeable’ wistly ‘intently’ dol ‘pain, grief’ bale ‘evil influence’

Suffixation to bound stems, although frequently attested in English, is generally tied to specific conditions. In the formations in (26a), bound stems are selected to avoid identical onsets; in (26b) the purpose is to yield output forms with alternating stress. The respective examples to the right show that the relevant suffixes attach to morphologically complex free stems otherwise. (26) a.

ámput-àte+ée áppet-ìte+ìze b. sénsit-ive+ìze áccur-ate+ìze

(àmput-ée)ω (áppet-ìze)ω (sénsit-ìze)ω (áccur-ìze)ω

éduc-àte+ée páras-ìte+ìze colléct-ive+ìze prív-ate+ìze

(èduc-àt-ée)ω (páras-ìt-ìze)ω (colléct-iv-ìze)ω (prív-at-ìze)ω

Bound stems are accessible only if a word contains a recognizable suffix that can be stripped off. Since suffixes are inherently tied to certain syntactic categories, the same phoneme sequences may function as suffixes in some words but not in others. Final /ənt/ or /i/ function as suffixes in abstract nouns and adjectives (delinquent, agent, modesty, scrutiny), but not in verbs. Consequently, the relevant bound stems are accessible in (27a), to satisfy constraints on word length in the relevant suffixations, but not in (27b). (27) a. néglig-ent]A astrónom-y]N

(néglig-able)ω (astrónom-er)ω

b. órient]V (órient-able)ω accómpany]V (accómpani-er)ω

A closer examination of the conditions for selecting bound stems shows that in English all verbal suffixation, be it verb-based or verb-deriving, satisfies paradigm uniformity. This generalization is obscured when comparing native suffixation as in (28a) with the corresponding free stems in (28b), as paradigm uniformity violations appear to be rampant. The claim that the suffixes satisfy phonological constraints (e.g., no adjacent stressed syllables for -ize-suffixation and no adjacent unstressed syllables for -ify-suffixation) by selecting the underlined bound stems in (28c), rather than by inducing stress shifts, is supported by the fact that words containing the relevant bound stems can always be shown to be attested in English prior to the coinage of verbs, often as foreign loanwords (cf. the earlier attested French loans Japonnais, sublimation, rigidité, angelique, bovarysme). The claim that the word-formation in (28a) involves the bound stems in (28c), not adjustments of the free stems in (28b), is further demonstrated by the observation that word-formation fails when there are no suitable stems, cf. (28d).

908

V. Rules and restrictions in word-formation II: Special cases (28)

a. (Jápan-ìze)ω b. Japán sublíme (súblim-ìze)ω rígid (rigíd-ify)ω ángel (angél-ify)ω Bóvary (Bóvar-ìze)ω

c. (Jàpan-ése)ω (súblim-àte)ω (rigíd-ity)ω (angél-ic)ω (Bóvar-ìsm)ω

d. Tibét − *Tíbet-ìze extréme − *éxtrem-ìze rándom − *randóm-ifỳ túnnel − *tunnél-ify Kénnedy − *Kénned-ìze

Paradigm violations in English native word-formation are accordingly found only when no verbs participate, as in (29); truly productive cases of word-formation involving paradigm uniformity violations are mostly limited to formations based on proper nouns. Suffixation which induces paradigm violations supplies novel bound stems, underlined in (29b), which can then be used in other suffixation when needed to satisfy phonological markedness constraints (e.g., Viétnam-ìze, Darwín-ifỳ). (29) a. Vietnám+ése Dárwin+ian Íceland+ic cónsonant+al téchnical+ity hórmòne+al phónème+ic

→ → → → → → →

(Viètnamése)ω (Darwínian)ω (Icelándic)ω (cònsonántal)ω (tèchnicálity)ω (hormónal)ω (phonémic)ω

b. {Vietnám-, Viétnam-} {Dárwin-, Darwín-} {Íceland-, Icelánd-} {cónsonant-, cònsonánt-} {téchnical-, tèchnicál-} {hórmòne-, hormón-} {phónème-, phoném-}

The last two correlations in (19) concern the combinability of suffixes, which is also sensitive to coherence. The observation that cohering suffixes cannot follow non-cohering suffixes, illustrated in (30a), indicates that cohering suffixes attach only when they can be integrated into a phonological word. This restriction does not concern non-cohering suffxes, which can follow one another. Apparent counter-examples are rare individual words, where the location of main stress indicates complete prosodic fusion (e.g., (devèlopméntal)ω, (gòvernméntal)ω, as opposed to ungrammatical *(púnish)ωment-al, *(encóurage)ωment-al, *(embárass)ωment-al). (30) a.

*(kind)ωness+y *(free)ωdom+al *(grate)ωful+ize *(establish)ωment+al *(suck)ωling+ish *(court)ωship+ous *(child)ωhood+ish *(gorm)ωless+ity *(man)ωly+ize

b. (grate)ωful+ness (reck)ωless+ness (world)ωly+ness (pen)ωman+ship (govern)ωment+less (gentle)ωman+ly (book)ωlet+less (lone)ωsome+ness (free)ωdom+less

The final characteristic of cohering suffixes to be illustrated here concerns the potential development of strong productivity with respect to specific other suffixes (cf. 19vi). The suffix -ation, for instance, occurs only in a few native suffixations based on simplexes, all of which end in an unstressed syllable or are monosyllabic (e.g., hỳphenátion, defòrestátion, flìrtátion, stàrvátion). The suffix is however very productive with stressed verbal suffixes, including -ize (victimization, computerization, vulgarization). Similarly, -itysuffixation occurs only sporadically in native formations based on simplexes with final

50. Phonological restrictions on English word-formation

909

stress (oddity, queerity), but prevails over the generally much more productive -nesssuffixation with respect to bases ending in -able (corruptibility, readability, drinkability). In English, such cases of hyperproductivity, where words automatically license the attachment of a specific, somewhat fossilized affix, involve cohering suffixes only. The force behind the development of such affinities among suffixes appears to be phonological: when a suffix causes allomorphy in a preceding suffix, without causing a stress clash or stress allomorphy in the preceding stem, the two suffixes are likely to “fuse” (cf. (31a), where the relevant allomorphy is boldfaced). This fusion is then apparently grounded in the preference for uniform stems, which favors the segmentation mèmor-izátion (cf. mémor-y, mémor-ìze) over mèmoriz-átion, as the latter would involve stem allomorphy ({mémorìze-, mèmoriz-}). Fused suffixes in relation to their first member may develop into fully productive correlative patterns, as is shown in (31b). (31) a.

→ (stàndard-iz-átion)ω → (respòns-ibíl-ity)ω

(stándard-ìze)ω (respóns-ible)ω

b. /aiz/ ↔ /ɪzέiʃən/ /əbəl/ ↔ /əbɪ´ləti/

The suffix sequences -izátion and -abílity will never cause a stress clash or stress allomorphy in the preceding stem, as they start with a stressless syllable. These suffix sequences thereby differ from those illustrated in (32a), where, due to the initial stress, constraint violations would be unavoidable whenever there is a stem with final stress, cf. (32b). The existence of such a conflict appears to inhibit the development of hyperproductivity even when there is no stem-final stress as in (32c). There is no automatic licensing of the second cohering suffix in such cases; instead non-cohering suffixes are often preferred (bóisterousness, náturalness, intúitiveness). (32) a.

sócial-ist → (sòcial-íst-ic)ω vírtu-ous → (vìrtu-ós-ity)ω whímsic-al → (whìmsic-ál-ity)ω sénsit-ive → (sènsit-ív-ity)ω c. théor-ist / *-íst-ic bóister-ous / *-ós-ity nátur-al / *-ál-ity intúit-ive / *-ív-ity

b. deféat-ist / *-íst-ic enórm-ous / *-ós-ity parént-al / *-ál-ity aggréss-ive / *-ív-ity

The development of hyperproductivity in English suffixation, while basically determined by stress, can have specific segmental restrictions. The data in (33a) demonstrate the automatic licensing of -ation-nominalization for -ize-suffixation in English, except when the stem ends in a coronal fricative, as in (33b). The avoidance of such stems appears to be due to the presence of a single unstressed vowel between the coronal fricatives in cases like *fànta/sɪz/átion, as opposed to the stressed diphthong in fánta/sàiz/. All -izeformations in (33) are native and involve bound stems, inferred by affix stripping (e.g., fántas-y, crític-ism, mémor-y, émphas-is). (33) a.

mémor-ìze hármon-ìze metábol-ìze plágiar-ìze

→ → → →

(mèmor-iz-átion)ω (hàrmon-iz-átion)ω (metàbol-iz-átion)ω (plàgiar-iz-átion)ω

b. fánta/s/-ìze / *-iz-átion críti/s/-ìze / *-iz-átion éxor/s/-ìze / *-iz-átion óstra/s/-ìze / *-iz-átion

910

V. Rules and restrictions in word-formation II: Special cases epítom-ìze mésmer-ìze súbsid-ìze monópol-ìze metrópol-ìze

→ → → → →

(epìtom-iz-átion)ω (mèsmer-iz-átion)ω (sùbsid-iz-átion)ω (monòpol-iz-átion)ω (metròpol-iz-átion)ω

émpha/s/-ìze / *-iz-átion públi/s/-ìze / *-iz-átion sýmpa/θ/-ìze / *-iz-átion éner/dʒ/-ìze / *-iz-átion apólo/dʒ/-ìze / *-iz-átion

The gap in nominalization illustrated in (33b) indicates the satisfaction of a markedness constraint against nuclei flanked by coronal fricatives. A rough classification of the phonological markedness constraints playing a role in English suffixation is given in (34): (34) a. Syllable wellformedness (onset, sonority thresholds, hiatus avoidance, identity avoidance) b. Foot well-formedness (*clash, *lapse, size constraints) c. Phonological word (size constraints) d. Alignment constraints Identity avoidance typically refers to phonemes in specific syllable positions, including the segments flanking the nucleus, the onsets in adjacent syllables, or the codas in adjacent syllables. English suffixes associated with dissimilatory constraints were listed in (7). The stem restrictions on the suffixes -en and -ive stated in (5) are examples for onset constraints. The restriction to obstruent-final stems associated with verbal -en satisfies a phonological constraint concerning sonority thresholds: due to their low sonority, obstruents constitute optimal onsets. (The constraint in question concerns only verbal -en, not the homophonous adjectival suffix -en, cf. woolen.) The restriction to stems ending in -s or -t associated with -ive satisfies an additional constraint on place of articulation: alveolar consonants are least marked. Apart from suffixes which are associated with constraints favoring particular unmarked onsets there are others, such as -eer and -ese, which accept any onset, requiring only that the stem end in a consonant. While the suffixes -en and -ive attach only to stems ending in a phoneme with low sonority, the suffix -th attaches only to stems ending in a segment with relatively high sonority, such as liquids or vowels, cf. (35). This difference is prosodically conditioned: in vowel-initial suffixation, the stem-final consonant forms an onset, where minimal sonority is preferred, whereas before a consonantal suffix, maximally sonorous stemfinal consonants are preferred to ensure a maximal sonority decrease among the two consonants. (35)

(illth)ω (< ill) (spilth)ω (< spill) (blowth)ω (< blow)

(tilth)ω (< till) (coolth)ω (< cool) (sloth)ω (< slow)

Apparent counter-examples have cognates in other Germanic languages (cf. Engl. depth, Gothic diupiþa, Dutch diepte; Engl. length, Swedish längd, Dutch lengte; Engl. breadth, Icelandic breidd, Dutch breedte), presumably originating at a time when the suffix was still vowel-initial. (Its reconstructed form in Germanic is -iþō.) As was noted above,

50. Phonological restrictions on English word-formation

911

inherited words are not expected to satisfy the constraints associated with synchronic rules for novel word-formation. Regarding sensitivity to foot structure, the stressed suffixes -ize and -ify differ in that -ize tolerates so-called stress lapses, that is sequences of unstressed syllables as in (36a), while -ify does not, cf. (36b). The other examples show that -ize-suffixation must not exhibit a stress clash: as a result the suffix attaches neither to monosyllabic nor to iambic stems, cf. (36c, d). (36)

a. (rádical-ìze)ω (prímitiv-ìze)ω (skéleton-ìze)ω (cánnibal-ìze)ω

b. *féudal-ifỳ *áctiv-ifỳ *cárbon-ifỳ *tríbal-ifỳ

c. *stríct-ìze *crísp-ìze *púr-ìze *bóld-ìze

d. *obscúr-ìze *concrét-ìze *cafféin-ìze *robúst-ìze

Size constraints can be illustrated by comparing the unstressed suffixes -en and nominal -al in (37). Both suffixes select stems with final stress, to yield a word which ends in a branching foot; they differ in that -en-suffixation has two syllables while -al-suffixation has three syllables: (37) a.

(tóugh-en)ω (swíft-en)ω (héight-en)ω (thréat-en)ω

b. (caróus-al)ω (survív-al)ω (perús-al)ω (appráis-al)ω

The restriction on -en-suffixation indicates a constraint requiring the alignment of foot and word boundaries in the output. The restriction to three syllables could be expressed by associating nominal -al-suffixation with a requirement for binary branching structure both within and above the foot level, cf. (6b). Alternatively, the number of syllables within the phonological word could also play a role. Evidence for relevant maximality constraints can be gleaned from the formations in (38), where bound stems are preferred to free stems if this serves to avoid words with more than four syllables: (38) a. déleg-àte+able ségreg-àte+able váccin-àte+able precípit-àte+able

(déleg-able)ω (ségreg-able)ω (váccin-able)ω (precípit-able)ω

b. dón-àte+able díl-àte+able nárr-àte+able vác-àte+able

(dón-àte-able)ω (díl-àte-able)ω (nárr-àte-able)ω (vác-àte-able)ω

5. Prefixes vs. modifiers The distinction between modifying and head prefixation in English is characterized by a number of correlating properties, concerning morphosyntactic, semantic and phonological structure. Modifying prefixes combine with words without affecting the category of the derived word and can therefore be freely omitted without affecting grammaticality (she was ✓unable/✓able to leave). In accordance with their optionality modifying prefixes have inherent meaning (e.g., un- ‘not’). By contrast, head prefixes determine the resulting category − mostly verbs, but also adverbs and prepositions (e.g., anew, ago, behind) − and cannot be omitted (they ✓enabled/*abled her). Head prefixes often lack

912

V. Rules and restrictions in word-formation II: Special cases inherent meaning. For instance, the causative meaning in the verb enable is not necessarily associated with the prefix en-, as this meaning component appears to be present whenever a transitive verb has a recognizable adjectival base (cf. becalm, deepen, slenderize, intensify, corrupt). Modifying prefixes, but not head prefixes, form separate phonological words. The prosodic organization of unable in (39a) is manifest both in the word-initial secondary stress and in the potential juncture before the stem-initial vowel. By contrast, the lack of initial stress together with the pronunciation of the prefix-final consonant as the onset of the stem-initial syllable indicates the integration of the head prefix en- into the phonological word of the stem, cf. (39b). In native word-formation, both modifying and head prefixes combine with free stems only. (39)

Whether head prefixes are always prosodically cohering, or whether they are integrated only when necessary for supplying an onset, is not clear. In either case, the representation, also that with an unintegrated head prefix as in (39c), differs from the prosodic organization of modifying prefixes, which form separate phonological words. These prosodic differences are reflected in how much the prefixes are sensitive to the phonological structure of the stem: there are no phonological restrictions on modifying prefixation (cf. (40a)), while head prefixation has phonologically conditioned gaps. In English, head prefixes attach only to stems consisting of a single foot consisting of one or two syllables as is illustrated in (40b): b. em(pówer)ω (40) a. (ùn)ω(fít)ω (ùn)ω(belíef)ω em(bódy)ω (ùn)ω(pópular)ω en(mésh)ω (ùn)ω(méntionable)ω en(nóble)ω (ùn)ω(nécessary)ω en(líst)ω

c. (ìl)ω(lógical)ω (ìr)ω(repláceable)ω (ìm)ω(bálance)ω (ìm)ω(méasurable)ω (ìm)ω(precíse)ω

Modifying prefixation never exhibits gaps nor stem allomorphy. There is one such prefix, the negative prefix in- illustrated in (40c), whose final sonorant assimilates fully to a following liquid and partially to a following stop or nasal; curiously this prefix assimilates even more strongly than the head prefix, cf. (40b, c). Assimilation does not indicate prosodic fusion here as is shown by the evidence from stress: secondary stress before

50. Phonological restrictions on English word-formation following main stress as in ìmbálance and final main stress in ìmprecíse indicate the organization of the prefix as a separate stress domain, with weaker stress than the stem (Wells 1990). The absence of phonologically conditioned gaps follows from this organization. There are various word-formation rules which are sensitive to the distinct prosodic organization between modifying and head prefixations. Verb-to-noun conversion is rather productive in modifing prefixation with a monosyllabic stem, cf. (41a, b), much less so in head prefixation, cf. (41c). The latter formations would involve a dilemma: the derived noun either has final main stress and thereby deviates from regular stress patterns in nouns or there is violation of paradigm uniformity in that the main stress in the noun corresponds to a syllable which is unstressed in the base. No such conflict is present in the conversions in (41a, b), which involve a mere reversal of relative prominence among phonological words. (41) a.

V (mìs)ω(prínt)ω (prè)ω(bóil)ω (rè)ω(chárge)ω (sùb)ω(léase)ω

b.

N (mís)ω(prìnt)ω (pré)ω(bòil)ω (ré)ω(chàrge)ω (súb)ω(lèase)ω

c.

V be(stów)ω en(rích)ω re(néw)ω de(fróst)ω

While generally organized as unstressed syllables, head prefixes can form separate phonological words when participating in antonymic relations as in (42b) (Eckert and Barry 2002: 115). (42) a. de(dúce)ω − in(dúce)ω ex(plóre)ω − im(plóre)ω

b. (dè)ω(créase)ω ↔ (ìn)ω(créase)ω (èx)ω(pórt)ω ↔ (ìm)ω(pórt)ω

Antonymic verbs are also more prone to be converted to nouns (with concomitant relative prominence reversal). The observation that, here, too, conversion correlates with the prosodic organization of the base verb suggests that word-formation is sensitive to prosodic organization itself, rather than the underlying morphosyntactic or semantic differences.

6. Implications for the mental lexicon The phonological restrictions observed in English word-formation lend themselves to an analysis based on interacting phonological markedness constraints and paradigm uniformity constraints referring exclusively to output forms. These restrictions moreover provide insight into the sort of structure and level of phonological abstractness to be associated with those output forms. The presence of prosodic organization in the relevant representations is reflected in various restrictions on word-formation, including sensitivity to the location of feet and to the number of syllables in a phonological word. Regarding the level of abstractness of segmental structure, phonological constraints requiring segmental non-identity need to refer to phonemic representations, from which all contextually determined allophony has been abstracted. For instance, the ungrammaticality of

913

914

V. Rules and restrictions in word-formation II: Special cases formations in which a nucleus is flanked by identical segments such as *appéalal indicates reference to the phonemic level, where the two liquids are indeed identical (i.e. /ləl/), not the phonetic level, where the liquids may differ quite strongly in many varieties of English as a result of their appearance in onset vs. coda position (i.e. “clear [l]” vs. “dark [ɫ]” in [ləɫ]). The evidence for reference to phonemic representation in word formation, along with the desirability of limiting such reference to a single level, calls for a close inspection of alleged counter-evidence. Dressler (1977) argues that the regular truncation of the sequence -er in German -in-suffixation as in Zauberin ‘sorceress’ (*Zaubererin), but not in -er-suffixation as in Zauberer ‘sorcerer’, can only be explained with reference to phonetic representations. Specifically, he claims that the constraint against sequences of identical CV-strings would be violated only in *Zaubererin, where both instances of /ʀ/ occur in onset position and therefore are pronounced alike. Consequently truncation applies. By contrast, in Zauberer the final /ʀ/ occurs in coda position, being subject to /ʀ/-vocalization. Consequently the relevant strings are phonetically distinct, obliterating the need for truncation. (43) Phonemic: Phonetic:

/tsáubəʀəʀ/ [tsáubəʁɐ]

/tsáubəʀɪn/ */tsáubəʀəʀɪn/ [tsáubəʁɪn] *[tsáubəʁəʁɪn]

Dressler’s analysis, if correct, would indeed refute the generalization that word-formation is sensitive to phonemic representation only, as /ʀ/-vocalization is clearly allophonic. However, his conclusion is not cogent because the relevant words also exhibit a difference at the phonemic level: unlike Záuberer, *Záubererin includes a sequence of three syllables following the main stress. Similar restrictions are observed in English -esssuffixation (e.g., sórcerer − sórceress (*sórcereress), múrderer − múrderess (*múrdereress)). All of these instances of “truncation” lend themselves to an analysis where bound stems are selected if necessary to satisfy phonological constraints in word-formation. Significantly, bound stems are inferred by stripping affixes from the phonemic representations of complex words (cf. the examples in (6) or (11)), with no reference to phonetic form. The question of which level of abstractness is referred to in word-formation is of considerable theoretical significance. If there is no evidence for any reference to allophonic structure in word-formation, this would support a model of the lexicon, which recognizes a separate level of “post-lexical” phonetic implementation. This conclusion is consistent with standard assumptions in lexical phonology (cf. Kiparsky 1982; Mohanan 1982). The exclusive reference to (prosodically organized) phonemic representations in word-formation has also been challenged by those who posit more abstract morphophonemic representations. The alleged claims are based on putative correlations between affixes which are “stress-shifting” and combine with bound stems (so-called “level 1” affixes) vs. those which combine with free stems and are “stress-neutral” (so-called “level 2” affixes) (cf. Newman 1948; Chomsky and Halle 1968; Siegel 1974; Kiparsky 1982; Mohanan 1982). The relevant distinction is captured by representing bound stems abstractly, such that phonemic form and prosodic structure are created only in the course of stress-shifting affixation. Empirically, this approach is flawed in that stress neutrality and stem selection (free vs. bound) do not in fact correlate. Indeed there are many

50. Phonological restrictions on English word-formation suffixes which never shift stress, yet select bound stems when necessary to satisfy phonological constraints, e.g., -able, -ee, -er, -ize, -ify, cf. (26), (36), (38). Moreover, the twolevel models predict that stress-neutral suffixes can be sensitive to prosodic stem structure; yet consonant-initial suffixes, all of which are stress-neutral, are in fact never sensitive to such structure. Those models further predict that stress-shifting suffixes cannot follow stress-neutral suffixes; yet the cases of hyperproductivity identified in (31) are precisely of that kind. The evidence from phonological restrictions on English wordformation hence supports a “one-level” model where affixes and stems, free or bound, are represented phonemically, including prosodic organization. Phonological restrictions on word-formation also shed light on the acquisition of morphological rules. Here, the idea that novel word-formation is best described as a proportional analogy (Becker 1990), where pairs of related words serve as a concrete model for new coinages (e.g., explain : explanation = compláin : X) is not supported. There is no reason to doubt that learners recognize the relatedness between the inherited loans in -ation and the relevant iambic verbs in (44a) and yet analogous stems are strictly avoided in native -ation-suffixation. (44) a.

èxplanátion − expláin pèrturbátion − pertúrb ìnspirátion − inspíre àdorátion − adóre cònsultátion − consúlt còndensátion − condénse èxhalátion − exhále

b. compláin / *complanation distúrb / *disturbation desíre / *desiration ignóre / *ignoration insúlt / *insultation incénse / *incensation assáil / *assalation

The rejection of the starred forms in (44) indicates the association of -ation-suffixation with necessary paradigm uniformity in the minds of English learners, despite the dozens of apparent exceptions to this constraint in existing noun-verb pairs as in (44a). As was noted above, the recognition of the paradigmatic relations in (44a) and the rejection of analogous formations are easily reconciled: one concerns analysis of the given while the other concerns synthesis of the new. Regarding the latter, it appears that paradigm uniformity functions as some kind of default constraint on English word-formation, which is “demoted” only for certain affixes, in fact never for word-formation involving verbs. It appears then that the potential for analogy is restricted by specific structural constraints, including those on prosodic organization. For instance, citing the examples beatnik, beachnik, peacenik, Bauer (1983: 260) notes the original prevalence of the phonetic pattern /i/ + obstruent in the base, which however soon dissolved (e.g., jazznik, folknik, so-whatnik, draftnik, fashionnik, computernik). Bauer’s remark that the pattern /i/ + obstruent “can never have provided more than a preference” makes sense in view of the prosodic structure: being consonant-initial and therefore non-cohering, the suffix -nik is prosodified outside of the phonological word of the stem and therefore is unlikely to be sensitive to stem phonology. Hence the pattern can only function as a fleeting model for sporadic coinages, but cannot become encoded in grammar.

915

916

V. Rules and restrictions in word-formation II: Special cases

7. Summary Phonological restrictions on English word-formation manifest both as allomorphy and as gaps, that is, abstention from word-formation for phonological reasons. The occurrence of such restrictions can be linked to certain phonological properties of suffixes and to morphosyntactic properties of prefixes. The observation that in English phonological restrictions are characteristic for vowel-initial or vowelless suffixes, which are integrated into the phonological word of the stem, but are not characteristic for consonant-initial syllabic suffixes, which are not integrated, indicates the relevance of prosodic organization. Modifying prefixes, which form separate phonological words, and consonant-initial suffixes, which are also prosodically organized separately from the phonological word of the stem, are attached in virtual absence of any phonological restrictions; presumably this is a consequence of their being comparatively separate prosodically. The range of phonological restrictions indicates the satisfaction of certain constraints, including phonological markedness constraints, in output forms. The reference to outputs also explains the reason behind the relevant restrictions: the goal is to yield phonologically wellformed words. Phonological wellformedness is thereby assessed neither at the phonetic level, nor is there any reference to putative abstract morphophonemic representations. The evidence from phonological restrictions on English word-formation supports a model of the lexicon, where all morphological units − words, stems, and affixes − are represented phonemically and word-formation is completed prior to phonetic implementation.

8. References Aronoff, Mark 1976 Word Formation in Generative Grammar. Cambridge, MA: MIT Press. Bauer, Laurie 1983 English Word-formation. Cambridge: Cambridge University Press. Becker, Thomas 1990 Analogie und morphologische Theorie. München: Fink. Booij, Geert 1985 The Phonology of Dutch. Oxford: Clarendon Press. Chapin, Paul 1970 On affixation in English. In: Manfred Bierwisch and Karl-Erich Heidolph (eds.), Progress in Linguistics, 51−63. The Hague: Mouton. Chomsky, Noam and Morris Halle 1968 The Sound Pattern of English. New York: Harper & Row. Dixon, Robert M. W. 1977 Some phonological rules in Yidiny. Linguistic Inquiry 8(1): 1−34. Dressler, Wolfgang U. 1977 Phono-morphological dissimilation. In: Wolfgang U. Dressler and Oskar E. Pfeiffer (eds.), Phonologica 1976, 41−48. Innsbruck: Institut für Sprachwissenschaft. Eckert, Hartwig and William Barry 2002 The Phonetics and Phonology of English Pronunciation. Trier: Wissenschaftlicher Verlag. Flemming, Edward 1995 The analysis of contrast and comparative constraints in phonology. Talk presented at MIT Phonology Circle lecture. Huddleston, Rodney and Geoffrey K. Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press.

50. Phonological restrictions on English word-formation Jespersen, Otto 1942 A Modern English Grammar on Historical Principles. Part VI: Morphology. London: Allen & Unwin. Jones, Daniel 2011 Cambridge English Pronouncing Dictionary. Ed. by Peter Roach, James Hartman and Jane Setter. 8th ed. Cambridge: Cambridge University Press. Kiparsky, Paul 1982 From cyclic phonology to lexical phonology. In: Harry van der Hulst and Norval Smith (eds.), The Structure of Phonological Representations. Part 1, 131−175. Dordrecht: Foris. Lappe, Sabine 2007 English Prosodic Morphology. Dordrecht: Springer. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. Mohanan, Karuvannur Puthanveettil 1982 Lexical phonology. Ph.D. dissertation, MIT. Muthmann, Gustav 1999 Reverse English Dictionary. Based upon Phonological and Morphological Principles. Berlin/New York: Mouton de Gruyter. Newman, Stanley S. 1948 English suffixation: A descriptive approach. Word 4, 24−36. Oxford English Dictionary online: http://www.oed.com [last access 9 Mar 2012] Plag, Ingo 2003 Word-Formation in English. Cambridge: Cambridge University Press. Prince, Allen and Paul Smolensky 1993 Optimality Theory. Constraint Interaction in Generative Grammar. Technical Report CU-CS-696-96. RuCCS-TR-2. [2002, as ROA-537]. Quirk, Randolph, Sydney Greenbaum, Geoffrey Leech and Jan Svartvik 1972 A Grammar of Contemporary English. London: Longman. Raffelsiefen, Renate 1999 Phonological constraints on English word formation. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1998, 225−288. Dordrecht: Kluwer. Raffelsiefen, Renate 2004 Absolute ill-formedness and other morphophonological effects. Phonology 21: 91−142. Raffelsiefen, Renate 2005 Paradigm uniformity effects versus boundary effects. In: Laura Downing, Tracy Allan Hall and Renate Raffelsiefen (eds.), Paradigms in Phonological Theory, 211−262. Oxford: Oxford University Press. Raffelsiefen, Renate 2007 Morphological word structure in English and Swedish: The evidence from prosody. In: Geert Booij, Luca Ducceschi, Bernard Fradin, Emiliano Guevara, Angela Ralli and Sergio Scalise (eds.), On-line Proceedings of the Fifth Mediterranean Morphology Meeting, Fréjus, 15−18 September 2005, 209−268. University of Bologna. (http://mmm.lingue. unibo.it/mmm-proc/MMM5/MMM5-Proceedings_full.pdf) Siegel, Dorothy C. 1974 Topics in English morphology. Ph.D. dissertation, MIT. Wells, John C. 1990 Longman Pronunciation Dictionary. Harlow: Addison Wesly Longman.

Renate Raffelsiefen, Mannheim (Germany)

917

918

V. Rules and restrictions in word-formation II: Special cases

51. Morphological restrictions on English word-formation 1. 2. 3. 4. 5.

Introduction Morphological restrictions on compounding and derivation Combinability, selectional restrictions and parsing in (multiple) affixation Conclusion References

Abstract Morphological restrictions have been dealt with in traditional accounts of English wordformation as well as in more recent and theoretically-oriented studies. The present article gives an overview of the history of the treatment of morphological restrictions and the major theoretical questions and concepts involved. Its focus is on studies pertaining to the complex issue of restrictions on English multiple affixation, or, more precisely, multiple suffixation. The linguists’ debate suggests that reliable findings are more likely when a combination of morphological, phonological and speech perception parameters as well as the parameter of frequency of occurrence are applied in the investigation process.

1. Introduction The study of morphological restrictions (or constraints) in the system of English wordformation has been the subject of scholarly research, with varying approaches, for more than sixty years. Among the early studies mentioning problems of restriction, we find Jespersen (1961 [1942]) and Marchand (1969). However, neither of these works deals with restrictions in a systematic way. Koziol (1937, 1972), a diachronic and more formally oriented description of English word-formation, includes no theoretical discussion of the issue of productivity and restrictions at all. Nonetheless, the early scholars were all inspired by Paul’s (e.g., 1896, 1920) pioneering works. In Paul (1896: 694−697), for example, we find an elaboration of the idea that competing word-formation patterns interfere with each other’s productivity. He also establishes the fact that agent nouns of a specific type (e.g., *Weiner ‘crier’) cannot be derived from certain verbs of Modern German (Paul 1896: 698). Paul’s treatment of word-formation thus already points towards correlations between different levels of language − an idea he exemplifies by making reference to Germanic languages. As for English, Jespersen (1961 [1942]: 418−420), in his discussion of the suffix -ful, also mentions the aspect of competition between English -ful and the foreign suffix -able. Such findings, obviously, lead Hansen et al. (1982: 33−34), when outlining the principles that underlie the production of new complex words, to assume that morphological restrictions are intertwined with etymological and semantic restrictions. They also say that (systematic) morphological restrictions are predominantly related to the morphological structure of derivational bases and thus explain the combina-

51. Morphological restrictions on English word-formation bility of, e.g., -ful with, almost exclusively, simple stems and that of -ness with both simple and complex stems, the use of which, however, blocks any subsequent formation (Hansen et al. 1982: 34). Pertaining to the same patterns, we find an analogous passage in Kastovsky (1982: 161). The aspect of blocking in -ness derivations also raises the question of whether this is caused by the morphological structure or, e.g., by the prosodic structure of the complex word. More recently, this phenomenon was explained by the concepts of “closure” in Stump (1998) and “closing morphemes” in Szymanek (2000) (cf. also article 54 on closing suffixes and Stein 2009: 238). Hansen et al. (1982: 34) conclude their short general introduction to morphological restrictions by saying that (a) the potential of derivational options are the fewer, the more derivational stages are already contained in a stem, and (b) the upper limit of derivation is formed by the result of four successive derivational processes, such as in the word decentralization. Apart from these structure-related aspects, the issue of etymology is discussed fairly early on with a view towards affixation. Marchand (1969: 318−321) and Jespersen (1961 [1942]: 318−320), for example, examine whether the suffix -ize/-ise can only be combined with bases of Latinate and Greek (or other foreign) origin and, after examination, find a number of counter-examples. Jespersen (1961 [1942]: 319), however, fails to give evidence for his statement that “[a] great many derivatives formed from Engl[ish] roots are in common use”. The study of origin and diachronic development seems to be one of the keys to a number of questions involved in the study of word-formation constraints. However, with massive, deep-reaching foreign inputs entering the lexicon of English in the Middle English and Early Modern English periods, the potential for productive word-formation patterns was enormously widened and equipped with a wealth of new options (cf. article 110 on historical word-formation in English). Early and Late Modern English, at least until ca. 1750, had no strict rules as to the use of certain affixes. This often resulted in the existence of two, three or even four formations expressing the same concept with the same degree of complexity. Many of the rules and restrictions of interest in our realm are, then, the result of later standardisation/rationalisation (e.g., the principle of blocking) and the course of “pure” morphological development. With regard to the process commonly referred to as “standardisation”, I agree with Kastovsky (2006: 170) when he says, “[…] the rivalry of these competing patterns and their sorting out in the eighteenth and nineteenth centuries are in need of further investigation, especially with regard to their distribution among text types and the influence of prescriptive grammar”. However, this article will be concerned with the aspect of synchrony only, i.e. it will focus on the problem of morphological restrictions in Present-Day English.

2. Morphological restrictions on compounding and derivation 2.1. Defining and delimiting morphological restrictions Traditionally, morphological restrictions are conceived of as structural restrictions that are due to morphological properties of the constituents of a word-formation, e.g., those

919

920

V. Rules and restrictions in word-formation II: Special cases of a (potential) derivational base or those of an affix (potentially) attaching to some base. In the history of the study of English word-formation, the notion of what morphological properties comprise has varied to some extent. This is especially true of the relationship between morphology and etymology. Thus, in certain publications on (English) wordformation, the etymological aspect is discussed as a separate phenomenon as in Marchand (1969, sections 3.1.2, 3.1.3 and 4.1.4−8). Sometimes restrictions that are caused by the etymology of, for instance, specific affixes are considered to be outside morphological restrictions (cf., e.g., Schmid 2011: 119 and Hansen et al. 1982: 34−35). In this article, however, morphological restrictions are understood to include issues of the origin (source language/s) of affixes as well as formal development of affixes. We find this to be in accordance with most linguistic accounts of the issue of morphological restrictions. Another general issue concerns the questions of where exactly these restrictions are “located”. In some publications, we find that morphological restrictions are similarly addressed as pattern-related (e.g., in Schröder and Mühleisen 2010) or “modal-specific” (Schmid 2011: 118). In other words, they only consider the morphological combination of, e.g., a base and an affix in the sense of the affix attaching or not attaching, for one reason or the other, to some base. Schmid (2011), in this context, discusses Aronoff’s (1976) example of the adjectival suffix -al, which can only be attached to complex nouns ending in -ment if -ment itself is not a suffix but part of a morpheme. However, one of the most researched problems in this respect is multiple affixation and the order in which affixes occur therein, particularly when the same affixes vary their order in different complex words (cf. Hay and Plag 2004, also Schmid 2011: 119). In some earlier studies, we find that a distinction is made between restrictions relating to individual words and others concerning a pattern. Thus Neuhaus (1971, ch. 3) develops a model of level I to level III restrictions, in which, e.g., on level II various systematic conditions are shown (Neuhaus 1971: 35). A similar approach is developed in Hansen et al. (1982: 33), in which a division was introduced between non-systematic (itemspecific) and systematic (pattern-related) restrictions. Consequently, they explain the impossibility of the formation *friendlinessful on the basis of two systematic restrictions: (1) -ness blocks any further formation, and (2) -ful can only (almost exclusively) be attached to simple stems.

2.2. Compounding and derivation The study of the relevant literature shows that the two major fields of English wordformation, compounding and derivation, have attracted interest in restrictions in an unbalanced way. As regards compounding, no paper has been written which systematically deals with morphological restriction. It is, therefore, not surprising that all recent handbook-type publications address affixation only, or, more precisely, suffixation. This is why such comments cannot be found in Schmid (2011: 115−120), Lieber and Štekauer (2009) or Štekauer and Lieber (2005). In Lieber and Štekauer (2009), a handbook of compounding, any reference in its index to restrictions or constraints is even missing altogether.

51. Morphological restrictions on English word-formation However, since the publication of Gold’s (1969) article on the competing patterns with a (de)verbal first constituent as represented by the compounds frypan and frying pan, there has been some awareness of the implications. Apart from different preferences in the various national varieties of English, there seem to be certain restrictions on the use of verbal or nominal constituents. We notice, for example, that only the pattern V + N is used to form a compound with verbal blow as determinant; cf., e.g., blow-pipe, blowhole and blowlamp. Conversely, compounds with a (de)verbal determinant and the noun method as the determinatum only follow the (V + -ingN) + N pattern: cf., e.g., learning method, sampling method and teaching method. As mentioned before, most of the research work concerned with morphological restrictions on English word-formation has been done in the field of affixation in general and multiple affixation in particular. Here only a small number of publications are concerned with special issues of prefixation; cf. from among the established works, e.g., Marchand (1969: 134−137) and Hansen et al. (1982: 67−71). One of the few topics addressed is the morphological and semantic interpretation of denominal verbs, e.g., unbutton (with a reversative meaning), encage (directional/locative) and disarm (privative), in connection with which, according to Marchand (1969: 134), “the formula AB is B does not apply”. A similar pattern is discussed by Hansen et al. (1982: 68−70) with regard to complex words such as postcentral in the meaning ‘related to something behind the centre’ or transcontinental ‘related to something across (the) continents’. For these formations, they assume a structure with a “prepositional” prefix, a nominal stem and an adjectival suffix, determining the word-class. Because a regular morphological analysis of the formations and their semantic analysis contradict each other, they consider most of the formations in question pseudo-prefixations and categorise them as zeroderivations. Among the more recent publications, there are very few on problems of prefixation. Lehrer (1995) is one of greater general interest in that she raises several questions pertaining to the general status of prefixes, their combinability with one another and the role of semantic and pragmatic restrictions relevant to their use. Schröder (2011) is a study combining the issue of productivity of selected verbal prefixation patterns and their occurrences in several corpora as well as the statistical measurement thereof. When reflecting on the theoretical basis of her study, she discusses restrictions as “limitations of productivity” in general (section 2.4) and language-specific and rule-specific aspects in particular (Schröder 2011: 43−45). A result of her analysis with relevance to our problems is that prefixed verbs, in relation to (in certain respects) functionally equivalent particle verbs (e.g., underperform vs. go under), are morphologically more versatile than the latter in that they can generate structurally more diverse derivations or derivations at all (Schröder 2011: 81−82). In her conclusion (sections 7.3 and 7.4), she revisits such problems, saying, “[a]lthough […] productivity cannot be defined on the basis of rule restrictions alone, these nevertheless have an influence on the productivity of a wordformation rule” (Schröder 2011: 246). Otherwise, she reports no morphological restrictions (also cf. section 7.3). Returning to Lehrer (1995), she examines, in the context of prefixation, two major issues: selectional restrictions and whether affixes can be regarded as having the status of signs. However, her study is also concerned with combinations of prefixes. In addition, she raises the point of “bracketing paradoxes” (Lehrer 1995: 143), referring to the problem of whether the principle that all affixation takes place before compounding

921

922

V. Rules and restrictions in word-formation II: Special cases found elsewhere is applicable to prefixation as well (cf., e.g., anti-animal rights with the meaning ʻagainst animal rights’), an idea she rejects concerning the pattern in question. She also discusses combinations of prefixes and suffixes as in anti-racism and antiliberialism, revisiting the aspects dealt with by Marchand (1969) and Hansen et al. (1982). To turn to her initial issues again, Lehrer (1995: 147) argues in her summary that “many limitations on recursion and combination of word-formation processes should be explained in terms of pragmatic and processing factors, as well as in terms of the semantics of the affix and the base” and adds that, in tendency, “affixes are similar to lexemes”. The latter idea is emphasised by her observation of the obvious parallelism among certain prefixes, on the one hand, and certain prepositions, adjectives and adverbs on the other. This short discussion may serve to show why the problem of morphological restrictions on English prefixation is so difficult to deal with. It is apparent that the use of prefixes is not only determined by simple structural rules and restrictions, but also by a combination of linguistic structure and meaning properties as well as by mental processing and, possibly, extralinguistic factors. As pointed out earlier, however, compounding and prefixation have not been in the focus of most studies which have been concerned with morphological restrictions on English word-formation. Relevant publications in the field, in different traditions and stages, have, on other hand, been written about suffixation. The following section will deal with various approaches towards restrictions on suffixation.

3. Combinability, selectional restrictions and parsing in (multiple) affixation Apart from the concepts of the earlier publications discussed here in sections 1 and 2 (e.g., Jespersen 1961 [1942]; Marchand 1969; Hansen et al. 1982), the problem of restrictions on multiple affixation, particularly suffixation, has attracted much attention over the last four decades. As will be shown, this issue is not easy to clarify. The first studies addressing this problem in a more or less systematic way are Stein (1971) and Neuhaus (1971). Other or new aspects of multiple affixation are discussed by, e.g., Burgschmidt (1978), Siegel (1979), Kiparsky (1982), Fabb (1988), Giegerich (1999), Plag (1996, 1999, 2002, 2003), Hay (2000, 2002), Hay and Baayen (2002), Aronoff and Fuhrhop (2002), Hay and Plag (2004), Rainer (2005), Kastovsky (1986, 2005), Bauer (2001, 2005), Plag and Baayen (2009), and Stein (2009). Most of the more recent investigations are enriched and made more reliable concerning the results they yield by the use of large electronic corpora and/or databases, such as the British National Corpus or the Oxford English Dictionary in one of its electronic versions.

3.1. Early studies on English multiple affixation Stein (1971) is probably the first comprehensive and systematic study of French and English adjectival suffixation and prefixation, as well as of English word-formations

51. Morphological restrictions on English word-formation with a very high degree of morphemic complexity. In this work, she introduces the distinction between primary or non-derived and derived adjectives as secondary, tertiary, quaternary and quinary word-formations. This work is largely based on theoretical notions developed by Marchand and Coseriu. The basic question posed is that of whether secondary adjectives behave in the same way as primary adjectives. Another goal was to identify typological differences between the two languages. Also (and with special regard to English), the study is designed to give more detailed information about the kinds of words (derived and non-derived) in which further word-formation is not possible. A corresponding issue is to develop a theoretical basis for a new classification of prefixes and suffixes. This classification is presented as a detailed listing of (adjectival) prefixes and affixes in word-formations with indices stating the degrees of complexity in the use of the respective affix and thus its highest possible degree (cf., e.g., -able4 in depolarizable vs. -wise1 in clockwise, i.e. fourth stage of formation vs. first stage). The lists also contain information about their combinability with stems of varying complexity (nouns, verbs and adjectives) and other affixes. Thus, through implication, the reader gains access to information concerning restrictions as well. With regard to her basic question, Stein (1971: 267) summarises that “the morphological and conceptual structures of primary and secondary adjectives are different. The difference in their behaviour becomes most obvious at the level of word-formation: While derivational options of primary adjectives are relatively open, those of secondary adjectives become more and more restricted. The more word-formation stages an adjective has undergone, the fewer the chances of subsequent affixation” (my translation). However, much of the value of her study lies in the detailed findings pertaining to specific affixes. Neuhaus (1971) is designed to develop a framework for a generative morphology of English affixations and for finding restrictions on English word-formation, using the example of -ish suffixations. In analogy to phonotactic rules, he uses the term morphotactic rules and seeks to describe the combination of certain affixes (Neuhaus 1971: 47− 50). He thus shows that the suffixes -al, -ist and -ic and their combination with nouns like nature, terror and period may occur in different “cycles” of word-formation in various ways. The major part of his work is an exemplary grammar of the system of suffixations ending in -ish. It uses corpus-oriented and computer-based methods developed to process data that were gained from the Finkenstaedt, Leisi and Wolff (1970) chronological dictionary of English. Before the suffixations are examined, they are semantically classified, in an approximative way, as to their derivational bases. By comparing his data with those of other dictionaries on historical principles, he is able to show that the productivity rates of the sub-types vary according to historical periods and sociocultural foci of society (Neuhaus 1971: 159−161). In addition, he stresses the mutual influence of competing word-formation models. A typical example here is the current use of the words Indian and Greek that have rendered suffixations like Indish and Greekish obsolete, although the latter clearly fulfil the criterion of morphological well-formedness (and display a greater regularity than the former). Based on this and the discussion of language change, i.e. typological change of the meanings of -ish suffixations, he stresses that chronological indicators are absolutely essential for an adequate description of the various types of restriction (Neuhaus 1971: 166). However, this also raises the general question of the interplay between morphological and extralinguistic factors.

923

924

V. Rules and restrictions in word-formation II: Special cases

3.2. Principles of affix-ordering The first studies examining restrictions on multiple affixation within the framework of generative linguistics appeared in the 1970s using a stratification theory with a distinct phonological orientation. Drawing largely from principles described by Chomsky and Halle (1968), Siegel (1979) is an early attempt to combine aspects of derivational morphology and generative phonology, involving issues of stress rules and their influence on affixation. One of its aims is to describe how lexical entities, i.e. stems, affixes, etc., are represented in the lexicon (Siegel 1979: 101). Furthermore, affixes are classified and discussed with regard to differences in their properties. Siegel (1979: 111−148) identifies and elaborates two classes of affixes, which then, in turn, break down into prefixes and suffixes. The classification is based on the criterion of whether the affix is introduced with a + boundary (= morpheme boundary) or with a # boundary (= word boundary). Accordingly, she distinguishes Class I suffixes and Class II suffixes. Class I suffixes are introduced with the + boundary and Class II suffixes with the # boundary. According to Siegel, Class I suffixes are “all suffixes which satisfy the environment of the cyclic stress assignment rules and influence the placement of primary stress assignment rules” (Siegel 1979: 112). In other words, they cause and are affected by a shift of primary stress, such as nominal -y and -ation. Class II suffixes were considered to not be involved in such processes, e.g., adjectival -y and -less or nominal -al. In analogy, Class I prefixes and Class II prefixes were distinguished. These two classes of affixes are also called, with the notion of the corresponding strata, level 1 and level 2 affixes. In essence, the distinction implied a grouping of non-Germanic (i.e. mostly Latin or French) affixes as level 1 and Germanic affixes as level 2 with different morphological and phonological properties. With respect to these criteria, Spencer (1991: 79) lists the most mentioned affixes, in accordance with Siegel, in the following table: Tab. 51.1: Class I and class II affixes Class Class Class Class

I suffixes: I prefixes: II suffixes: II prefixes:

+ion, +ity, +y, +al, +ic, +ate, +ous, +ive re+, con+, de+, sub+, pre+, in+, en+, be+ #ness, #less, #hood, #ful, #ly, #y, #like re#, sub#, un#, non#, de#, semi#, anti#

Among the studies influenced by Siegel’s theory are Aronoff (1976) and Kiparsky (1982 and later). Aronoff’s book is the first published generative study of English word-formation. Siegel’s stratification model was included in Kiparsky’s version of lexical phonology. He assumed that the lexicon of English, in correspondence to Siegel, has primary morphology (involving word-stress and trisyllabic shortening) and secondary morphology, including compounding (with compound stress). With reference to the suffix -ism, for instance, Kiparsky (1983: 4) states that it “does not participate in the assignment of word stress and is not followed by primary suffixes because it is added at level 2, where word stress does not apply and primary suffixes are not available” and concludes that -ism “could not be sensitive to the difference between an underived base and a primary derivative”.

51. Morphological restrictions on English word-formation Although Siegel’s theory was relatively influential, it was soon criticised because it was considered too restrictive and prone to make the wrong predictions. Fabb (1988) reconsiders Siegel’s level-ordering hypothesis and the potential outcomes of the rules based on this theory. He examines them with regard to the behaviour of 43 frequently used suffixes, studying their combinability. Using the assumptions of levelordering, he predicts 459 suffix pairs. He finds, however, that there are, in fact, only about 50 attested pairs of suffixes (combinations in complex words). Thus he concludes that “[s]ome other constraint must be at work ruling out the other ca 400 suffix pairs. It can be seen that level-ordering of suffixes achieves relatively little in predicting which suffix pairs exist and which do not” (Fabb 1988: 530). Also, he identifies four groups of suffixes: (a) suffixes which never attach to an already-suffixed word (e.g., deverbal -al), (b) suffixes which attach outside one other suffix (adjective-forming -ary), (c) freely attaching suffixes and (d) according to him, a problematic group of six semi-productive suffixes (e.g., noun-selecting -al combining with -ion, -ment and -or). Based on his findings, he rejects the feasibility of level-ordering with the argument: “In each case a suffix seems to have a particular affinity for another suffix; we are not dealing with affinities between sets of suffixes, as in level-ordering” (Fabb 1988: 535). He exemplifies his reasoning by postulating selectional restrictions and chooses the categorical statement “English suffixation is constrained only by selectional restrictions” as the title of his article. Nevertheless, various more recent publications relativise Fabb’s findings. One reason for this criticism is the small empirical basis of his study because he only uses Walker’s (1924) rhyming dictionary and his own personal collection of word-formations. The first response to Fabb (1988) is by Plag (1996). Plag formulates his fundamental criticism of Fabb on two levels, theoretical and empirical. For his own study, he uses a large database that draws from Lehnert’s (1971) reverse dictionary of the English language and the second edition of the Oxford English Dictionary on CD-Rom as well as from handbooks of English word-formation. This allows him to identify a number of counter-examples to Fabb’s claims. In addition, three general aspects established by Plag are of interest: 1. The regular distinction between an unrestricted domain (-ness suffixations with a high combinability) and specific restricted domains limiting the applicability of the suffixes to a linguistically determined set of bases; 2. The postulation of “base-driven” restrictions in contrast to “affix-driven” restrictions, the latter forming Fabb’s only perspective in his analysis; 3. The notion of “lexical rules” for rules of (multiple) affixation and the “fact that lexical rules are often subject to exceptions” (Plag 1996: 794). In sum, he dismisses many of Fabb’s generalisations, particularly the ones pertaining to Fabb’s four groups of suffixes, as having no theoretical significance. Plag (1996: 795) concludes that the incapability of specific suffixes to combine with certain stems is mainly due to some important property of the putative base rather than to a feature of the suffix itself and adds: “Taking this point seriously means to question any strictly affix-driven approach as put forward by standard lexical morphology or other stratumoriented approaches.” The last aspect is emphasized again in Plag (1999: 91): “These constraints, in particular base-driven selectional restrictions and general morphological constraints like the

925

926

V. Rules and restrictions in word-formation II: Special cases Latinate constraint and blocking, regulate the applicability of derivational processes to given domains, ruling out a great many logically possible combinations of stems and affixes.” In spite of its weaknesses, Fabb (1988) and some of his observations have continued to attract attention from researchers, particularly his thoughts about certain affixes that do not attach to already affixed words. This was because his findings and their explanation implied a number of important issues concerning the roles of the base, the affix(es) and the new complex word as a whole in the over-all picture of restrictions.

3.3. Corpus evidence, psycholinguistics and parsing Hay (2002), methodologically based on Hay (2000) and Hay, Pierrehumbert and Beckman (2003) and corresponding experimental work, offers a new approach to the understanding of suffixal restrictions by including aspects of speech perception and the frequency with which the given stems and affixes occur. Within the framework of her model of morphological processing she raises the question of how a morphologically complex word is accessed (perceptively analysed). Hay (2002: 528) thus identifies two perceptive “access routes”, “a decomposed access route (e.g., climber may be accessed via the representations for climb and -er), or a nondecomposed whole-word access route (e.g., climber)”. Which of the two really applies, according to Hay, has to do with, among more factors, “the phonotactics across the morpheme boundary, and the relative frequency of the derived form and the base”. She assumes the existence of a phonological prelexical processor with sensitivity to distributional cues. Consequently, she regards certain suffixations to have different probabilities for either of the access routes, depending on whether or not the boundary between base and suffix is more likely to be perceived as a word-internal morpheme boundary and the respective suffixation to be accessed via decomposition. Such a case is given in the word pipeful because the phonemic sequence /pf/ has to be interpreted as a signal of transition from the base pipe /paɪp/ to the suffix -ful /f(ə)l/, whereas bowlful is more often interpreted as less complex. Hay’s second point is that frequent words are accessed faster (Hay 2002: 529). These assumptions together with others form a framework of hypotheses. Among the others, we find that suffixes beginning with a consonant tend to be more separable than those beginning with a vowel and that more separable affixes occur outside less separable affixes. She discusses these hypotheses in the light of her and other linguists’ experiments and relevant publications and concludes (Hay 2002: 552) that the issue of restrictions on English affix-ordering is mainly one of parsability: The overall result is that the less phonologically segmentable, the less transparent, the less frequent, and the less productive an affix is, the more resistant it will be to attaching to already affixed words. This prediction accounts for the pattern of the original affix-ordering generalization it was intended to explain. Importantly, the prediction also extends to the parsability of affixes as they occur in specific words. This accounts for the so-called duallevel behavior of some affixes. An affix may resist attaching to a highly decomposable complex word but be acceptable when it attaches to a comparable complex word that favors

51. Morphological restrictions on English word-formation the direct route in access. Understanding affix ordering, then, requires a full understanding of factors influencing the parsing and storage of individual words. […] it appears clear that the properties of affixes cannot be sensibly detached from the properties of the specific words in which they appear.

Hay and Baayen (2002) complement these generalisations by identifying a critical ratio of base frequency and the frequency of corresponding derived forms, which can be calculated and represented by a parsing line, above which (in a graph) derived forms are likely to be parsed. Hay and Plag (2004), then, is meant to confirm the applicability of Hay’s parsability theory. This research report is based on Hay (2002) and Plag (2002), the latter of which captures the essence of Hay’s theory as “complexity based ordering” (CBO). In their article, Hay and Plag test the following contrary hypotheses: (1) the hierarchy hypothesis (based on hierarchy of juncture strength) and (2) the selectional restriction hypothesis. The hypotheses are examined with regard to the combinability behaviour of 15 English suffixes (13 of which are taken from Aronoff and Fuhrhop 2002), e.g., to check all 210 potential two suffix-combinations for attestations in several very large databases, such as the British National Corpus, and on the Internet. According to their findings, Hay and Plag (2004: 591−592) summarise that Hay’s (2000, 2002) CBO predictions are largely compliant with the attestations. They find suffixes with weaker boundaries located closer to the base and those with stronger boundaries further away from the base. They also see a clear correspondence between selectional restrictions and parsing restrictions and add: “Overall, it was observed that only combinations that are well processable are possible combinations, and that this range of possible combinations is further curtailed by suffix-particular phonological, syntactic and semantic restrictions.” However, they relativise this statement by saying that the correlation between boundary strength and affix-ordering is very strong, but not absolute, as shown in the cases of -less and -ness: “-less is more separable than -ness by all measures considered. And yet -ness is very clearly positioned after -less in the ordering hierarchy” (Hay and Plag 2004: 590). In a follow-up study, Plag and Baayen (2009) basically confirm the Hay and Plag (2004) findings and also show that there is a correlation to the productivity of the suffixes in question. In addition, they focus on the issue of perception, i.e. (time-related) processing costs in the mental analysis of complex words, to find that “constituent-driven processing is not necessarily the most time-efficient way of processing. In particular, constituent-driven processing (as gauged in terms of rank in the hierarchy) does not stand in a linear relationship with processing costs” (Plag and Baayen 2009: 146). All of these research projects have been focused on English. However, Hay and Plag (2004) express the need to compare their findings to those in other languages. In this respect, Aronoff and Fuhrhop (2002) is the first attempt to examine, within the framework of the theories discussed in this article, restrictions on suffix combination in two languages, German and English. Their study is largely based on assumptions in Aronoff (1976) and uses the data set of the Oxford Dictionary of English on CD-Rom. The major finding is that German and English are typologically different with respect to suffix combination: English has the monosuffix constraint, i.e. roughly the tendency for a word to have no more than one suffix, whereas German has closing suffixes. More precisely,

927

928

V. Rules and restrictions in word-formation II: Special cases in English a suffix only combines with a Germanic base when the base has no suffix as a constituent. Concerning typology, they hold that “German morphology is closer to the Latinate morphology than to the Germanic morphology of English” (Aronoff and Fuhrhop 2002: 487−488). The study has been strongly criticised by several publications because of its theoretical and empirical weaknesses as well as its descriptive inadequacies (cf. Hay and Plag 2004: 592−593 and Stein 2009: 239−247). Stein (2009), similar to, e.g., Hay (2002) and Hay and Plag (2004), also proposes a multiple-parameter approach. She favours a combination of ten factors, such as language origin, linguistic form, position of the affix, word class (as determined by the suffix), base structure, meaning, effect (whether the base remains intact or is caused to change), productivity, valency (combining properties of affixes) and closure. In sum, these combined approaches towards a problem that is so complex and difficult to pin down at least raises the question of whether the concept of morphological restrictions in the traditional sense can still be upheld. It has become clear that the explanation routes via the morphological pattern, i.e. via base and/or affix properties, do not suffice to solve the issue adequately.

4. Conclusion First and foremost, it can be said that the concept of morphological restrictions has become blurred as a consequence of the general finding that the factors constraining English affixation are not purely morphological, even if one extends the notion of morphology in word-formation to matters of etymology and morphonology. A number of restrictions traditionally considered to be of the morphological type can be thought of as having a combination of causes. It has been shown that recent multiple-parameter studies involving morphological, phonological and speech-perception aspects as well as frequency of occurrence of wordformation elements (as attested in large electronic databases) have produced more detailed and more reliable results in comparison to the findings in traditional word-formation research, most of which were based on observation rather than systematic empirical investigation. These studies also yield more than the theories that are primarily or exclusively based on phonologically-based stratification models. In addition, the degree of “granularity” has been enhanced in the recent studies by going beyond mere attestations of the given formations in the corpora in that they include psycholinguistic methods, such as native speakers’ parsing of potential words. However, the exceptions remain despite the high degree of granularity and complexity of conditioning factors applied in the examination of English affixations. We thus agree with Schmid (2011: 119) that “the reasons for gaps in principally productive word-formation patterns have not been definitively clarified”. However, in contrast to, e.g., Schmid (2011) and Mühleisen (2010) we do not think that this is caused by the English language being flexible or elusive. Rather, we hold the view that “natural morphology” (in the extended sense above, including competition between functionally equivalent patterns) is intertwined with historical decisions ruling out certain word-formations (cf. Kastovsky’s remark discussed in section 1) and other extralinguistic factors. Nevertheless, it remains to discover whether linguistically predictable cases of restriction have increased in new

51. Morphological restrictions on English word-formation complex words of Present-Day English in comparison to those that are the result of word-formation over several previous developmental stages of English. Also, it is to be hoped that the research work discussed in this article is extended, with a similar degree of intensity, to the study of restrictions on English compounds.

5. References Aronoff, 1976 Aronoff, 2002

Mark Word Formation in Generative Grammar. Cambridge, MA: MIT Press. Mark and Nanna Furhop Restricting suffix combination in German and English: Closing suffixes and the monosuffix constraint. Natural Language and Linguistic Theory 20: 451−490. Bauer, Laurie 2001 Morphological Productivity. Cambridge: Cambridge University Press. Bauer, Laurie 2005 Productivity: Theories. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 315−334. Dordrecht: Springer. Burgschmidt, Ernst 1978 Wortbildung im Englischen. Dortmund: Lensing. Chomsky, Noam and Morris Halle 1968 The Sound Pattern of English. New York: Harper and Row. Fabb, Nigel 1988 English suffixation is constrained only by selectional restrictions. Natural Language and Linguistic Theory 6: 527−539. Finkenstaedt, Thomas, Ernst Leisi and Dieter Wolff 1970 A Chronological Dictionary of the English Language. Heidelberg: Winter. Giegerich, Heinz J. 1999 Lexical Strata in English. Morphological causes, phonological effects. Cambridge: Cambridge University Press. Gold, David L. 1969 Frying pan versus frypan: A trend in English compounds? American Speech 44(4): 299− 302. Hansen, Barbara, Klaus Hansen, Albrecht Neubert and Manfred Schentke 1982 Englische Lexikologie. Einführung in Wortbildung und lexikalische Semantik. Leipzig: Verlag Enzyklopädie. Hay, Jennifer 2000 Causes and consequences of word structure. Ph.D. dissertation, North Western University. Hay, Jennifer 2002 From speech perception to morphology: Affix ordering revisited. Language 78(3): 527− 555. Hay, Jennifer and Harald Baayen 2002 Parsing and productivity. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 2001, 203−235. Dordrecht: Kluwer. Hay, Jennifer, Janet Pierrehumbert and Mary Beckman 2003 Speech perception, wellformedness, and the statistics of the lexicon. In: John Local, Richard Ogden and Rosalind Temple (eds.), Phonetic Interpretation, 58−74. Cambridge: Cambridge University Press.

929

930

V. Rules and restrictions in word-formation II: Special cases Hay, Jennifer and Ingo Plag 2004 What constrains possible suffix combinations? On the interaction of grammatical and processing restrictions in derivational morphology. Natural Language and Linguistic Theory 22(3): 565−596. Jespersen, Otto 1961 [1942] A Modern English Grammar on Historical Principles. Part VI: Morphology. London: Allen & Unwin. Kastovsky, Dieter 1982 Wortbildung und Semantik. Düsseldorf: Bagel-Francke. Kastovsky, Dieter 1986 The problem of productivity in word formation. Linguistics 24: 585−600. Kastovsky, Dieter 2005 Hans Marchand and the Marchandeans. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 99−124. Dordrecht: Springer. Kastovsky, Dieter 2006 Typological changes in derivational morphology. In: Ans van Kemenade and Bettelou Los (eds.), The Handbook of the History of English, 151−176. Oxford: Blackwell. Kiparsky, Paul 1982 Explanation in Phonology. Dordrecht: Foris. Kiparsky, Paul 1983 Word-formation and the lexicon. In: Frances Ingemann (ed.), Proceedings of the MidAmerica Linguistics Conference, 3−29. Lawrence: University of Kansas. Koziol, Herbert 1972 Handbuch der englischen Wortbildungslehre. 2nd ed. Heidelberg: Winter. Lehrer, Adrienne 1995 Prefixes in English word formation. Folia Linguistica 29(1−2): 133−148. Lieber, Rochelle and Pavol Štekauer (eds.) 2009 The Oxford Handbook of Compounding. Oxford: Oxford University Press. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. Mühleisen, Susanne 2010 Heterogeneity in Word Formation Patterns. A Corpus-Based Analysis of Suffixation with -ee and its Productivity in English. Amsterdam/Philadelphia: Benjamins. Neuhaus, H. Joachim 1971 Beschränkungen in der Grammatik der Wortableitungen im Englischen. Ph.D. dissertation, Universität des Saarlandes. Paul, Hermann 1896 Ueber die Aufgaben der Wortbildungslehre. In: Sitzungsberichte der philosophisch-philologischen und historischen Classe der k. b. Akademie der Wissenschaften zu München, Jahrgang 1896, Heft 4, 692−713. Paul, Hermann 1920 Deutsche Grammatik. Band V. Teil V. Wortbildungslehre. Halle/S.: Niemeyer. Plag, Ingo 1996 Selectional restrictions in English suffixation revisited: A reply to Fabb (1988). Linguistics 34: 769−798. Plag, Ingo 1999 Morphological Productivity. Structural Constraints in English Derivation. Berlin/New York: Mouton de Gruyter. Plag, Ingo 2002 The role of selectional restrictions, phonotactics and parsing in constraining suffix ordering in English. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 2001, 285−315. Dordrecht: Kluwer.

51. Morphological restrictions on English word-formation Plag, Ingo 2003 Word-Formation in English. Cambridge: Cambridge University Press. Plag, Ingo and Harald Baayen 2009 Suffix ordering and morphological processing. Language 85(1): 109−152. Rainer, Franz 2005 Constraints on productivity. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 335−352. Dordrecht: Springer. Schmid, Hans-Jörg 2011 English Morphology and Word-Formation. An Introduction. 2nd ed. Berlin: Schmidt. Schröder, Anne 2011 On the Productivity of Verbal Prefixation in English. Synchronic and Diachronic Perspectives. Tübingen: Narr. Schröder, Anne and Susanne Mühleisen 2010 New ways of investigating morphological productivity. Arbeiten aus Anglistik und Amerikanistik 35(1): 43−58. Siegel, Dorothy 1979 Topics in English Morphology. New York/London: Garland. Spencer, Andrew 1991 Morphological Theory. An Introduction to Word Structure in Generative Grammar. Oxford: Blackwell. Stein, Gabriele 1971 Primäre und sekundäre Adjektive im Französischen und Englischen. Tübingen: Narr. Stein, Gabriele 2009 Classifying affixes and multiple affixation in Modern English. Zeitschrift für Anglistik und Amerikanistik 57(3): 233−253. Štekauer, Pavol and Rochelle Lieber (eds.) 2005 Handbook of Word-Formation. Dordrecht: Springer. Stump, Gregory 1998 Inflection. In: Andrew Spencer and Arnold M. Zwicky (eds.), The Handbook of Morphology, 13−43. Oxford: Blackwell. Szymanek, Bogdan 2000 On morphotactics: Closing morphemes in English. In: Bożena Rozwadowska (ed.), PASE Papers in Language Studies. Proceedings of the 8th Annual Conference of the Polish Association for the Study of English, 311−320. Wrocław: Aksel. Walker, John 1924 The Rhyming Dictionary. Revised by Lawrence H. Dawson. London: Routledge and Kegan.

Lothar Peter, Berlin (Germany)

931

932

V. Rules and restrictions in word-formation II: Special cases

52. Semantic restrictions on word-formation: the English suffix -ee 1. 2. 3. 4. 5. 6. 7. 8.

Introduction A brief historical overview The morpho-syntactic “blindness” of -ee A suffix-specific theta-role and its components Ranking of constraints and the principle of co-indexation Prototypes − a recent trend in derivational word-formation A symbolic approach to deverbal -ee derivation References

Abstract English derivatives ending in -ee are of particular morphological and semantic interest because their suffix has developed from a Romance inflectional ending to a productive element of English word-formation. Moreover, the referents of -ee derivatives, which are typically, though not exclusively human beings, play a variety of roles in the events denoted by or associated with the bases. As a result, the semantic description of these derivatives constitutes a challenge for representatives of different linguistic schools. In this article, substantial theories which aim at a restriction of derivational processes involving -ee will be presented after a brief historical overview.

1. Introduction In classical generative approaches to word-formation (e.g., Williams 1981; Selkirk 1982; Lieber 1983; Olsen 1986; Burzio 1986; Randall 1988; Rappaport Hovav and Levin 1992) it is assumed that argument structure and thematic relations play a role not only in the syntax, but also in word-formation processes. In particular, it is claimed that derivational suffixes which preferably occur in the context of verbs are sensitive to the argument structure and theta-grid of their input. A frequently discussed example is the suffix -er, which prototypically realizes the external argument of its verbal base − i.e. the argument which occupies the position of the subject in the syntax − and absorbs the theta-role associated with this argument position. The theta-role which is most likely to be absorbed by -er is the role of the AGENT (e.g., driver, singer, producer). The suffix -ee, which may be considered the passive equivalent to -er (and -or) in many contexts, frequently refers to an internal argument of the base verb and thus to an argument which surfaces as the direct or indirect object in the syntax (e.g., employee, examinee, congratulatee; sendee, addressee, explainee). However, these observations reflect strong tendencies rather than absolute properties of the suffixes concerned. A number of recent approaches to word-formation (e.g., Ryder 1999; Lieber 2004, 2006; Plag 2004; Barker 1998) convincingly show that there is no constant one-to-one relation between a suffix and a particular argument position of the

52. Semantic restrictions on word-formation: the English suffix -ee input. As far as -er is concerned, examples like roaster, wilter, diner or laugher (Ryder 1999) suggest that this suffix is by no means restricted to an agentive reading. The discrepancy between the semantics of a derivational suffix and the argument structure of its input is even more striking in the case of -ee. As stated already by Jespersen (1974 [1942]: 221−222), a number of -ee derivatives are not interpretable as direct or indirect objects of their verbal bases, e.g., absentee, debauchee, devotee, fusee or refugee. This view is shared by Bauer (1993: 245−250), who identifies four grammatically defined patterns involved in -ee derivation: Pattern 1 and 2 capture the “direct” and “indirect object” reading exemplified above. Pattern 3 gives rise to derivatives which realize the object of a preposition (e.g., laughee ‘someone who is laughed at’), and pattern 4 applies to derivatives which make reference to the subject (e.g., dilutee ‘unskilled worker used to dilute a skilled work-force’). Moreover, Bauer recognizes that there are forms which do not fit either of these patterns (e.g., biographee, amputee) or which are ambiguous between a passive and an active reading (e.g., embarkee, retiree). The examples presented so far clearly show that the reference of -ee goes beyond the semantics of the most common derivatives belonging to pattern 1 (type employee) or pattern 2 (type addressee).

2. A brief historical overview The suffix -ee goes back to the inflectional ending -é (extended to -ée in feminine forms), which marks the participe passé in French, and entered Middle English via legal terms such as appellee (< French appelé), feoffee (< Anglo-French feoffé), lessee (< Old French lessé) or assignee (< Old French a(s)signé). Passive nouns like these, some of which have an agentive pendant in -or (e.g., feoffor, lessor, assignor), still reflect the strong influence which French exerted on legal English after the Norman Conquest (Koziol 1972 [1937]: 227; Marchand 1969: 267−268; Baeskow 2002: 539−543). In Early Modern English, -ee began to assume the status of an English derivational suffix. According to Jespersen (1905: 111), the derivative vendee is symbolic of this process because it imitates the French pattern regardless of the fact that the second participle of vendre ‘to sell’ is vendu(e) and not *vendé(e). Initially, the formation of -ee derivatives was restricted to bases of Romance origin, and derivatives with native bases are not attested before the 17th century (e.g., trustee, pawnee). However, in the 19th century, -ee became a very popular suffix which was no longer confined to the legal vocabulary. Among the 19th century formations, there are derivatives like sendee, addressee, employee, visitee, kickee, jokee, conferee, laughee, biographee and many more. Likewise, the 20th century gave rise to numerous neologisms ending in -ee such as recoveree, meetee, relaxee, pickpocketee, lovee, followee or inquisitee. Quite a few of these neologisms have the character of creative ad hoc-formations and were introduced in the context of corresponding nouns in -er. The following examples are provided by Bauer (2001: 72). (1) (2) (3)

“Maybe”, he offered, “they didn’t want the follower and the followee to meet.” The inquisitor becomes the inquisitee. If there was any conning to be done, Jack was supposed to be one of the conners, not one of the connees.

933

934

V. Rules and restrictions in word-formation II: Special cases According to Bauer, -ee derivatives like these result from paradigmatic pressure, i.e. they are required as equivalents to established -er derivatives, and their paradigmatic relation to these derivatives allows for easy processing. The paradigmatic dimension of wordformation has always played a key role in the Netherlands School (cf. especially Marle 1985). Because of its semantic flexibility and its compatibility with non-native and native bases, -ee has replaced the Germanic suffix -ling in its passive reading. In Modern English, there is only a small set of patient nouns ending in -ling, including hireling, foundling, suckling and starveling.

3. The morpho-syntactic “blindness” of -ee A comprehensive analysis of -ee derivatives performed by Barker (1998) on the basis of data from the Wall Street Journal confirms the observation formulated for example by Jespersen (1974 [1942]) and Bauer (1993) that this suffix realizes not only the direct or indirect object of its verbal input. It also refers to the object of a governed preposition (e.g., conferee, consultee, gazee), to the subject of intransitive verbs (e.g., escapee, arrivee, resignee) and even to the subject of transitive verbs (e.g., attendee, forgettee, signee ‘someone who has signed a contract or register’). Moreover, -ee derivatives occasionally denote individuals which are in a part-whole relation to an unexpressed entity (e.g., amputee, erasee ‘someone whose mind has been erased’, drainee ‘person involved in brain drain’). Although the majority of -ee derivatives have a verbal input, a few denominal forms are attested as well (e.g., biographee, festschriftee, blind datee, asylee ‘person granted political asylum’). According to Barker’s (1998: 705) statistics, 53 % of the -ee derivatives under consideration realize the direct and 16 % the indirect object of transitive verbs, whereas the other types mentioned above are less central or even marginal. Nevertheless, the statistics indicate that there is not generally a one-to-one relation between an -ee derivative and a particular argument position of the base. Deverbal -ee derivatives are problematic because their referents do not form a natural class from a syntactic point of view. The following sentences from Barker (1998: 705) illustrate this point: (4)

a. b. c. d. e.

The city employed the employee. She addressed the letter to the addressee. The psychologist experimented on the experimentee. The retiree retired. The attendee attended the concert.

Derivatives of the type amputee, erasee do not realize a syntactic argument of their verbal base at all although, (parts of) their referents are definitely involved in the events denoted by the verbs. In the case of denominal derivatives like biographee, asylee, there is no syntactic argument to be realized by -ee. These are Barker’s core arguments against models of word-formation which associate suffixes with particular syntactic argument positions.

52. Semantic restrictions on word-formation: the English suffix -ee

4. A suffix-specific theta-role and its components The problems addressed in the preceding section caused Barker (1998) to describe person-denoting nouns ending in -ee from a purely semantic perspective and thus independently of syntactic argument positions projected by the input. Significantly, he postulates an individual theta-role for -ee. Influenced by Dowty’s (1991) view that theta-roles are cluster concepts rather than discrete categories, Barker defines the suffix-specific role via three interacting semantic constraints: sentience, episodic linking and lack of volitional control. The first constraint, “sentience”, requires the referent of an -ee derivative to be human. This requirement holds for most, though not for all derivatives ending in -ee. Exceptions listed by Barker (1998: 710) are actee, causee, controllee and governee, all of which constitute linguistic terms. Moreover, catapultee ‘a catapulted aeroplane’ and raisee ‘a war-ship or vessel reduced in height by the removal of her upper deck or decks’ do not refer to human beings either. Another exception not mentioned by Barker is settee ‘a long, soft seat like a sofa, with sides and back, for two or more persons’. However, since most of the -ee derivatives referring to non-sentient entities are domain-specific because they belong to the linguistic vocabulary, they are not an obstacle to the “sentience” requirement. The second constraint, “episodic linking”, requires that the referent of an -ee derivative has participated in the event denoted by or associated with the base for a certain period of time. Analogously to Carlson’s (1977) description of individual-level predicates (e.g., intelligent, tall, brown-haired) and stage-level predicates (e.g., drunk, sad, nervous), the temporally determined involvement of an individual x in the event e denoted by the base is represented in the form of a stage, i.e. an ordered pair , in the following definition of episodic linking (Barker 1998: 712): A derived noun N is EPISODICALLY LINKED to its stem S iff for every stage in the stage set of N, e is a member of the set of events that characterizes S.

For instance, John qualifies for the noun trainee if and only if e is a training event and if John has participated in this event − which is referred to as a qualifying event − in the appropriate manner. A significant aspect of “episodic linking” is its reference to time because the extension of -ee derivatives is determined by aspectual properties of the base verb. For example, the referent of employee remains in the extension of this noun only for the duration of employment. On the other hand, the referents of nouns like adoptee or retiree are permanently involved in the qualifying events denoted by the base verbs from a particular point in time. It is precisely the correspondence between the duration of involvement in an event denoted by the base and the extension of -ee derivatives which gives rise to the episodic character of these person-denoting nouns. Since the argument structure of the base is inaccessible to this purely semantic constraint, episodic linking ensures a uniform description of all -ee derivatives independently of the morpho-syntactic properties of their bases. Moreover, episodic linking applies not only to concrete events and participants, but also to implicit constituents, which are conceptually present although they are not spelled out. This property makes it a suitable device for the description of -ee derivatives which do not correspond to any of the

935

936

V. Rules and restrictions in word-formation II: Special cases arguments projected by the verbal base (e.g., amputee, erasee) and for the description of denominal -ee derivatives (e.g., biographee, asylee). A further characteristic of the referents of -ee derivatives is that they are involved rather passively in the qualifying events. At any rate, they have no control over the activities performed. This property is expressed by the third constraint defining the thetarole for -ee, namely “lack of volitional control”. Evidence for this constraint comes from the observation that numerous nouns ending in -ee fall within the domain of compulsion and obligation. Barker’s corpus comprises not only legal terms such as debtee, evictee, warrantee, but also nouns relating to personal violence (e.g., beatee, hittee, knockee), crime and police work (e.g., abductee, arrestee, blackmailee), prisons and punishment (e.g., escapee, offendee, releasee) or military and war (e.g., bombee, enlistee, invadee). As far as apparent agent nouns like escapee, enlistee or arrivee are concerned, Barker (1998: 718) argues that the referents act volitionally, but do not have control over the immediate consequences of the qualifying events. However, Barker does not consider this solution to be optimal because it relativizes the efficiency of one constraint, namely “lack of volitional control”. This problem is solved by Lieber (2004), whose account of -ee derivatives will be presented in the following section.

5. Ranking of constraints and the principle of co-indexation The semantic analysis of -ee derivatives proposed by Barker (1998) is compatible with Lieber’s (e.g., 2004, 2006) lexical semantics approach to word-formation, which is discussed in some detail in article 45 on rules, patterns and schemata in word-formation. In Lieber’s approach, lexical items are defined via sets of cross-categorial semantic features such as [+material] (man, chair, book), [−material] (peace, time, love), [+dynamic] (sing, write, teach), [−dynamic] (know, own; tall, intelligent), which instantiate the ontological categories SUBSTANCES/THINGS/ESSENCES and SITUATIONS. Another feature, namely [+IEPS] “inferable eventual position or state”, is introduced for verbs denoting a change of state (e.g., evaporate, forget, grow) or a change of position (e.g., descend, fall, go). Semantic feature constellations are also involved in the description of affixes. As illustrated in (5) and (6), both -er and -ee are lexically specified for [+material] because the derivatives they form denote concrete entities. In traditional terminology, these suffixes serve to form concrete nouns. Moreover, their entries display a non-binary feature [dynamic], which signals that the referents of nouns ending in -er and -ee are typically involved in some activity. This privative feature is absent in non-processual nouns, i.e. in nouns which do not imply activities (e.g., happi-ness, modern-ity, child-hood). The crucial difference between -er and -ee is that only the latter imposes semantic restrictions on its input, namely sentience and non-volitionality. In principle, these restrictions, which are integrated into the suffix’s semantic skeleton, correspond to Barker’s (1998) constraints “sentience” and “lack of volitional control”. (5)

-er Syntactic subcategorization: attaches to V, N Skeleton: [+material, dynamic ([ ], )]

52. Semantic restrictions on word-formation: the English suffix -ee (6)

-ee Syntactic subcategorization: attaches to V, N Skeleton: [+material, dynamic ([sentient, non-volitional], )]

The adequate interpretation of -er and -ee derivatives naturally follows from the principle of co-indexation (Lieber 2004: 61), which is defined in article 45. Applied to suffixation, this principle states that the highest argument of the non-head (i.e. of the base) is coindexed with the highest argument of the head-forming suffix provided that these arguments are semantically compatible. As demonstrated in article 45, this mechanism straightforwardly accounts for the well-formedness of a derivative like driver, whose suffix does not impose a semantic restriction on the non-head argument it is co-indexed with. The external argument of the dynamic verb drive is co-indexed with the referential argument of the suffix, and the derivative assumes an agentive interpretation. The referential argument occupies the placeholder in the semantic skeleton of -er. (7)

driver [+material, dynamic ([i ], [+dynamic] ([i ], [ ]))] -er drive

In the case of -ee (e.g., employee) the situation is different because this suffix requires the referent of the non-head argument it is co-indexed with to be sentient and nonvolitional. This requirement is specified in the position for the referential argument of -ee. As a consequence, co-indexation with the highest argument of the base (e.g., of employ) is blocked because the referent of this argument is generally sentient and acts volitionally. Instead, -ee is co-indexed with the internal argument, which meets the requirements specified by the suffix, e.g., (8)

employee [+material, dynamic ([sentient, non-volitional i ], [+dynamic ([ ], [i ])])] -ee employ

As far as the noun escapee is concerned, there is a mismatch between the requirement “non-volitional” of -ee and the base escape, whose first argument refers to a volitional being and whose second argument generally denotes a non-sentient entity such as an institution. Nevertheless, according to Lieber (2004: 65−66), the acceptability of escapee is accounted for if we assume that the requirement of non-volitionality is weaker than the requirement of sentience and therefore violable. Since the first argument of escape at least meets the requirement of sentience, it is this argument which is co-indexed with the referential argument of -ee. Mismatches of this kind, which were considered problematic by Barker (1998), are idiosyncratic and hence dealt with in the lexicon. (9)

escapee [+material, dynamic ([sentient, non-volitional i ], [+dynamic ([i ], [+Loc ([ ])])])] -ee escape

A semantic mismatch is also observable for the derivative amputee, which refers to an implied argument of the base verb amputate. The referent of the highest argument of

937

938

V. Rules and restrictions in word-formation II: Special cases this verb is a sentient being who acts volitionally (i.e. the surgeon). The second argument position is occupied by a noun denoting a limb (e.g., arm, leg). Thus, both positions are incompatible with the semantic requirements specified by -ee at first sight. However, nouns denoting body parts open semantic argument structures themselves. For example, since the referent of leg is in a part-whole relation with a human being, this noun has not only a referential argument , but also an argument position for the possessor of the leg. It is this argument which is eventually co-indexed with the referential argument of the suffix -ee in order to yield the appropriate reading for amputee. In Lieber’s feature-based model, denominal -ee derivatives of the type biographee are described analogously to deverbal ones. Significantly, biography is conceived of as a processual abstract noun, which is typically associated with the activity of writing. Furthermore, biography opens two argument positions. The first position is occupied by the referential argument . Since a biography is always written about someone, the second position is opened for the argument which would be realized as the complement of the preposition of in a syntactic phrase like a biography of Dr. Samuel Johnson. If -ee combines with the noun biography, the referential argument of the suffix, which is semantically specified for “sentience” and “non-volitionality”, is co-indexed not with the first, but with the second argument of biography, which fulfils at least the stronger requirement of “sentience”. (10) biographee [+material, dynamic ([sentient, non-volitional i ], [−material, dynamic ([ ], [i ])])] -ee biography Importantly, -ee derivatives involving constraint ranking (e.g., escapee), multi-layered semantic skeletons (e.g., amputee) or nominal bases (e.g., biographee) are not a result of productive word-formation but of paradigmatic pressure. Like Bauer’s (2001) examples followee, inquisitee and connee in (1)−(3), they were formed in order to denote concepts for which no words have existed so far. The semantically oriented works of Barker (1998) and Lieber (2004) provide an attractive and well-founded alternative to the morpho-syntactic view that derivational suffixes like -er or -ee are inextricably linked to specific argument positions of their bases. In both models, idiosyncrasies such as denominal -ee derivatives or deverbal -ee derivatives which do not correspond to any of the arguments of their bases are accounted for by general principles, namely episodic linking on the one hand and the principle of co-indexation on the other.

6. Prototypes – a recent trend in derivational word-formation The semantic approaches presented in the previous sections reveal that there is a tendency to accept the lexical diversity of productive suffixes like -er and -ee and to relativize the significance of the argument structure associated with the bases. The decreasing relevance of morpho-syntactic information is particularly obvious in approaches to wordformation which incorporate aspects of prototype theory as developed by Eleanor Rosch (e.g., Rosch 1977). The meaning of complex concepts involved in compounding has

52. Semantic restrictions on word-formation: the English suffix -ee been constructed on the basis of prototypes for quite a long time (e.g., Cohen and Murphy 1984; Smith et al. 1988; Štekauer 2005). In derivation, too, prototype theory is beginning to gain importance because it helps to account for the numerous deviations from regular word-formation patterns. Evidence comes from various works, which were developed independently of each other. Until recently, the application of prototypes to derivation was restricted to -er derivation (cf. Ryder 1991; Panther and Thornburg 2002; Baeskow 2010). A highlight in the analysis of -ee derivatives is the work of Susanne Mühleisen (2010; cf. also article 75 on patient nouns), whose formal synchronic description is based on the notion of “semantic prototype” as defined by Lyons (1995). According to Mühleisen, the prototypical -ee derivative of the type interviewee has the following properties: verbderived, with existing correlative -er noun, direct-object relation to the verb, sentient and probably human, role participant, non-volitional and non-active part in the event, can be used in a legal as well as more general contexts (2010: 57). Formations such as retiree, festschriftee or benefactee are less central examples, but nevertheless belong to the class of -ee derivatives. By tracing the development of -ee suffixation from Middle English legal terms to 20th century neologisms from the world wide web, Mühleisen convincingly shows that the synchronic heterogeneity of -ee derivatives is diachronically motivated. The anglicized French past participles, which belonged to the legal vocabulary (cf. section 2), are described as a result of a language contact situation “in which English gradually took over domains like the legal sphere from the erstwhile hegemonic French” (Mühleisen 2010: 13). Once the nativized French loanwords were established in the 14th/ 15th century, -ee began to assume the status of an English derivational suffix. At this early stage, -ee derivatives were formed analogously to agent nouns in -or with verbal bases. This kind of correlative noun formation smoothed the way to deverbal -ee derivation without mediating nouns in -or, which prevailed in the 17th century. During the process of nativization, -ee was subject to significant morpho-syntactic and semantic changes, which are responsible for the synchronic diversity of -ee derivatives and which may be summarized as follows: Most of the derivatives formed by the speakers of English in the 15th and 16th century realize the indirect object or the prepositional object of their base verbs, e.g., lessee, grantee, donee, vendee, vowee, debtee. In this respect, they differ from the Anglo-French participial nouns, which made reference to the direct object. Moreover, the effective use of the coinage absentee (< French sʼabsenter ‘to withdraw’) first attested in 1537 gave rise to a new but weak pattern: the agentive -ee derivative, which is further instantiated by later formations like submitt(i)e, refugee, escapee, standee, etc. In the 17th century, further borrowings from French induced a shift towards direct object formations (e.g., seducee, cheatee, challengee, transplantee, nominee). The meanings of these derivatives began to generalize, and the bases were no longer restricted to non-native verbs. In 1691, the sense of the established noun patentee ‘person to whom letters patent have been granted’ (1442), ‘the inventor, proprietor of something’ (1616) was extended to include the invention itself. This is the first instance of a non-human -ee derivative. After a century of lexical stagnation, a wealth of -ee derivatives was produced by the speakers of British and American English in the 19th century. Apart from semantically neutral derivatives (e.g., addressee, employee, consultee, pardonee), there are quite a few playful, humorous or ironic imitations of the established French borrowings. Examples

939

940

V. Rules and restrictions in word-formation II: Special cases of the creative use of -ee are kissee, drivee, kickee, gazee or shavee. From a morphosyntactic point of view, most of the 19th century coinages refer to the direct object of their verbal input. Semantically, many new words in -ee formed in the 20th century may be ascribed to the military jargon because they were coined during the two World Wars (e.g., enlistee, internee, bombee, selectee, returnee). The inventory of humorous -ee derivatives initiated in the 19th century was extended (rushee, crackupee, squeezee, quizzee, etc.), and a morpho-syntactic innovation is the formation of -ee nouns on the basis of compounds (e.g., handshakee, blind datee, moneylendee, return addressee). The fact that representative late-twentieth-century dictionaries of new words fail to provide a sufficient number of more recent neologisms in -ee caused Mühleisen to perform an efficient internet-based search of -ee derivatives of the late 1990s and early 2000s. The aim of this empirical study was to test the productivity of contemporary patterns of -ee formation on the basis of 1,000 potential -ee words which are listed neither in the Oxford English Dictionary nor in other reference books. 748 words turned out to be successful new -ee formations displaying different degrees of frequency. Mühleisen’s analysis, which is founded on a careful distinction between actual words, possible words, creativity and various types of productivity, reveals that the coinages provided by the world wide web largely adhere to the synchronically defined -ee prototype described above. Mühleisen (2010: 147−149) also provides reasons for the non-occurrence of certain -ee words. For example, bases ending in or are prevented from combining with -ee for orthographic and phonological reasons (*see-ee, *free-ee, *lie-ee, *eyeee). Semantically, some -ee derivatives are blocked because they would refer to either non-events (e.g., *uneducatee, *untamee) or negative events (e.g., *discouragee, *misguidee). Apart from presenting formal, historical and innovative empirical aspects of -ee derivation, Mühleisen also examines the distribution of -ee words in certain varieties of English and thus accounts for the fact that language contact, which once brought about the English derivational suffix -ee, still plays a role in word-formation processes involving this suffix.

7. A symbolic approach to deverbal -ee derivation A unified cognitive account of -ee derivatives is offered by Heyvaert (2006), who identifies a relationship between the different types of deverbal -ee derivatives discussed in the previous sections and the functions of the English past-participle morpheme -ed. It is a well-known fact that the past participle may occur in adjectival, passive or perfect constructions, as shown by the sentences in (11)−(13). In its adjectival use, it attaches either to intransitive verbs (11a) or to transitive verbs (11b). (11) a. Maturing and matured birds were on free range. b. Psychotherapy and analysis can sometimes help the more motivated patient who is curious to understand him- or herself. (12) They were captured on the 5th day. (13) She has painted all three of her children.

52. Semantic restrictions on word-formation: the English suffix -ee Following Langacker (1982, 1991), Heyvaert distinguishes four uses of the past participle. Adjectival PERF1, which attaches to intransitive verbs, profiles a stative relation in which the agentive participant finds himself/herself after he/she has undergone an (internal) change of stage, e.g., (14) Diana Tether is a newly enlisted Friend who makes good use of Kew gardens. PERF2, which is adjectival in nature as well, profiles a relation in which the non-agentive participant finds himself/herself through the intervention of another, agentive entity, e.g., (15) Tens of thousands of refugees, including Rwandans, and displaced persons are in fear of their safety. According to Langacker (1991: 203), PERF2 enhances the salience of a “terminal participant” or a participant “that lies downstream from another with respect to the flow of energy”. PERF3 is the passive variant of the past-participle morpheme -ed. Unlike PERF2, it profiles not only a final state, but all the stages of a process. This difference is reflected by the following sentences, which include PERF2 (16a) and PERF3 (16b): (16) a. The town was (already) destroyed (when we got there). b. The town was destroyed (house by house). PERF4, which is the perfect variant of the past-participle morpheme -ed, shares properties with PERF1 and PERF3. On the one hand, it behaves like PERF1 in that it does not imply that a participant is acted upon by another participant. On the other hand, it is like PERF3 in that it profiles all the stages of a process, including its termination, e.g., (17) Damon Hill already has been out testing the new Renault engine […]. According to Heyvaert (2006: 352), the crucial characteristic of PERF4 is that “it invokes a temporal reference point and indicates that the process designated by the content verb is prior to that reference point”. Heyvaert’s theory is based on the assumption that the participial constructions presented above constitute “agnates” of -ee nominalizations. Following Gleason (1965), she defines agnation as systematic or paradigmatic relations among large numbers of constructions. Significantly, all the functions identified for the past-participle are reflected by deverbal -ee derivatives. The most common -ee derivatives, i.e. those which realize the direct object of their verbal base in traditional terminology, correlate with clausal PERF3 − the passive use of the past participle. (18)

a. b. c. d. e.

employee detainee payee experimentee nominee

(s)he (s)he (s)he (s)he (s)he

is is is is is

employed detained payed experimented on nominated

941

942

V. Rules and restrictions in word-formation II: Special cases Because of their conceptual relation to PERF3, these nouns profile all the stages of the underlying processes in which the non-agentive participants are involved. In this respect, they differ from nouns like adoptee ‘an adopted child’ or electee ‘an elected member’, which behave like adjectival PERF2 in that they emphasize the final state in which the participants find themselves as the result of some change. However, some -ee derivatives are ambiguous between a PERF3 and a PERF2 reading. For example, the noun abusee may be interpreted as ‘a woman who is (being) abused’ (passive interpretation in the sense of PERF3) or ‘an abused woman’ (adjectival interpretation in the sense of PERF2, which focuses on the final state). In Heyvaert’s approach, agentive -ee derivatives do not constitute a problem because they also correlate with functions of the past-participle morpheme. Heyvaert distinguishes between two types of agentive -ee derivatives. The first type shifts the profile to a state in which the agentive participant finds himself/herself and thus correlates with PERF1, the adjectival variant of the past participle: (19)

a. b. c. d.

retiree escapee enlistee enrollee

a retired officer an escaped prisoner an enlisted soldier an enrolled student

The second type of agentive -ee derivatives correlates with PERF4, the perfect variant of the past-participle morpheme. These nouns, some of which are listed in (20), do not have an adjectival equivalent. (20)

a. b. c. d.

resignee returnee forgettee deferee

(s)he (s)he (s)he (s)he

has has has has

resigned returned forgotten deferred

To summarize, Heyvaert’s analysis of deverbal -ee derivatives does not rely on a matching relation between the meaning of the suffix and a particular syntactic argument position of the base. In this respect, it differs from classical generative approaches like Bauer (1993), who was the first to develop concrete patterns for -ee derivation, or Rappaport Hovav and Levin (1992), who concentrated on the suffix -er. Unlike the theories proposed by Barker (1998) and Lieber (2004), Heyvaert’s analysis does not make use of independent semantic constraints or principles either. Instead, it reveals a symbolic, i.e. a formal and semantic relation between the functions of the past-participle morpheme -ed and four types of deverbal -ee derivatives, which is diachronically motivated because -ee originated as an anglicized version of the French past participle -é(e). In cognitive terminology, non-agentive -ee derivatives profile a participant that is downstream with respect to the flow of energy (e.g., employee, experimentee) or with respect to the flow of time (e.g., adoptee, electee). On the other hand, agentive -ee derivatives are downstream with respect to the flow of time and profile the final state in which the referent finds himself/herself after some change (e.g., enlistee, returnee). Of course, the approaches presented in this article are not mutually exclusive, but rather reflect the complexity of English -ee nominalization and the interaction of morpho-syntactic, semantic and cognitive constraints involved in the restriction of the derivational processes under consideration.

52. Semantic restrictions on word-formation: the English suffix -ee

Acknowledgement I am much indebted to the Deutsche Forschungsgemeinschaft (DFG), whose support of my project on word-formation enabled me to write this article.

8. References Baeskow, Heike 2002 Abgeleitete Personenbezeichnungen im Deutschen und Englischen. Kontrastive Wortbildungsanalysen im Rahmen des Minimalistischen Programms und unter Berücksichtigung sprachhistorischer Aspekte. Berlin/New York: de Gruyter. Baeskow, Heike 2010 Derivation in generative grammar and neo-construction grammar: A critical evaluation and a new proposal. In: Susan Olsen (ed.), New Impulses in Word-Formation, 21−59. Hamburg: Buske. Barker, Chris 1998 Episodic -ee in English: A thematic role constraint on new word formation. Language 74(4): 695−727. Bauer, Laurie 1993 English Word-Formation. Cambridge: Cambridge University Press. Bauer, Laurie 2001 Morphological Productivity. Cambridge: Cambridge University Press. Burzio, Luigi 1986 Italian Syntax. A Government-Binding Approach. Dordrecht: Reidel. Carlson, Gregory Norman 1977 Reference to kinds in English. Ph.D. dissertation, University of Massachusetts, Amherst. Cohen, Benjamin and Gregory L. Murphy 1984 Models of concepts. Cognitive Science 8: 27−58. Dowty, David R. 1991 Thematic proto-roles and argument selection. Language 67: 547−619. Gleason, Henry A. 1965 Linguistics and English Grammar. New York: Holt, Rinehart and Winston. Heyvaert, Lisbeth 2006 A symbolic approach to deverbal -ee derivation. Cognitive Linguistics 17(3): 337−364. Jespersen, Otto 1905 Growth and Structure of the English Language. Leipzig: B.G. Teubner. Jespersen, Otto 1974 [1942] A Modern English Grammar on Historical Principles. Part VI: Morphology. London: Allen & Unwin. Koziol, Herbert 1972 [1937] Handbuch der englischen Wortbildungslehre. Heidelberg: Winter. Langacker, Ronald W. 1982 Space grammar, analysability, and the English passive. Language 58(1): 22−80. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar. Descriptive Application. Stanford: Stanford University Press. Lieber, Rochelle 1983 Argument linking and compounds in English. Linguistic Inquiry 14(2): 251−285. Lieber, Rochelle 2004 Morphology and Lexical Semantics. Cambridge: Cambridge University Press.

943

944

V. Rules and restrictions in word-formation II: Special cases Lieber, Rochelle 2006 The category of roots and the roots of categories: What we learn from selection in derivation. Morphology 16: 247−272. Lyons, John 1995 Linguistic Semantics. An Introduction. Cambridge: Cambridge University Press. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. Marle, Jaap van 1985 On the Paradigmatic Dimension of Morphological Creativity. Dordrecht: Foris. Mühleisen, Susanne 2010 Heterogeneity in Word-Formation Patterns. A Corpus-Based Analysis of Suffixation with -ee and its Productivity in English. Amsterdam/Philadelphia: Benjamins. Olsen, Susan 1986 Wortbildung im Deutschen. Stuttgart: Kröner. Panther, Klaus-Uwe and Linda Thornburg 2002 The roles of metaphor and metonymy in English -er nominals. In: René Dirven and Ralf Porings (eds.), Metaphor and Metonymy in Comparison and Contrast, 279−319. Berlin/ New York: Mouton de Gruyter. Plag, Ingo 2004 Syntactic category information and the semantics of derivational morphological rules. Folia Linguistica 38(3−4): 193−225. Randall, Janet 1988 Inheritance. In: Wendy Wilkins (ed.), Syntax and Semantics 21. Thematic Relations, 129−146. San Diego: Academic Press. Rappaport Hovav, Malka and Beth Levin 1992 -Er nominals: Implications for the theory of argument structure. In: Tim Stowell and Eric Wehrli (eds.), Syntax and Semantics 26. Syntax and the Lexicon, 127−153. San Diego: Academic Press. Rosch, Eleanor 1977 Human categorization. In: Neil Warren (ed.), Studies in Cross-Cultural Psychology. Vol. 1, 1−49. London/New York: Academic Press. Ryder, Mary Ellen 1991 Mixers, mufflers, and mousers: The extending of the -er suffix as a case of prototype reanalysis. In: Laurel A. Sutton, Christopher Johnson and Ruth Shields (eds.), Proceedings of the Seventeenth Annual Meeting of the Berkeley Linguistics Society, February 15−18, 1991: General session and parasession on the grammar of event structure, 299− 311. Berkeley: Berkeley Linguistics Society. Ryder, Mary Ellen 1999 Bankers and blue-chippers: An account of -er formation in present-day English. English Language and Linguistics 3(2): 269−297. Selkirk, Elizabeth 1982 The Syntax of Words. Cambridge, MA: MIT Press. Smith, Edward E., Daniel N. Osherson, Lance J. Rips and Margaret Keane 1988 Combining prototypes: A selective modification model. Cognitive Science 12: 485−527. Štekauer, Pavol 2005 Meaning Predictability in Word Formation. Novel, Context-Free Naming Units. Amsterdam/Philadelphia: Benjamins. Williams, Edwin 1981 Argument structure and morphology. Linguistic Review 1: 81−114.

Heike Baeskow, Wuppertal (Germany)

53. Dissimilatory phenomena in French word-formation

53. Dissimilatory phenomena in French word-formation 1. 2. 3. 4. 5. 6.

Introduction Phonological reactions Morphological reactions Dissimilatory phenomena in javanais Conclusion References

Abstract The present article provides an overview of the ways in which the universal tendency to avoid the repetition of similar consonants manifests itself in French word-formation. Phonological reactions include consonant dissimilation, truncation, haplology and epenthesis. Morphological reactions may consist in the choice of another suffix or stem, or in the insertion of an interfix. Dissimilation is also shown to play a major role in word games such as “javanais”.

1. Introduction The main finding of Grammont’s doctoral thesis on consonant dissimilation in IndoEuropean and Romance languages (1895) was the awareness that all observed cases of dissimilation can be traced back to “a unique and very general principle: If in a word (or a closely knit word group) one and the same articulatory gesture has to be repeated twice, one of these gestures must be omitted” (Meillet 1904: 461; our translation). By conferring on this principle of dissimilation the status of a law − or a series of laws − governing exclusively phonetic changes, Grammont has restricted its scope and oriented the debate towards a discussion of the formulation of these laws (cf. Posner 1961) and their validity for historical phonetics (cf. Togeby 1964). It is obvious, however, that in its general form such a principle may have other phonological and morphological effects, both in synchrony and in diachrony, than that of changing the words of the lexicon (cf. Maiden 1997). As a modern descendant of this principle, the obligatory contour principle (Goldsmith 1976), conceived of originally as a constraint on underlying representations, has turned out to be useful for blocking or, on the contrary, triggering certain phonological processes (McCarthy 1986; Yip 1988). The present contribution follows this modern line of research by establishing the range of effects for which the principle of dissimilation can be held responsible in the formation of new words in Modern French. Due to space limitations, we can only illustrate each of these effects with a short analysis of some cases of consonantal dissimilation. The reader can find more detailed accounts and more numerous references in Roché et al. (2011: chapter 4). The study of dissimilatory phenomena is impeded by the small number of relevant data in dictionaries. The main effect of the principle of dissimilation is to dissuade speakers from

945

946

V. Rules and restrictions in word-formation II: Special cases coining lexemes which would be rendered awkward or in need of repair by a cacophonic repetition. In the dictionaries, for example, the prefix re- or ré- expressing repetition does not precede a verb that begins with an r. This restraint does not work for all speakers alike, however, which is why many find themselves in a situation where they have to avoid the immediate repetition of two similar consonants. Systematic investigations on the Internet allow one to find a number of new forms which in one way or another avoid an expected repetition. These, mostly occasional, neologisms have little chance of ever entering the dictionary. However, due to their large numbers and homogeneity, they constitute a solid starting point for a synchronic study of dissimilatory phenomena. We will classify the numerous and varied effects of the principle of dissimilation in French according to the stage at which they interfere with word-formation. Certain phonological processes correct repetitions which arise from the concatenation of elements in morphology. These phonological changes, however, are not the only possibility: An unexpected morphological combination of elements may also preempt the appearance of cacophonic repetitions. Finally, the principle of dissimilation can be seen at work even in secret languages after the phonology has given a word its canonical shape.

2. Phonological reactions A priori, there are several ways of avoiding a sequence containing two similar consonants in neighbouring onsets or in the onset and rhyme of one and the same syllable. One can act on the presence or the identity of the neighbouring consonants by altering or deleting one of them; alternatively, one can take advantage of the distance which separates them and avoid a deletion that would bring them closer together by inserting, for example, an epenthetic consonant between them.

2.1. Progressive and regressive dissimilations Dissimilation proper is rarely found in complex words. Grammont (1895: 16) has even claimed that “no dissimilation occurs when the etymology of the different parts of the word is transparent for the speakers”. Nevertheless, we have found two good cases of dissimilation, one concerning progressive dissimilation, the other regressive dissimilation. Both manifest themselves in the suffixation of -esque ‘-esk’, which derives adjectives from nouns (e.g., Courteline (name of a playwright) → courtelinesque, funambule ‘tightrope walker’ → funambulesque). (1)

a. Bakayoko /bakajoko/ (name of a football player) → bakayokeste, cirque /sirk/ ‘circus’ → cirqueste, blague /blag/ ‘joke’ → blagueste, blog /blɔg/ ‘id.’ → blogueste, bling-bling /bliŋbliŋ/ ‘ostentatious’ → blingeste, jogging /dʒɔgiŋ/ ‘id.’ → joggineste b. Bakayoko /bakajoko/ → bakayotesque, Facebook /fεsbuk/ → facebootesque, gloubi-boulga /glubibulga/ (imaginary food) → gloubibouldesque, novlangue /nɔvlɑ˜g/ (imaginary language) → novlandesque, bling-bling /bliŋbliŋ/ → bling-blinesque, jogging /dʒɔgiŋ/ → jogginesque

53. Dissimilatory phenomena in French word-formation If the final consonant of the base is a velar stop (/k/, /g/ or /ŋ/), the ending -esque (/-εsk/) is sometimes replaced by -este (/-εst/), cf. (1a), or the velar stop of the base may be replaced by the corresponding dental stop (/t/, /d/, /n/), cf. (1b). In both cases the feature [+velar] of one of the velar stops disappears, and the consonant concerned takes on the unmarked place of articulation, which is the normal case in dissimilation. Maybe this is why dental suffixes, which are the most numerous, seem to shun dissimilation. Note that the large majority of cases of regressive dissimilation affect polysyllabic stems, while half of the derivatives in -este are formed from monosyllabic stems which an alteration would render unrecoverable.

2.2. Deletions and haplologies The deletion of one of two similar consonants constitutes a radical remedy against repetition, but its consequences with respect to the syllabification and intelligibility of the derivative limit its application. Let us use again for illustration suffixation by -esque. This suffix does not tolerate the presence of a sibilant at the end of the stem any better than that of a velar stop. (2)

a. Rolls /rɔls/ (trade mark) → rollsesque, pin’s /pins/ ‘id.’ → pinsesque, lynx /lɛ˜ks/ ‘id.’ → lynxesque, ronce /ro˜s/ ‘blackberry bush’ → roncesque b. Beatles /bitœls/ → beatelesque, Dickens /dikεns/ → dickeneste, Bill Gates /bilgεts/ → bill-gatesque, Brassens /brasɛ˜s/ → brassinesque, Camoëns /kamɔɛ˜s/ → camoïnesque c. bidasse /bidas/ ‘soldier’ → bidassesque, saucisse /sosis/ ‘sausage’ → saucissesque, Pangloss /pɑ˜glɔs/ (figure in Voltaire) → panglossesque, Nimbus /nɛ˜bys/ (comic figure) → nimbusesque d. Juliénas /ʒyljenas/ (wine from Burgundy) → juliénesque, Toutatis /tutatis/ (god of the Gauls) → toutatesque, tétanos /tetanos/ ‘tetanus’ → tétanesque, Cosinus /kɔsinys/ (comic figure) → cosinesque

When the final sibilant of the base is preceded by another consonant or a nasal vowel that can give rise to a nasal consonant, its deletion does not affect syllabification. However, this deletion (sometimes) only takes place if the base contains at least two syllables (cf. 2b); the very rare monosyllabic bases of this type always retain their sibilant (cf. 2a). When the sibilant is preceded by an oral vowel, its deletion would provoke the appearance of a hiatus if it were not accompanied by the deletion of the vowel. But generally this double deletion only takes place if the base has at least three syllables (cf. 2d); if it has only two, the last rhyme is normally maintained intact (cf. 2c). This kind of behaviour is not infrequent. The deletion of the final rhyme of a polysyllablic string for reasons of dissimilation affects consonants other than sibilants (for example stops, cf. Goldorak → goldorakeste ~ goldoresque, cosmonaute ‘cosmonaut’ → cosmonette ‘female cosmonaute’), and may be triggered by suffixes other than -esque (for example suffixation of -issime ‘very’, cf. rigoureux ‘rigorous’ → rigourissime, or suffixation of -iste ‘-ist’, cf. rhinocéros ‘id.’ → rhinocériste). The two-syllable threshold below which the dissimilatory constraint does not apply is normally rigorously respected (cf. Chirac → chiraquesque, heureux ‘happy’ → heureusissime).

947

948

V. Rules and restrictions in word-formation II: Special cases There is nevertheless one condition under which the two-syllable threshold often seems to be ignored, viz. when the final rhyme of the base is identical with the suffix or its initial part. (3)

(Louis de) Funès /fynεs/ → funesque, DS /deεs/ (game console) → déesque, Lewis /lewis/ → lewissime, propice /prɔpis/ ‘favourable’ → propissime

It is tempting to analyze these forms as cases of haplology rather than as cases of deletion by postulating, for example, that in propissime (cf. section 3) the sequence /is/ represents at the same time the end of the stem /prɔpis/ and the beginning of the suffix /isim/. In this case, the derivative would faithfully respect its base and the two-syllable threshold, only infringing the constraint of a one-to-one correspondence which requires that each phoneme of the derivative corresponds to one of the phonemes of its constituents.

2.3. Non-truncation The strength of the principle of dissimilation is not only sensitive to the number of similar phonemes concerned, but also to the distance which separates them: The more distant the two phonemes are from one another, the better their presence is tolerated. We have conducted two experiments showing that the difficulty of deleting the first of two vowels in a hiatus is dependent on the degree of similarity between the two consonants surrounding it. In the first study (Plénat 1996: 590 ff.), we asked 12 speakers to choose between the derivative in -esque with a truncated stem and the derivative in -esque with a nontruncated stem corresponding to di- or trisyllabic bases ending in /u/ or /o/. The subjects had to choose, for example, between gourou-esque and gour-esque (← gourou ‘guru’), or between kangourou-esque and kangour-esque (← kangourou ‘kangaroo’). The 18 consonants constituting possible onsets in French were represented twice each in front of the two vowels both in the disyllables and in the trisyllables. In the second experiment (Roché et al. 2011: chapter 4), we collected from the Internet all the derivatives we could find in -esque derived from a base ending in a velar consonant + a or a sibilant + a with two and three syllables (907 items containing a velar consonant + a and 161 items with a base ending in a sibilant + a), and for each of the other consonants the derivatives from a di- and trisyllabic base were chosen at random (705 derivatives in all), in order to determine for each category the percentage of cases of truncation, hiatus and epenthesis. Neither of the two experiments was entirely satisfactory. In the first one, the speakers were placed in a rather unnatural situation; in the second one, the origin of the derivatives was too diverse (there were numerous loan words). However, in spite of these shortcomings, these experiments show clearly not only that the presence of a sibilant or of a velar stop hinders the deletion of the final vowel of the base (for example, while yakuzaesque is attested, one only finds loyolesque), but also that this deletion is sensitive to the degree of resemblance between the final consonant of the base and the consonants of the suffix: Sonorants tolerate deletion better than obstruents, and among these, voiced obstruents tolerate deletion better than unvoiced ones.

53. Dissimilatory phenomena in French word-formation

949

2.4. The choice of epenthetic consonants If, for whatever reason, the final vowel of a word is retained before a vowel-initial suffix, a hiatus is very often avoided by the epenthesis of a consonant. It seems likely that the choice of this consonant is determined by dissimilatory constraints, but lexical constraints sometimes work against this tendency. Consider the distribution of the two most frequent epenthetic consonants of French, /t/ and /z/ (as shown in Table 53.1), when they are inserted between the disyllabic bases gaga ‘senile’, bébé ‘baby’, neuneu ‘stupid’, bobo ‘bourgeois’ and the suffixes -ité ‘-ity’, -isme ‘-ism’, -esque and -itude (Google search, December 18, 2010): Tab. 53.1: Epenthetic consonants before -ité, -isme, -esque and -itude -s-ité

-t-ité

gaga

0

0

0

2690

0

bébé

1

0

0

25

46

0

1

7

0

0

neuneu bobo

-s-isme

-t-isme

-s-esque

-t-esque

-s-itude

-t-itude

353

9

960

0

50

0

40

307

0

57

4

1680

215

0

32

8

554

As one can see, the distribution of /t/ and /z/ is not random. /t/ almost never appears in front of -ité, nor does /z/ occur in front of -isme or -esque: One normally says neuneusité, but neuneutisme and neuneutesque. This distribution can only be attributed to the dissimilatory constraints prohibiting the repetition of two onsets or of an onset and a similar coda. However, as one can also see, -itude is regularly preceded by /t/: One says neuneutitude rather than neuneusitude. This anomaly is undoubtedly due to the frequency of the ending -titude in the mental lexicon. The majority of lexicalized nouns in -itude (aptitude, beatitude, certitude, exactitude, promptitude, etc.), which are often borrowings from Latin, have a stem in /-t/. This explains the “gang effect” imposing the choice of /t/ as an epenthetic consonant to prevent a hiatus. Such gang effects sometimes appear locally even with suffixes normally sensitive to the dissimilatory constraints. For example, the epenthetic /s/ of bétacisme, deltacisme or zétacisme is borrowed from iotacisme and la(m)dacisme, which in turn were borrowed from Greek through Latin. The French pronunciation of these words has introduced a repetition of /s/ into the final sequence /sism/, which has been extended to the other members of the small series in spite of the dissimilatory constraints (cf. Roché 2007: 50).

3. Morphological reactions Languages often prefer prevention to cure. In such a case, a sequence of two similar articulations may be avoided by choosing an unexpected suffix or stem. Alternatively, a meaningless segmental chain may be inserted between the base and suffix.

950

V. Rules and restrictions in word-formation II: Special cases

3.1. The choice of another suffix The choice of a suffix other than the one required by the type of base and the meaning of the derivative is not infrequent (cf. already Roché 1997). It can be found with verbs, nouns and adjectives. (4)

a. tunisien ‘Tunisian’ → tunisifier, rather than tunisianiser /tynizjanize/ vs. algérien ‘Algerian’ → algérianiser, marocain ‘Maroccan’ → marocaniser, and mauritanien ‘Mauritanian’ → mauritaniser b. champignon ‘mushroom’ → champignonniste, and not champignonnier /ʃɑ˜piɲɔnje/ vs. betterave ‘sugar beet’ → betteravier, céréales ‘cereals’ → céréalier, houblon ‘hop’ → houblonnier c. Molière → moliéresque, rather than moliérien /mɔljerjɛ˜/ vs. Corneille → cornélien, Pascal → pascalien, Racine → racinien

French has a series of verbs in -iser ‘-ize’ derived from the Latinate variant of relational adjectives corresponding to names of countries and conveying the meaning that some institutions are given a national character which lacked it before (cf. 4a). In the case of Tunisia, the two sequences /niz/ of tunisianiser, though distant, are subjected to the principle of dissimilation. Choosing the noun instead of the adjective (tunisiser) would not make things better. Therefore, the only other verb-forming suffix, -ifier ‘-ify’ is resorted to. French also possesses a series of nouns in -ier derived from plant names and yielding nouns referring to people who grow them (cf. 4b). But when mushroom growers needed to be designated, champignonniste was preferred over champignonnier, which is awkward because of the sequence of a palatal nasal and an onset formed by a nasal and a palatal. Since it is used for productively deriving agent nouns from nouns, -iste was the most suitable suffix for replacing -ier in its function as an agentive suffix. The most common way of deriving a relational adjective from a proper name in French consists in adding -ien to its Latinate stem (cf. 4c). However, if the base noun contains a palatal in its final onset, the risk of having to pronounce a palatal immediately following another is reason enough for speakers to prefer another suffix. That is why one says le théâtre cornelien ‘Corneille’s plays’, le théâtre racinien ‘Racine’s plays’, but le théâtre moliéresque ‘Molière’s plays’. Due to its frequent use in Italian loanwords such as caravagesque, dantesque or pétrarquesque, -esque is a good choice of replacement for -ien. The replacement of one suffix by another only constitutes a remedy for the risk of repetition of similar articulations if the language has recourse to a suffix similar in function to the one that causes the difficulties. The suffix -iste, for example, which serves to designate human beings, cannot replace -ier if this suffix is used for deriving a place noun (cf. une champignonnière ‘mushroom bed’) or a noun designating a tree (if mushrooms grew on trees, these would be called champignonniers, cf. brugnons ‘white nectarines’ → brugnoniers). At the same time, French, the “language of Molière”, which is sometimes jokingly called moliérien, could not be dubbed moliéresque. That is why one and the same suffix can sometimes be replaced by another suffix or can itself sometimes replace another suffix. So -ien, which gives way to -esque after a palatal in relational adjectives derived from names of persons, itself replaces -ais and -ois after a sibilant in

53. Dissimilatory phenomena in French word-formation inhabitant names (cf. Parisien, Calaisien vs. Marseillais, Lillois) and -iste in agent nouns derived from a stem ending in a sibilant (cf. physicien ‘physicist’ vs. chimiste ‘chemist’, see Lignon and Plénat 2009).

3.2. The choice of another stem The choice of a stem different from the expected one can take on two forms. Sometimes a sequence of two similar articulations is avoided by violating the restrictions on the affix that selects the stem on which the derivative is based; on other occasions, a substitute stem is borrowed from the derivational paradigm of the base lexeme. Lexemes − written here in small caps − are represented in the lexicon, not by a single form, but by a set of stem-forms each allotted to one or more slots in their inflectional or derivational paradigm. Adjectives, for example, have a Latinate stem-form which by default takes over the stem-form used for the feminine gender, but which can also be deduced from the default stem-form or can be unpredictable. Since the suffix -ité requires such a stem-form, the adjective RIGOUREUX ‘rigorous’ will yield, depending on the speaker, rigorosité /rigɔroz-ite/ (borrowed from Latin), rigourosité /riguroz-ite/ (backing of the last vowel), or rigoureusité /rigurøz-ite/ (default form). The superlative suffix -issime is one of those suffixes which require a Latinate stem (cf. RIGOUREUX → rigorissime, according to certain speakers). Let us consider now the form taken on by the superlatives of adjectives in -ique /-ik/, whose specific Latinate stem-form is assibilated, yielding /-is/ (e.g., ÉROTIQUE ‘erotic’ → érotic-ité /erɔtis-ite/): (5)

a.

DRAMATIQUE ‘dramatic’ → dramatissime ~ dramatiquissime, ÉROTIQUE → érotissime ~ érotiquissime, PSYCHÉDÉLIQUE ‘psychedelic’ → psychédélissime ~ psychédéliquissime b. COMIQUE ‘comical’ → comiquissime, LUBRIQUE ‘sensual’ → lubriquissime, PRATIQUE ‘practical’ → pratiquissime c. MAGIQUE ‘magical’ → magissime ~ magiquissime, MERDIQUE ‘lousy’ → merdissime ~ merdiquissime, MYTHIQUE ‘mythical’ → mythissime ~ mythiquissime

When the base contains three or more syllables (5a), the superlative, as expected, selects either the default stem-form or a stem derived by truncation or haplology from the stem-form ending in a sibilant (ÉROTIQUE cannot yield erotic-issime /erɔtisisim/, only érot-issime /erɔtisim/). When, on the contrary, the base has only two syllables − and is not transparent − (5b), the stem selects the default form in /-ik/ (COMIQUE → comiquissime /kɔmikisim/). Neither the stem ending in a sibilant nor its shortened variant are attested (neither comicissime /kɔmisisim/, nor comissime /kɔmisim/). Assibilation − and, as a consequence, haplology − are blocked due to the existence of the default form. When, on the contrary, the disyllabic base is transparent, the superlative can again take either a short form or the default form (cf. MAGIE ‘magic’ → MAGIQUE ‘magical’ → magissime /maʒisim/ ~ magiquissime /maʒikisim/). Assibilation and shortening are not sufficient explanations in this case for the short form. This form is somehow licensed

951

952

V. Rules and restrictions in word-formation II: Special cases by the existence in the family of the base lexeme (MAGIQUE) of a lexeme with a different lexical category but the same meaning as the short form (MAGIE /maʒi/). In the case of magissime, it is the noun which provides the stem of the derivative of the corresponding relational adjective. In the same way MAURITANIEN yields mauritaniser (← Mauritanie ‘Mauritania’) instead of mauritanianiser (cf. 4a). But the inverse case also exists: (6)

a. CLAUSEWITZ → clausewitzianisme, KEYNES → keynésianisme, LEIBNIZ → leibnizianisme, MALTHUS → malthusianisme b. BERGSON → bergsonisme, CALVIN → calvinisme, ERASME → érasmisme, MONTAIGNE → montaignisme

Roché (2007) has shown that the names of philosophical and religious systems − but not political and artistic ideologies − derived from the name of a philosopher take the stem of the corresponding relational adjective if this name ends in a sibilant (6a): malthusianisme /maltyzjanism/ avoids the repetition of sibilants that would arise in malthusisme /maltyzism/. Inversely, the names of philosophers ending in a nasal phoneme directly take the suffix -isme (6b): bergsonisme /bεrgsɔnism/ avoids the repetition of nasals that would arise in bergsonianisme /bεrgsɔnjanism/. In these cases, the risk of having two sibilants is not the only determining factor (cf. marxisme, debussysme, etc.), the selection of the relational adjective also obeys a lexical constraint favouring -ianisme at the expense of -isme in the philosophical and religious sphere (cf. luthérianisme, zwinglianisme, hégélianisme, etc.). But this lexical logic must yield to the risk of an accumulation of nasals.

3.3. Shifted suffixation A last remedy against the repetition of similar consonants consists in inserting a segmental string between base and suffix which takes on the form of a suffix but is devoid of meaning, a process referred to as “shifted suffixation” (suffixation décalée, in French). We will limit ourselves here to some short remarks about the diminutive suffix -ette (cf. Plénat 2005). Shifted suffixation is treated more in detail in article 31 on interfixes in Romance. The suffix -ette is shifted to the right if the base is short, generally monosyllabic. But in addition to size constraints, which tend to impose a disyllabic stem, dissimilatory constraints also come into play. In fact, 1) the frequency of the phenomenon depends on the degree of similarity between the final consonant of the base and the consonant of the suffix. Shifting is rare after sonorants, but dominates after bases ending in /d/ and occurs regularly after bases in /t/ (e.g., goutte ‘drop’ → gout-el-ette, etc.); 2) the consonant of the inserted segmental string is as different as possible from the consonants separated by this chain. In the case of the derivation in -ette, it is always a sonorant (cf. boîte ‘box’ → boît-el-ette, boît-in-ette, boît-oun-ette); 3) the only disyllabic bases which admit shifting are those whose final consonant is identical or almost identical to the consonant of the suffix. Hence we find derivations such as patate ‘potato’ → patat-inette, patat-oun-ette, salade ‘salad’ → salad-in-ette, salad-oun-ette. Shifted suffixation ensures more distance between similar articulations.

53. Dissimilatory phenomena in French word-formation

4. Dissimilatory phenomena in javanais “Javanais” refers to language games which are often described as relying on infixation, but should rather be conceived of, as we will see, as associating the sounds of the source word with partly prespecified schemata. If the prespecified schema contains one or several phonemes identical to phonemes of the source word, dissimilatory phenomena can often be observed. These language games are of particular interest to us since they do not operate over lexemes, but over inflected words. This shows that the principle of dissimilation is also operative after morphology and phonology have provided inflected words with their canonical form. We will study here a javanais in -av- for which we have recordings, and a javanais in -guede- which we have found on the web.

4.1. Haplologies and reduplications in the javanais in -avFrench has a javanais associating the sounds of the source word with a series of disyllabic schemata of the form /C(C)avV(C(C))/. (7)

a. Frédo /fredo/ (hypocoristic) → /frave davo/, maton /matõ/ ‘warder’ → /mava tavõ/ b. garde-à-vous /gardavu/ ‘attention’ (in the military sense of the word) → /gavar davu/, désespoir /dezεspwar/ ‘despair’ → /davεs pavwar/, relevé /rœlœve/ ‘raised’ → /ravœ lave/, à un /aɛ˜/ ‘to one’ → /avɛ˜/ c. gradé /grade/ ‘NCO’ → /grave dave/, locataire /lɔkatεr/ ‘tenant’ → /lavɔ kavo tavε ravœ/, Follenfant /fɔlɑ˜fɑ˜/ (family name) → /favɔ lavo favɑ˜/, souvenance /suvœnɑ˜s/ ‘memory’ → /savu navœ navɑ˜ savœ/

If this javanais relied on infixation, the coded word should always have twice as many syllables as the original word and, apart from the occurrences of the coding sequence -av-, contain the same number of phonemes as the source word, as in Frédo → Fr-av-é d-av-o (cf. 7a). However, this is not always the case. Sometimes (cf. 7b) the coded word has fewer disyllabic schemata than the source word has syllables. This happens when the latter contains the sequence /av/ (cf. garde-à-vous), or when it contains two similar consecutive VC sequences (cf. désespoir), two identical consecutive vowels and a /v/ (cf. relevé), or two identical consecutive onsets and an /a/ (cf. à un, with two empty onsets). In other cases (cf. 7c), one of the phonemes of the source word appears twice in the javanais. In most cases, we find a vowel preceding or following an /a/ (cf. gradé and locataire), or followed by two identical vowels (cf. Follenfant); but some examples have also been found where a consonant is preceded by a /v/ which is repeated (cf. souvenance). It is clear (cf. Plénat 1991) that all these apparent deviations allow one and the same general description: pairs of identical phonemes figuring both in the source word or one in the source word and the other in the coding sequence /av/ are represented by a single phoneme in the javanais. When a pair of vowels and a pair of consonants are merged in this way into one unit, the javanais can dispense with one disyllabic schema. In /gavar davu/, for example, the disyllable /davu/ simultaneously codes the syllable /da/ and the syllable /vu/ of garde-à-vous, at the same time respecting the schema /CavV/. When, on

953

954

V. Rules and restrictions in word-formation II: Special cases the contrary, only a pair of vowels or a pair of consonants are merged into a unit, this reduction necessarily provokes the reduplication of a consonant or of a vowel in the javanais. So in /grave dave/ the two vowels of gradé, /a/ and /e/, both figure in the first disyllable /grave/, but since the /d/ is not represented, coding must have recourse to a second disyllable which takes up the /e/. These haplologies are often subject to multiple conditionings. Metrical constraints can play a role in explaining cases where haplology allows the reduction of the number of syllables. If, for example, ô désespoir! becomes avô davespavoir!, in our materials this serves to yield the second hemistich of an alexandrine. But this certainly is not the kind of constraint which explains the cases where the simplification is compensated by reduplication. It is reasonable to assume that the main raison d’être of these haplologies is to diminish the number of repetitions or to distribute them in a more harmonious manner. If all speakers of a javanais produce /gavar davu/ instead of /gavar dava vavu/, this is probably due to the fact that they avoid in this way having to pronounce a sequence of three /a/ and three /v/ one after the other. This certainly constitutes another manifestation of the principle of dissimilation.

4.2. Metathesis in the javanais in -degue There is a second type of javanais whose coding sequence is based on a consonantal skeleton containing a /d/ and a /g/. These two consonants can appear in the order /d ... g/ or in the order /g ... d/ and be separated or not by a vowel. All the vowels of one and the same trisyllable or of one and the same disyllable are borrowed from the coded syllable. The syllable /la/ can therefore be coded as /ladaga/, /lagada/, /ladga/ or (probably) /lagda/, but when coding an utterance the speaker normally will stick to one of these variants. (8)

a. jeguede trouguoudou veguede quedegue tugudu èguèdè condonguon (corresponding to: je trouve que tu es con ‘I think that you are stupid’) b. sadaga vadaga tudugu [faidaigais] kiodiogio [sic] degede bodogo (corresponding to: Ça va ? Tu fais quoi de beau ? ‘How are you? What are you doing?’) c. tugdu peudgeu endgan chandgan geandgan lesdges autgo togolangan (corresponding to: Tu peux en changeant les autocollants ‘You can by changing the stickers’)

However, if the coded syllable starts with a velar stop, one often observes that the order of the sequence /g ... d/ is reversed (cf. que → quedegue instead of queguede, con → condongon instead of congondon in (8a), and, in a parallel fashion, that the order of the sequence /d ... g/ is reversed if the onset is a dental stop (cf. de → degede instead of dedegue in (8b) and tu → tugdu instead of tudgu in (8c). Not all speakers resort to this kind of metathesis, and those who practice it do not do so consistently (cf. tu → tudugu in (8b). But, as far as I know, one never observes inversions in contexts other than those described. It is furthermore noteworthy that other anomalies appear if the normal coding would lead to a repetition of two similar onsets. So, for example, the disyllable togo in (8c) codes at the same time the syllables to and co of autocollants (which ought to be

53. Dissimilatory phenomena in French word-formation coded as audgo todgo codgo landgants): A sort of haplology conflates the expected syllables with a velar onset go and co, while the d of the coding sequence, which would partially repeat the initial t, is omitted. Both the distribution of the metatheses and their co-occurrence with anomalies analogous to those which we find in the javanais in -avspeak in favour of the idea that they are also a consequence of the principle of dissimilation. They do not suppress the repetitions of velars and dentals, but by widening the distance which separates the occurrences of these phonemes they render these repetitions more tolerable. We have not found dissimilatory metatheses outside the javanais in -degue-. This is probably due to the fact that, since the infix is devoid of semantic content, it does not matter whether the order of its elements is reversed or not.

5. Conclusion In French it is frequently the case that, when an articulatory gesture in a derivative has to be repeated immediately, these two gestures are kept at a distance from one another or merged into a single articulation. The facts can be described in process terms: On the one hand, we observe progressive and regressive dissimilations, cancellations and haplologies, unexpected choices of suffixes and stems; on the other, the absence of truncations, epentheses, shifting of suffixes, and metatheses. But this conspiracy speaks in favour of the existence of the general principle already vaguely alluded to by Grammont. The strength of this principle of dissimilation, which seems to affect all consonants, varies according to the degree of similarity between the phonemes involved, the distance which separates them in the word, and their number. Frequently, one and the same derivative may take on several forms. Although we could not go into more detail here, it nevertheless turns out that the distribution of these forms is not due to chance: Short stems are more resistant than long ones and favour the widening of the distance between the similar articulations over the suppression of one of them; it is also only in case the suffix is somehow unique that the stem-building rules are infringed upon; and it is no coincidence that metathesis only affects elements devoid of meaning. These facts taken together speak against a serial approach of morphology and phonology. The principle of dissimilation not only affects the contextual adaptation of the elements concatenated in word-formation, but also, beforehand, the very choice of these elements, as well as, further down, the deformation of words in word games. In order to be successful, the principle of dissimilation not only has to outdo other phonological principles, but also lexical regularities and the mechanics of javanais; in this task, it finds allies in other principles and rules. The complexity of these interactions speaks in favour of the idea that phonology and morphology proceed in parallel.

6. References Goldsmith, John 1976 Autosegmental phonology. Ph.D. dissertation, MIT. Grammont, Maurice 1895 La Dissimilation consonantique dans les langues indoeuropéennes et dans les langues romanes. Dijon: Darantière.

955

956

V. Rules and restrictions in word-formation II: Special cases Lignon, Stéphanie and Marc Plénat 2009 Échangisme suffixal et contraintes phonologiques (Cas des dérivés en -ien et en -icien). In: Bernard Fradin, Françoise Kerleroux and Marc Plénat (eds.), Aperçus de morphologie du français, 65−81. Saint-Denis: Presses Universitaires de Vincennes. Maiden, Martin 1997 La dissimilation à la lumière des pronoms clitiques en roman. Zeitschrift für romanische Philologie 113(4): 531−562. McCarthy, John J. 1986 OCP effects: Gemination and antigemination. Linguistic Inquiry 17(2): 207−263. Meillet, Antoine 1904 Note sur quelques recherches de linguistique. L’année psychologique 11: 457−467. Plénat, Marc 1991 Le javanais: Concurrence et haplologie. Langages 25: 95−117. Plénat Marc 1996 De l’interaction des contraintes: Une étude de cas. In: Jacques Durand and Bernard Laks (eds.), Current Trends in Phonology. Models and Methods, 585−615. Salford: ESRI, University of Salford. Plénat, Marc 2005 Rosinette, cousinette, starlinette, chipinette: Décalage, infixation et épenthèse devant -ette. In: Injoo Choï-Jonin, Myriam Bras, Anne Dagnac and Magali Rouquier (eds.), Questions de classification en linguistique. Méthodes et descriptions. Mélanges offerts au Professeur Christian Molinier, 275−298. Bern: Lang. Posner, Rebecca 1961 Consonantal Dissimilation in the Romance Languages. Oxford: Blackwell. Roché, Michel 1997 Briard, bougeoir et camionneur: Dérivés aberrants, dérivés possibles. In: Danielle Corbin, Bernard Fradin, Benoît Habert, Françoise Kerleroux and Marc Plénat (eds.), Mots possibles et mots existants. Actes du colloque de Villeneuve d’Ascq (Forum de morphologie, 1res rencontres, 28−29 avril 1997) [= Silexicales 1], 241−250. Lille: SILEX, Université Lille 3. Roché, Michel 2007 Logique lexicale et morphologie: La dérivation en -isme. In: Fabio Montermini, Gilles Boyé and Nabil Hathout (eds.), Selected Proceedings of the 5th Décembrettes. Morphology in Toulouse, 45−58. Somerville, MA: Cascadilla Press. Roché, Michel, Gilles Boyé, Nabil Hathout, Stéphanie Lignon and Marc Plénat 2011 Des unités morphologiques au lexique. Paris: Hermès-Lavoisier. Togeby, Knud 1964 Qu’est-ce que la dissimilation? Revue de philologie 17: 642−667. [Also in: Togeby, Knud 1968 Immanence et structure. Recueil d’articles publié à l’occasion de son cinquantième anniversaire. Etudes Romanes 2: 96−121]. Yip, Moira 1988 The obligatory contour principle and phonological rules: A loss of identity. Linguistic Inquiry 19(1): 65−100.

Marc Plénat, Valence d’Albigeois (France)

54. Closing suffixes

54. Closing suffixes 1. 2. 3. 4. 5. 6. 7. 8. 9.

Introduction History of research Delimiting the phenomenon and providing a precise definition Form and semantics in closing suffixation Closing suffixation and word-class specification Closing suffixes and diachrony Are conversion and subtraction closing rules? Conclusion References

Abstract Closing suffixes are a topic related to affix ordering. A closing suffix closes the word to further suffixation. The article discusses a number of theoretical issues relevant to closing suffixation and considers examples from different languages. It is shown that there are closing suffixes in inflection and closing suffixes in derivation and that in derivation attention should also be paid to evaluative suffixes, since they are not always trivially closing.

1. Introduction Closing suffixes or closing suffixation (I use the two terms as interchangeable) is a topic related to affix ordering. The latter is a central issue in linguistics and many theories have been suggested to account for the way affixes combine (see the overviews in Manova and Aronoff 2010 and Rice 2011). According to Manova and Aronoff (2010), there are eight types of approaches to affix order: 1) phonological; 2) morphological; 3) syntactic; 4) semantic; 5) typological (or statistical); 6) psycholinguistic; 7) cognitive; and 8) templatic. Phonological affix ordering uses phonological information, morphological ordering relies on morphological information, etc., and templatic ordering means that there is some order but it is (usually) inexplicable. Following this classification, closing suffixes are a subtype of morphological ordering, since the definition of closing suffixation relies on morphological information such as the existence of suffixes (morphemes). The idea of closing suffixation is thus an additional attempt to explain restrictions on affix ordering. The logic behind it is the following. On the one hand, diagrammatic affixation (i.e. affixation by addition of an overt affix) is the most natural way of expressing addition of meaning (i.e. new semantics) in a natural language; on the other hand, the words are not always formed by the addition of affixes to already affixed words. This fact provides evidence, among other things, that some of the affixes a language possesses may serve to stop further affixation. Such affixes seem to “close” the word to the addition of further affixes and are therefore termed closing. In the literature, the issue has been discussed only with respect to suffixation, therefore the term closing suffixes. However, as we will see below, prefixation also appears compatible with the idea of closing morphemes.

957

958

V. Rules and restrictions in word-formation II: Special cases

2. History of research The first mentioning of the term closing with respect to morphological material is usually referred to Eugene Nida’s book Morphology (1946, 1949 2nd ed.). Nida (1949: 85) differentiated closing and non-closing morphemes and used the label “closing” to describe the role of the inflectional suffixes in morphology, i.e. the fact that after an inflectional suffix no derivational suffixes can be added: Certain morphemes close the construction to further formation. For example, in English the use of a genitive suffix closes the noun to further suffixation. No suffix may follow the genitive. […] [T]he addition of the plural -s closes any form to further derivation by such suffixes as -ment, -ity, -ence, -ion, -ian, -ize, -er. A genitive suffix does the same thing. This break in structure in English coincides with the division between inflectional and derivational formations.

Thus, Nida’s definition of a closing suffix is a version of Greenberg’s “Universal 28: If both the derivation and the inflection follow the root, or they both precede the root, the derivation is always between the root and the inflection” (Greenberg 1963: 93). In his 1985 Ph.D. dissertation, van Marle reported the “inability” of some Dutch derivational suffixes “to constitute the starting point for further morphological coining” (p. 236). However, van Marle did not use the term closing suffix but spoke of suffixes the morphological valence of which is either low or zero (cf. Schultink 1962: 132 f.). The characteristic trait of low-valence [derived] words is that their morphological valence is (i) either highly restricted, or (ii) zero. In the case that their morphological valence is not equal to zero, there is a proviso that further coining is restricted to categories with a predominantly − or even exclusively − ‘syntax-directed’ nature. (van Marle 1985: 236)

In a similar fashion, in the more recent literature on affix ordering closing suffixes are usually defined as derivational suffixes that do not allow addition of further derivational suffixes. A number of linguists have reported the existence of closing suffixes in various languages: Szymanek (2000) is on closing morphemes (the term Szymanek uses) in English and Polish; Aronoff and Fuhrhop (2002) report a phenomenon that bans the further derivation in German and explain it in terms of closing suffixes; Manova (2008, 2010) provides evidence for closing suffixes in Bulgarian and Russian; Plungian and Sitchinava (2009) speak of closing suffixes in Russian; Melissaropoulou and Ralli (2010) acknowledge the existence of closing suffixes in Greek derivational morphology; and Manova and Winternitz (2011) discuss closing diminutive suffixes in Bulgarian and Polish. (Note, however, that these studies, though dealing with closing suffixes, are not thorough investigations of the phenomenon of closing suffixation.) It should also be mentioned that the term closing suffix is difficult to find in reference sources. Even reference books devoted exclusively to morphological terminology (e.g., Laurie Bauer’s A Glossary of Morphology from 2004) do not have an entry for “closing suffix”. I could find “closing suffix” only in the glossary of Aronoff and Fudeman’s book What is Morphology? (2005).

54. Closing suffixes Finally, some approaches to affix ordering, without explicitly mentioning closing suffixes, are perfectly compatible with the phenomenon of closing suffixation. Approaches of this type are: the monosuffix constraint (Aronoff and Fuhrhop 2002); the parsability hypothesis (Hay 2001, 2002, 2003) and the elaboration of it termed “complexity-based ordering” (Plag 2002; Hay and Plag 2004; Plag and Baayen 2009). Finally, an analysis that relies on morphological selectional restrictions (Plag 1996) is also compatible with closing suffixes. According to the monosuffix constraint, in English “suffixes that select Germanic bases select unsuffixed bases” (Aronoff and Fuhrhop 2002: 473), i.e. the Germanic part of the English derivational morphology allows only one derivational suffix and that single derivational suffix is thus a closing suffix. This issue is also discussed in Szymanek (2000) who illustrates it with copious examples and who, in contrast to Aronoff and Fuhrhop (2002), provides an analysis in terms of closing morphemes. Aronoff and Fuhrhop’s (2002) argument as to why they speak of the monosuffix constraint in English and of closing suffixes in German is the following: “If English Germanic suffixes were all closing suffixes, then all the adjectival suffixes would have to be viewed as exceptionally non-closing only when they are followed by -ness. When not followed by -ness, the same suffixes would be closing suffixes. So the exceptionality of -ness cannot be expressed properly within the closing suffix framework” (p. 475). Note that the English suffix -ness is closing. The parsability hypothesis claims that a set of factors are responsible for morphological parsing, such as phonology, productivity, regularity, semantic transparency, and relative frequency. Since parsability depends on a number of factors, it is a gradual notion and allows affixes to be ordered hierarchically according to their degree of parsability. Parsability determines affix order in the sense that a more parsable affix should occur outside a less parsable affix because this order is easier to process. As a parsable affix adds morphological structure to a base, making the latter more complex morphologically, Plag (2002) termed affix ordering that depends on parsability complexity-based ordering. Thus the parsability hierarchy (or complexity-based ordering) of the suffixes A, B, C, D and E (see Table 54.1), where A is the least parsable suffix and E is the most parsable one, predicts that all combinations in which A is followed by the suffixes B, C, D and E should be possible (e.g., ACE), whereas combinations such as *CAD and *EAB should be impossible. This way of ordering of affixes assigns to the suffix E the status of a closing suffix. Table 54.1 illustrates this type of affix ordering with a hypothetical example of a parsability hierarchy: Tab. 54.1: A hypothetical parsability hierarchy (“+” marks existing combinations)

959

960

V. Rules and restrictions in word-formation II: Special cases Note that in the parsability hierarchy more than one suffix can be closing. For example, in Table 54.1 both D and E are closing. Plag (2002) and Hay and Plag (2004) also demonstrate that parsability works in conjunction with selectional restrictions on affix order. An example of a selectional rule is the fact that the English suffix -ization always selects the suffix -al. On the role of selectional restrictions in English word-formation, see Plag (1996). Thus, a selectional rule may also state that a particular suffix is never followed by another suffix, i.e. that it is a closing suffix.

3.

Delimiting the phenomenon and providing a precise definition

3.1. Delimiting closing suffixation 3.1.1. Terminal suffixes A terminal suffix is the last suffix in a word form. However, since there may be a suffix that is terminal in one word and followed by a suffix in another word, not all terminal suffixes are closing. For example, the English suffix -ation is terminal in organiz-ation but followed by -al in organiz-ation-al.

3.1.2. Blocking Another phenomenon related to closing suffixation is blocking. We speak of blocking if the existence of one lexeme prevents the derivation of another lexeme with the same or similar semantics (Aronoff 1976, and many others). For example, the existence of glory in English blocks the derivation of *gloriousity (Aronoff 1976: 44) and thus also the suffix combination -ous + -ity in this particular case. However, if one of the combinations of the suffix -ous is blocked, it does not mean that -ous is closing. Consider, for example, gloriousness, where -ous is followed by -ness. Blocking refers to a single combination of two particular suffixes, often only in a single word (on blocking, see also Rainer 1988). Closing suffixation refers to the general combinability of a suffix, that is closing suffixation accounts for the non-combinability of a suffix with all other suffixes in a language. Additionally, while there is a clear semantic explanation of blocking, a closing suffix cannot be always successfully defined with the help of semantics only. The role of semantics in closing suffixation is discussed in section 4.2.

3.2. Closing suffixes and morphological organization In this subsection I discuss closing suffixation in relation to various issues that I label “morphological organization”: base-driven vs. affix-driven morphology, morphological language types, derivation vs. inflection, evaluative morphology and suffixation vs. prefixation.

54. Closing suffixes

3.2.1. Base-driven vs. affix-driven morphology In morphological theory, affixation is seen as being either (i) base-driven, i.e. it is the base that selects the affix and the direction of the derivation is thus from the base to the affix, or as (ii) affix-driven, i.e. it is the affix that selects the base and the direction of the derivation is from the affix to the base. Classical descriptive sources are affix-driven − they list suffixes and explain to what bases a suffix attaches to express a particular semantic meaning associated with that suffix. An example of a statement of this type is: the English suffix -(at)ion attaches to verbs derived by the suffix -ize to form abstract nouns, as in nasalize → nasalization. Studies on affix ordering are often base-driven, i.e. they would describe the same combination as starting with the suffix -ize (a verb derived with -ize, that is a base) to which then the suffix -ation is attached. Since closing suffixes stop the attachment of further suffixes but a closing suffix can follow other suffixes and bases, closing suffixation is base-driven by definition. This could be the explanation of why many sources, especially classical grammars and textbooks that provide affix-driven descriptions of morphology, do not mention closing suffixes.

3.2.2. Closing suffixes and morphological language types In the literature so far, closing suffixes have been discussed primarily in relation to the inflecting languages (recall the history of research). Closing suffixes have not been reported in agglutinating and incorporating languages, though they should be compatible with these morphological types too. This conclusion has the following motivation. There are two ways to produce morphological structure − template morphology and layered morphology. A template is a linear set of slots and each slot accommodates particular suffixes which can be substituted in that slot but never co-occur. Thus, the last slot in the template hosts the closing suffixes. On templates, see Stump (2006). Layered morphology is usually semantically compositional and also scopal by nature. In a scopal affix order, a suffix with a broader semantic scope follows a suffix with a narrower scope. Thus, the suffix(es) with the broadest scope in a language will be the closing one(s). On semantic scope in affix ordering, see Rice (2000). An accessible explanation of template and layered morphology with a comparison of these two types of morphological organization can be found in Manova and Aronoff (2010).

3.2.3. Derivation vs. inflection A language that distinguishes between derivational and inflectional suffix slots usually allows more than one derivational and more than one inflectional suffix. Such a situation is found in the Slavic family. (1) is a schema of the structure of the Slavic word. In (1), a slot that is associated with more than one arrow can host more than one affix. The term evaluative is used after Scalise (1984) and denotes diminutive and augmentative suffixes.

961

962

V. Rules and restrictions in word-formation II: Special cases (1)

The structure of the Slavic word

(cf. Manova 2011b) For German, a language that, like Slavic languages, distinguishes between derivational and inflectional suffixes, Aronoff and Fuhrhop (2002: 468) argue for two types of closing suffixes − closing suffixes in derivation and closing suffixes in inflection. We will also adopt this differentiation of closing material. A derivational suffix that is never followed by another suffix of the same type (i.e. another derivational suffix) is a closing derivational suffix, whereas an inflectional suffix that is never followed by another inflectional suffix is a closing inflectional suffix. A closing derivational suffix can be followed by inflection. As regards inflection, in Slavic languages a word has either a single inflectional suffix or if there is more than one inflectional suffix, the order of the inflectional suffixes is fixed. (2) is an example of a fixed (also called templatic) order of inflectional morphemes. The template in (2) gives the order of the adjectival inflectional suffixes in Bulgarian; note that GEND and NUM share the same slot, i.e. are cumulatively expressed (Matthews 1972): (2)

BASE−GEND/NUM−DEF a. krasiv-ø-ø ‘beautiful’ (masculine) krasiv-ø-ijat ‘beautiful-DEF’ b. krasiv-a-ø ‘beautiful-FEM/SG’ krasiv-a-ta ‘beautiful-FEM/SG-DEF’ c. krasiv-o-ø ‘beautiful-NEUT/SG’ krasiv-o-to ‘beautiful-NEUT/SG-DEF’ d. krasiv-i-ø ‘beautiful-PL’ krasiv-i-te ‘beautiful-PL-DEF’

These examples show that the definite article (DEF) is a closing suffix in inflection. Let us now look into the order of the suffixes that occupy the evaluative derivational subslot in (1).

3.2.4. Evaluative suffixes As indicated in (1), Slavic derivational suffixes are of two types − non-evaluative and evaluative; and as already mentioned, the latter type includes diminutive and augmenta-

54. Closing suffixes

963

tive suffixes. Aronoff and Fuhrhop (2002) exclude diminutive suffixes from their investigation of closing suffixes in German because according to these authors diminutive suffixes are closing by definition, which is due to the fact that a German noun can have only one diminutive suffix which is also always the last derivational suffix in the word form. In the Slavic languages, however, only the augmentative suffixes are like the German diminutive suffixes. The Slavic diminutive suffixes combine with each other (Szymanek and Derkach 2005; Manova 2011c; Manova and Winternitz 2011). Consider the following double diminutives: Russian kartina ‘picture’ → DIM1 kartin-ka ‘small picture’ → DIM2 kartin-oč-ka ‘very small picture’; Polish dom ‘house’ → DIM1 domek ‘small house’ → DIM2 dom-ecz-ek ‘very small house’; and Ukrainian dub ‘oak’ → DIM1 dub-ok ‘small oak’ → DIM2 dub-oč-ok ‘very small oak’. Thus, in Slavic languages not all diminutive suffixes are closing. Table 54.2 gives the exact combinations of diminutive suffixes in Bulgarian, a language in which also triple diminutives are possible: dete ‘child’ → DIM1 det-ence ‘little child’ → DIM2 det-enc-ence ‘very little child’ → DIM3 det-enc-enc-ence ‘very very little child’. From Table 54.2, we see that there are diminutive suffixes that are clearly closing: -ec, -čica and -ička. These suffixes are never followed by another diminutive (or other derivational) suffix. It is, however, not the case with the suffix -ence that is never followed by another diminutive suffix but can be repeated on adjacent cycles. (Recall also the examples of double diminutives from Russian, Polish and Ukrainian cited above.) In order to account properly for the repetition of a diminutive suffix, we should allow the recursive use of a closing derivational suffix. Tab. 54.2: Combinability of the Bulgarian diminutive suffixes (from Manova and Winternitz 2011) Nouns in in -C

DIM1 suffixes

DIM2 suffixes

DIM3 suffixes

-ec (unproductive) -le (unproductive) -če

-ence

-ence

-čica (unproductive) in -a

-ica

-ka

-ka

-ica

-ica

-ence

-ence

-ička (unproductive) in -o

-ce

in -e

-ence -ice (unproductive)

Clearly, the suffixes -ica and -ka that occur in both orders, -ica + -ka and -ka + -ica, are both non-closing.

3.2.5. Are there closing prefixes? In the literature, there is no information on closing prefixes but there are studies on prefixation that seem to offer evidence for the existence of closing prefixes. A closing

964

V. Rules and restrictions in word-formation II: Special cases prefix will be the last prefix, outward from the root, in a sequence of prefixes. The account of the English prefixes in Zirkel (2010) where a complexity-based ordering hierarchy of prefixes is provided (recall Table 54.1) gives the closing prefixes in English. The assignment of the Slavic prefixes into different groups, such as lexical, super-lexical and perfectivizing (Babko-Malaya 1999; Svenonius 2004, among others) also seems to speak for the existence of closing prefixes, since the prefixes of the different groups combine in a specific way and there are prefixes that are always in the last position, outward from the root, in a sequence of prefixes. Thus, closing prefixation is an issue for future research.

3.3. Precise definition of closing suffixation The following definition accounts for all instances of closing suffixes we discussed above: Closing suffixation is a base-driven morphological phenomenon whereby a suffix closes the word to the addition of further suffixes of the same type: a closing derivational suffix is never followed by another derivational suffix, except by itself, i.e. a closing derivational suffix can be repeated on adjacent cycles or followed by inflection. A closing inflectional suffix is never followed by another suffix, be it derivational or inflectional.

4. Form and semantics in closing suffixation 4.1. The role of morphotactics Homophonous suffixes can be used to check the role of morphotactics in closing suffixation. Put differently, if homophonous suffixes are always closing, morphotactics defines the feature [+/-closing]. The Polish suffixes -k1-a and -k2-a in (3) below are an instance of homophonous suffixes. The suffix -k1-a derives objects, whereas the suffix -k2-a subserves the formation of nouns for female humans from male humans. Intriguingly, the nouns derived by -k1-a can be further derived (3a), whereas nouns derived by the suffix -k2-a do not serve as bases for further derivation (3b). (3)

a. kołys-k1-a ‘cradle’ kołys-k1-a ‘cradle’ b. trener-k2-a ‘female trener-k2-a ‘female

→ ADJ kołys-k1-ow-y ‘cradle-’ → DIM kołys-ecz1-k-a trainer’ → ADJ ø trainer’ → DIM ø

Thus, (3) provides evidence that morphotactics cannot define closing suffixation. However, Aronoff and Furhop’s (2002) account of German closing suffixes relies on morphotactics. Following the tradition of generative morphology, they related the German closing suffixes to a non-semantic fact − the presence of a linking element in a compound with a first constituent a derived noun that terminates in a closing suffix. To exemplify:

54. Closing suffixes (4)

965

Üb-ung-s-sache to train-ing-linking element-matter ‘a matter of training’

In this example, -ung is a closing suffix and -s- is a linking element. Note that German is a language with very productive compounding and almost every noun can be used as a first constitutent of a compound. Therefore, Aronoff and Fuhrhop (2002) had to provide an explanation for why a closing suffix is followed by word-formation material in compounds, and they claimed that the German linking elements “reopen” closing suffixes. However, linking elements are not a perfect diagnostic criterion for the feature [+/− closing]. For example, they cannot explain the closing character of the suffix -ismus that derives abstract nouns, as in the German noun Real-ismus ‘realism’, but does not require a linking element in a compound, e.g., Real-ismus-streit ‘polemic on realism’. Intriguingly, the suffix -ismus (English -ism, Russian -izm, Polish -i/yzm, Bulgarian -izăm, etc.) is one of the very few instances of a suffix that seems to be cross-linguistically closing. Thus, we will conclude that morphotactics is not a sufficient diagnostic for the definition of a suffix as [+/−closing].

4.2. The role of morphosemantics In order to check the role of semantics in closing suffixation, Manova (2009) compares the German closing suffixes from Aronoff and Fuhrhop (2002) with their semantic homologues from two Slavic languages, Bulgarian and Russian. In what follows, I will use the same strategy and some of Manova’s examples of closing suffixes in Bulgarian. I will not give Russian examples because there is a special article on closing suffix patterns in Russian (see article 55). According to Aronoff and Fuhrhop (2002: 461), six German suffixes are closing: -e, -heit/-keit/-igkeit, -in, -isch, -ling, and -ung. With respect to the categories of the base and the output, we can define these German suffixes in the following way: (5)

V ADJ N males N person V, ADJ V

+ + + + + +

-e -heit/-keit/-igkeit -in -isch -ling -ung

→ → → → → →

N N N females ADJ N N

The German pattern V + -e → N, as in pflegen ‘to care for’ → Pflege ‘care’, can be compared with Bulgarian derivations of the type griža se ‘(I) care for’ → griž-a ‘care’. Such derivations are, however, conversions in Bulgarian where the Bulgarian suffixes parallel to the German -e are purely inflectional (consider Bulgarian griž-a − PL griži), which explains why the suffix -a cannot be followed by derivational suffixes in Bulgarian. The German pattern ADJ + -heit/-keit/-igkeit → N, e.g., schön ‘beautiful’ → Schönheit ‘beauty’, has a clear parallel in Bulgarian derivations such as xubav ‘beautiful’ → xub-ost ‘beauty’. In the two languages, N is an abstract noun expressing a property.

966

V. Rules and restrictions in word-formation II: Special cases However, the Bulgarian parallels to the German closing suffix -heit (-keit/-igkeit) allow further suffixation. Consider, Bulgarian cjal ‘whole’ → cjal-ost ‘wholeness’ → cjalost-en ‘complete’ → cjalostn-ost ‘completeness’. The Bulgarian suffix -ost can also be followed by the suffix -nik. This unproductive pattern derives human nouns with negative characteristics, e.g., Bulgarian xubav ‘beautiful’ → xubost ‘beauty’ → xubost-nik ‘rascal’. Thus, the suffix -ost is not closing. Note that also the German suffix -heit can be followed by other suffixes, e.g., ein-heit-lich ‘uniform’. Such instances are, however, exceptions in German (Aronoff and Fuhrhop 2002: 460). The German closing suffix -in derives female humans from male humans. Over ninety percent of all female humans in German are derived by the attachment of -in. Additionally, formations with other suffixes often have -in doublets or allow the addition of -in, e.g., Baron-esse and Baron-in, as well as Prinzess-in (cf. Wellmann 1975: 107 ff.). In other words, the fact that the suffix -in is closing is sufficient to determine the semantic meaning ‘a female human derived from a male human’ as closing in German. Bulgarian, however, possesses a set of suffixes for deriving female humans from male humans: -ka, -inja, -kinja, -ica, -esa, -isa, and -va (cf. Stojanov 1993). Manova (2008) establishes that the Bulgarian suffixes are closing only if the suffix is native and added to a base denoting a male person (except for the unique suffix -va which only forms a single noun). Suffixes deriving female animals are not closing, e.g., magare ‘donkey’ → FEM magar-ica → DIM magarič-ka. Female humans derived from foreign bases differ from those derived from native bases and can also be diminutivized, e.g., princ ‘prince’ → FEM princ-esa → DIM princes-ka, poet ‘poet’ → FEM poet-esa → DIM poetes-ka, etc. Another exception to the closing character of the pattern of female humans constitute lexicalizations, e.g., daskal ‘teacher (archaic)’ → daskal-ica ‘female teacher and female pupil’ → DIM daskalič-ka ‘little female teacher and little female pupil’. The German -isch, as in Schriftsteller ‘writer’ → schriftsteller-isch ‘literary; lit. writer-REL.ADJ’, corresponds to Bg. -ski, e.g., pisatel ‘writer’ → pisatel-ski ‘literary; lit. writer-REL.ADJ’. The Bulgarian suffix -ski is closing, but since the same suffix is not closing in Russian (R. -skij), recent borrowings from this language could be analyzed as exhibiting the suffix combination -ski + -ost = -skost. Consider: rus-sk-ost ‘Russian-like style’, svet-sk-ost ‘worldly-minded style’, det-sk-ost ‘child-like style’. However, as such nouns are borrowings and used only in highly specialized texts, it is difficult to decide whether they should count as evidence for changing combinability of -ski in Bulgarian (see the discussion in Manova 2010). The German -ling has no parallel in Bulgarian. It is also unclear whether the suffix is closing in German. On the Internet one can find: prüfen ‘to examin’ → Prüfling ‘the examined person’ → FEM Prüflingin, lehren ‘to teach’ → Lehrling ‘the taught person’ → FEM Lehrlingin, Haft ‘prison’ → Häftling ‘prisoner’ → FEM Häftlingin. For an alternative explanation of the peculiar behavior of the German -ling, see Plank (2012). The last German closing suffix -ung derives action nouns as in bewegen ‘to move’ → Bewegung ‘moving, movement’. This German suffix has two corresponding suffixes in Bulgarian, the suffix -Vne, e.g., dviža (se) ‘(I) move’ → dviž-ene ‘moving’, and the suffix -Vnie, e.g., dviža (se) ‘(I) move’ → dviž-enie ‘moving, movement’. Curiously, only -Vnie nouns can be diminutivized, e.g., DIM dviž-eni-jce whereas the suffix -Vne cannot be followed by other suffixes, dvižene ‘moving’ → *DIM. This different behavior of the two suffixes with respect to diminutivization can be explained as due to the more lexicalized character of the -nie pattern, i.e. -nie nouns often exhibit lexicalized semantics

54. Closing suffixes and denote objects instead of actions (cf. Radeva 1991: 139), e.g., piša ‘(I) write’ → pis-anie ‘a piece of writing’ (cf. pis-ane ‘writing’). In other words, the Bulgarian -Vne is closing while -Vnie is not. Lexicalizations of -ne nouns are seldom. The reverse dictionary of Bulgarian (Andrejčin 1975) lists only prane ‘laundry’, piene ‘drink’, jadene ‘food’, and imane ‘wealth’. Some native speakers use diminutivized forms of these nominalizations. (For lexicalizations of action nouns in English and Polish and how they relate to closing suffixation, see Szymanek 2000.) In sum, meanings that are closing in German are not always closing in Bulgarian, which thus speaks against the existence of semantic patterns that are universally closing. However, we can define a specific closing suffix in a particular language by its semantics, e.g., the suffix -Vne that derives action nouns in Bulgarian is closing, as is the suffix -ung in German. We can even define sets of semantically related closing suffixes in one language and in different languages, e.g., native suffixes deriving female humans from male humans are closing in Bulgarian (except the suffix -va) and in German. However, due to the exceptions found, we cannot conclude that closing suffixes can be defined properly on the basis of their semantics only. So far, we have seen that closing suffixes could have exceptions. However, the exceptions we had to deal with above were primarily due to borrowing and lexicalization. The role of lexicalization in closing suffixation is discussed in some detail in Szymanek (2000) who systematically differentiates between a semantics that is compositional and a semantics that is lexicalized (i.e. non-compositional). Szymanek argues that if only instances of compositional semantics are considered, closing suffixes do not have exceptions. However, I will now give examples of closing suffixes from Bulgarian that allow exceptions and which share the semantics of the closing pattern, i.e. exceptions that are also semantically compositional. The Bulgarian suffixes -ina and -ota both derive abstract nouns from adjectival bases, as in dobrina ‘good deed’ and dobrota ‘goodness’, both formed from the adjective dobăr ‘good’. Nevertheless, while the suffix -ina easily diminutivizes, one even does not need a basic word in order to produce the corresponding suffix combination -in-ka (as in dobrinka ‘little good deed’), the suffix -ota is hard to diminutivize. Perhaps the only exception is krasotička ‘little beautiful place’ (found on the Internet), which is, however, a diminutive form of the lexicalized krasota ‘beautiful place’. In addition, despite the existence of the suffix -oten, e.g., straxoten ‘terrific’ (note that there is no *straxota, i.e. -oten is not a combination of -otaN + -enADJ), abstract nouns in -ota do not adjectivize, i.e. there is no *dobroten in Bulgarian. In other words, the suffix -ota seems to be a good example of a closing suffix. Nevertheless, two words that are clearly semantically compositional are exceptions to the closing character of -ota: (sam ‘alone’ →) sam-ota ‘loneliness’ → sam-ot-en ‘lonely’ and (čest ‘frequent’ →) čest-ota ‘frequency’ → čest-ot-en ‘frequency-’. Thus, semantic compositionality is not a perfect diagnostic for the feature [+/−closing] either. In sum, closing suffixes allow exceptions. If one of a set of suffixes with synonymous semantics is closing, the other suffixes of the set are not all closing. Lexicalizations always behave as exceptions but also semantically compositional derivations, including borrowings, that match a closing semantic pattern could allow further derivations. Nevertheless, there seems to exist some relation between closing suffixes and semantics. Suffixes that derive female humans as well as action and abstract nouns tend to be closing even in languages (such as Bulgarian and German) that do not belong to the same language family.

967

968

V. Rules and restrictions in word-formation II: Special cases

5. Closing suffixation and word-class specification Of the six German closing suffixes in Aronoff and Fuhrhop (2002), five derive nouns; and this fact is neither because of the suffixes selected nor by chance. In what follows I will try to explain why nominal suffixes tend to be closing, while verbal and adjectival ones do not. For illustration of my argument, I will refer to word-formation in two Germanic languages, English and German, and three Slavic languages, Bulgarian, Russian, and Polish. Germanic and Slavic languages have very productive patterns for derivation of deverbal nouns and suffixes that derive verbs always combine with at least one nominalizing suffix, which thus explains why Slavic and Germanic verbal suffixes are not closing. Of the above listed five languages, Bulgarian is the only language with a closing verbal suffix, the suffix -Vsa-m that is of a Greek origin. Then, Germanic and Slavic languages possess very productive patterns for derivation of abstract nouns from adjectives and in English, Russian and Polish every adjectival suffix combines with at least one nominalizing suffix (E. -ness, R. -ost’ and P. -ość). Bulgarian and German, compared above, are rather exceptional in this respect; both languages have an adjectival suffix that is closing (-isch in German and -ski in Bulgarian). Thus, we come to suffixes that derive nouns in Germanic and Slavic. Nouns are the only word class with a potential for closing suffixes, which explains why closing suffixes usually have this syntactic specification.

6. Closing suffixes and diachrony A language’s morphology changes over time and one of the visible results of this process is the changed morphological status of some affixes. For example, in the diachronic development of a language, an inflectional suffix may be reanalyzed as a derivational formative. Such neo-derivational suffixes are closing. A good illustration of this type of closing suffixes is provided by the Slavic family. The example below is from Bulgarian, the situation is, however, the same in the other Slavic languages (Manova 2011a). In Bulgarian, the suffix -in that is found in singular forms of ethnicity terms, as in bălgarin ‘a Bulgarian’ (PL bălgari ‘(the) Bulgarians’) is classified as a derivational suffix in the grammars (Stojanov 1993: 174). However, the suffix -in was originally a singulative (Georgiev 1969: 111−115; see also article 65 on singulatives); and since there is no other singulative suffix in Bulgarian, -in has been assigned to the class of derivational suffixes. This is the explanation for why -in is never followed by another derivational suffix in Bulgarian (and in the other Slavic languages).

7. Are conversion and subtraction closing rules? In this section, I briefly discuss whether there is some relation between conversion (when new meaning is expressed by the same form) and subtraction (when new meaning is

54. Closing suffixes expressed by some deletion of form) and closing suffixation. On conversion and subtraction as morphological rules, see Manova (2011a). We devote attention to this issue, since when morphology “re-uses” the same form (as in conversion) or shortens a form (as in subtraction), it, like in closing suffixation, avoids further suffixation; and one thus expects that conversion and subtraction could be closing operations too. There are linguists who do claim that conversion (or zero suffixation) closes the word to further derivation in the way closing suffixation does (Sitchinava and Plungian 2009 for Russian; Szymanek 2000 based on Myers 1984 for English). This claim, however, seems unjustified. There are enough Russian conversions that allow further derivation, but since Russian is an inflecting language, Russian conversions involve deletion and addition of inflection and this makes conversion in Russian more difficult to identify than in English. Thus, in the limited space of this article, I will discuss only English conversions, as such examples are well known from the literature. In English, the levelordering or stratal approach (Siegel 1979; Allen 1978; Kiparsky 1982) divides the lexicon in levels or strata that are ordered in a specific way: level-II affixes can follow both level-I and level-II affixes, but after a level-II affix no level-I affix can be attached. According to Kiparsky (1982), English verb-to-noun derivations are level-I, while nounto-verb derivations are level-II. This analysis thus assigns conversions to nouns and verbs to different levels, conversions to nouns being level-I, which means that at least conversions to nouns should serve as bases for further derivation. This is indeed the case and can be illustrated with the following examples: alarm-ist, escap-ism, segment-al, etc. (Kiparsky 1982). In these examples, the bases alarm, escape and segment are derived by verb-to-noun conversion and then suffixed by an overt suffix. Examples such as these undoubtedly show that conversion is not a closing rule. For a discussion of these and similar issues within level-ordering, I refer the reader to Don (1993: 29−35). As regards subtraction, the situation is similar to that for conversion. There are subtractions that allow further derivations and such that do not. Consider the following examples from Bulgarian (see Manova 2011a): biologija → biolog (derived by subtraction) → biolož-ki ‘biologist’s’ and mečka ‘bear’ → meči ‘bear-’ (with subtraction of -k-) → 0/, i.e. no further derivation is possible. However, we cannot attribute the missing further derivations from meči ‘bear-’ to the subtractive origin of this adjective, as the adjective bioložki, which is formed by affixation with an overt affix, cannot be further derived either.

8. Conclusion Closing suffixation is a specific instance of affix ordering. There are closing suffixes in derivation and closing suffixes in inflection. A closing suffix closes the word to the addition of further suffixes of the same type but a closing derivational suffix may be repeated on adjacent cycles. There is evidence that closing suffixes exist in a number of languages, but it is hard to define semantic patterns that are cross-linguistically closing, though semantics is of importance to closing suffixation. Moreover, closing suffixation allows for exceptions. Therefore, closing suffixes seem best describable as a tendency, i.e. as suffixes that tend to prohibit further suffixation.

969

970

V. Rules and restrictions in word-formation II: Special cases

Acknowledgements This article uses data collected for the project (De)composing the Slavic word that was carried out at the University of Vienna (2007–2011), PI Stela Manova. The project was supported by the Austrian Science Fund (FWF), Grant V64-G03. The support is hereby gratefully acknowledged.

9. References Allen, Margaret 1978 Morphological Investigations. Ph.D. dissertation, University of Connecticut. Andrejčin, Ljubomir (ed.) 1975 Obraten rečnik na săvremennija bălgarski ezik. Sofija: BAN. Aronoff, Mark 1976 Word Formation in Generative Grammar. Cambridge, MA: MIT Press. Aronoff, Mark and Kirsten Fudeman 2005 What is Morphology? Malden, MA: Wiley-Blackwell. Aronoff, Mark and Nanna Fuhrhop 2002 Restricting suffix combinations in German and English: Closing suffixes and the monosuffix constraint. Natural Language and Linguistic Theory 20: 451−490. Babko-Malaya, Olga 1999 Zero Morphology. A Study of Aspect, Argument Structure and Case. Piscataway, NJ: Rutgers University Press. Bauer, Laurie 2004 A Glossary of Morphology. Washington, D.C.: Georgetown University Press. Don, Jan 1993 Morphological Conversion. Utrecht: OTS. Georgiev, Vladimir 1969 Osnovni problemi na slavjanskata diaxronna morfologija. Sofija: BAN. Greenberg, Joseph 1963 Some universals of grammar with particular reference to the order of meaningful elements. In: Joseph Greenberg (ed.), Universals of Language, 40−70. Cambridge, MA: MIT Press. Hay, Jennifer 2001 Lexical frequency in morphology: Is everything relative? Linguistics 39: 1041−1070. Hay, Jennifer 2002 From speech perception to morphology: Affix ordering revisited. Language 78(3): 527− 555. Hay, Jennifer 2003 Causes and Consequences of Word Structure. London: Routledge. Hay, Jennifer and Ingo Plag 2004 What constrains possible suffix combinations? On the interaction of grammatical and processing restrictions in derivational morphology. Natural Language and Linguistic Theory 22: 565−596. Kiparsky, Paul 1982 Lexical morphology and phonology. In: In-Seok Yang (ed.), Linguistics in the Morning Calm, 3−91. Seoul: Hanshin. Manova, Stela 2008 Closing suffixes and the structure of the Slavic word: Movierung. Wiener Slavistisches Jahrbuch 54: 91−104.

54. Closing suffixes Manova, Stela 2009 Closing suffixes in Bulgarian, Russian and German: The role of semantics. In: Mónika Farkas Baráthi (ed.), Bulgarian Language and Literature in Slavic and non-Slavic Contexts, 286−292. Szeged: JATE Press. Manova, Stela 2010 Suffix combinations in Bulgarian: Parsability and hierarchy-based ordering. Morphology 20(1): 267−296. Manova, Stela 2011a Understanding Morphological Rules. With Special Emphasis on Conversion and Subtraction in Bulgarian, Russian and Serbo-Croatian. Dordrecht: Springer. Manova, Stela 2011b A cognitive approach to SUFF1-SUFF2 combinations: A tribute to Carl Friedrich Gauss. Word Structure 4(2): 272−300. Manova, Stela 2011c Affix ordering in Bulgarian, Russian and Polish double diminutives. Paper presented at the 6th Annual Meeting of the Slavic Linguistics Society, Aix-en-Provence, France, 1− 3 September 2011. Manova, Stela and Mark Aronoff 2010 Modeling affix order. Morphology 20(1): 109−131. Manova, Stela and Kimberley Winternitz 2011 Suffix order in double and multiple diminutives: With data from Polish and Bulgarian. Studies in Polish Linguistics 6: 117−140. Matthews, Peter H. 1972 Inflectional Morphology. A Theoretical Study Based on Aspects of Latin Verb Conjugation. Cambridge: Cambridge University Press. Melissaropoulou, Dimitra and Angela Ralli 2010 Greek derivational structures: Restrictions and constraints. Morphology 20(2): 343−357. Myers, Scott 1984 Zero derivation and inflection. In: Margaret Spears and Richard Sproat (eds.), MIT Working Papers in Linguistics. Papers from the January 1984 MIT Workshop in Morphology. Vol. 7, 53−69. Cambridge, MA: Department of Linguistics and Philosophy, MIT. Nida, Eugene 1949 Morphology. 2nd ed. Ann Arbor: The University of Michigan Press. Plag, Ingo 1996 Selectional restrictions in English suffixation revisited: A reply to Fabb (1988). Linguistics 34: 769−798. Plag, Ingo 2002 The role of selectional restrictions, phonotactics and parsing in constraining suffix ordering in English. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 2001, 285−314. Dordrecht: Kluwer. Plag, Ingo and Harald Baayen 2009 Suffix ordering and morphological processing. Language 85: 109−152. Plank, Frans 2012 Why *-ling-in? The pertinacity of a wrong gender. Morphology 22(2): 277−292. Radeva, Vasilka 1991 Bălgarskoto slovoobrazuvane. Sofija: Universitetsko izdatelstvo “Sv. Kl. Oxridski”. Rainer, Franz 1988 Towards a theory of blocking: The case of Italian and German quality nouns. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1988, 155−186. Dordrecht: Foris. Rice, Keren 2000 Morpheme Order and Semantic Scope. Cambridge: Cambridge University Press.

971

972

V. Rules and restrictions in word-formation II: Special cases Rice, Keren 2011 Principles of affix ordering: An overview. Word Structure 4(2): 169−200. Scalise, Sergio 1984 Generative Morphology. Dordrecht: Foris. Schultink, Henk 1962 De morfologische valentie van het ongelede adjectief in modern Nederlands. The Hague: Van Goor. Siegel, Dorothy 1979 Topics in English Morphology. New York: Garland. Sitchinava, Dmitri and Vladimir Plungian 2009 Closing suffixation patterns in Russian, with special reference to the Russian National Corpus. Paper presented at the 2nd Vienna Workshop on Affix Order: Affix Order in Slavic and Languages with Similar Morphology, Vienna, 5−6 June 2009. Stojanov, Stojan 1993 Gramatika na bălgarskija knižoven ezik. 5th ed. Sofija: Universitetsko izdatelstvo “Sv. Kl. Oxridski”. Stump, Gregory T. 2006 Template morphology. In: Keith Brown (ed.), Encyclopedia of Language and Linguistics. Vol. 12, 559−563. Oxford: Elsevier. Svenonius, Peter 2004 Slavic prefixes and morphology. Introduction to Nordlyd 32(2). Special issue on Slavic prefixes, 177−204. Tromsø: University of Tromsø. Szymanek, Bogdan 2000 On morphotactics: Closing morphemes in English. In: Bożena Rozwadowska (ed.), PASE Papers in Language Studies, 311−320. Wrocław: Aksel. Szymanek, Bogdan and Tetyana Derkach 2005 Constraints on the derivation of double diminutives in Polish and Ukrainian. Studies in Polish Linguistics 2: 93−112. van Marle, Jaap 1985 On the Paradigmatic Dimension of Morphological Creativity. Dordrecht: ICG Printing. Wellmann, Hans 1975 Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Eine Bestandsaufnahme des Instituts für deutsche Sprache. Forschungsstelle Innsbruck. Zweiter Hauptteil: Das Substantiv. Düsseldorf: Schwann. Zirkel, Linda 2010 Prefix combinations in English: Structural and processing factors. Morphology 20(1): 239−266.

Stela Manova, Vienna (Austria)

55. Closing suffix patterns in Russian 1. 2. 3. 4. 5.

Introduction The Russian National Corpus Closing status of Russian semantic homologues of German and Bulgarian suffixes Conclusions References

55. Closing suffix patterns in Russian

Abstract The article discusses, with special reference to the material of the Russian National Corpus, semantic parallels in Russian to the closing derivational suffixes of German and Bulgarian. The putative Russian parallels are tested to see whether they are also closing (as the semantic typology may predict), and some of their combinatory properties are discussed.

1. Introduction Closing derivational suffixes are those that do not allow the addition of any further derivational suffixes to the word form. Hereafter we use the term without derivational, keeping in mind that closing inflectional suffixes also exist. The term closing suffix was originally introduced by Nida (1949) for both inflectional and derivational affixes in English. Aronoff and Fuhrhop (2002) identify the set of closing derivational affixes for German, viz. the following: -esuff, -heit/-keit/-igkeit, -in, -isch, -ling, and -ung. Manova (2008, 2011) examines their semantic homologues in Bulgarian testing the hypothesis of universal semantic constraints. (For further details see also article 54 on closing suffixes.) Our article continues Manova’s approach, aligning Russian suffixes with their German correlates and Bulgarian cognates. Their semantic equivalents in Russian show both typological parallels and differences compared to their German and Slavic counterparts, displaying a certain variation across semantic constraints and the diachronic timeline. Thus the article provides not only insight into the morphological properties of semantically analogous units (as is the case between the suffixes in Germanic and Slavic) but also into the historically cognate units in closely related languages (as is the case within the Slavic group). Bulgarian and Russian share a common Slavic heritage belonging, however, to different subgroups, viz. Bulgarian being a South Slavic language, and Russian an East Slavic language, and they have secondarily influenced each other. South Slavic words and derivational patterns were borrowed extensively into Standard Russian via Church Slavonic, and later, during the Bulgarian national revival in the 19th century, Russian became a source for Bulgarian loan words. Thus, the discrepancies between two languages with so close a history could be an additional test for hypotheses about the semantic mechanisms governing derivation. The data of the Russian National Corpus are used consistently for the evaluation of productivity and diachronic details; we begin with a brief presentation of the possibilities this corpus opens up for gaining insights into morphological structure.

2. The Russian National Corpus The Russian National Corpus (henceforth: RNC), under development primarily by the Vinogradov Institute of Russian Language of the Russian Academy of Sciences since 2001, was placed online on April 29, 2004. Its present size is more than 100 thousand

973

974

V. Rules and restrictions in word-formation II: Special cases texts, and its main subcorpus (of written texts) contains over 200 million tokens, encompassing the chronological span from the 18th to the 21th century. All the subcorpora of the RNC have different levels of linguistic tagging, depending largely upon the properties of the relevant subcorpus (for example, metrical annotation in the subcorpus of poetry, or stress in the subcorpus of oral speech), but the main types of annotation that run through the bulk of the RNC are word-by-word morphological and semantic annotations. Morpheme-by-morpheme annotation is currently being introduced as a test markup (see also below). The morphological characteristics are ascribed to the word form in full; the lemma is provided, as well as its part-of-speech, its lexical (as, e.g., gender or transitivity), and its inflectional categories. A word-by-word glossing like golovami (GOLOVA ‘head’.N.F.INANIMATE.PL.INSTR) is largely equivalent to the morphological annotation found in the RNC. However, in only a relatively small portion of the corpus is the homonymy between different word forms disambiguated so far. Otherwise, all the possible variants for a given orthographical representation are provided, like in peč’ (PEČ’ ‘stove’.N.F.INANIMATE.SG.NOM|ACC| PEČ’ ‘to bake’.V.TRANS.IPF.INF). The semantic markup is made for each word form, and its semantic categories are given according to a dictionary; a separate set of semantic categories exists for all the part-of-speech-categories. For example, concrete nouns signifying animals, abstract nouns signifying qualities, adjectives of colour or of intensity, and verbs of motion are all tagged and separately searchable. Additionally, some derivation features are indicated in the semantic markup, including morpho-semantic features (e.g., diminutive, caritive, semelfactive) or taxonomic type of the motivating word (adverb derived from adjective of size). In the context of the present article, the possibilities that the RNC offers for linguistic research into Russian morphology are the most relevant. Obviously, the main instrument of morphological research is the grammatical search in the corpus that helps to yield the search results on different lexemes, grammemes and their combinations. Currently, there is no affix markup available for searching (except for the orthographical affixes ne- ‘un-’ in participles and pol- ‘half’ that exhibit some word properties, in the subcorpus with manually resolved homonymy). At present a project on affix markup is under development, relying on the corpora dictionary with tagged prefixes and suffixes (see Tagabileva et al. 2009). A test version of morpheme search has been available in the RNC since 2011 but it will not be treated here in detail. It should be noted that, both in the search for lexemes and for word forms, the RNC allows also for wildcards, which can partially substitute for morpheme-by-morpheme tagging. For example, the combination of prefixes po-raz- that yield a semantic combination of distributive activity and centrifugal (ablative) movement, as po-raz-bežat’sja ‘to run away, one by one, by groups’ is searchable using the wildcard poraz*. Likewise, circumfixes involving both a prefix and a suffix can be set up as in do*sja ‘to arrive at a (usually, undesired) result due to a continuous activity’ (cf. do-igrat’-sja ‘to arrive at an unpleasant result while playing’). The actual semantic tagging of the RNC also includes some information about derivation, in particular, the separate treatment of nominalizations, diminutives, and secondarily derived nouns signifying females. Combination of both search types is permitted; therefore, it is possible to select specific word types from a larger group of words with a polysemic affix. Thus, the advanced search for nouns denoting female persons among

55. Closing suffix patterns in Russian words in -nica will yield učitel’nica ‘female teacher’ (← učitel’ ‘teacher’) but not bol’nica ‘hospital’ (← bol’noj ‘sick’). The RNC allows for quick sorting of the results according to token-frequency; it is possible to attest the productive or marginal status of affix combinations quickly. The diachronic dimension of the RNC includes sorting of the search results by date and customizing subcorpora to monitor changes in productivity/frequency. Software for offline statistical analysis of the query results (including collocations and other information data) is under preparation.

3. Closing status of Russian semantic homologues of German and Bulgarian suffixes 3.1. Suffixes denoting female persons In German, the feminine -in as in Lehrer-in ‘female teacher’ is a closing suffix and does not allow for further suffixation, e.g., diminutivization: *Lehrerin-chen ‘little female teacher’. A linking element -en- in compound words “reopens” it, however, in Aronoff’s and Fuhrhop’s terminology, for combination with further elements on its right: Lehrerinnenzimmer ‘room for female teachers’. Compounds, however, are a different subclass of word-formation, particularly in German, where they are highly productive and governed more by syntax-like incorporating patterns than by word-internal morphology. In Bulgarian, the feminine suffixes -k-, -inj-, -ic- are closing when applied to native stems denoting human beings (Manova 2008). In Russian, the properties of the native feminine -nic- are somewhat more complex, at least diachronically. The suffix -nicdisallows -sk-ij (adj.) and diminutive -k-a: cf. colloquial and pejorative učilka or rarely učitel’ka ‘female teacher’, the latter possibly additionally influenced by the same (standard) word from Ukrainian, or plemjaš-ka (with the n/š alternation fairly common in Russian, cf. Itkin 2007) ‘niece’ formed from plemjan-nic-a, both times with truncation of the -nic- suffix. This suffix, however still rarely, allows the addition of the relational possessive suffix -in-/-yn-: učitel’nicyn ‘of a teacher-FEM’ attested in RNC for 2000 (cf. also plemjannicyn for 1995–1999), being otherwise closing. According to Manova (2011), Russian -in- is not a derivational suffix sensu stricto as it is synonymous with the inflectional genitive (syn učitel’nic-y = učitel’nic-yn syn ‘teacher-FEM’s son’). Although it is not clear whether this is the case synchronically, diachronically in Old Russian and Church Slavonic the relational adjectives were a far more productive pattern that certainly had a status closer to inflection (while the possessive genitive was hardly used except in some special cases). See also Corbett (1995) who compares this phenomenon to double case marking (so-called Suffixaufnahme). This status can offer a good historical (if not synchronic) explanation for the peculiar status of -in-/-yn-. Indeed, in earlier Russian -nic+yn- was more productive. There are only three texts using -nicyn (slightly stylized fiction by Palej, Slavnikova, and Ulickaja of 1990–2000) in the RNC after 1930, whereas a greater degree of productivity is witnessed by a considerable number of examples in the 18th and 19th centuries (such as izmennicyn ‘of a traitor-FEM’ or volšebnicyn ‘of a fairy’). The same pattern has been preserved in a

975

976

V. Rules and restrictions in word-formation II: Special cases number of Russian surnames as Kormilicyn (< ‘[son] of a wet-nurse’) or Solženicyn (< ‘[son] of a woman who trades in malt, Russ. solod’). There are also other feminine suffixes like -in- (bog-in-ja ‘goddess’), -ic- (car-ic-a ‘tsarina’), -ess- (poėt-ess-a ‘poetess’) or -š- (in standard language with the slightly obsolete meaning ‘wife of X’, colloquial ‘female X’, as direktor-š-a formed from direktor ‘director’). They all do accept relational -in- even better than -nic- (cf. forms like knjagin-in ‘of a princess’ ← knjag-in-ja, or graf-in-in ‘of a countess’ ← graf-in-ja, direktoršin ‘of director’s wife’ ← direktorša), but generally not -k- (hypocoristic or depreciative), with the exception of the non-native suffixes as poėt-es-k-a ‘poetess’ (← poėtessa) or aktr-is-k-a ‘actress’ (← aktr-is-a), just as in Bulgarian (the two words coincide in both languages). It seems also that in Russian, unlike Bulgarian, there is no exception for nouns signifying animals (where feminine suffixes are not closing but still allow for diminutivization, cf. Bulg. magar-ic-a ‘female donkey’ → magarič-k-a), or that this exception, if it exists, is very weak. Words like Russ. l’vička ‘lionness’ (← l’v-ica ‘lioness’) are attested only as occasional nicknames of people (in some Internet forums found by Google), and lisička ‘little, cute, etc. fox’ is derived from lisica, where -ic- is not, at least synchronically, an obvious feminine suffix (both words, lisa and lisica, grammatically belonging to the feminine gender, are names of the whole species of Vulpes, the masculine lis being a marked term (poetic language)).

3.2. (Next-to-)zero nominalization German closing -e forming abstract nouns, as in pfleg-en ‘to look after sb., to care’ → Pfleg-e ‘care (noun)’, has according to Manova (2011) no equivalent in Bulgarian; likewise, it has no single directly comparable equivalent in Russian either. In Russian (as in Bulgarian) there is a zero suffixing model forming abstract nouns. For example, it is attested in words like beg (with a zero masculine inflection) ‘running’ formed from beg-at’ or bež-at’ ‘to run’, or in words like ssor-a (with feminine inflection) ‘quarrel’ formed from ssor-it’sja ‘to quarrel’. Although the zero suffix is indeed always (technically) closing according, for example, to the analysis proposed by a Russian morpheme dictionary (Kuznecova and Efremova 1986), it seems more careful not to include a model which is by definition elusive on the surface (and postulated by definition where no other suffixes can be found) into the discussion of the closing patterns. Indeed, both Russian and Bulgarian models are analyzed by Manova (2011) as instances of conversion rather than affixal patterns. The nominalizing suffix -k1- (stroj-k-a ‘(process of) construction’ from stro-it’ ‘to build’) is not closing (cf. stro-eč1-k2-a ‘little/nice process of construction’, with a posterior diminutive -k2-). However, this seems to be accounted for by the formal rather than the semantic properties of this morpheme, because nearly all the homonymous -k-(a) suffixes allow a second, always diminutive (interpreted by default also as caritative) -k2a, yielding a reduplicative pattern -(o/e)č1-k2-a: a) diminutive: knig-a ‘book’ → kniž-k-a ‘little book’ → kniž-eč-k-a ‘(very) little book’; b) suffix used for metaphorical derivation: spin-a ‘back’ → spin-k-a ‘back (clothes, furniture)’ → spin-oč-k-a; strel-a ‘arrow’ → strel-k-a ‘image resembling an arrow’ → strel-oč-k-a;

55. Closing suffix patterns in Russian c) univerbation, in Slavic word-formation defined as a procedure for forming a colloquial synonym of an adjectival group by omitting the noun and adding -k- to the adjective stem: samovol’naja otlučka ‘absence of a soldier without official leave; lit. unauthorized leave’ → samovol-k-a (formed from the adjective samovol’nyj ‘unauthorized’) → samovol-oč-k-a. The same pattern of reduplicating diminutive -k- exists in Bulgarian.

3.3. Nominalization of adjectives German -heit/-keit/-igkeit, forming abstract nouns from adjectives, as Schön-heit ‘beauty’ from schön ‘beautiful’, is closing. Russian -(n)ost’, an equivalent to the German closing suffix -heit/-keit/-igkeit, is n o t closing (cel-ost-n-yj ‘coherent’ from cel-ost’ ‘integrity’, from cel-yj ‘complete’, the same holds for Bulgarian, with the exact cognate cjal-osten). In Russian, the suffix shows a reduplication pattern of (-n)ost-nost’ that is much rarer (though not unknown) in other Slavic languages. It yields abstract nouns with the same or slightly different semantics than the ones with single -ost’. The majority of the stems are derived through an adjectival stage. Cel-ost-n-ost’ ‘integrity’ is attested in about 70 % of the examples of this model in the RNC, which may be related to the fact that it supersedes the synonymous celost’ which is attested mainly as a part of the fixed phraseological locution v celosti (i nevredimosti/soxrannosti) ‘intact, safe and sound’ and described only as such in dictionaries. According to the RNC data, the old form is still sometimes used in the meaning ‘integrity’ (largely the same as celostnost’) in different texts (some, but not the majority, of them contain certain archaic traits or use church vocabulary) in the 2000s, cf. bespokojstvo ... za celost’ tovara ‘anxiety about the goods being intact’ (corpus example from a story by Olga Slavnikova, 2001). Semantically, there is mainly a group expressing the evaluative personal qualities as blag-ost-n-ost’ ‘benevolence’ ← blag-ost-n-yj ‘benevolent’ ← blag-ost’ ‘gentleness, kindness’ ← blag-oj archaic or ironical: ‘good’, zl-ost-n-ost’ ‘malignancy’ ← zl-ost-nyj ‘malicious’ ← zl-ost’ ‘anger’ ← zl-oj ‘wicked, angry’. In these cases the whole set of two adjectives and two nouns coexist in Russian. A different case is bez-žal-ost-nost’ ‘ruthlessness’ from bez-žal-ost-n-yj ‘ruthless’, where no stage with a single suffix (*bez-žal-ost’) has ever existed; the adjective is formed with the help of the prefix bez‘without’ from the word žal-ost’ ‘pity’ that has no direct primary adjective as a counterpart (*žal-yj); the suffixal žal-k-ij ‘pitiful, poor’ has its own, also suffixal, abstract noun žal-k-ost’ ‘pity’, though the latter is rare. A similar model, with bez-/bes- forming a secondary adjective, is attested in bes-xitr-ost-n-ost’ ‘simplicity, ingenuousness’ ← besxitr-ost-n-yj ‘simple, ingenuous’ ← xitr-ost’ ‘trick’ ← xitr-yj ‘cunning’, bez-rad-ost-nost’ ‘dullness, sadness’ ← bez-rad-ost-n-yj ‘sad’ ← rad-ost-n-yj ‘joyful’ ← rad-ost’ ‘joy’ ← rad ‘happy’. In the latter case counterparts without bez- are also attested, that is rad-ost-n-ost’ ‘being joyful’, with a primary noun rad-ost’, and žizn-e-rad-ost-n-ost ‘being joyful/optimist’ (incorporating žizn’ ‘life’), with no such word as *žizn-e-radost’. The model without -k- in the primary adjective is exemplified also by gad-ost-nost’ ‘filthiness’ ← gad-ost-n-yj ‘filthy’ (practically semantically the same as gadkij) ← gad-ost’ ‘filthy thing’ ← gad-k-ij ‘filthy’, its synonym merz-ost-n-ost’ ← merz-ost-n-yj

977

978

V. Rules and restrictions in word-formation II: Special cases ← merz-ost’ ← merz-k-ij, among many others. Yet another type is instantiated by poverx-n-ost-n-ost’ ‘superficiality’ from po-verx-n-ost-nyj ‘superficial’ and po-verx-n-ost’ ‘surface’ without a simple adjective in Modern Russian (*poverxnyj). So, different links of this quadripartite chain can be missing. Both in Russian and Bulgarian -ost’/-ost can also be followed by the suffix -nik signifyinghumanbeings.InBulgariantheyarepejorativenounslikexub-ost-nik‘rascal’(Manova 2011) formed, with a considerable semantic shift, from xub-ost ‘beauty’. Russian nouns in -ost-nik are not productive and rather marginal, often with little or no synchronic relation to the underlying adjectives, like krep-ost-nik ‘partisan of serfdom’ (krepost’ no longer exists in the sense of ‘serfdom’, the meaning of which is expressed by an adjective phrase: krepostnoe pravo). There are also some neologisms from the 1990s found in the RNC and formed from abstract nouns in -ost’ signifying ‘a professional who works with X-ost’’ like nov-ost-nik ‘news journalist’ from nov-ost’ ‘(piece of) news’, ličn-ost-nik ‘psychologist who studies personality’ from ličn-ost’ ‘personality’; the Soviet-era term skor-ost-nik ‘a professional worker remarkable for his speed of performance’ from skor-ost’ ‘speed’ is also close to this semantic type. It is worth mentioning that the pejorative nouns in -ost-nik like derz-ost-nik ‘insolent man’ or gad-ost-nik ‘wretch’ belong to an archaic Church Slavonic style (they are attested in the RNC in the works of the 19th century writer Leskov in the speech of priests), and so they can have common roots with the Bulgarian pattern. The sole commonly used noun of this kind, pakost-nik ‘rascal, some who plays little mean tricks’ from pakost’ ‘mean trick’, already has no suffix -ost’ synchronically (due to the lack of a corresponding adjective).

3.4. Relational/qualitative adjectives In Aronoff and Fuhrhop’s material, German -isch forming relational and/or qualitative adjectives from noun stems, is a closing suffix: Hund ‘dog’ → hünd-isch ‘doggish’. Russian relational -sk-(ij), a homologue of German closing -isch, is not a totally closing suffix, unlike -sk-(i) in Bulgarian analyzed by Manova (2008, 2011). A -sk-ost’pattern exists combining the adjective suffix with the suffix of secondary nominalization. The pattern yields nouns signifying ‘the fact of having some properties of X’ as svet-skost’ ‘being of high society’ or ‘secularism’ (according to the two meanings of svet-sk-ij stemming from two different meanings of svet ‘world’), det-sk-ost’ ‘childishness’, russk-ost’ ‘Russianness’, evropej-sk-ost’ ‘European style’ (dozens of stems in the RNC). About 1860 the model -sk-ost’ became productive in Russian. Before this date only two words belonging to this pattern − svet-sk-ost’ and obsolete ljud-sk-ost’ − are attested in the RNC. Bulgarian probably reflects an earlier situation in Slavic languages when the suffix -sk- was closing. Surnames are a separate subsystem of morphology, often reflecting an earlier stage of development. In Russian surnames (ending in -skij and obeying the adjectival declension), the suffix -sk- is generally closing. This creates some gaps in derivational paradigms. For example, Russian surnames without -skij (as Puškin) usually do allow adding of -skij to form an adjective: Puškinskie čtenija ‘Pushkin conference’. This is not generally the case with surnames in -skij like Dostoevskij that categorically disallow reduplication, even following an additional linking morpheme (*Dostoevskij-skij or *Dostoevskov-skij) and resist a single suffix with an adjectival reduplication: ?Dostoevskie čtenija.

55. Closing suffix patterns in Russian Note, however, some invented formations with no surface redundancy: Puškin-sk-aja (name of a Metro station) but also Dzeržin-sk-aja < Dzeržin-sk-ij and Majakov-sk-aja < Majakov-sk-ij (stations since the 1930s) − normally, both would be female surnames. They can be analyzed as cases of haplology (that is, an omission of one of the identical affixes or phonological segments on the surface), but another type of analysis seems more plausible. In adjectives from city names like kur-sk-ij ‘of Kursk’, according to a recent comprehensive treatment of Russian morphophonemics (Itkin 2007: 254), there seems to be no overlap of the suffixes but just absence of the first -sk- (cf. kur-jan-in ‘a person from Kursk’). The same is probably true here (cf. dostoev-ščina ‘Dostoevskystyle passions’ where the pejorative -ščina is added directly to the stem). So the closing status of -sk- in surnames persists. The only exceptions are rare surnames in -sk-ov (mainly names of the Don Cossacks), where -sk- is not a closing suffix although diachronically the same as in -skij; -ov- “reopens”, in Aronoff’s and Fuhrhop’s terminology, the form for further suffixes (galansk-ov-sk-ij ‘belonging to [a Soviet dissident] Galanskov’, RNC). The closing status of -sk-ij in surnames is incidentally violated in the invented word dostoevskij-mo ‘≈ Dostoevskyness’ with -mo as in pis’-mo ‘letter’, coined by the Russian Futurist poet Velimir Khlebnikov in a poem of 1908. Another Russian suffix seems to be a better candidate for a Russian morphological equivalent to the relational -isch in German than its Bulgarian cognate -ski, as it is closing without exceptions. This is the case of the Russian relational suffix -in which is productive with words in -a/-ja yielding attributive forms of a mixed declension (papin ‘daddy’s’ from pap-a, plemjannic-yn ‘niece’s’ from plemjannic-a, Petin ‘Pete’s’ from Petj-a) and also with some words ending in a soft consonant, when it yields long adjective forms (golub-in-yj from golub’ ‘pigeon’s’, sobolinyj from sobol’ ‘sable’s’).

3.5. Nominalizations In German, -ung (like Prüf-ung ‘examination’ from prüf-en ‘to examine’) yielding nominalizations is a closing suffix. Bulgarian, according to Manova (2011), has two semantically distinct suffixes: -(V)n-e (pisane ‘writing’) and -(V)ni-e (pisanie ‘a piece of writing’). The latter allows for a -cdiminutive − pisanijce, whereas the former is a closing suffix. In Russian, the situation basically corresponds to what occurs in Bulgarian, but the facts are slightly more complicated, as the two cognates of -(V)n-e and -(V)ni-e do not display a clear semantic and formal opposition but either intermix or lexicalize completely. There exist two deverbal nominalizing suffixes in Russian that largely overlap in function: a) -(V)ni- as in posl-ani-e ‘message’ from posl-a-t’ ‘to send’, predstavl-eni-e ‘presentation’ from predstav-i-t’ ‘to present’. Upon first glance it seems that it is not closing, losing -i- before nominalizing -ec (like upravl-en-ec ‘manager’ from upravl-en-ie ‘management’), but see below; b) -(V)n’j- under the same circumstances as in posl-an’[j]-e ‘message’, predstavlen’[j]e ‘presentation’. Typically, each word with -(V)nie potentially has a -(V)n’je counterpart that is used, for instance, in poetry metri causa or in some highly individual prose style (characteristic,

979

980

V. Rules and restrictions in word-formation II: Special cases e.g., for Pasternak). They have the same etymology (diachronically the second is a phonological reduction of the first), so their co-existence in the language offers a kind of “layering” according to the theoretical assumptions of grammaticalization proposed by Hopper (1991). Phonetically, in allegro speech, the two are often indistinguishable, due to a proximity between a reduced unstressed [i] and the semivowel [j]. From a semantic point of view, they also largely coincide. Both suffixes form action nouns (sočine-nie/-n’je knigi ‘composing a book’) and object/result nouns depending on the situation expressed by the verb (škol’noe sočinenie/-n’je ‘composition/essay as a school task’). However, in some words -(V)n’j- is lexicalized, marking a distinction between action nouns and concrete nouns: varen’je ‘jam’ vs. varenie ‘cooking’ from varit’ ‘to cook’, pečen’je ‘cookies’ vs. pečenie ‘baking’ from peč’ ‘to bake’. This distinction is directly related to the problem of closing status and further derivation, as only the (concrete nouns) take diminutive -ic-: varen’jice ‘little jam’, but not the latter, which is clearly related to semantic constraints (diminutives from action nouns are less plausible semantically). Moreover, in words where there is no opposition, the diminutive preferably refers to a concrete noun and not to an action noun and takes obligatorily -(V)n’j- and not -(V)nor -(V)ni-: poslanie − poslan’jice, *poslani(i)ce ‘message’, predstavlenie − predstavlen’jice, *predstavleni(i)ce ‘performance’. But the most interesting aspect is that the same holds for all Russian nouns ending in -i-e and -’j-e, not only after -n- and not only formed from verbs. We may take a phrase illustrating Pasternak’s -’je-style from his autobiography “Oxrannaja gramota” [Safe conduct]: Zaglav’je skryvalo genial’no prostoe otkryt’je ‘The title concealed a genially simple discovery’. Both -’je-words in the phrase correspond to Standard Russian nouns in -ie that have nothing to do with the -(V)n- pattern: otkrytie ‘discovery’ ← otkry-t-yj ‘open’ (with another participial suffix), zaglavie ‘title’ without a corresponding verb. In the web we have found a handful of examples both of zaglav’j-ic-e ‘little title’ and otkryt’j-ic-e ‘little discovery’ showing that both (but not zaglavie and otkrytie without reduction) can add -ic-. We propose the following morphological analysis for nouns in -(V)nie and -(V)n’je: -(V)n- can be considered an affix used in both cases, whereas the full and reduced forms of -i- and -j- are separated from it by a morpheme boundary (and they are also used in words like zaglav-i-e/zaglav’-j-e and otkryt-i-e/otkryt’-j-e). In this case their status as closing affixes is the following: a) -(V)n- is not closing, as upravl-en-i-e ‘management’ yields upravl-en-ec ‘manager’, upravl-en-č-esk-ij ‘managerial’, etc.; b) -j- is not closing as it allows for diminutives both in nominalizations and in other nouns (upravlen’-j-ic-e, zaglav’-j-ic-e); c) -i- is closing, allowing no further derivation. The proposed decision is not unlike the situation in Bulgarian, where -ne is closing and -nie is not. The two languages have developed from the same diachronic source, different suffixes for action nouns and concrete nouns and the former suffixes have become closing, obeying the same semantic constraints as German -ung. Note, however, that in Bulgarian the action noun closing suffix (-ne) is phonologically simpler and more reduced as compared to the concrete noun suffix which is not closing; in Russian the situation is inverse.

55. Closing suffix patterns in Russian

981

3.6. Varia (-ling, -izm) This section offers brief remarks about two other closing suffixes that do not show parallels among the three languages. Aronoff and Fuhrhop (2002) consider German -ling to be closing, e.g., lehr-en ‘to teach’ → Lehr-ling ‘one who is taught’ → *Lehr-ling-in ‘a woman who is taught’, although there are numerous counterexamples, especially on the Internet, cf. Manova (2011: 289): Prüflingin ‘examinee-FEM’, Lehrlingin ‘apprentice-FEM’, Häftlingin ‘prisoner-FEM’). It has no equivalent either in Bulgarian or in Russian, showing that this type of derivation is uncommon in Slavic which prefers inflectional passives (cf. Engl. -ee, which is productive, closing and borrowed from a French passive participle suffix). A borrowed suffix that is closing in Russian is -izm belonging to the Greek-based handful of international affixes, cf. -ismo in Italian which is also closing. (Relational adjectives in -ičesk-ij are derived from personal nouns in -ist, e.g., Russ. turističeskij ← turist ‘tourist’, but semantically they can refer to abstract nouns, e.g., turizm ‘tourism’ as well.) A dubious counterexample mexanizm-ik ‘a little piece of machinery’ formed from mexanizm ‘machinery’ is found in the RNC. Adjectives like kommunizmennyj are sometimes attested in newspapers and the Internet as puns. (In this case, however, the alleged suffixation goes back to a portmanteau of kommunizm ‘communism’ and nizmennyj ‘vile’.)

4. Conclusions A summary juxtaposing the German closing suffixes with the Bulgarian and Russian ones can be found in Table 55.1. Tab. 55.1: Closing suffixes in German, Bulgarian, and Russian German (Aronoff and Fuhrhop 2002)

Bulgarian (Manova 2008, 2011)

Russian

-e

NONE

zero − (closing but elusive) -k- − not closing

-heit/-keit/-igkeit

-ost (Bulg.) / -ost’ (Russ.) allows for reduplication and adjectives, not closing

-in

-k-a, -(k)inj-a, -ic-a, -es-a, -is-a, -v-a closing only with native stem denoting humans

-nic-a, -inj-a, -ic-a, -ess-a, -š-a closing save the relational adjectives (with -nic- productivity is shrinking), diminutives impossible, no human “constraint”

-isch

-ski (closing)

-sk-ij (not closing) -in (closing)

-ling

NONE

-ung

-(V)n-e/-(V)ni-e (Bulg.) / -(V)ni-e/-(V)nj-e (Russ.) The first suffix signifies action nouns and is closing, the second is not closing.

982

V. Rules and restrictions in word-formation II: Special cases The following general conclusions can be drawn. Russian predictably demonstrates semantic patterns of closing suffixes more similar to the Bulgarian suffixes than to the German ones. However, in both Slavic languages there may exist suffixes of the same semantic field that are both closing but still non-cognate (whereas cognates behave differently). Evidence is also presented that the “closingness” of a suffix, depending on its semantics and the combinatory force of other suffixes, is not a discrete feature but can rather be described in terms of principal trends that have exceptions and change diachronically (for example, the status of -in). The use of a corpus enhances detecting rare/marginal models and tracing diachronic ways of developing closing patterns (and even some recent divergence between languages in this point).

Acknowledgements The article was written with the support of the Russian Academy of Sciences (Research programme Corpus Linguistics − Корпусная лингвистика), the Russian Foundation of Humanities (РГНФ 10-04-00256а), and the Humboldt Foundation (Research Fellowship for Postdoctoral Researchers at Würzburg University, 2010).

5. References Aronoff, Mark and Nanna Fuhrhop 2002 Restricting suffix combinations in German and English: Closing suffixes and the monosuffix constraint. Natural Language and Linguistic Theory 20: 451−490. Corbett, Greville G. 1995 Slavonic’s closest approach to Suffixaufnahme: The possessive adjective. In: Frans Plank (ed.), Double case. Agreement by Suffixaufnahme, 265−282. Oxford: Oxford University Press. Hopper, Paul J. 1991 On some principles of grammaticalization. In: Elizabeth Closs Traugott and Bernd Heine (eds.), Approaches to Grammaticalization. Vol. 1, 17−36. Amsterdam/Philadelphia: Benjamins. Itkin, Il’ja 2007 Russkaja morfonologija. Moskva: Gnozis. Kuznecova, Ariadna and Tat’jana Efremova 1986 Slovar’ morfem russkogo jazyka. Moskva: Russkij jazyk. Manova, Stela 2008 Closing suffixes and the structure of the Slavic word: Movierung. Wiener Slavistisches Jahrbuch 54: 91−104. Manova, Stela 2011 Closing suffixes in Bulgarian, Russian and German: The role of semantics. In: Mónika Farkas Baráthi (ed.), Bulgarian Language and Literature in Slavic and Non-Slavic Contexts, 286−292. Szeged: JATE Press. Nida, Eugene 1949 Morphology. Ann Arbor: The University of Michigan Press. Russian National Corpus online http://ruscorpora.ru

55. Closing suffix patterns in Russian

983

Tagabileva, Maria, Elena Grishina, Ilya Itkin and Olga Lyashevskaya 2009 O zadačax i metodax slovoobrazovatel’noj razmetki v korpuse tekstov. Poljarnyj vestnik. Reports from Tromsø University Department of Russian 12: 5−25. < (http://munin.uit.no/ bitstream/handle/10037/3167/article.pdf?sequence=1> [last access 8 Dec 2014].

Dmitri Sitchinava, Moscow (Russia)

VI. Semantics and pragmatics in word-formation I: General aspects 56. Motivation, compositionality, idiomatization 1. 2. 3. 4.

Introduction and preliminary definitions Motivation, compositionality, idiomatization and word-formation Summary: The importance of human perception References

Abstract “Motivation”, “compositionality” and “idiomatization” are closely connected notions that are sometimes used as synonyms (motivation and compositionality) or antonyms (motivation and compositionality vs. idiomatization). The present article provides a short overview of the recent history of research on these phenomena, discusses the central questions and problems linked to them and places them in the context of word-formation research.

1. Introduction and preliminary definitions English preacher (cf. Ullmann 1962: 91) is generally believed to be a motivated word for essentially two reasons: First of all, because its form most obviously consists of the verbal element preach and the agent suffix -er; second, because every speaker of English (including L2 learners) who is familiar with the meaning of both preach and -er is also able to infer the semantics of preacher from the semantics of its parts when hearing it for the first time. The word preacher can thus easily be understood as ‘someone who preaches’ (cf. also speaker, reader, singer, thinker in Ullmann 1962: 91). From this perspective, English deverbal agent nouns in -er can also be regarded as “compositional” (cf. Barz 1982: 8). Still, in some cases, compositionality seems to be less straightforward than in the examples cited above. Barz (1982: 18, following Uluchanov 1977) observes that German Zuschauer ‘spectator’ (verbal stem zuschau- ‘to watch’ + agent suffix -er) and Bäcker ‘baker’ (verbal stem back- + agent suffix -er) are not compositional to the same degree: While a Zuschauer is just any person carrying out the action of zuschauen, a Bäcker typically carries out the action of backen professionally. In her view, instances such as Bäcker are still motivated, but slightly idiomatized, because their sense deviates from the “expected” compositional sense. In other cases idiomatization seems to be even stronger than in the case of G. Bäcker. Though the form of G. Bauer ‘farmer’, e.g., is composed of the verbal stem bau- ‘to construct’ and, again, the agent suffix -er, the meaning of Bauer is related more closely to G. anbauen ‘to grow something’ than to bauen ‘to construct’ (cf. Marzo 2014).

56. Motivation, compositionality, idiomatization The main purpose of this article is to provide an overview of the recent history of research on the phenomena of motivation, compositionality and idiomatization as sketched above (as well as related notions such as iconicity, transparency and lexicalization) and to discuss them in the context of word-formation (cf. section 2). Special attention will be given to the role of human perception as a driving force behind all issues related to motivational phenomena (cf. section 3).

2. Motivation, compositionality, idiomatization and word-formation 2.1. Do the formal properties of words correspond to what they express? In the various shapes it may take, this question is at the core of any linguistic semiotic reflection. At the latest since Plato’s Cratylus (cf. Plato 1996), throughout centuries and across philosophical traditions (cf. the contributions in Simone 1990; Coseriu 2004), two extreme positions on this issue have been discussed: (i) the form of words corresponds “by nature” to what they express (naturalist position) and are therefore motivated; (ii) the form of words is purely conventional and does not have anything to do with what they express (conventionalist position), which is why words are arbitrary. This article focuses on the intermediate position that has, in modern times, been formulated by Saussure (1972 [1916]: 182−183) and researched by modern lexicologists and morphologists of different theoretical backgrounds (cf. sections 2.2−2.3): Linguistic signs are only “relatively motivated” (Saussure 1972 [1916]: 182; cf. section 2.2.1), as are the agent nouns in section 1. In the following subsections I will first discuss some approaches to motivation in the lexicon (2.2) as well as its relation to iconicity (2.3), and then take a closer look at how motivation relates to the notions of compositionality and idiomatization (2.4).

2.2. Cornerstones in lexical motivation research 2.2.1. Saussure and the notion of “relative motivation” Though linguistic signs are, in Saussure’s opinion (1972 [1916]: 100), fundamentally arbitrary, he argues that every language has at least some relatively motivated lexemes (1972 [1916]: 182−183). Even if onomatopoeic words and exclamations tend to play a relatively marginal role in a language’s lexicon, they could − at first sight − be a valid objection to the principle of arbitrariness (1972 [1916]: 100−102). However, the simple fact that they differ across languages demonstrates that they are only an approximate imitation of the concept they designate, meaning that they are at least partly conventional and thus only relatively motivated (cf. Fr. ouaoua vs. G. wauwau ‘bow-wow; sound of a dog’s barking’; 1972 [1916]: 102). The same is true for linguistic signs that can be analysed syntagmatically and are paradigmatically related to other signs, such as Fr. poirier ‘pear-tree’, that evokes not only the morphologically simple word poire ‘pear’,

985

986

Semantics and pragmatics in word-formation I: General aspects but whose suffix -ier calls to mind paradigmatically related words such as pommier ‘apple-tree’, cerisier ‘cherry-tree’, etc. (1972 [1916]: 181). Since the morphologically simple parts of which motivated words consist are themselves still fundamentally arbitrary, the latter can only be considered as relatively motivated (for a discussion of different interpretations of Saussure’s notion of arbitrariness cf. Marzo 2013: 33−34 and the literature cited therein).

2.2.2. Ullmann’s “types of motivation” Ullmann, who terms arbitrary words “opaque” and motivated words “transparent” (1962: 80−115), distinguishes three types of motivation, i.e. phonetic, morphological and semantic motivation. Phonetic motivation concerns a direct relation between the form of a word and its meaning (cf. also section 2.3) in the sense that the meaning of the sign somehow motivates its form. This type of motivation is most typically represented by onomatopoeia (for the distinction of different subtypes of onomatopoeia, cf. Ullmann 1962: 84). Morphological motivation concerns, in Ullmann’s view (1962: 91), morphologically complex words as the English agent nouns in -er or compounds such as penholder or penknife. Semantic motivation, in turn, is primarily characterised by metaphors and metonymies and mainly concerns polysemous words, such as E. bonnet, which is not only a special type of headdress, but also the cover of a car’s engine (cf. Ullmann 1962: 91−92; on semantic motivation, cf. also Bally 1965 [1932]: 137−139 and the critique in Scheidegger 1981; Fill 1980b; Augst 1996; Geeraerts 2003). Although Ullmann distinguishes these three types of motivation, especially his discussion of compounds suggests that words are not necessarily motivated in one way only, but that morphological and semantic motivation often come along in combination (cf. 1962: 92): E. blue-bell is not only motivated morphologically (decomposable in blue and bell), but also semantically (i.e. metaphorically), because the flower actually resembles a blue bell. He thus foreshadows what most motivational and word-formation researchers after him will agree upon: the principled interaction between morphological and semantic motivation (cf. Sauvageot 1964: 57; Marchand 1969: 2; Gauger 1971; Rufener 1971; Bartoszewicz 1974: 71−73; Shaw 1979; Fill 1980b; Rettig 1981; Bellmann 1988; Swanepoel 1992: 52; Rainer 1993: 16; Hiraga 1994; Gruaz 2002: 700; Radden and Panther 2004; Panther and Radden 2011a).

2.2.3. Koch’s “dimensions of motivation” Koch (2001; cf. also Koch and Marzo 2007) eventually systematizes the interaction between morphological and semantic motivation from a cognitive perspective and shows that they are not two types of motivation that can occasionally combine, but that they are two “dimensions” of non-onomatopoetic lexical motivation. Consequently, they cooccur systematically: “a lexical item L1 […] expressing a concept C1 is motivated with respect to a lexical item L2, if there is a cognitively relevant relation between C1 and C2, paralleled by a recognizable formal relation between the signifiants of L1 and L2” (Koch 2001: 1156). The “cognitively relevant relations” are the seven universal conceptual

56. Motivation, compositionality, idiomatization relations “conceptual identity”, “contiguity”, “metaphorical similarity”, “taxonomic similarity”, “taxonomic superordination” and “subordination” as well as “co-taxonomic contrast” (cf. Koch 2001: 1158−1159). The formal dimension is language-specific and manifests itself not only in a language’s set of word-formation devices (e.g., affixation, composition, reduplication, etc.), but comprises also other types of complex lexical units, such as idioms, lexicalized syntagms, etc. (cf. Koch 2001: 1159−1161). What is most original in this approach is the inclusion of polysemy as a “formal” motivational device (via “formal identity”; cf. Koch 2001: 1157−1159): In this respect, not only English deverbal agent nouns such as preacher, etc. are motivated with respect to their verbal base (here preach; recognizable formal relation: suffixation; cognitively relevant relation: contiguity), but also morphologically simple lexical units (i.e. units consisting of one form and one meaning; Cruse 1986: 49, 80) such as E. bonnet ‘cover of a car’s engine’ are motivated with respect to other lexical units (here bonnet ‘special type of headdress’; recognizable formal relation: formal identity; cognitively relevant relation: metaphorical similarity; for an extensive analysis of the role of polysemy in lexical motivation, cf. Marzo 2008; 2011; 2013).

2.2.4. Rettig’s notion of “motivatability” The fact that Koch (2001: 1156) labels formal relations as “recognizable formal relations” hints at the importance of the language users’ perception of lexical motivation (cf. also section 3). As a matter of fact, word-formation and motivation researchers from different theoretical backgrounds agree upon the dependence of motivation on human perception and do not necessarily consider it only a static relation, but also a mental process (cf. Gauger 1971: 9, 45; Ernst 1981: 68; Rainer 1993: 16−22; Monneret 2003: 49; Ungerer 1991: 161−162). Willems (2005: 266) even claims that Saussure already considered motivation as an intentional phenomenon (cf., similarly, Radden and Panther 2004: 1). A very explicit stance on this issue is expressed by Rettig (1981: 75; 153− 156). In his view, words such as Fr. poirier ‘pear-tree’, are not motivated “per se”, but only “motivatable”, even if they have been formed by productive word-formation rules, meaning that motivation always depends on a conscious act of (metalinguistic) reflection. However, even if the importance of metalinguistic reflection for motivation is nowadays generally acknowledged, especially psycholinguistically oriented research on lexical motivation (and, more generally, on the structure of and the access to the mental lexicon) distinguishes two aspects of motivation, i.e. (conscious) “motivatability” on the one hand and (unconscious) “motivatedness” on the other (for a more detailed discussion of this distinction and an overview of relevant research, cf. Rainer 1993: 17−22; Marzo 2013: 45−48).

2.2.5. “Consociation” and “dissociation” according to Leisi Leisi (1955) does not study lexical motivation explicitly. However, his definitions of the notions of consociation and dissociation are absolutely relevant for lexical motivation research. Words are “consociated”, if they are formally and semantically related to other

987

988

Semantics and pragmatics in word-formation I: General aspects words in a language’s lexicon, such as G. mündlich ‘oral’ to Mund ‘mouth’. They are “dissociated”, if they do not have such a relation to any other lexeme: E. oral, e.g., is only semantically, but not formally related to mouth; the direct formal relation to the conjunction or, in turn, is of no semantic relevance. “Consociation” thus corresponds to motivation and transparency, whereas “dissociation” equals arbitrariness and opacity. What is interesting and has had a fruitful impact on modern motivation research (e.g., Sanchez 2008; Umbreit 2013) is the fact that from Leisi’s perspective mündlich is as consociated with its derivational basis Mund as Mund is with its derivative mündlich, i.e. that the relation has no preferred direction. Especially from the perspective of speaker-dependent motivatability (cf. Rettig 1981: 171 on the principled non-directedness of motivation) and, on a more general level, from the perspective of cognitive network approaches to the lexicon (cf. the overviews in Umbreit 2010; 2011 and article 10 on word-formation in cognitive grammar), there is no reason to assume that native speakers always motivate formally more complex words with respect to less complex words, though there often are − under certain circumstances − directionality preferences (cf. Umbreit 2013 on an extensive study of this issue in French and Italian). Consider, e.g., the relation between toponyms and the corresponding relational adjectives in Italian discussed by Rainer (2004b: 405): In the majority of cases the adjective is derived from the noun by suffixation as in tirol-ese (‘Tyrolese’ ← Tirolo ‘Tyrol’). However, toponyms containing the suffix -ia (such as Lombardia ‘Lombardia’) are diachronically derived from an adjective (here lombardo ‘lombardic’). Rainer (2004b: 405) argues that despite the diachronic suffixation relation native speakers are − from a strictly synchronic perspective − more likely to motivate the adjective on the basis of the noun than vice versa because they perceive the morphological relation between toponyms and their adjectives as a uniform category.

2.2.6. Radden and Panther’s “language-independent factors of motivation” Although this article focuses, from the perspective of word-formation research, on motivational relations between lexemes on the one hand and between lexemes and their parts on the other, and does not systematically take into account extra-linguistic sources of motivation (for a discussion of these, cf. the contributions in Radden and Panther 2004 and Panther and Radden 2011b), Radden and Panther’s theory of motivation deserves a closer look, because it is directly relevant for any non-compositional understanding of motivation (for compositionality, cf. section 2.4). Radden and Panther’s goal is to find a common ground for rather diverging approaches to ecological, genetic, perceptual, experiential and cognitive motivation that embrace motivational phenomena in grammar as well as in the lexicon (cf. Radden and Panther 2004: 34−42; Panther and Radden 2011a: 13−24; for an overview of other authors who discuss these phenomena, cf. Ungerer 2002: 379). According to Panther and Radden (2011a: 9) “a linguistic sign (target) is motivated to the extent that some of its properties are shaped by a linguistic or nonlinguistic source and language-independent factors”. In their view, the linguistic source can be a whole sign, its content, or its form. Language-independent factors that contribute to motivation are, e.g., experience, perceptual gestalt principles, salience, economy, etc. (cf. Radden and Panther 2004: 3, 8, 23−32). These factors might lead to differences

56. Motivation, compositionality, idiomatization across languages, as they show in their analysis of the expression of the concept SCREWin different European languages: The tool itself is associated to a complex idealized cognitive model (ICM) that contains representations of actions that are typically carried out with this specific tool, objects involved in the actions such as the tool itself, etc. (cf. Radden and Panther 2004: 5−8). If languages have a transparent designation for the tool, we can observe that they do not encode exactly the same elements of the ICM in the designations for the tool, though cross-linguistically there are still considerable similarities: While It. cacciavite literally means ‘screwsticker/screwpuller’ (cacciare ‘to stick something in, to pull something out’) and Fr. tournevis ‘screwturner’ (tourner ‘to turn’; cf. also It. giravite), G. Schraubenzieher focuses on the action of pulling (‘screwpuller’). All these languages − as well as many others taken into account by Radden and Panther − chose, among the objects represented in the ICM, the same and most salient object as a component of the designation: the screw. In order to account for the slight differences between the verbal parts of the compounds, Radden and Panther (2004: 7) argue that “none of the specific actions performed with a screwdriver stands out as particularly salient so that each of the actions is equally appropriate to stand metonymically for the whole range of actions”.

DRIVER

2.3. Motivation and iconicity The phenomenon of “motivation” is, of course, closely connected to the notion of “iconicity”, though the focus of interest of motivation and iconicity research is not exactly the same: In Saussure’s spirit, motivation research generally concentrates on relations between words and on relations between words and their parts, whereas the direct relation between a word form and the concept it expresses traditionally plays a minor role (but cf. section 2.2.6 as well as research on phonetic motivation, e.g., Lu 1998). In contrast, linguistic iconicity research mainly studies − on a more general level and inspired by Peirce (1960) − the direct relation between linguistic forms of any kind and size (including, e.g., phonaesthemes, morphemes, words as well as sentences) and the concepts they express or are associated with (cf. Pusch 2001: 371). By contrast, the relation between different linguistic forms seems to be of lesser interest (but cf. below for Hiraga’s 1994 “relational diagrams”). The most obvious parallel between motivation and iconicity research being the joint interest in onomatopoeia (cf. also Peirce’s “images”; 1960: 157), there are other points of contact, viz. in the realm of “diagrammatic” iconicity (cf. also Ungerer 2002: 376− 379). Peirce (1960: 157) defines “diagrams” as “those [icons; D. M.], which represent the relations, mainly dyadic, or so regarded, of the parts of one thing by analogous relations in their own parts”. With regard to linguistic signs, this definition has two important implications: First, the structure of a linguistic form reflects the structure of what it expresses (cf. also Pusch 2001: 373; Ungerer 2002: 374−375). Consequently, and in contrast to onomatopoeia, the form of a diagram does not necessarily resemble the concept it expresses phonologically, though there is a certain structural resemblance. Second, as Haiman (1980: 515) argues, in an ideal diagram every distinct component of the represented “reality” should, in addition, correspond to one distinct component of the form and vice versa (cf. the “isomorphism” principle). Hiraga (1994) complements

989

990

Semantics and pragmatics in word-formation I: General aspects this structural perspective on the direct relation between form and meaning of diagrammatic linguistic signs by an additional relational perspective on the link between different linguistic signs by distinguishing “structural” diagrams from “relational” diagrams: The latter are characterized by the fact that they not only represent their own content (as the former do, cf. examples below), but reflect, in addition, to a certain extent also the meaning of other formally and semantically related signs, converging thus with the perspective of most motivation research (cf. sections 2.2.2−2.2.6). A notorious syntactic example for structural iconicity is Caesar’s sentence veni − vidi − vici that has been analyzed by Jakobson (1971: 351) as an instance of sequential iconicity, because the “temporal order of speech events tends to mirror the order of narrated events in time or rank” (cf. the overviews in Ungerer 2002: 374−375 and Van Langendonck 2007: 405−413 for other types of structural iconicity in syntax). In this example, the form of the sentence not only reflects the relations between the events by mirroring their temporal order, but meets, in addition, the principle of isomorphism in that the number of distinct events is signaled by the number of words in the sentence. In morphology, diagrammatic iconicity is often exemplified by inflection, especially by pluralization: “The signans of the plural tends to echo the meaning of a numeral increment by an increased length of the form” (Jakobson 1971: 352; cf. also the overview in Van Langendonck 2007: 403−405). Spanish and English nouns, e.g., are usually pluralized by the addition of an -s, like in Sp./E. perro/dog vs. perros/dogs (cf. also the term “constructional iconism” in natural morphology as, e.g., in Mayerthaler 1981: 23−24 and article 9 on word-formation in natural morphology, esp. section 5.1 iconicity). From both the structural and the relational perspective, diagrammatic iconicity also plays a role in word-formation research: Compounds of the kind G. Apfelbaum ‘appletree’ have been studied from the perspective of the isomorphism principle (cf. Ungerer 1999; 2002). Ungerer (1999: 312; 2002: 377) argues that even compounds that aren’t at first sight entirely compositional because they mean more than just the sum of their parts (e.g., G. Apfelsaft *‘apple and juice’, but ‘juice made out of apples’), are structurally iconic and isomorphic because the fusion of the two words to one word reflects the fusion of two concepts to one concept only, that is no longer perceived as having different subcomponents (cf. also E. wheelchair *‘wheel and chair’, *‘chair with wheels’, but ‘special chair with relatively large wheels used by people who cannot or should not walk’; Ungerer 1999: 310−311). Another example is Koch’s discussion (cf. Koch 1999) of words for trees and their respective fruits in typologically different languages. Koch (1999: 334) argues that in many languages the morphological relation between the words for trees and those for the corresponding fruits iconically mirrors the importance we attribute to the respective trees and fruits in our daily experience: If the tree is more important in our everyday life than its fruit, the word for the latter is constructed on the basis of the word for the former. In cases such as G. Eiche ‘oak tree’ vs. Eichel ‘acorn’, e.g., the timber is more important than the fruit, because it is a popular construction material, whereas the fruit is of no particular interest. On the contrary, if the fruit is more relevant to us than the tree, this is paralleled by the fact that the word for the tree is constructed on the basis of the word for the fruit, as in the case of G. Apfel ‘apple’ vs. Apfelbaum ‘apple-tree’. In this case, the fruit is a very common comestible and as such much more important than the tree that has, in turn, no particular significance in our everyday experience (cf. similarly Fr. poirier ‘pear-tree’ in section 2.2.1; cf. also Sp. bellota ‘acorn’, a very

56. Motivation, compositionality, idiomatization popular feed for animals in general and pigs in particular: the word is opaque with respect to, e.g., roble ‘oak’ and encina ‘durmast oak’, but has itself given rise to belloto designing a Chilean tree whose nut-like fruit is also used as animal feed, though belloto cannot be regarded, of course, as morphologically more complex than bellota). From the structural iconicity point of view, the primacy of the more salient concepts is thus paralleled by the “primacy” of their morphologically simple(r) forms with respect to the morphologically more complex forms that designate the less salient concepts. Though Koch explicitly speaks of iconicity in this case, the parallel to his motivation model (cf. section 2.2.3) is self-evident: From a relational perspective, the morphologically more complex form not only expresses its own content, but also points to a more salient concept by containing the latter’s linguistic form.

2.4. The compositionality hypothesis: motivation vs. idiomatization? 2.4.1. The compositionality hypothesis: a brief overview of the state of the art Compositionality is nowadays most commonly defined as the principle that “the meaning of every syntactically complex expression of a language (save, maybe, for idioms) be determined by the meanings of its syntactic parts and the way they are put together” (cf. Hinzen, Werning and Machery 2012: 1; similarly Gayral, Kayser and Lévy 2005: 83). However, while relative motivation is nowadays generally considered to be a universally valid linguistic principle (cf. especially sections 2.2−2.3), the status of compositionality is much less agreed upon: Depending on the theoretical framework, it can have (among others) the status of an “a priori principle” (cf. the discussions in Cohnitz 2005; Klos 2011: 44−52), a “natural law of semantics” or simply “an analytic issue” (for both cf. the discussion in Peregrin 2005). Accordingly, its origins as well as the reasons in favour of it are still controversial: While many authors attribute its first formulation to Frege (cf. Bartsch 2002: 571; Hinzen, Werning and Machery 2012: 1), others profoundly question Frege’s interest in and his commitment to the principle (cf. Klos 2011: 69; Cohnitz 2005: 23−25). In his review of “right and wrong reasons for compositionality”, Werning (2005) shows in addition that most reasons commonly cited in favour of compositionality are problematic from various respects. He thus concludes: “The prospects of a justification of compositionality are not entirely bleak, but less than comfortable” (Werning 2005: 307). In a complementary perspective, Zimmermann studies the problems semantic approaches that rely on compositionality might encounter (Zimmermann 2012: 86−93) and carves out some strategies of how to solve them (Zimmermann 2012: 93−106). From what has been said so far, it should be clear that compositionality is not understood in exactly the same way across research areas and theoretical frameworks. In effect, Pelletier (2012: 149−153) shows that researchers have so far committed to at least two diverging types of compositionality: While the “building block version” of compositionality says that “a whole in a compositional system is built up solely from the materials of its parts” (Pelletier 2012: 150), “the functional version of compositionality” defines the “µ of a whole” as “a function of the µs of its parts and the ways those parts are combined” (Pelletier 2012: 151). In the following sections I will not present in detail the manifold approaches that rely on these and other definitions, but concentrate on their implications

991

992

Semantics and pragmatics in word-formation I: General aspects for motivation in word-formation (for a more detailed overview, cf. the contributions in Werning, Hinzen and Machery 2012; Werning, Machery and Schurz 2005; Machery, Werning and Schurz 2005; for their significance for specific theoretical frameworks, cf. also Mudersbach 2002; Dupuy-Engelhardt 2002; Goddard 2002).

2.4.2. Compositionality, motivation and idiomatization in word-formation Word-formation is the area par excellence in which compositionality is fundamentally questioned (cf. Bartsch 2002: 574 and in general Blank 2001; Fill 1980a and 1980b). While most authors believe that morphologically complex words have at least some kind of relatively compositional “word-formation meaning” that is usually considered to be fully motivated, opinions differ with regard to the motivational status of not fully compositional idiomatized “lexicon senses” of words. Since word-formation research that is interested in the definition of rules (for different theoretical approaches, cf. article 45 on rules, patterns and schemata in word-formation) mainly studies productive and regular components of word-formation, a motivated word is first of all a word whose form and meaning can successfully be analyzed on the grounds of the form and the meaning of its parts. From this perspective, G. Zuschauer is a fully motivated word, because its meaning can be deduced from the combination of G. zuschauen ‘to watch’ and the agent suffix -er, a Zuschauer thus being ‘someone who watches’, i.e. a ‘spectator’. In contrast, words such as G. Bauer ‘farmer’ cannot be decoded directly on the basis of the meaning of G. bauen ‘to construct’ and the agent suffix -er, as a Bauer is not ‘someone who constructs’. In a strictly compositional approach to word-formation such words would be considered as idiosyncratic, lexicalized or simply opaque, because “their lexicon sense deviates considerably from their derivational or ‘expected’ sense” (Laca 2001: 1222), i.e. from the sense that is predicted by the applied word-formation rule. From the compositional point of view, idiomatization is thus considered as a process that lowers the degree of motivation and compromises compositionality (cf. Closs Traugott 2002: 1706; Schröder 1981: 454; Barz 1982: 9; Rainer 2004a: 13−14; Booij 2007: 207). In contrast, the motivation researchers cited in sections 2.2 and 2.3 (and, in general, research on schemata in word-formation) do not see any harm in considering them as motivated (on predictability vs. arbitrariness cf. also Panther and Radden 2004: 2; Booij 2007: 207− 208; Laca 2001: 1222; Harley 2001: 161; Lüdtke 2001: 768−769; Lehmann 1989: 15− 18; Günther 1987: 188−189; Ermakova 1984; Dokulil 1968; on the relations between word-formation rules and compositionality of meaning cf. also article 45 on rules, patterns and schemata in word-formation). While in the case of German deverbal nouns in -er an expected word-formation sense can be formulated rather easily, in other cases of word-formation, such as, e.g., composition, it is not always very clear what the expected word-formation meaning should be and where idiomatization starts. Consider Blank’s (2001: 1599) series of German nouns for different types of Kuchen ‘cake, pie’, e.g., Apfelkuchen ‘apple pie’ and Hundekuchen ‘dog biscuit’, to mention only two of them. While their morphological form is exactly the same, the semantic relation between the determinant and the determinatum of these compounds is not: In the case of Apfelkuchen (as in many other Kuchencompounds) the determinant is an ingredient, whereas it specifies whom the Kuchen is

56. Motivation, compositionality, idiomatization for in the case of Hundekuchen. Even if there are, of course, other German compounds whose determinants specify the beneficiary of the determinatum (such as Frauenparkplatz ‘parking area for women’, Kindergetränk ‘drink for children’, etc.), in the moment of the word-formation it is neither the semantics of the single parts nor their combination that predetermine the outcome ‘biscuit for dogs’, as the word would also have been possible for a cake with dog in it (cf. Apfelkuchen) or a cake in the form of a dog (cf. G. Lammkuchen ‘cake in the form of a lamb’) and probably many others. From the perspective of the interpretation (or motivatability) of the word we can moreover observe exactly the same thing: Without a certain amount of world knowledge and prior experience with prototypical food, pets, and cakes, as well as other linguistic contexts in which the words Kuchen and Hund are typically used, we would not be able to infer the “right” meaning of Hundekuchen only from the meaning of its parts (cf. similarly Marzo 2013: 200−201 and in general Gayral, Kayser and Lévy 2005 on the importance of world knowledge; Dunbar 2005 on the role of pragmatic factors; Klos 2011: 7 on the importance of prior contextual experience with the semantics of the parts; Nagórko 2002 on the role of inference in meaning comprehension of morphologically complex words in Polish).

2.4.3. Degrees of compositionality, idiomatization and motivation Compositionality as well as idiomatization and motivation are, as can be seen from the examples cited in section 1., often a matter of degree: While G. Zuschauer in the sense of ‘someone who watches’ (i.e. ‘spectator’) might be considered as fully compositional, motivated and transparent, full compositionality seems a little blurred in the case of G. Bäcker, because the most salient and prototypical meaning of Bäcker is not ‘someone who bakes’, but ‘someone who bakes professionally’ and is definitely disturbed in the case of G. Bauer that does not mean ‘someone who constructs’ (though it could), but ‘farmer’ (on the predictability of the semantic drift from the expected word-formation sense, cf. especially Hay 2003: 57−61). Accordingly, the degree of idiomatization is generally said to increase from Zuschauer to Bäcker and then to Bauer (cf. Barz 1982; Uluchanov 1977). Moreover, most motivational researchers would agree in saying that these words are not motivated to the same degree, even if they are all motivated. While degrees of compositionality and idiomatization usually refer exclusively to semantic transparency of words, scales of graded motivation also take into account degrees of formal transparency of words (e.g., rider with respect to ride vs. conclusion with respect to conclude vs. decision with respect to decide, cf. Dressler 1985: 330−331) as well as the interaction of formal and semantic aspects (cf. Marzo 2013: 191−253, also for a detailed overview of research on degrees of compositionality, idiomatization and motivation). In this context, it is important to note that the degree of formal transparency of a word and the degree of its semantic compositionality do not necessarily have to be on a par in order for the word to be motivated and motivatable (cf. Fill 1980a and 1980b in general). These findings are corroborated by psycholinguistic research on the mental representation of words: E.g., in a study on morphologically complex German verbs Smolka, Preller and Eulitz have most recently shown “that lexical representation of complex verbs refers to the base regardless of meaning compositionality” (Smolka, Prel-

993

994

Semantics and pragmatics in word-formation I: General aspects ler and Eulitz 2014: 16; for a general overview on other psycholinguistically oriented research in this area, cf. Marzo 2013: 191−237 and article 13 on word-formation in psycholinguistics and neurocognitive research).

3. Summary: The importance of human perception As was foreshadowed by the preceding sections, human perception plays a major role in motivational issues. On the basis of what has been said so far, we can distinguish different, but closely connected aspects of motivation (cf. also Marzo 2013: 46−47): First of all, motivation is indispensable when new words are formed. In this sense, motivation is an active, not necessarily conscious process guided among others by language-independent perceptual factors such as, e.g., salience and gestalt principles (cf. Radden and Panther’s understanding of motivation in section 2.2.6). This aspect of motivation can be exemplified by Radden and Panther’s explanation of the cross-linguistically different, but similar words for SCREWDRIVER (cf. section 2.2.6 as well as Radden and Panther 2004: 4−8; Panther and Radden 2011a: 12) and Koch’s discussion of the relation between trees and their fruits (cf. section 2.3 and Koch 1999). Motivation of this kind also plays a role in phenomena such as folk etymology (cf. also article 57 on word-formation and folk etymology), i.e. when opaque “orphaned” words are (re)motivated by giving them a familiar structure (cf. Maiden 2008: 315 and the contributions in Harnisch 2010; Blank 1997: 309) and connecting them to a new word-family (on word-families, cf. Hundsnurscher 2002). Similarly, it might be of some relevance for language learners (L1 and L2), if they try to memorize newly acquired words (cf. Hausmann 2002 and articles 120 and 121 on word-formation in first and second language acquisition, respectively). Second, this process of motivation results in a state of motivatedness (cf. section 2.2.4), a phenomenon that contributes to the structure of a language’s lexicon. In contrast to the above mentioned process of motivation, the static motivatedness-relation is not unidirectional, because a motivational relation between two words can, in principle, be perceived in both directions (cf. also Leisi’s notions of consociation and dissociation in section 2.2.5 as well as Umbreit 2013 on directionality in general). Third, as motivatedness can be blurred and eventually lost (e.g., through total idiomatization or other phenomena of language change, cf. Seebold 2002: 1333), a lexical unit might at some point lose its motivatability (cf. Rettig’s notion in section 2.2.4). In this case, no motivational relation whatsoever can be perceived between the lexical unit in question and any other unit of the lexicon (cf. also Koch’s definition of motivation in section 2.2.3). Motivatability is thus a conscious phenomenon that guarantees the success of processes of (re)motivation and that can even restructure the lexicon. Recent research has shown that motivatability is not limited to word-formation devices, especially if we take into account the perception of native speakers: Marzo (2013) has empirically studied the contribution of polysemy to motivation in the lexicon. Marzo, Rube and Umbreit (2011: 384−387) complement, on the basis of results of questionnaire studies with native speakers, Koch’s formal dimension of motivation (cf. Koch 2001 and section 2.2.3) by phenomena such as affix-alternation (cf. It. piccino ‘small’ and piccolo ‘small’), affiliation to the same word family without direct relation (cf. Fr. homme ‘male human’ and

56. Motivation, compositionality, idiomatization humanité ‘humanity’) and even simple graphic similarity (cf. G. Blatt ‘sheet of paper’ with respect to platt ‘flat, even’). In a similar vein, but focusing on the semantic dimension of lexical motivation, Ising (2014) argues in a typological perspective that words such as the English compound blue beret (cf. also Ullmann’s example blue-bell in section 2.2.2) and the Sakha (an Altaic language) suffixed buhar ‘to cook’ are motivated holistically and not segmentally (i.e. compositionally). In this sense, motivation goes far beyond compositionality. The decisive factor for motivation not being the degree of compositionality, but the simultaneous perception of a semantic as well as of a formal relation to another item in the lexicon, idiomatization does not automatically block motivation, though it might, in some cases, decrease the degree of transparency.

4. References Augst, Gerhard 1996 Motivationstypen und diasystematische Differenzierung der semantischen Motiviertheit. In: Ernst Bremer and Reiner Hildebrandt (eds.), Stand und Aufgaben der deutschen Dialektlexikographie. II. Brüder-Grimm-Symposion zur Historischen Wortforschung. Beiträge zu der Marburger Tagung vom Oktober 1992, 17−28. Berlin/New York: de Gruyter. Bally, Charles 1965 [1932] Linguistique générale et linguistique française. 4th ed. Bern: Francke. Bartoszewicz, Albert 1974 On the problems of divisibility, motivation and derivation of words. Bulletin phonographique 15: 67−74. Bartsch, Renate 2002 Kompositionalität und ihre Grenzen. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen. Vol. 1, 570−577. Berlin/New York: de Gruyter. Barz, Irmhild 1982 Motivation und Wortbildungsbedeutung: Eine Diskussion sowjetischer Forschungsergebnisse. Beiträge zur Erforschung der deutschen Sprache 2: 5−21. Bellmann, Günter 1988 Motivation und Kommunikation. In: Horst Haider Munske, Peter von Polenz, Oskar Reichmann and Reiner Hildebrandt (eds.), Deutscher Wortschatz. Lexikologische Studien. Ludwig Erich Schmitt zum 80. Geburtstag von seinen Marburger Schülern, 3−23. Berlin/New York: de Gruyter. Blank, Andreas 1997 Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. Tübingen: Niemeyer. Blank, Andreas 2001 Pathways of lexicalization. In: Martin Haspelmath, Ekkehard König, Wulf Oesterreicher and Wolfgang Raible (eds.), Language Typology and Language Universals. An International Handbook. Vol. 2, 1596−1608. Berlin/New York: de Gruyter. Booij, Geert 2007 The Grammar of Words. An Introduction to Morphology. 2nd ed. Oxford: Oxford University Press.

995

996

Semantics and pragmatics in word-formation I: General aspects Closs Traugott, Elizabeth 2002 Lexicalization and grammaticalization. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen. Vol. 2, 1702−1712. Berlin/New York: de Gruyter. Cohnitz, Daniel 2005 Is compositionality an a priori principle? In: Markus Werning, Edouard Machery and Gerhard Schurz (eds.), The Compositionality of Meaning and Content. Vol. 1: Foundational Issues, 23−58. Frankfurt/M.: Ontos. Coseriu, Eugenio 2004 Der Physei-Thesei-Streit. Sechs Beiträge zur Geschichte der Sprachphilosophie. Tübingen: Narr. Cruse, Dean Alan 1986 Lexical Semantics. Cambridge: Cambridge University Press. Dokulil, Miloš 1968 Zur Theorie der Wortbildung. Wissenschaftliche Zeitschrift der Karl-Marx-Universität Leipzig. Gesellschafts- und sprachwissenschaftliche Reihe 17(2/3): 203−211. Dressler, Wolfgang U. 1985 On the predictiveness of natural morphology. Linguistics 21: 321−337. Dupuy-Engelhardt, Hiltraud 2002 Lexikalische Dekomposition I: Strukturalistische Ansätze. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen. Vol. 1, 245− 256. Berlin/New York: de Gruyter. Dunbar, George 2005 The Goldilocks scenario: Is noun-noun compounding compositional? In: Edouard Machery, Markus Werning and Gerhard Schurz (eds.), The Compositionality of Meaning and Content. Vol. 2: Applications to Linguistics, Psychology and Neuroscience, 217− 228. Frankfurt/M.: Ontos. Ermakova, Ol’ga P. 1984 Leksičeskie značenija proizvodnych slov v russkom jazyke. Moskva: Nauka. Ernst, Gerhard 1981 Ein Blick durch die durchsichtigen Wörter: Versuch einer Typologie der Wortdurchsichtigkeit und ihrer Einschränkungen. Linguistica 21: 47−70. Fill, Alwin 1980a Durchsichtige Wörter im Englischen: Betrachtungsweisen und Forschungsansätze. Arbeiten aus Anglistik und Amerikanistik 5(1): 13−35. Fill, Alwin 1980b Wortdurchsichtigkeit im Englischen. Eine nicht-generative Studie morphosemantischer Strukturen. Mit einer kontrastiven Untersuchung der Rolle durchsichtiger Wörter im Englischen und Deutschen der Gegenwart. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Gauger, Hans-Martin 1971 Durchsichtige Wörter. Zur Theorie der Wortbildung. Heidelberg: Winter. Gayral, Françoise, Daniel Kayser and François Lévy 2005 Challenging the principle of compositionality in interpreting natural language texts. In: Edouard Machery, Markus Werning and Gerhard Schurz (eds.), The Compositionality of Meaning and Content. Vol. 2: Applications to Linguistics, Psychology and Neuroscience, 83−105. Frankfurt/M.: Ontos. Geeraerts, Dirk 2003 The interaction of metaphor and metonymy in composite expressions. In: René Dirven and Ralf Pörings (eds.), Metaphor and Metonymy in Comparison and Contrast, 435− 465. Berlin/New York: Mouton de Gruyter.

56. Motivation, compositionality, idiomatization Goddard, Cliff 2002 Lexical decomposition II: Conceptual axiology. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen. Vol. 1, 256−268. Berlin/New York: de Gruyter. Gruaz, Claude 2002 The analysis of word-families and their motivational relations. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen. Vol. 1, 700− 704. Berlin/New York: de Gruyter. Günther, Hartmut 1987 Wortbildung, Syntax, be-Verben und das Lexikon. Beiträge zur Geschichte der deutschen Sprache und Literatur 109: 179−201. Haiman, John 1980 The iconicity of grammar: Isomorphism and motivation. Language 56(3): 515−540. Harley, Trevor 2001 The Psychology of Language. From Data to Theory. 2nd ed. Hove/New York: Psychology Press. Harnisch, Rüdiger (ed.) 2010 Prozesse sprachlicher Verstärkung. Typen formaler Resegmentierung und semantischer Remotivierung. Berlin/New York: de Gruyter. Hausmann, Franz Josef 2002 Nur nützliche Wörter lernen: Durchsichtigkeit des Wortschatzes und Optimierung der Wortschatzarbeit. Französisch heute. Informationsblätter für Französischlehrerinnen und -lehrer in Schule und Hochschule 3: 256−269. Hay, Jennifer 2003 Causes and Consequences of Word Structure. New York/London: Routledge. Hinzen, Wolfram, Markus Werning and Edouard Machery 2012 Introduction. In: Markus Werning, Wolfram Hinzen and Edouard Machery (eds.), The Oxford Handbook of Compositionality, 1−16. Oxford: Oxford University Press. Hiraga, Masako K. 1994 Diagrams and metaphors: Iconic aspects in language. Journal of Pragmatics 22: 5−21. Hundsnurscher, Franz 2002 Das Wortfamilienproblem in der Forschungsdiskussion. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen. Vol. 1, 675−680. Berlin/New York: de Gruyter. Ising, Markus 2014 Motivation via holistic conceptual shifts: Examples from cross-linguistic COOKING. Philologie im Netz 69: 51−76. Jakobson, Roman 1971 Quest for the essence of language. In: Roman Jakobson, Selected Writings. Vol. 1, 345− 359. Den Haag/Paris: Mouton [First published in Diogenes 51 (1965): 21−37]. Klos, Verena 2011 Komposition und Kompositionalität. Möglichkeit und Grenzen der semantischen Dekodierung von Substantivkomposita. Berlin/New York: de Gruyter. Koch, Peter 1999 Tree and fruit: A cognitive-onomasiological approach. Studi italiani di linguistica teorica e applicata 28(2): 331−347. Koch, Peter 2001 Lexical typology from a cognitive and linguistic point of view. In: Martin Haspelmath, Ekkehard König, Wulf Oesterreicher and Wolfgang Raible (eds.), Language Typology

997

998

Semantics and pragmatics in word-formation I: General aspects and Language Universals. An International Handbook. Vol. 2, 1142−1178. Berlin/New York: de Gruyter. Koch, Peter and Daniela Marzo 2007 A two-dimensional approach to the study of motivation in lexical typology and its first application to French high-frequency vocabulary. Studies in Language 31(2): 259−291. Laca, Brenda 2001 Derivation. In: Martin Haspelmath, Ekkehard König, Wulf Oesterreicher and Wolfgang Raible (eds), Language Typology and Language Universals. An International Handbook. Vol. 2, 1214−1227. Berlin/New York: de Gruyter. Lehmann, Christian 1989 Grammatikalisierung und Lexikalisierung. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 42(1): 11−19. Leisi, Ernst 1955 Das heutige Englisch. Wesenszüge und Probleme. Heidelberg: Winter: Lu, Angela Yi-chün 1998 Phonetic Motivation. A Study of the Relationship between Form and Meaning. München: Hieronymus. Lüdtke, Jens 2001 Morphologie II: Wortbildungslehre. In: Günter Holtus, Michael Metzeltin and Christian Schmitt (eds.), Lexikon der Romanistischen Linguistik. Vol. 1.1, 765−781. Tübingen: Niemeyer. Machery, Edouard, Markus Werning and Gerhard Schurz (eds.) 2005 The Compositionality of Meaning and Content. Vol. 2: Applications to Linguistics, Psychology and Neuroscience. Frankfurt/M.: Ontos. Maiden, Martin 2008 Lexical nonsense and morphological sense: On the real importance of ‘folk etymology’ and related phenomena for historical linguists. In: Thórhallur Eythórsson (ed.), Grammatical Change and Linguistic Theory. The Rosendal Papers, 307−328. Amsterdam/ Philadelphia: Benjamins. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. Marzo, Daniela 2008 What is iconic about polysemy? A contribution to research on diagrammatic transparency. In: Ludovic De Cuypere and Klaas Willems (eds.), Naturalness and Iconicity in Language, 167−187. Amsterdam/Philadelphia: Benjamins. Marzo, Daniela 2011 Intrinsic or extrinsic motivation? The implications of metaphor- and metonymy-based polysemy for transparency in the lexicon. In: Klaus-Uwe Panther and Günter Radden (eds.), Motivation in Grammar and the Lexicon, 251−267. Amsterdam/Philadelphia: Benjamins. Marzo, Daniela 2013 Polysemie als Verfahren lexikalischer Motivation. Theorie und Empirie am Beispiel von Metonymie und Metapher im Französischen und Italienischen. Tübingen: Narr Francke Attempto. Marzo, Daniela 2014 Warum der Bauer Bauer heißt: Zur Bedeutung perzeptionsbasierter Datenerhebung für die Motivationsforschung. In: Thomas Krefeld and Elissa Pustka (eds.), Perzeptive Linguistik. Phonetik, Semantik, Varietäten, 153−167. Stuttgart: Steiner. Marzo, Daniela, Verena Rube and Birgit Umbreit 2011 Similarité sans contiguïté − la dimension formelle de la motivation lexicale dans la perspective des locuteurs. In: Sarah Dessì Schmid, Ulrich Detges, Paul Gévaudan, Wil-

56. Motivation, compositionality, idiomatization trud Mihatsch and Richard Waltereit (eds.), Rahmen des Sprechens. Beiträge zu Valenztheorie, Varietätenlinguistik, Kreolisitik, Kognitiver und Historischer Semantik. Peter Koch zum 60. Geburtstag, 381−392. Tübingen: Narr. Mayerthaler, Willi 1981 Morphologische Natürlichkeit. Wiesbaden: Athenaion. Monneret, Philippe 2003 Le sens du signifiant. Implications linguistiques et cognitives de la motivation. Paris: Champion. Mudersbach, Klaus 2002 Struktur und Strukturierung in der Lexikologie. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen. Vol. 1, 45−58. Berlin/New York: de Gruyter. Nagórko, Alicja 2002 Über die Rolle der Inferenz in der Wortbildungslehre (anhand polnischen Sprachmaterials). Zeitschrift für Slawistik 47(4): 410−422. Panther, Klaus-Uwe and Günter Radden (eds.) 2004 Studies in Linguistic Motivation. Berlin/New York: Mouton de Gruyter. Panther, Klaus-Uwe and Günter Radden 2011a Introduction: Reflections on motivation revisited. In: Klaus-Uwe Panther and Günter Radden (eds.), Motivation in Grammar and the Lexicon, 1−26. Amsterdam/Philadelphia: Benjamins. Panther, Klaus-Uwe and Günter Radden (eds.) 2011b Motivation in Grammar and the Lexicon. Amsterdam/Philadelphia: Benjamins. Peirce, Charles Sanders 1960 Collected Papers. Vol. 2. Cambridge, MA: Harvard University Press. Pelletier, Francis Jeffry 2012 Holism and compositionality. In: Markus Werning, Wolfram Hinzen and Edouard Machery (eds.), The Oxford Handbook of Compositionality, 149−174. Oxford: Oxford University Press. Peregrin, Markus 2005 Is compositionality an empirical matter? In: Markus Werning, Edouard Machery and Gerhard Schurz (eds.), The Compositionality of Meaning and Content. Vol. 1: Foundational Issues, 231−246. Frankfurt/M.: Ontos. Plato 1996 Cratylos. In: Plato, Vol. 4, 3−191. English translation: Harold N. Fowler. Cambridge, MA: Harvard University Press. Pusch, Claus D. 2001 Ikonizität. In: Martin Haspelmath, Ekkehard König, Wulf Oesterreicher and Wolfgang Raible (eds.), Language Typology and Language Universals. An International Handbook. Vol. I, 369−384. Berlin/New York: de Gruyter. Radden, Günter and Klaus-Uwe Panther 2004 Introduction: Reflections on motivation. In: Klaus-Uwe Panther and Günter Radden (eds.), Studies in Linguistic Motivation, 1−46. Berlin/New York: Mouton de Gruyter. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rainer, Franz 2004a Premesse teoriche. In: Maria Grossmann and Franz Rainer (eds.), La formazione delle parole in italiano, 4−23. Tübingen: Niemeyer. Rainer, Franz 2004b Etnici. In: Maria Grossmann and Franz Rainer (eds.), La formazione delle parole in italiano, 402−408. Tübingen: Niemeyer.

999

1000

Semantics and pragmatics in word-formation I: General aspects

Rettig, Wolfgang 1981 Sprachliche Motivation. Zeichenrelationen von Lautform und Bedeutung am Beispiel französischer Lexikoneinheiten. Frankfurt/M.: Lang. Rufener, John 1971 Studies in the Motivation of English and German Compounds. Zürich: Juris. Sanchez, Christina 2008 Consociation and Dissociation. An Empirical Study of Word-Family Integration in English and German. Tübingen: Narr. Saussure, Ferdinand de 1972 [1916] Cours de linguistique générale. Édition critique, préparée par Tullio de Mauro. Paris: Payot. Sauvageot, Aurélien 1964 Portrait du vocabulaire français. Paris: Larousse. Scheidegger, Jean 1981 Arbitraire et motivation en français et en allemand. Examen critique des thèses de Charles Bally. Bern: Francke. Schröder, Marianne 1981 Zur Rolle des Motivationsbegriffes für Wortbildungskonstruktionen und feste Wortverbindungen. Wissenschaftliche Zeitschrift der Karl-Marx-Universität Leipzig. Gesellschafts- und sprachwissenschaftliche Reihe 30(5): 453−458. Seebold, Elmar 2002 Etymologie und Wortgeschichte IV: Wurzeletymologie. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen. Vol. 2, 1333−1346. Berlin/New York: de Gruyter. Shaw, Howard J. 1979 Motivierte Komposita in der deutschen und englischen Gegenwartssprache. Tübingen: Narr. Simone, Raffaele 1990 The body of language: The paradigm of arbitrariness and the paradigm of substance. In: René Amacker and Rudolf Engler (eds.), Présence de Saussure. Actes du colloque international de Genève (21−23 mars 1988), 121−141. Genève: Droze. Smolka, Eva, Katrin H. Preller and Carsten Eulitz 2014 ‘Verstehen’ (‘understand’) primes ‘stehen’ (‘stand’): Morphological structure overrides semantic compositionality in the lexical representation of German complex verbs. Journal of Memory and Language 72: 16−36. Swanepoel, Piet H. 1992 Linguistic motivation and its lexicographical application. South African Journal of Linguistics 10(2): 49−60. Ullmann, Stephen 1962 Semantics. An Introduction to the Science of Meaning. Oxford: Blackwell. Uluchanov, Igor’ S. 1977 Slovoobrazovatel’naja semantika v russkom jazyke i principy ee opisanija. Moskva: Nauka. Umbreit, Birgit 2010 Does love come from to love or to love from love? Why lexical motivation has to be regarded as bidirectional. In: Sascha Michel and Alexander Onysko (eds.), Cognitive Perspectives on Word-Formation, 301−333. Berlin/New York: De Gruyter Mouton. Umbreit, Birgit 2011 Motivational networks: An empirically supported cognitive phenomenon. In: Klaus-Panther and Günter Radden (eds.), Motivation in Grammar and the Lexicon, 269−286. Amsterdam/Philadelphia. Benjamins.

56. Motivation, compositionality, idiomatization

1001

Umbreit, Birgit 2013 Zur Direktionalität der lexikalischen Motivation. Motiviertheit und Gerichtetheit von französischen und italienischen Wortpaaren auf der Basis von Sprecherbefragungen. Ph.D. dissertation, University of Tübingen (Germany). Ungerer, Friedrich 1991 What makes a linguistic sign successful? Towards a pragmatic interpretation of the linguistic sign. Lingua 83: 155−181. Ungerer, Friedrich 1999 Diagrammatic iconicity in word-formation. In: Max Nänny and Olga Fischer (eds.), Form Miming Meaning, 307−324. Amsterdam/Philadelphia: Benjamins. Ungerer, Friedrich 2002 Arbitrarität, Ikonizität und Motivation. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen. Vol. 1, 371−380. Berlin/New York: de Gruyter. Van Langendonck, Willy 2007 Iconicity. In: Dirk Geeraerts and Hubert Cuyckens (eds.), The Oxford Handbook of Cognitive Linguistics, 394−418. Oxford: Oxford University Press. Werning, Markus 2005 Right and wrong reasons for compositionality. In: Markus Werning, Edouard Machery and Gerhard Schurz (eds.), The Compositionality of Meaning and Content. Vol. 1: Foundational Issues, 285−309. Frankfurt/M.: Ontos. Werning, Markus, Wolfram Hinzen and Edouard Machery (eds.) 2012 The Oxford Handbook of Compositionality. Oxford: Oxford University Press. Werning, Markus, Edouard Machery and Gerhard Schurz (eds.) 2005 The Compositionality of Meaning and Content. Vol. 1: Foundational Issues. Frankfurt/ M.: Ontos. Willems, Klaas 2005 Die Grenzen der Ikonizität der Sprache: Saussures Konzeption des “fait linguistique” revisited. Ars Semeiotica 28(3−4): 243−272. Zimmermann, Thomas Ede 2012 Compositionality problems and how to solve them. In: Markus Werning, Wolfram Hinzen and Edouard Machery (eds.), The Oxford Handbook of Compositionality, 81−106. Oxford: Oxford University Press.

Daniela Marzo, Munich (Germany)

1002

Semantics and pragmatics in word-formation I: General aspects

57. Word-formation and folk etymology 1. 2. 3. 4. 5. 6. 7. 8.

Introduction Historical outline What is (not) folk etymology? Folk etymology and language typology Folk etymology and neighbouring categories/processes Attempt at a systematization − typology Summary and desiderata References

Abstract Folk etymology is a process that adapts unknown words or parts of words to known ones in certain languages, thus integrating them into the lexical system and making them more transparent. This article aims to shed some light on the history of folk etymology research. It defines folk etymology and discusses it in the context of word-formation as well as language typology, separates it from neighbouring categories, presents a classification/typology and sketches desiderata.

1. Introduction From a synchronic point of view, the connection between New High German Würgengel (‘angel of death; lit. throttle angel’) and Old High German wargengil (warg = ‘avenger’) is no longer visible. The same applies, for example, to English rosemary from Latin ros marinus, French choucroûte ‘sauerkraut’ (chou = ‘cabbage’, croûte = ‘crust’) from German Sauerkraut and Spanish vagamundo ‘vagrant’ (mundo = ‘world’) from Latin vagabundus. From a diachronic point of view, these examples represent a very special process of lexical change expressing speakers’ preference for (more) iconicity in language. This process has come to be known as “folk etymology”, which is still highly disputed: On the one hand, this term is established internationally, focusing on the actions (processes) of the language users as the driving force of this kind of change. On the other hand, the connotations that go along with the traditional concept of “folk” seem to be outdated. What is also still unclear is the categorial and typological status of this phenomenon: Is folk etymology a lexical, i.e. phonetic/semantic, or a word-formation change? This comprises the question of whether folk etymology represents a word-formation process and − if so − how it can be distinguished from neighbouring categories/processes. Among others, the following questions will be dealt with in this article. First of all, a historical outline will sketch the major steps in the history of research of folk etymology (section 2) before the question of what is (or is not) folk etymology is tackled (section 3). Following this, folk etymology will be focused on in the light of language typology (section 4) and what distinguishes it from neighbouring categories/processes (section 5). This is followed by the attempt at a systematization and typology of folk

57. Word-formation and folk etymology

1003

etymological phenomena (section 6). A summary and an overview of research desiderata conclude the article (section 7).

2. Historical outline Traditionally, research in folk etymology has its origins in Germany and most subsequent literature refers to German examples and applies them to other languages. The beginning of this research dates back to Ernst Förstemann’s initial article Ueber deutsche volksetymologie [On German folk etymology] in 1852. The core of his understanding of folk etymology is formulated in this sentence: “Oft naemlich glaubt der volksgeist irrthuemlicherweise in einem worte das etymon eines andern gefunden zu haben und da das volk als solches nie bei der theorie stehen bleibt, sondern gleich in die praxis hinuebergeht, so wandelt es dann das abgeleitete wort so um, daß es eine dem angeblichen etymon angenaeherte form enthält” (1852: 2 ff.) [Often, the spirit of the folk wrongly believes to find the etymon of one word in another one and since the folk as such never stops at theory, but moves on immediately to practice it changes the derived word in a way that brings it closer to the supposed form of the alleged etymon]. Apart from the fact that Förstemann’s article is the first comprehensive and focused treatment of this phenomenon, defining and characterizing folk etymology as a separate process, it had been preceded by Johann Andreas Schmeller’s research Die Mundarten Bayerns grammatisch dargestellt [The Bavarian dialects in grammatic representation] in 1821 and Christoph W. Heinzelmann’s Probe einer Sprachenverähnlichung an den fremden Wörtern im Teutschen [Probe of a language similarization on the foreign words in German] in 1798 (cf. Harnisch 1998: 142, 2001 and Rainer 2012). After Förstemann’s introduction of the term folk etymology, Karl Gustaf Andresen’s book Ueber deutsche Volksetymologie [On German folk etymology] from 1876 established this phenomenon academically. In addition to a theoretical treatment and systematization of folk etymology, Andresen’s book discusses a considerable number of single examples. In general, research in folk etymology in the nineteenth century is dominated by the collection and analysis of single examples (cf. Olschansky 1996: 21). Nevertheless, a more theoretical perspective is presented in Hermann Paul’s Principien der Sprachgeschichte [Principles of language history] and Otto Behaghel’s Die deutsche Sprache [The German language] from 1886. Both authors consider folk etymology a general phenomenon of language perception and don’t attribute it to the concept of folk as a homogenous social stratum. At the beginning of the twentieth century, Wilhelm Wundt (1900: 477) introduces the term lautlich-begriffliche Wortassimilation [phonic and conceptual word assimilation] which, for him, is a purely associative process belonging to the psychophysical mechanism of language function. Two main tendencies are characteristic for this century: 1. folk etymology is perceived as a negative, “pathologic” process (Saussure 1967: 210), 2. folk etymology is perceived as a positive process belonging to everyday language (Gilliéron 1919). Saussure’s dichotomy between “synchrony” and “diachrony” as well as his notion of motivation highly influenced the forthcoming theoretical approach to this phenomenon. To sum up, twentieth century research in folk etymology is − under structuralistic impact − concerned with more fundamental theoretical issues on the one

1004

Semantics and pragmatics in word-formation I: General aspects

hand and with special or single characteristic aspects, often in special contexts, on the other.

3. What is (not) folk etymology? 3.1. Definition Different historical perspectives on and approaches to folk etymology led to many different term proposals: “associative etymology”, “false etymology”, “pseudoetymology”, “secondary semantic motivation”, “Umdeutung” [semantic/etymologic reinterpretation], “metaphysic etymology”, “lautlich-begriffliche Wortassimilation”, etc. Surprisingly, none of these terms could prevail over the compound folk etymology which obviously combines two aspects of the respective process in a handy way: the determinatum etymology categorizes it as a certain learnèd linguistic process, the determinant folk makes clear who is responsible for this process so that there is a clash between the connotation “sophisticated” of etymology and the connotation “unsophisticated” of folk. The term folk etymology thus ranges between the rather academic poles of, e.g., associative etymology, which might be too general and abstract on the one hand and, e.g., secondary semantic motivation, which may be too specific and restrictive on the other hand. The fact that folk etymology − the process and, above all, the examples − has become popular even outside the academic community may have helped in sustaining this term (cf. Olschansky 1999). The term proposals mentioned at the beginning of this section highlight special aspects of this phenomenon, which − taken as one − culminate in the following comprehensive definition by Olschansky (1996: 107, my translation): Folk etymology is a process by which a synchronically isolated and unmotivated word, or a word constituent, is − in an etymologically and diachronically incorrect way − newly or secondarily motivated, interpreted and de-isolated by following a phonetically similar or (partially) identical, non-isolated, well-known word (word family) without considering phonetic-phonological and morphological regularities. The lexeme, which is the product of this process, acquires a new morphological, morpho-semantic or semantic interpretation or interpretability.

The complexity of this definition makes it necessary to separately discuss and evaluate certain central aspects: a) Synchronic isolation This aspect comprises two essential characteristics of folk etymology: 1. In contrast to academic etymology, which aims at the diachronic development of the meaning and origin of words, folk etymology occurs at a certain stage, i.e. is a synchronic process. 2. From this synchronic point of view, words that are affected by folk etymology must be isolated. This means that they are not motivated and transparent since they cannot be linked to any existing word or word family (cf. Bebermeyer 1974: 157).

57. Word-formation and folk etymology

1005

b) Phonetic similarity and phonetic change/spoken language Phonetic similarity between the isolated word and another word/word family seems to be the conditio sine qua non for folk etymology. It functions as a “bridge” between synchronically unmotivated and motivated words. As far as phonetic change is concerned, one might distinguish between a group of examples with and a group without such a change. As Ronneberger-Sibold (2002: 106) points out, folk etymology with phonetic change represents the largest group. This aspect plays a significant role for typology and will be discussed in section 6 in more detail. c) Secondary motivation/de-isolation and semantic relatability Since the isolated word is annexed to a well-established word or word-family, it adopts certain semantic characteristics and thus can be considered as “secondarily motivated”. This motivation can comprise free and bound morphemes and − according to Olschansky (1996: 143 ff.) − should be distinguished from “secondary interpretation” by the fact that the latter requires a semantic/referential reason for choosing certain words/ morphemes (e.g., Middle High German wet(t)erleichen ‘weather dancing/leaping/playing’ > New High German Wetterleuchten ‘heat/summer lightning’). However, most examples do not show any such semantic connections between sourceand target-words, e.g., Middle High German mürmendīn, murmedīn (from Latin mūs montānus lit. ‘mountain mouse’) > New High German Murmeltier ‘marmot; lit. mutter animal’. Very often, it seems quite difficult to decide post hoc whether an example has to be considered as secondarily motivated or secondarily interpreted so that this must be seen as a continuum with two different poles. The status of these phonetic and semantic characteristics is highly disputed. Ducháček (1964), for example, distinguishes between “attraction lexicale” which only covers examples with phonetic similarity and “étymologie populaire” which also includes examples with semantic relatability. Interestingly, some characteristics often adduced in research in folk etymology do not occur in Olschansky’s definition, which implies that she considers them as being less relevant for this phenomenon. These are: a) Word structure, word length, word category Words that are affected by folk etymology are mostly morphologically complex, i.e. compounds or pre- and suffixations. Complexity often goes along with the length of the words which for Paul (1886: 183) is a matter of perceptibility and memorability. As far as word structure is concerned, nouns are far more susceptible to folk etymology than verbs, adjectives or particles. Rundblad and Kronenfeld (2000, 2003) discover that the majority of their corpus examples are nouns, “[m]any of the nouns are compounds, and the folk etymology can either affect the whole word or just one of its elements” (2003: 132). This cross-linguistic discovery might be due to the greater number of nouns along with certain word-formation preferences (compounding) in the lexicon on the one hand and the greater lexical/referential semantics that is carried by nouns on the other. b) Word frequency According to Mayer (1962: 13 ff.), less frequent words are more prone to folk etymology than more frequent ones. However, the factor of word frequency lacks a precise

1006

Semantics and pragmatics in word-formation I: General aspects

description in most studies of folk etymology. Does it comprise type- or token-frequency and how many items are necessary to be considered more or less frequent? A direct correlation between low token frequency and the probability of a word undergoing folk etymology is proven in the corpus analysis by Girnth, Klump and Michel (2007: 54) where items with comparatively low token frequency are more often folk-etymologized than those with relatively high token frequency. This study of the word family diffam-, which is currently folk-etymologized to defam- in German, can also show that there are even frequency differences according to different word categories and word forms of the same word family (Girnth, Klump and Michel 2007: 54). Nevertheless, Olschansky (1996: 141−142) restricts the importance of word frequency as a constitutive characteristic of folk etymology by adducing counterexamples that are prone to folk etymology in spite of their high frequency. c) (Ortho-)graphy, written language Folk etymology is mostly considered as a phenomenon of spoken language, which only secondarily manifests itself in written language. This is why definitions and typologies tend to neglect the question of orthographic change in processes of folk etymology and highlight phonetic aspects instead. Nevertheless, it should not be ignored that most examples of folk etymology that can be found in literature reflect the final status, i.e. when they are already lexicalized and written (cf. Eichinger 2010: 72 ff.). Studies of these examples often rely on (ortho-)graphic comparisons which can provide hints as to their phonetic development (cf. Klump 2014). How exactly this development started in spoken language and proceeded from spoken to written language, however, is left to speculation. Furthermore, many folk etymologies in statu nascendi can only be verified by orthographic analyses. As is shown for the development of diffamieren to defamieren in German (and similarly in Romance languages), there’s only a small vocalic difference perceptible that could also be misinterpreted as false perception (cf. Mondegreens in section 5) or articulatory ease if no written corpora would prove otherwise (cf. Girnth, Klump and Michel 2007; Klump 2014). As this discussion shows, the characteristics that define folk etymology do not apply in a binary way and are obviously not all relevant in the same way. Girnth, Klump and Michel (2007: 55) postulate instead that − considering more and more marginal items − on the parole-level a continuum of prototypical and unprototypical examples (with a respective grading and weighing of the characteristics) should be assumed that even influences the langue-level. Accordingly, a more dynamic and flexible definition considering prototypicality of characteristics, centre and periphery and combining langue with parole could complement a rather (too) comprehensive langue-based definition (even though for folk etymology the difference between langue and parole is not completely undisputed, cf. Harnisch 1998: 143).

3.2. Folk etymology and word-formation Is folk etymology a word-formation process and − if so − what role does it play? These two questions cannot be answered clearly because many textbooks and overviews on

57. Word-formation and folk etymology

1007

word-formation don’t explicitly mention this aspect. One might get some hints by distinguishing two different categories: 1. textbooks that don’t register folk etymology and 2. textbooks that do register folk etymology. The first group is represented − for example − by Marchand (1969), Fleischer and Barz (1995), Altmann and Kemmerling (2000), Eichinger (2000), Naumann (2000), and Motsch (2004). That folk etymology is not mentioned might be due to the explicit synchronic orientation of these books: “Volksetymologisch neu geschaffene Motivationsbedeutungen lassen sich nur diachron erklären” [New semantic motivations that are created by folk etymology can only be explained diachronically] (Barz et al. 2007: 103). To the second group of textbooks − those registering folk etymology − belong, for example, Henzen (1965), Rainer (1993), Erben (2000), Donalies (2005), and Elsen (2012). For Henzen (1965: 256), folk etymology belongs to the “Besondere Arten von Wortbildung” [Special kinds of word-formation], Erben (2000: 21) discusses it shortly in the chapter on motivation, Donalies (2005: 152) wants to abolish the term folk etymology by interpreting the examples as “wortbildende Neumotivierung” [Remotivation based on word-formation] and Elsen (2012: 302) is restrained when she states that folk etymology “kann als Wortbildungsprozess verstanden werden” [can be understood as word formation process]. To conclude, the phenomenon in question plays a rather minor role in these books so that it can be considered as a marginal phenomenon of word-formation. Two aspects are important at this point which refer to the different perspectives of synchronic and diachronic word-formation: 1. Most lexicalized examples are folk etymologies from a diachronic point of view. Synchronically, they represent one of the major established word-formation processes. German Friedhof ‘cemetery’, for example, has to be categorized as N+N-compound with the meaning ‘yard of peace’ nowadays. The diachronic remotivation process (Old High German frīten ‘to foster’ > Friede ‘peace’, is no longer perceived and is not important for the formation and use of this compound (the products of folk etymology are often determinative compounds; sometimes they belong to the subclass of “verdeutlichende/explikative Komposita” [elucidating/explicative compounds] such as German Turteltaube ‘turtle dove’ (< Old High German turtura ‘id.’) or German Kichererbse ‘chickpea’ (< Latin cicer ‘pea’), cf. Fleischer and Barz 2012: 146 ff. In contrast to these formations, unique morphemes are examples of cases where synchronically unmotivated morphemes/words can persist in certain compounds and are not subject to folk etymology. Eichinger (2010: 71) explains this with paradigmatic semantic sets the affected − mainly first − elements of these compounds are integrated into. 2. As far as diachronic word-formation (this also pertains to ongoing processes of folk etymology in New High German) is concerned, only those examples of folk etymology represent a word-formation process that use existing morphemes (affixes and words) on the one hand and show typical changes of patterns in form and/or meaning on the other (cf. Fleischer and Barz 2012: 18−22). Again, the respective products must be interpreted as realizations of the word-formation processes of compounding or derivation (for Ronneberger-Sibold 2002: 110, examples of folk etymology are “lautnachahmende Wortschöpfungen” [word creations that emulate sounds] and obviously don’t belong to word-formation, cf. also Rainer 1993: 15 for a similar position).

1008

Semantics and pragmatics in word-formation I: General aspects

The second aspect makes clear that folk etymologies that only show either phonetic or semantic changes (cf. section 4) do not belong to diachronic word-formation but are special categories of language change, i.e. phonetic or semantic change. Nevertheless, the boundary between neighbouring categories seems rather fuzzy as will be demonstrated in section 5.

4. Folk etymology and language typology It is a well-known fact that language change often goes along with phonetic and semantic reduction. With respect to grammar, grammaticalization processes show that the reduction of phonetic as well as semantic complexity are typical stages of heightening the grammatical status of constructions. As for lexicalization, phonetic and semantic reduction lead to the loss of motivation of formerly analytic, transparent and motivated words and result in morpho-semantic opaque/non-transparent and holistic formations (cf. Harnisch 2004: 228). Both processes, grammaticalization and lexicalization, are language universals and represent the prototype of language change. According to Lüdtke (1988), such morphologically reduced forms tend to combine with neighbouring syntactic units in order to realize what markedness theory and naturalness theory call constructional iconism: semantic complexity must correspond to formal complexity and vice versa (cf. Mayerthaler 1981; Wurzel 1994). If there is an unequal relationship between form and meaning/function, language change tends to abolish this kind of unnaturalness/markedness. While grammaticalization and lexicalization restore constructional iconism by giving form to semantics/function, Harnisch (2000, 2004, 2010) analyzes processes where degrammaticalization and de-lexicalization restore constructional iconism by giving semantics/function to form. Here, de-lexicalization goes along with resegmentation and remotivation of opaque morphological and lexical units − a process that is traditionally and prototypically represented by folk etymology (cf. Harnisch 2010: 8). Harnisch adduces examples that represent typical folk-etymology processes where isolated unmotivated words are compositionally structured and remotivated, but also examples that represent forms of “sublexematischer Volksetymologie” [sublexematic folk etymology] (2010: 10; cf. also Harnisch 2000). This means that some characteristic phonetic patterns are morphologically reanalyzed − a process that Jespersen (1925) called “secretion”. Although folk etymology and secretion share certain features, there is one fundamental difference that makes it necessary to consider both processes as two opposite poles on a continuum: While folk etymology annexes isolated words or morphemes to existing words or morphemes − thus linking them up to word families −, secretion reanalyzes words by semanticizing certain sound sequences either with wordinternal semantics (e.g., Terminator > -minator, as in German Klausurminator ‘someone who manages a number of Klausuren [‘tests’] in a short period of time’ or Entertainment > -tainment as in Politainment, cf. Michel 2009, alcoholic > -holic as in workaholic) or by establishing paradigmatic functional relationships between similar sound sequences (cf. Haspelmath 1995).

57. Word-formation and folk etymology

1009

In other words: while folk etymology may − sometimes − link up a (unmotivated) sound sequence to existing affixes, secretion − regularly − creates new bound morphemes (affixes or combining forms) out of (unmotivated) sound sequences. This assumption should not be understood as denying the possibility that some secreted (productive) items may have started as folk etymologic reanalyses. To conclude, grammaticalization and lexicalization, as realizing speakers’ needs on the one hand, as well as de-grammaticalization and de-lexicalization, as realizing hearers’ needs on the other, are complementing ways of restoring constructional iconism that Harnisch (2004: 222) defines as “re-constructional iconism”. In the first case, semantic/ functional excess tends to be expressed formally, in the latter case, formal excess needs semantics/function. Accordingly, folk etymology can be considered as one way of abolishing markedness in agglutinating languages and restoring a natural relationship between form and meaning.

5. Folk etymology and neighbouring categories/processes As has been mentioned before, folk etymology is not a distinct phenomenon with clearcut boundaries to neighbouring categories, which will be demonstrated in this section: a) Etymology Often, folk etymology is contrasted with scientific etymology on the one hand and so-called false “gelehrte Etymologie” [learnèd etymology] (Olschansky 1996: 152) on the other by considering the aspects of intention and perspective. Accordingly, folk etymology does not belong to etymology proper if it is considered from a synchronic perspective and if its immediate non-intentionality is valid. It is obvious that the distinction between etymology and folk etymology is a matter of definition and hence a question of which parameter is ranked most important and sufficient to postulate such a boundary. This is supported by the fact that characteristics such as intentionality are graded phenomena and difficult to determine post hoc. Thus, it is open to speculation from a synchronic point of view, whether August Ludwig Schlözer, for instance, intentionally linked Russian knjaz’ ‘prince’ to German Knecht ‘churl’ with fatal personal consequences (the correct link to Proto-Germanic *kuningaz ‘king’ was established subsequently). Examples like this are the main reason that the assumption of maintaining such a distinction is contrasted by the assumption of levelling it. Vennemann, for instance, denies the view that folk etymology has to be separated from etymology as an intended and naive process. For him, each etymologic action serves the explanation of the meaning of words and is thus a scientific (linguistic) process: “Insoweit bei dieser Tätigkeit Volksetymologie auftritt, ist sie also Volksetymologie als Etymologie” [As far as folk etymology occurs in this process, it is thus folk etymology as etymology] (1999: 274; bold print in the original; cf. also Antos 1996). Augst (1975: 184) points in a similar direction by postulating that the difference between synchrony and diachrony doesn’t justify a separation of these two processes. From a synchronic point of view, Augst argues, a difference between historically true etymology and so-called folk etymology doesn’t exist.

1010

Semantics and pragmatics in word-formation I: General aspects

Referring to Brückner’s assumption that there’s no need “[um] Etymologie und Volksetymologie gegeneinander auszuspielen” [to play etymology and folk etymology against one another] (2006: 144), we receive evidence for assuming shifting transitions between these two poles. b) Language acquisition For Vennemann, language acquisition plays a dominant role in folk etymology so that he considers “folk etymology as language acquisition” (1999: 274, my translation) as the complementary part to the afore-mentioned “folk etymology as etymology”. This kind of language acquisition − i.e. with respect to folk etymology − refers to the normal language acquisition process of children, but can also occur as an adult while acquiring foreign words and names as well as during second language acquisition. The importance of first language acquisition for folk etymology is also pointed out by Harnisch (2007: 16) as mirroring the speaker’s need to create natural form-meaning relationships and to display such “good” signs in chain structures that avoid marked sounds by choosing segments that are easier to pronounce. Here, language acquisition represents one important proof of the tendency for languages to restore constructional iconism (cf. section 4), cf. Russian budil’nik ‘alarm clock’ > gudil’nik (gudeť ‘to noise’), Russian vazelin ‘vaseline’ > mazelin (mazať ‘to slather’); cf. Vvedenskaja and Kolesnikov (2003). Nevertheless, the general problem of the children’s limited power to initiate language change must also be considered as a restrictive force behind the establishment of these specific folk etymologies. c) Phonetic/semantic change One important difference between word-formation change and phonetic/semantic change has already been sketched above: while word-formation change comprises changes of patterns in form and/or meaning, phonetic and semantic change either comprise form (phonetics) or meaning of word(familie)s. This difference, however, does not pertain to the distinction between folk etymology and phonetic/semantic change, since − as will be demonstrated in section 6 − there are many phonetic and/or semantic changes that are due to folk-etymology processes. But there is a difference between “natural” language change, i.e. phonetic and semantic change, and changes that go back to folk etymology: while phonetic and semantic changes operate within certain domains (certain word-internal sound changes or semantic developments), folk etymology operates between two domains (certain changes due to word-external annexation to different word fields). In many cases, regular phonetic and/ or semantic change even lead(s) to folk etymology by creating the necessary isolation of word(element)s or leading to homonymy with already existing words (this argument might be supported by the fact that the importance of analogy − as a driving force behind “natural” language change − is highly disputed for folk etymology, cf. Mayer 1962: 331 ff., Rundblad and Kronenfeld 2003: 121): e.g., Old High German einōti ‘loneliness’ (ein ‘lonely’ + -ōti) changes due to the umlaut caused by the “i” to New High German Einöde ‘wasteland’, Öde ‘dreariness’. Depending on the importance (grading) of this difference, the examples in question might either constitute categories that cover word-formation changes, special (sub-)cat-

57. Word-formation and folk etymology

1011

egories of phonetic and semantic change, or − from a prototypical point of view − represent less prototypical cases of these changes. d) Onomastics Folk etymology covers both common nouns and proper names. Especially proper names (most of them are non-native) are affected by folk etymology as can be seen in family names (e.g., German Balthasar > Waldhauser ‘forest + house + -er’, German Pokorny > Bockhorn ‘bock + horn’), the names of saints (e.g., German Valentin > Fallent-hin ‘falling sickness + down’ as patron against epilepsy), names of places (e.g., German Suderland ‘southern country’ > Sauerland ‘sour + land’, German struot berg (Old High German struot ‘bushes’) > Streitberg ‘dispute + mountain’, English Selevan > St Levan) or names of plants (e.g., English buck bean > bogbean, English feverfew > featherfew/featherfold/featherfowl) (cf. Bebermeyer 1974: 181 ff.; Olschansky 1996: 204 ff.; Panagl 2005: 1347−1348; Rundblad and Kronenfeld 2000, 2003: 130 ff.; Vennemann 1999; Fetzer 2011). Traditionally, onomastics analyses the origin, development, semantics and geographic distribution of names. Nevertheless, folk etymologic processes are mostly and systematically demonstrated and characterized by examples that cover names. This special “relationship” between folk etymology and onomastics is characterized by Vennemann (1999: 287) as follows: “Wir haben es hier mit einer besonderen Abteilung der Volksetymologie zu tun, der Volksonomastik” [Here, we deal with a special section of folk etymology, namely folk onomastics] (bold-print in original). Although the difference between folk etymology and a neighbouring category is not the crucial point here, stating the difference addresses the status of folk etymology within onomastics − and not onomastics within folk etymology as Vennemann’s quote might suggest − as an established linguistic branch. Following the afore-mentioned difference between folk etymology and phonetic/semantic change, the differences between “regular” phonetic and semantic changes of proper names on the one hand and changes due to folk etymology on the other have to be considered. This may either lead to a purely langue-based special and separate category “folk onomastics” or to a langue- and parolebased prototypical continuum. e) Word play (pun)/paronomasia Word play, more precisely paronomasia as one subtype that is important here, is commonly considered as an intended process which aims at a certain − mostly funny, satiric or polemic − effect. Often, it can be found in literary texts (cf. Panagl 2005: 1348). Paronomasias such as German Durststillstation ‘station to satisfy your thirst’ < Destillation ‘destillation’, German Kater ‘hangover’ < Katarrh ‘catarrh’, English non-prophet < non-profit, Spanish es pera ‘is a pear’ < espera ‘to wait’ or Russian guvernjan’ka ‘nanny; lit. governanny’ < guvernantka ‘governess’ are − in contrast to folk etymology − highly context-dependent. They are often restricted to parole, i.e. are occasional creations that have to be interpreted pragmatically, if the recipient can detect the intentionally false association. This is only possible when linguistic and extra-linguistic contexts reveal this kind of causal pragmatics (cf. Olschansky 1996: 172). For Ronneberger-Sibold (2002: 117−118), creative processes that underly paronomasia and folk etymology are part of the speaker’s competence. For her, the crucial difference between paronomasia and folk etymology is motivation: all paronomasias are moti-

1012

Semantics and pragmatics in word-formation I: General aspects

vated, since the emulated words are still visible in paronomasias − in contrast to folk etymologies with sound change. This visibility is realized by certain sound shapes so that the differences are rather gradual. Again, it depends on the importance of different factors such as intention or motivation whether word play is considered as a separate category or − as the last quote suggests − the examples are located on a continuum of prototypical and non-prototypical examples of folk etymology and between the two poles with word play on the one end and folk etymology on the other. f) Blending Comparable to the distinction between folk etymology and word play, the distinction between folk etymology and blending (contamination) is mostly based on the question of whether the examples are produced intentionally or not − something that is according to Panagl (2005: 1351) often difficult to decide. A more fundamental difference pertains to the status of the words involved: blending usually comprises two free and motivated words (smoke + fog → smog, motor + hotel → motel) that fuse formally and semantically, while in folk etymology one word (part) must be isolated before it becomes linked to an existing word (part). Furthermore, in folk etymology the words involved don’t fuse phonetically and/or semantically (RonnebergerSibold 2002: 107 also mentions the aspect that − in contrast to folk etymology − blending doesn’t lead to transparency so that typical blends are simplexes; cf. also article 21 on blending). Olschansky (1996: 224), however, points out that it is difficult to differentiate between contamination and folk etymology if one of the two constituents is isolated. Ronneberger-Sibold (2002: 108) even discusses examples that show traits of both processes and defines them as “volksetymologische Kontaminationen” [folk-etymological contaminations]: e.g., Late Middle High German krûsp ‘confused’ < Middle High German krûs + krisp (both with the same meaning and krisp being a vaguely known loanword). This shows that blending represents another process that cannot be separated from folk etymology in a discrete way. g) Mondegreen The term Mondegreen covers words that are the results of false reconstruction and misinterpretation either because the hearer doesn’t perceive a word correctly or re-interprets it differently. The most popular example of this is Mondegreen itself that the Scottish writer Sylvia Wright created by misinterpreting They ha’e slain the Earl of Murray And laid him on the green from the ballad “The Bonny Earl of Murray” as They ha’e slain the Earl of Murray And Lady Mondegreen. Such examples − others are German der weiße Nebel wunderbar ‘the white fog wonderful’ > der weiße Neger Wumbaba ‘the white Negro Wumbaba’, French il ferait encore envie ‘he would still be envied’ > il serait encore en vie ‘he would still be alive’ or

57. Word-formation and folk etymology

1013

Russian O kak velik, velik Napoleon ‘O how great, great is Napoleon’ > O kak velik, velik Na-pole-on ‘O how great, great [is] On-[the] (battle)field-he’ (author: Gavrila R. Deržavin) − are similar to folk etymology because words that are perceived as unmotivated are motivated by linking them up to established and well-known word(familie)s. There are some differences, however, that should not be disregarded (cf. RonnebergerSibold 2010: 91−92): First of all, folk etymologies are usually not restricted to certain texts and contexts, whereas Mondegreens are mostly dependent on textual knowledge to be interpreted which makes them a phenomenon of parole. Secondly, Mondegreens are often longer than folk etymologies and can depart semantically from their origin to a greater extent. Accordingly, it is a matter of grading the parameters of context-sensitivity, length and semantic departure in deciding whether the examples are folk etymologies or Mondegreens, or whether they are to be situated on a continuum between these two poles. h) Malapropism Malapropisms are words that are used instead of other words − with the intended meaning − due to their phonetic similarity: German kosmisch ‘cosmic’ for kosmetisch ‘cosmetic’, German insolvent ‘insolvent’ for insolent ‘insolent’, English obtuse for abstruse or German (dialectal) sublimieren ‘to sublimate’ for supplieren ‘to give a replacement lesson’. For the producers of these Malapropisms, both words are not motivated (or at least similarly more or less motivated), their meaning is vaguely known (cf. Olschansky 1996: 221−222, Ronneberger-Sibold 2010: 92). This lack of motivation of both words − supplemented by the fact that Malapropisms are rather a phenomenon of parole which makes them comparable to Mondegreens − is the main reason why these formations cannot be considered prototypical folk etymologies where motivated words replace unmotivated ones.

6. Attempt at a systematization – typology Focusing on selected criteria, various researches on folk etymology try to systematize the examples and propose different ways to typologize them. Referring to Bebermeyer (1974), Olschansky (1996: 181 ff.) proposes Figure 57.1 as a modified and extended

Fig. 57.1: Typology according to Olschansky (1996: 181 ff.)

1014

Semantics and pragmatics in word-formation I: General aspects

typology [my translation]. This hierarchizes the criteria “phonetic change”, “semantic change” (sc) and “conventionality” in a binary way. The criterion “phonetic change” is considered as most important so that on the uppermost level two categories are established: folk etymologies (FEs) with phonetic change on the one hand and those without phonetic change on the other. Each category is further subdivided by the criterion “semantic change”, folk etymologies without phonetic and without semantic change are additionally divided into the criteria “conventional” and “not conventional”. This categorization yields four different groups: 1. Folk etymologies with phonetic change and without semantic change are, for instance, German Bienenkorb ‘beehive’ (< Middle High German binenkar − kar ‘vessel’), Russian polusadik ‘half little garden’ < palisadnik ‘front yard’ and English dormouse (< Anglo-Norman dormeus ‘sleepy (one)’). 2. This rather small group covers examples that are changed phonetically and semantically, such as German schmutzig lachen lit. ‘to laugh beastly’ (< Middle High German smutzelachen ‘to chuckle’) or English penthouse (< Middle English pentis ‘attached building’). 3. German mundtot (jmdn. mundtot machen ‘to silence sb.; lit. to make sb. mouthdead’ (< Middle High German munt ‘hand, protection’; Early New High German mundtot ‘incompetent to carry out legal acts’) and irritieren ‘to confuse sb.’ (associated with irren ‘to confuse’ < Latin irrītāre ‘to irritate’) are examples without phonetic but with semantic change. 4. The fourth group − folk etymology without phonetic and semantic change − plays a comparatively minor role, covering only a minority of examples, whereas all other groups − those with formal and/or semantic change − “are by far the most common ones” (Rundblad and Kronenfeld 2003: 122). According to Olschansky (1996: 188), subgroups of this group are constituted by the fact that the examples are tendentially considered as folk etymology by the majority of the speakers (= conventional) on the one hand and only by some, even only one speaker (= not conventional) on the other. Conventional examples that fit in this subgroup of the typology are German betucht ‘monied’ (associated with Tuch ‘cloth’ < Yiddish betūche ‘financially safeguarded’) or English quai/key (< Old French kay, cay ‘sand-bank, bar’), a non-conventional example might be German Laster ‘vice’ (< Middle High German laster ‘disgrace, dishonour, blemish’). Following a usage- and parole-based prototype theory, this typology seems problematic and deficient the cause of which is mainly reducible to the lack of transparency as far as certain procedural aspects are concerned: 1. The selection of the different possible criteria seems rather arbitrary. Numerous criteria that also pertain to the definition of folk etymology (e.g., word structure, word frequency, (ortho)graphy) are completely neglected. 2. Similarly, the evaluation/weighting of the different criteria appears rather intuitive. Why is the criterion “phonetic change” ranked higher than the criterion “semantic change”? 3. As one important tenet of prototype theory, a binary division of criteria disregards the dynamics and flexibility of language (cf. Geeraerts 2006). Accordingly, the grading of these criteria cannot be reduced to an either-or-decision, but must be considered as a more-or-less phenomenon. The criterion of conventionality, for instance, shows the difficulty in drawing a statistics-based line between majority (conventional) and

57. Word-formation and folk etymology

1015

minority (not conventional) and should be replaced by a more sophisticated account of folk etymologies in different varieties, such as dialects, youth language, technical language, etc.

7. Summary and desiderata As has been demonstrated, folk etymology is a process that connects isolated word(segment)s to established word(-familie)s thus making them more motivated and transparent, at least from a purely structural point of view (cf. Maiden 2008 who postulates that the morphological/structural analysis of words often prevails over the lexical analysis in language change, i.e. that the recognition of a recurrent structure does not necessarily imply the association with certain semantics/functions). This process, however, cannot be considered as a clear-cut phenomenon of langue, since there are fluent transitions with respect to: a) parameters of definition Criteria such as phonetic isolation, phonetic change, secondary motivation and semantic relatability, word structure, word length, word category, word frequency and orthography play a role in folk etymology. These criteria, however, can only be evaluated and graded by considering parole. b) word-formation Only those examples of folk etymology belong to word-formation that show changes of patterns in form and/or meaning. The products often are subtypes of determinative compounds (“elucidating compounds”) or suffixations. On the one hand, the question arises whether these subtypes are necessary considering the fact that a prototypical classification of determinative compounds seems more adequate. On the other hand, the transition of word-formation to either formal or semantic change of words is sometimes hard to define. c) language typology Folk etymology is one way to restore constructional iconism in agglutinating languages by giving meaning/function to form. However, there are also other possibilities, such as secretion, which go in the same direction and which only marginally differ from folk etymology. d) neighbouring categories and processes Certain categories and processes such as etymology, language acquisition, phonetic and semantic change, onomastic, word play (paronomasia), blending, Mondegreen and malapropism often differ from folk etymology in only one single criterion. Depending on the status of these criteria within the different categories and processes as one pole, they rather create continua to folk etymology as the other pole. e) classification/typology A typology of folk etymologies cannot rely on binary distinctions of arbitrarily evaluated and graded criteria. Instead, definition and typology should fit together so that the prototypicality of certain criteria must also be used to classify the examples.

1016

Semantics and pragmatics in word-formation I: General aspects

As this overview shows, langue-based binary and clear-cut definitions, typologies and distinctions of word-formation categories from neighbouring phenomena begin to totter when they are faced with language use, i.e. parole. What is thus necessary is a corpusbased complementation by evaluating and grading the criteria and establishing a continuum of prototypical and non-prototypical examples (cf. Elsen and Michel 2007, 2011). One way of yielding such prototype effects is the analysis of type- and token-frequency. The gradation of frequency results of different criteria may give a clue to their prototypicality. First of all, the origin and development of already lexicalized folk etymologies can be analyzed via medially and conceptually written corpora (cf. Koch and Oesterreicher 1985, 1996). Secondly, ongoing processes in corpora that are on the continuum between conceptual orality and literacy − but already medially written − can be analyzed in the internet. So, certain forms of communication and sorts of texts in the internet can make intermediary stages of folk etymological processes − from conceptual nearness to distance − accessible which do not come into view by narrowing the focus down to lexicalised and well-established examples.

8. References Altmann, Hans and Silke Kemmerling 2000 Wortbildung fürs Examen. Studien- und Arbeitsbuch. Wiesbaden: Westdeutscher Verlag. Andresen, Karl Gustaf 1876 Ueber deutsche Volksetymologie. Heilbronn: Henninger. Antos, Gerd 1996 Laien-Linguistik. Studien zu Sprach- und Kommunikationsproblemen im Alltag. Am Beispiel von Sprachratgebern und Kommunikationstrainings. Tübingen: Niemeyer. Augst, Gerhard 1975 Untersuchungen zum Morpheminventar der deutschen Gegenwartssprache. Tübingen: Narr. Barz, Irmhild, Marianne Schröder, Karin Hämmer and Hannelore Poethe 2007 Wortbildung − praktisch und integrativ. Ein Arbeitsbuch. Frankfurt/M.: Lang. Bebermeyer, Renate 1974 Zur Volksetymologie: Wesen und Formen. In: Jochen Möckelmann (ed.), Sprache und Sprachhandeln. Festschrift für Gustav Bebermeyer zum 80. Geburtstag am 16. 10. 1970. Arbeiten aus seinem Freundes- und Schülerkreis, 156−187. Hildesheim: Olms. Behaghel, Otto 1886 Die deutsche Sprache. Leipzig: Freytag and Tempsky. Brückner, Dominik 2006 Etymologie versus Volksetymologie: Ein Fall für die Sprachkritik? Muttersprache 116(2): 140−146. Donalies, Elke 2005 Die Wortbildung des Deutschen. Ein Überblick. 2nd ed. Tübingen: Narr. Ducháček, Otto 1964 L’attraction lexicale. Philologica Pragensia 46: 65−76. Eichinger, Ludwig M. 2000 Deutsche Wortbildung. Eine Einführung. Tübingen: Narr. Eichinger, Ludwig M. 2010 ‘… es müsse sich dabei doch auch was denken lassen.’ Remotivationstendenzen. In: Rüdiger Harnisch (ed.), Prozesse sprachlicher Verstärkung. Typen formaler Resegmentierung und semantischer Remotivierung, 59−86. Berlin/New York: de Gruyter.

57. Word-formation and folk etymology

1017

Elsen, Hilke 2012 Grundzüge der Morphologie des Deutschen. Berlin/Boston: de Gruyter. Elsen, Hilke and Sascha Michel 2007 Wortbildung im Sprachgebrauch: Desiderate und Perspektiven einer etablierten Forschungsrichtung. Muttersprache 117(1): 1−16. Elsen, Hilke and Sascha Michel 2011 Wortbildung im Deutschen zwischen Sprachsystem und Sprachgebrauch. Perspektiven − Analysen − Anwendungen. Stuttgart: ibidem. Erben, Johannes 2000 Einführung in die deutsche Wortbildungslehre. 4th ed. Berlin: Schmidt. Fetzer, This Michel 2011 Aspekte toponymischer Volksetymologie. Das Beispiel des Kantons Bern (deutschsprachiger Teil). Tübingen: Narr Francke Attempto. Fleischer, Wolfgang and Irmhild Barz 1995 Wortbildung der deutschen Gegenwartssprache. Unter Mitarbeit von Marianne Schröder. 2nd ed. Tübingen: Niemeyer. Fleischer, Wolfgang and Irmhild Barz 2012 Wortbildung der deutschen Gegenwartssprache. Unter Mitarbeit von Marianne Schröder. 4th ed. Berlin/Boston: de Gruyter. Förstemann, Ernst 1852 Ueber deutsche volksetymologie. Zeitschrift für vergleichende Sprachforschung auf dem Gebiete des Deutschen, Griechischen und Lateinischen 1: 1−25. Geeraerts, Dirk 2006 Prospects and problems of prototype theory. In: Dirk Geeraerts (ed.), Cognitive Linguistics. Basic Readings, 141−167. Berlin: de Gruyter. Gilliéron, Jules 1919 Études sur la Défectivité des Verbes. La Faillite de l’Étymologie phonétique. Résumé de conférences faites à l’école pratique des hautes études. Neuveville: Beerstecher. Girnth, Heiko, Andre Klump and Sascha Michel 2007 ‘Du defamierst somit die Verfasser der Gästebucheinträge, wo wir wieder bei den Beleidigungen wären’. Volksetymologie gestern und heute im Romanischen und Germanischen. Muttersprache 117(1): 36−59. Harnisch, Rüdiger 1998 Rezension zu: Olschansky, Heike (1996): Volksetymologie. Tübingen: Niemeyer. Zeitschrift für Sprachwissenschaft 17(1): 140−145. Harnisch, Rüdiger 2000 Morphosemantische Remotivierung verdunkelter Nominalkomposita im Englischen und Deutschen. Arbeiten aus Anglistik und Amerikanistik 25(1): 71−88. Harnisch, Rüdiger 2001 Johann Andreas Schmeller und die Frühgeschichte der volksetymologischen Forschung. In: Ilona Schern (ed.), Miscellanea Schmelleriana, 44−48. Bayreuth: Rabenstein. Harnisch, Rüdiger 2004 Verstärkungsprozesse. Zu einer Theorie der ‘Sekretion’ und des ‘Re-konstruktionellen Ikonismus’. Zeitschrift für Germanistische Linguistik 32: 210−232. Harnisch, Rüdiger 2007 Herstellung von Konstruktionalität − eine Strategie im Erstspracherwerb. Alkalmazott Nyelvtudomány (Hungarian Journal of Applied Linguistics) 7: 5−16. Harnisch, Rüdiger 2010 Zu einer Typologie sprachlicher Verstärkungsprozesse. In: Rüdiger Harnisch (ed.), Prozesse sprachlicher Verstärkung. Typen formaler Resegmentierung und semantischer Remotivierung, 3−23. Berlin/New York: de Gruyter.

1018

Semantics and pragmatics in word-formation I: General aspects

Haspelmath, Martin 1995 The Growth of Affixes in Morphological Reanalysis. In: Gert Booij and Jaap van Mark (eds.), Yearbook of Morphology 1994, 1−29. Dordrecht: Kluwer. Heinzelmann, Christoph W. 1798 Probe einer Sprachenverähnlichung an den fremden Wörtern im Teutschen. Stendal: Franzen and Grosse. Henzen, Walter 1965 Deutsche Wortbildung. 3rd ed. Tübingen: Niemeyer. Jespersen, Otto 1925 Die Sprache. Ihre Natur, Entwicklung und Entstehung. Heidelberg: Winter. Klump, Andre 2014 Volksetymologische Prozesse in der deutschen, französischen und spanischen Sprachgeschichte. In: Sascha Michel and József Tóth (eds.), Wortbildungssemantik zwischen Langue und Parole. Semantische Produktions- und Verarbeitungsprozesse komplexer Wörter, 311−319. Stuttgart: ibidem. Koch, Peter and Wulf Oesterreicher 1985 Sprache der Nähe − Sprache der Distanz. Mündlichkeit und Schriftlichkeit im Spannungsfeld von Sprachtheorie und Sprachgeschichte. Romanistisches Jahrbuch 36: 15−43. Koch, Peter and Wulf Oesterreicher 1996 Sprachwandel und expressive Mündlichkeit. Zeitschrift für Literaturwissenschaft und Linguistik 26(H. 102): 64−96. Lüdtke, Helmut 1988 Grammatischer Wandel. In: Ulrich Ammon, Norbert Dittmar and Klaus J. Mattheier (eds.), Soziolinguistik. Ein internationals Handbuch zur Wissenschaft von Sprache und Gesellschaft. Vol. 2, 1632−1642. Berlin/New York: de Gruyter. Maiden, Martin 2008 Lexical nonsense and morphological sense: On the real importance of ‘folk etymology’ and related phenomena for historical linguists. In: Thórhallur Eythórsson (ed.), Grammatical Change and Linguistic Theory, 307−328. Amsterdam/Philadelphia: Benjamins. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachcronic Approach. 2nd ed. München: Beck. Mayer, Erwin 1962 Sekundäre Motivation. Untersuchungen zur Volksetymologie und verwandten Erscheinungen im Englischen. Ph.D. dissertation, Universität Köln. Mayerthaler, Willi 1981 Morphologische Natürlichkeit. Wiesbaden: Athenaion. Michel, Sascha 2009 Das Konfix zwischen Langue und Parole. Ansätze zu einer sprachgebrauchsbezogenen Definition und Typologie. In: Peter O. Müller (ed.), Studien zur Fremdwortbildung, 91− 140. Hildesheim: Olms. Motsch, Wolfgang 2004 Deutsche Wortbildung in Grundzügen. 2nd ed. Berlin/New York: de Gruyter. Naumann, Bernd 2000 Einführung in die Wortbildungslehre des Deutschen. 3rd ed. Tübingen: Niemeyer. Olschansky, Heike 1996 Volksetymologie. Tübingen: Niemeyer. Olschansky, Heike 1999 Täuschende Wörter. Kleines Lexikon der Volksetymologien. Stuttgart: Reclam. Panagl, Oswald 2005 Volksetymologie und Verwandtes. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur

57. Word-formation and folk etymology

1019

und Struktur von Wörtern und Wortschätzen. Vol. 2, 1346−1353. Berlin/New York: de Gruyter. Paul, Hermann 1886 Principien der Sprachgeschichte. 2nd ed. Halle/S.: Niemeyer. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rainer, Franz 2012 étymologie populaire. TLF-Étym 2012 [http://www.atilf.fr/tlf-etym − last access 25 Mar 2014]. Ronneberger-Sibold, Elke 2002 Volksetymologie und Paronomasie als lautnachahmende Wortschöpfung. In: Mechthild Habermann, Peter O. Müller and Horst Haider Munske (eds.), Historische Wortbildung des Deutschen, 105−127. Tübingen: Niemeyer. Ronneberger-Sibold, Elke 2010 “… und aus der Isar steiget der weiße Neger Wumbaba”. Lautgestaltprägende Elemente bei der Schöpfung von Mondegreens. In: Rüdiger Harnisch (ed.), Prozesse sprachlicher Verstärkung. Typen formaler Resegmentierung und semantischer Remotivierung, 87− 106. Berlin/New York: de Gruyter. Rundblad, Gabriella and David B. Kronenfeld 2000 Folk-Etymology: Haphazard Perversion or Shrewd Analogy. In: Julie Coleman and Christian J. Kay (eds.), Lexicology, Semantics and Lexicography, 19−34. Amsterdam/ Philadelphia: Benjamins. Rundblad, Gabriella and David B. Kronenfeld 2003 The inevitability of folk etymology: A case of collective reality and invisible hands. Journal of Pragmatics 35: 119−138. Saussure, Ferdinand de 1967 Grundfragen der allgemeinen Sprachwissenschaft. Ed. by Charles Bally and Albert Sechehaye. 2nd ed. Berlin/New York: de Gruyter. Schmeller, Johann Andreas 1821 Die Mundarten Bayerns grammatisch dargestellt. München: Thienemann. Vennemann, Theo 1999 Volksetymologie und Ortsnamenforschung: Begriffsbestimmung und Anwendung auf ausgewählte, überwiegend bayerische Toponyme. Beiträge zur Namenforschung N.F. 34: 269−322. Vvedenskaja, Ljudmila A. and Nikolaj P. Kolesnikov 2003 Priemy osmsylenija vnutrennej formy (vidy nenaučnoj ėtimologii). Respectus philologicus 3(8). http://filologija.vukhf.lt/3−8/vved.htm [last access 25 Mar 2014]. Wundt, Wilhelm 1900 Völkerpsychologie. Eine Untersuchung der Entwicklungsgesetze von Sprache, Mythos und Sitte. Vol. 1: Die Sprache. Theil 1. Leipzig: Engelmann. Wurzel, Wolfgang Ulrich 1994 Skizze der natürlichen Morphologie. Papiere zur Linguistik 94(1): 23−50.

Sascha Michel, Koblenz (Germany)

1020

Semantics and pragmatics in word-formation I: General aspects

58. Categories of word-formation 1. 2. 3. 4. 5. 6. 7. 8. 9.

Categories of word-formation and function-oriented approaches Describing categories of word-formation Modifying categories Profiling categories Categories of recategorisation Uniting categories Conceptual innovation Historical aspects References

Abstract Categories such as agent noun, place noun, or gender marking, are the oldest, most common and most widely used semantic categories in word-formation, providing a suitable onomasiological basis for crosslinguistic comparison. Among the proposals to group such categories into more general semantic sets, the most well-known − especially in Slavic linguistics − is the one proposed by the Czech linguist Miloš Dokulil, who distinguished transposition, mutation and modification. In the present article, a more refined classification will be proposed.

1. Categories of word-formation and function-oriented approaches The main objective of this article is to present the typologically most prominent categories of word-formation, grouping sets of examples from English and from Slavic, the language subgroup which uses derivation to a greater extent than any other language group at least among the Indo-European languages. Categories of word-formation, i.e. sets of types or rules of word-formation with the same function, are the oldest and to this day the most common function-oriented descriptive tool used in word-formation. We define a function as a meaning component of a constituent form F, which is substituted or added or removed if F is substituted or added or removed. For example, the function ‘the referent is the agent’ is common to English deverbal derivatives with the suffixes -ant (applicant), -er (teacher), -or (demonstrator, sailor), -ar (liar), or Russian derivatives with the suffixes -ant (konsul’tant ‘adviser’), -ar (povar ‘cook’), -ar’ (tokar’ ‘turner’), -ator (restavrator ‘restorer’), -ač (tkаč ‘weaver’), -nik (pomoščnik ‘assistant’), -tel’ (pisatel’ ‘writer’), etc. Most of the articles in chapter VII of this handbook (“Semantics and pragmatics in word-formation II: Special cases”) are dedicated to categories of wordformation of this type. Malkiel (1978: 141−142) called derivational categories the best “technique” for classifying derivational suffixes. A full description of categories of word-formation in Spanish along these lines is given in Rainer (1993: 193−244), who defines a category of word-formation as a set of rules of word-formation having identical meaning. For various definitions of categories of word-formation in Slavic linguistics see Ohnheiser (2000).

58. Categories of word-formation

1021

Bauer (2002) gave a survey of major categories of derivation in his typological sample of more than 40 languages from different language families. He found that a) the most frequent nominal derivational categories are deverbal abstract nouns and deverbal personal/agent nouns, followed by denominal diminutives and abstract nouns derived from nouns or adjectives (Bauer 2002: 40), b) the most frequent verbal derivational categories are deverbal transitives, causatives, intensives, intransitives and denominal verbs (Bauer 2002: 41), c) the most frequent adjectival and adverbial derivational categories are transpositions of other word classes (denominal adverbs and adjectives, etc., cf. Bauer 2002: 42). Not surprisingly, the most frequent derivational categories, especially the verbal, adjectival, and adverbial ones, come close to inflectional categories, in other words, they show a high degree of grammaticalisation. Just as rules of word-formation are grouped together in categories of word-formation, these categories can be further grouped into more general classes on the basis of functional similarities. A first important step in this direction was Kuryłowicz’s (1936) distinction between dérivation lexicale and dérivation syntaxique. Dérivation syntaxique is a change in the primary syntactic function, which is, according to Kuryłowicz, part of the meaning of any content word. This change can be effected not only by derivational suffixes, but also by inflection (e.g., case endings) or context (e.g., word order), the lexical meaning (valeur lexicale) remaining unaffected. With dérivation lexicale additional semantic components come into play, changing the lexical meaning of a content word. It often presupposes dérivation syntaxique: “Quand on dit: la hauteur de cette montagne, il ne s’agit pas de la qualité d’être haut, mais de la dimension verticale, et nous nous trouvons […] en face d’une dérivation à deux étapes: 1 être haut → hauteur (= qualité d’être haut) représente la dérivation syntaxique; 2 hauteur (= qualité d’être haut) → hauteur (= dimension verticale) représente la dérivation lexicale” [When we say: the height of this mountain, we are not speaking about the quality of being high but about the mountain’s vertical dimension, and we are [...] confronted with a derivation in two steps: 1 to be high → height (= the quality of being high) represents the syntactic derivation, 2 height (= the quality of being high) → height (= vertical dimension) is the lexical derivation.] (Kuryłowicz 1936: 86). Notwithstanding the fact that Kuryłowicz is mainly concerned with the syntax of the parts of speech, his theory shows the various forms which a language and languages in general use to change the syntactical and lexical functions of content words. Hence this change is a first step to an onomasiological description of derivation in a broad sense, comprising word-formation. A problem of purely onomasiological descriptions is the degree of granularity of notional categories. Motsch’s “list of the most important elementary predicates” occurring in semantic patterns (“semantische Muster”) comprises 60 items (Motsch 2004: 455−458). The onomasiological descriptors proposed by Deltcheva-Kampf (2000: 321− 355) go down to categories like ‘exam’ (marked by Hungarian -beli and -i) or ‘disease’ (marked in Finnish, Hungarian, and Russian, e.g., R. -anka in vodjanka ‘hydropsy’ or -izm in alkogolizm). -ism in English, denoting a system of believe or theory (e.g., marxism) or a characteristic way of speaking (e.g., malapropism) is an analogous example, cited by Aronoff (1984). Another example is the prefix mag- (plus reduplication) in Tagalog which can denote (with reduplication) ‘vendor of the product designated by the base’ or (without reduplication) ‘two relatives, one of whom bears to the other the

1022

Semantics and pragmatics in word-formation I: General aspects

relation designated by the base’ (Aronoff 1984: 48−49; for more examples see Bauer 2002: passim). Derivational categories like these induce Aronoff to deny the possibility of establishing a neat delimination between the semantics of words (“meanings which are more typical of lexical categories like noun, verb or adjective”) and derivational semantics. As derivational categories cannot be confined to such notions as abstract/ concrete, mass/count, and thematic roles, “even the less restricted theory of the semantics of derivation, which allows reference to syntacticosemantic dimensions, must be untenable” (Aronoff 1984: 48). In an onomasiological description of a specific language, such morphemes figure in a list of idiosyncratic items which lack onomasiological generalisation.

2. Describing categories of word-formation 2.1. Dokulil’s categories The first comprehensive function-oriented theory with a corresponding description of Czech derivation was presented by Dokulil (1962, 1968). His approach, and in particular the derivational operations “modification” and “transposition”, were incorporated into the leading grammars of Polish (Grzegorczykowa, Laskowski and Wróbel 1998) and Russian (Švedova 1980). In a short article, Coseriu (1977) took an analogous path. Raecke (1999) presented Russian word-formation showing the terminological parallelism of Dokulil and Coseriu. Dokulil’s approach had an enormous influence on word-formation research in Slavic studies and in Eastern German linguistics. In Motsch’s (2004) function-oriented monograph about German word-formation, however, there is no mention of Dokulil at all, even though many parallel terms are used due to the analogy of the subject and the onomasiological direction of the description. Neither he nor Fleischer and Barz (1995) make any mention of Kuryłowicz’s (1936) relatively well-known French article. Motsch’s (2004) descriptive basis consists of “word-formation patterns” that are represented in the format of predicate logic and for which case roles have a central function (on pp. 455−458 he provides a list of elementary predicates). As usual in such descriptions, the actual presentation of German word-formation is not fully onomasiological. It is structured according to the word class of the results of word-formation (V, A, N), and within these chapters according to the word class of the bases, e.g., deverbal nouns. Only then are onomasiological categories like “pure nominalisation (action nouns)” and “recategorisation and semantic change” encountered. Dokulil’s onomasiological approach is well-suited to comparative investigations into word-formation (see Ohnheiser 1997, 1987 for a comparison of Russian and German). This possibility is, however, not used in Engel’s (1999) German-Polish contrastive grammar with its mixture of form-oriented and function-oriented descriptions. Nevertheless, derivational categories such as diminutives or action nouns are applied in lists of words and affixes. For Deltcheva-Kampf (2000), Dokulil’s theory provides the basis for the “contrastive-typological analysis” of Finnish, Hungarian and Russian word-formation, supplemented by the concept of functional operations. According to Dokulil (1962: 229), “[t]he onomasiological categories form a conceptual basis in which are grounded the categories of word-formation, language facts in the

58. Categories of word-formation

1023

full sense of the word”. He distinguishes the following types of onomasiological categories (Dokulil 1962: 229 f.; here cited after the English summary, pp. 220−250): “T h e t r a n s p o s i t i o n a l t y p e in which a phenomenon, usually conceived as a mark [in English texts referring to Dokulil, also translated as ‘[onomasiological] feature’ − V. L.] dependent on a substance (or, possibly, as a determination of the mark), becomes conceived as independent of it (or, possibly, as the mark itself). In other words, one has to do here with (a) an objectivisation of quality […]”, i.e. deadjective quality nouns like Cz. rychlý ‘quick’ → rychlost ‘quickness’, (b) “an objectivisation of action”, cf. deverbal action nouns (nominalisations) such as Cz. padat ‘to fall’ → pád, padnutí, padání ‘the fall, the falling’, (c) derivations such as Cz. statečně (žít) ‘(to live) bravely’ → statečný (život) ‘a brave (life)’. “T h e m o d i f i c a t i o n a l t y p e , in which the content of a given concept acquires a supplementary modifying mark”, comprises the following categories: “diminutive” and “augmentative”, “shift of gender”, “mark of minor age”, “collectiveness”, “measure or degree”, “supplementary marks denoting place, direction, time, phase, extent, and especially aspect”, etc. Dokulil’s “fundamental type” (1968: 209), later called m u t a t i o n a l t y p e , is defined by a concept similar to the relation between genus proximum and differentia specifica. The genus proximum can be very explicit (cf. compounds like business-plan), very general (cf. derivatives like writer) or implicit (cf. conversions like the green). In Fleischer and Barz (1995: 8) modification and transposition are accepted as onomasiological categories of word-formation, whereas mutation is not (see also article 134 on German). Dokulil’s examples of derivational categories already contain both purely grammatical types of derivation like verbal aspect and comparison (modificational type) and types with a lower degree of grammaticalisation like action nouns (pure nominalisation; transpositional type). Derivation forms a continuum from lexicon to grammar (see article 14 on the delimitation of derivation and inflection), especially in Slavic languages, where the transition from lexical to grammatical word-formation concentrates on what Dokulil called transposition, which is fully grammaticalised in some languages, e.g., participles or relational adjectives (see section 5), and deverbal nominalisations like Polish czytanie/ E. (the) reading. As shown by the typological findings of Kuryłowicz (1936), blending of syntactical and lexical processes in derivation is possible.

2.2. Functional operations As indicated above, Dokulil’s categories are onomasiological sets of derivational categories. For a full onomasiological system of word-formation his system must be adjusted by extension and differentiation in order to account also for compounding and other types of word-formation with more than one motivating word. This can be accomplished by functional operations, i.e. rules, describing changes of meaning. Whereas the categories “transposition” and “modification” can be looked upon as functional operations, the category “mutation” should be construed as a s e t o f d i f f e r e n t f u n c t i o n a l o p e r a t i o n s . With the help of functional operations, the way a new meaning is made out of one or two other meanings by word-formation or semantic extension (the development of polysemy) can be synchronically reconstructed. Functional operations are a tool

1024

Semantics and pragmatics in word-formation I: General aspects

for the description of not only synchronic, but also diachronic and ontogenetic wordformation and semantic extension (cf. Lehmann 1999: 229−252; some examples for semantic extension are given below). Functional operations are defined below in a traditional manner. If the definition of a functional operation is given as an instruction on how to change the definition of the meaning, we get for the operation of profiling applied to the agent noun writer the following rule: “Change in the definition of the motivating word the relation of subordination between the head (‘to write’) and a subordinate component (‘agent’)”. For more definitions in the form of instructions concerning functional operations see Lehmann (1999: section 4.3). On the one hand, functional operations include several categories of word-formation (or structural patterns in the sense employed by Motsch 2004). On the other hand, they differentiate Dokulil’s basic type, the onomasiological category of mutation. Analysing Dokulil’s category of mutation, we obtain three functional operations: profiling (cf. section 4), conceptual innovation (cf. section 7), and uniting (cf. section 6). Word-formation can result in motivated words corresponding to different categories of word-formation, i.e. in polysemous words: R. golubjatnik 1. ‘pigeon-fancier’, 2. ‘pigeon-hawk’, 3. ‘pigeon-house’ (← golub’ ‘pigeon’); kabotažnik 1. ‘coasting vessel’, 2. ‘coasting-trade sailor’ (← kabotaž ‘coasting trade, cabotage’). Besides such polyfunctional affixes there are, of course, numerous monofunctional affixes, cf. Russian place nouns with the sufix -l’nja ‘room for …’, e.g., spal’nja ‘bedroom’ (← spat’ ‘to sleep’). If a motivating word is polysemous, it is possible that derivation does not operate on all of its meanings. The set of meanings of the derivative writer is not identical with the set of to write. There is no derivation, e.g., from the meanings ‘to make a permanent impression of’, ‘to make evident or obvious’ (guilt written on his face), ‘to force, effect, introduce, or remove by writing’ (write oneself into fame and fortune), ‘to take part in or bring about (something worth recording)’. Derivational categories select specific senses of the motivating word, the agent noun writer selects the meaning ‘to author, to compose’. On the other hand, the motivated word can obtain a meaning that is not part of the motivating word (cf. writer ‘one who writes stock options’). Thus, a category of word-formation, being an onomasiological category, (i) gathers the types of word-formation, (ii) selects meanings in the motivating word, and (iii) functions in derivational morphemes. Especially with polyfunctional affixes, the derivational categories depend on the notional system of the description applied. A higher degree of objectivisation can be obtained when the meanings of monofunctional affixes are generalised and the differentiating power of distribution is used. It is even more objective yet when a cross-linguistic, typological or universal perspective is adopted. Furthermore, categories of word-formation, like any onomasiological description, presuppose a semasiological analysis (from form to meaning).

3. Modifying categories In descriptions of word-formation the term “modification” is widely used for operations that do not affect the essentials of the meaning of the motivating word. For Dokulil, it is by means of modification that “the content of a given concept acquires a supplementary

58. Categories of word-formation

1025

modifying mark [feature]” (Dokulil 1962: 229). The addition of a meaning component is the most frequent, but not the only way to realise modification. Defined as a functional operation, modification refers to the changes in meaning that arise by adding or substituting a component in the motivating meaning without altering the conceptual prototype. The relation between the motivating and the motivated word is therefore hyponymic (addition of meaning, e.g., ‘little, dear’) or co-hyponymic (substitution of meaning, e.g., ‘female’ for ‘male’). D i m i n u t i v e s , h y p o c o r i s t i c s , a u g m e n t a t i v e s , and the marking of degrees (i n t e n s i f i c a t i o n ) are products of modification. Diminutives are formed by adding the semantic component ‘little’, e.g., R. stolik ‘little table’ (← stol ‘table’), E. hillock (← hill), augmentatives by adding the semantic component ‘big, great’ (R. stolišče ‘big table’ ← stol ‘table’; E. superpower), hypocoristics by adding the pragmatic component ‘I want to be near to you (using this word)’ (cf. also Wierzbicka 1992: 251), e.g., R. anekdotec (← anekdot ‘joke’), matuška (← mat’ ‘mother’), bel’eco (← bel’ё ‘underwear, linen’). Pejoratives are formed by adding components like ‘I don’t like the referent of this word’, cf. redneck, faith-head, P. dziewczynisko ‘brat’ (← dziewczyna ‘girl’), babsko/babsztyl ‘jade’ (← baba ‘(old, country) woman’). Furthermore, it is not only in Slavic that diminutives are often functionally clustered with hypocoristic meaning, and augmentatives with pejorative meaning. Usually it is the context which underpins the semantic or the pragmatic component in these clusters. I n t e n s i f i c a t i o n (see article 77), the marking of degrees by adding components like ‘very’, ‘more’, ‘less’, in principle operates on verbs and adjectives, as in E. reddish (← red), R. grubovatyj ‘rather, somewhat rude’ (← grubyj ‘rude’). Some of the socalled aktionsart-derivatives in Slavic have an intensifying meaning. Aktionsarten in Slavic and German linguistics are mostly defined as a type of derivation; they are categories of deverbal verb derivatives, some aktionsarten having grammatical or semi-grammatical status (see section 4 and 5 on the operations of profiling and recategorisation). The other a k t i o n s a r t e n are lexical derivatives, as a rule by modification. There are analogous forms in German, but in Slavic they either have an aspectual partner of their own or don’t have one depending on the language and on the aktionsart involved. Here are some examples from Russian, giving an impression of the types of aktionsart (all derivatives are telic verbs, except for iterative and comitative verbs): a) Finitive action ‘to bring sth. to an end’, e.g., dopet’ (← pet’ ‘to sing’); b) Exhaustive action ‘to perform an action up to an exhausting degree’, e.g., ubegat’sja (← begat’ ‘to run’); c) Saturative action ‘to perform sth. up to a wholly satisfying degree’, e.g., nabegat’sja ‘to have one’s fill of running’ (← begat’ ‘to run’); d) Total action ‘to perform sth. covering all objects or all parts of the object’, e.g., izbegat’ (les) ‘to run all over (the forest)’ (← begat’ ‘to run’); e) Cumulative action ‘performing the action to achieve a considerable amount of sth.’, e.g., nabegat’ (40 km) ‘to cover a total distance (of 40 km)’ (← begat’ ‘to run’); f) Evolutive action ‘to come gradually to the process of doing sth.’, e.g., razbegat’sja ‘to run up, take a run’ (← begat’ ‘to run’); g) Attenuative action ‘to perform sth. to a lesser degree’, e.g., poprideržat’ (← prideržat’ ‘to hold (back)’); h) Intensive action ‘to perform sth. with a high degree of intensity’, e.g., derganut’ (← dergat’ ‘to pull’);

1026

Semantics and pragmatics in word-formation I: General aspects

i) Iterative action ‘to habitually realise the situation’, e.g., siživat’ (← sidet’ ‘to sit’); j) Comitative action ‘to perform a parallel action’, e.g., podpevat’ ‘to sing along (with)’ (← pet’ ‘to sing’). The lexical morphemes forming aktionsart-verbs do not have a local, but rather a nonlocal, qualitative or quantitative meaning. Local prefixes and their semantic derivatives as in G. eintreten 1. ‘to enter (a room, …)’, 2. ‘to join (the EU, …)’ modify the meaning of dynamic verbs, especially verbs of motion. In the Slavic languages they have, as a rule, a grammatical aspectual partner, e.g., R. vojti (perfective) ‘to enter’ → vchodit’ (imperfective), výrezat’ (perfective) ‘to cut out’ → vyrezát’ (imperfective). In the Slavic languages, or, e.g., in German, the local prefixes partly coincide with local prepositions and can denote all sorts of directions and locations. There are other operations of modification as well. Individual nouns can be changed to c o l l e c t i v e n o u n s (see article 66 on collectives), i.e. nouns for groups of persons or things, adding the component ‘group of’ and presupposing the plural, e.g., R./E. krest’janstvo/peasantry (← krest’jane/peasants), listva/leafage (← list’ja/leaves), dubnjak/oakery (← duby/oaks). If s t a t u s n o u n s (see article 73) are products of word-formation, a component ‘status or state of …’ is added, cf. R./E. korolevstvo/kingship (← korol’/king), detstvo/ childhood (← deti/child), družba/friendship (← drug/friend). N e g a t i o n as a device in word-formation (see articles 78 on negation and 79 on negation in the Slavic and Germanic languages) − e.g., E. illogical, discontented, R. nechorošij ‘not good’ (← chorošij ‘good’), bezporjadok ‘disorder’ (← porjadok ‘order’) − adds a negative or privative component to the motivated word and thereby changes the mostly positive default of the motivating word. An operation of modification by semantic substitution is the operation of g e n d e r m a r k i n g (see also article 64). The dynamics in the extension of gender marking differ depending on political and linguistic factors (cf., e.g., Łaziński 2006 for Polish). As the operation of modification has a stricter definition than Dokulil’s category “modification”, most of the “supplementary marks denoting place, direction, time, phase, extent, and especially aspect” (cf. section 2.1), with the exclusion of ‘extent’, do not belong to modification, but to extrinsic profiling (see section 4). Aspect belongs to the operation of recategorisation, profiling and modification (cf. Lehmann 2005), but as it has the status of a grammatical category, its classification is not commented on here. The operation of modification can also apply to semantic extension, but relatively seldom, e.g., E. quality in the meaning ‘high quality’ (cf. a man of quality), R. plavat’ 1. ‘to swim’, 2. ‘to be able to swim’.

4. Profiling categories Profiling categories change the highlighting of a component in the motivating lexical concept. The term profiling goes back to Langacker (1987); in this article it is not used in a manner totally equivalent to Langacker’s use of the term, but is also based on the gestalt concept of figure and ground. The functional operation of profiling consists in shifting the figure-status (the semantic focus, the highlighted component) from one component of a meaning to another component. It applies above all to parts of situations,

58. Categories of word-formation

1027

especially to the arguments of situations denoted by verbs. An example can be seen in agent nouns like E. writer/R. pisatel’/G. Schreiber. With a simple notation the formula for ‘to write’ is ‘WRITE (Agent, Effect)’ with the predicate ‘write’ profiled, while the formula for ‘writer’ is ‘Write (AGENT, Effect)’, where the ‘agent’ is profiled. While ‘WRITE (Agent, Effect)’ corresponds to the action meaning of writing or R. pisanie, the second, metonymical, meaning of the derivatives writing (cf. the writings of Chaucer) or R. pisanie (cf. Svjaščennoe Pisanie ‘Holy Scripture’) profiles the effect: ‘Write (Agent, EFFECT)’. Profiling corresponds to what has also been called topicalisation (cf. Brekle 1970: 127−140). Profiling in word-formation usually applies to the elements of a dynamic situation (including, as seen above, the predicate-argument-structure of a situation and its phases), of a script (time and place), or of a frame (parts and whole). In the typological sample of Bauer the hierarchy of thematic roles is: “agent is more frequent than instrument is more frequent than location” (Bauer 2002: 41). Categories of word-formation profiling arguments: A g e n t n o u n s (see article 74), e.g., E. writer, R. kuritel’ ‘smoker’ (← kurit’ ‘to smoke’); p a t i e n t n o u n s (see article 75), e.g., R. podarok ‘gift’ (← podarit’ ‘to make a present’), P. kochanek ‘lover (beloved)’ (← kochać ‘to love’), P. strata ‘loss’ (← stracić ‘to lose’); i n s t r u m e n t n o u n s (see article 74), e.g., R. ukazatel’ ‘pointer’ (← ukazat’ ‘to point’). Profiling can also apply to the argument structure of motivating adjectives and nouns, cf. R. fokusnik ‘conjurer, person who performs magic tricks’ (← fokus ‘hocus-pocus, trick performed by a magician or juggle’). Categories of word-formation profiling the b e a r e r o f a q u a l i t y : E. weakling, sweetling, etc., G . Fremdling ‘stranger’, Neuling ‘newcomer’, etc. (cf. Baeskow 2002: 624), P. głupiec ‘idiot’ (← głupi ‘stupid’). Deverbal nominalisations (abstract nouns) often develop a second, metonymic meaning by profiling the r e s u l t of the action, cf. E. work, G. Arbeit, R. rabota: ‘to be active in order to obtain a physical or mental product’ → (by word-formation) ‘activity directed toward the physical or mental production of something’ → (by semantic extension) ‘physical or mental product of an activity’; cf. section 5. Categories of word-formation profiling parts of situations (p h a s e m a r k i n g ): When pure phases are marked by affixes, the operation is considered by many authors to be semi-grammatical in Slavic, e.g., R. zaplakat’ ‘to begin to weep’. When it conveys additional meaning, the operation is lexical, in Slavic subsumed under aktionsarten: doigrat’ ‘to play to the end, finish playing’. The operation of profiling makes it possible to reconstruct word-formation as well as metonymic polysemy. In both cases there is a change of semantic focus, i.e. of the profile, on intrinsic or extrinsic components of the meaning, cf. Kremlin for ‘political power of Russia (residing in the Kremlin)’. Profiling by means of word-formation and semantic extension (metonymy) can be combined, cf. E. redneck or R. bel’ё ‘underwear, linen’ → ‘washing laundry’.

5. Categories of recategorisation Recategorisation involves a transfer into another category, whereby the original meaning is superimposed but not deleted. A prominent category consists of words traditionally

1028

Semantics and pragmatics in word-formation I: General aspects

referred to as a b s t r a c t n o u n s , also known as nominalisation (see article 70 on nominalization in Hungarian) or a c t i o n n o u n s (see article 67), e.g., action nouns such as the reading of x, or q u a l i t y n o u n s (see article 72) such as the brightness of x. It corresponds to Dokulil’s onomasiological category of transposition and Motsch’s (2004) “Umkategorisierung” (= recategorisation). Recategorisation also refers to semantic extension by metaphorisation, which in Russian is frequently − more often than in English or German − combined with derivational word-formation, cf. R. grib-ok ‘darning mushroom’ (← grib ‘mushroom’), syn-ok ‘young soldier’ (← syn ‘son’). Examples of transposition given in Dokulil (1962) show that the operation is also used for grammatical (inflectional) marking, e.g., for adjectival participles like E. reading (student)/R. čitajuščij (student) ‘(a) reading (student)’ and gerunds/adverbial participles like E. (sat) reading/R. (sidel) čitaja ‘(sat) reading’. In contrast, deverbal derivatives such as the nouns E. (the) beginning/R. načalo/P. początek ‘the beginning, start’ (← načat’/począć ‘to begin’) have a clear lexical status. Frequently there is a transitional zone between categories with lexical and grammatical status. In Polish, action nouns are grammaticalised, derivable from all verbs, including many aspect partners, cf. the imperfective czytanie (← czytać imperfective ‘to read’) and the perfective przeczytanie (← przeczytać perfective ‘to read’). In Russian there is only a corresponding tendency (Kukla 2013). Not only in Russian, but apparently cross-linguistically, these derivatives have a further tendency: abstract nouns easily form metonymies by implicit profiling. Examples include the result noun the writings (of Chaucer), R. rabota ‘work’ 1. ‘process’, 2. ‘result’; profiling of the bearer of an attribute: E. a beauty/R. krasota/ G. Schönheit. In Russian, 54 % of the 423 deverbal abstract nouns studied by Kukla (2013) show a metonymical secondary meaning, and 31 % of those do so with profiling of the result, so that the derivational suffixes are not purely recategorisation markers. A similar statement can be made about Slavic denominal r e l a t i o n a l a d j e c t i v e s (see article 43 on compounds and multi-word expressions in Slavic), e.g., R. gorodskoj ‘town (adj.), urban, municipal’ (← gorod ‘town’), G. städtisch (← Stadt ‘town’). Traditionally they are dealt with in connection with lexical derivation. However, in Russian, for example, they have a grammatical status since in principle they can be derived from any noun (as long as the nouns are not products of recategorisation themselves). Relational adjectives are often reinterpreted as qualitative adjectives that are then gradable, whereas “pure” relational adjectives are not. In general, products of recategorisation often lose certain grammatical categories which are typical of the respective part of speech: relational adjectives cannot form the comparative, abstract nouns cannot be pluralised, when used with their standard meaning. In Slavic languages, other types of recategorisation do not have a tendency towards grammaticalisation, e.g., verbs motivated by adjectives like belet’ ‘to be white’ (← belyj ‘white’). Recategorisation is also the functional operation that concerns homogeneity, the “count/mass distinction”. It is responsible for changing homogeneous nouns and verbs to heterogeneous ones. These changes are not caused by modification through substitution as is done with diminutives or augmentatives, gender or negation, where the prototype is preserved and only receives the additional attributes ‘little’ or ‘big’, ‘negative/ opposite’, or the alternative attribute ‘male’ or ‘female’, etc. The meanings of the homogeneous snow and heterogeneous snowflake belong to different categories; there is not a common prototype with additional or alternative attributes.

58. Categories of word-formation

1029

By deriving s i n g u l a t i v e s (see article 65) from mass nouns (words for homogeneous substances) the latter are changed to individual nouns (count nouns, nouns for heterogeneous substances, for individuals), R. solomina ‘culm’ (← soloma ‘straw’), snežinka ‘snowflake’ (← sneg ‘snow’), malinina ‘raspberry (singulative)’ (← malina ‘raspberry (mass noun)’). The English noun snowflake shows that an individualised part of a homogeneous mass can be designated by a composite noun. But as it is motivated by different nouns referring to two distinct concepts, the word-formation of snowflake is a case of uniting categories, cf. section 6, while R. snežinka refers to the same concept as the motivating sneg. Recategorisation can also apply to grammatical categories of the verb, cf. progressive aspect be sitting (← non-progressive aspect sit); R. perfective aspect posidet’ (← imperfective aspect sidet’ ‘to sit’). Formerly this process in Slavic was regarded as the lexical derivation of a “temporal aktionsart”, nowadays its grammatical status is widely acknowledged. The purely semantic analogues corresponding to derivational recategorisation in word-formation are metaphors. When mushroom is used for all sorts of artefacts that have the form of a mushroom, a meaning is shifted from the category of plants to the category of artefacts, whereby the original meaning is not deleted, but only superposed.

6. Uniting categories The operation of uniting (or semantic combination) consists in forming a new lexical unit by combining two lexical concepts. Various types of word-formation are available for the explicit procedure of uniting where both motivating concepts are conveyed by the motivating words, with compounding as the central type which has been the subject of much linguistic research (see articles 33 on synthetic compounds in German, 34 on verbal pseudo-compounds in German, 38 on noun-noun compounds in French, 39 on verb-noun compounds in Romance), including the rather complex meaning construction processes (see article 63 on noun-noun compounds). Consider the following examples: a) compounds such as E. keyboard, fire-engine, swimming pool; R. biznes-plan ‘business plan’, biznes centr ‘business centre’, biznesspecializacija ‘business specialisation’; b) formations at the boundary between syntax and word-formation, both constituents of which are inflected (in Slavic studies also called “binomina”), e.g., R. aktёr-direktor, cf. F. acteur-directeur (see Bergmann 2006, whose investigation is also based on the concept of functional operations); c) various forms of abbreviations (NATO, UK, Oxfam). The explicitly motivated operation of uniting can in principle be motivated by any type of content words, cf. E. goodlooking, everybody, output, runaway, tonight; R. sineglazyj ‘blue-eyed’. (a)−(c) are default-cases of uniting. When uniting is not motivated by more than one content word, an implicit lexical concept is added, cf. R. korovnik ‘cowshed’ (← korova ‘cow’), zadačnik ‘problem book’ (← zadača ‘problem, task’), spal’nja ‘bedroom’ (← spat’ ‘to sleep’). Sometimes it is difficult to decide between the operations of profiling and uniting, i.e. to decide whether or not a semantic component is part of the motivating concept. A possible empirically based solution can be found in association tests: the motivating stimulus words korova, zadača, spat’ don’t have reaction words with the meaning of shed, book, or room (cf. Russkij associativnyj slovar’).

1030

Semantics and pragmatics in word-formation I: General aspects

The operation of uniting is also applied to elliptical derivations from multi-word expressions that are called “univerbation” in Slavic word-formation research (see article 42), e.g., R. večerka (← večernjaja gazeta ‘evening newspaper’), maršrutka (← maršrutnoe taksi ‘route taxi’). The derivatives from numerals can have different implicit complements, e.g., R. dvojka (← dvoe collective numeral ‘two’) 1. ‘number’, 2. ‘various types of public transport such as a tram or a bus (according to their number)’, 3. ‘grade/ mark in school (in Russia: ‘poor’)’, 4. ‘type of boats (‘pair-oar boat’), etc.’, 5. ‘playing cards (deuce)’. It can be necessary to combine various functional operations in order to reconstruct products of word-formation. For example, E. smoker/R. kuril’ščik ‘a person who habitually smokes tobacco’ (← to smoke/kurit’) exhibit profiling for an agent noun together with the modification ‘habitually’. Many operations of uniting are combined with the concurrent operation of profiling, e.g., with the profiling of the agent (salesperson), patient (stockbroker), or instrument (boat trip). Profiling, then, is a secondary operation together with the primary operation of uniting. In crybaby, being derived from ‘some baby crying’, the agent baby is profiled (topicalised, see Brekle 1970: 131). For handling uniting in cases that require more than one functional operation (or a problematic decision between uniting and another operation), a form-oriented preference can be applied (see Lehmann 1999: section 4.4): For the description of compounds and other words motivated by two words, preference goes to uniting, for derivatives motivated by one word, preference goes to an operation other than uniting.

7. Conceptual innovation Conceptual innovation is one of three general types of l e x i c a l i n n o v a t i o n s with formal changes: 1. Formal innovations without conceptual changes: Only the formal structure of the lexical unit is changed by one of the types of abbreviation. 2. Conceptual alterations: Word-formation changes an existing lexical concept by the functional operations of modification, recategorisation, profiling, or by combining existing lexical concepts by operations of uniting. 3. C o n c e p t u a l i n n o v a t i o n s : New lexical concepts are introduced into the lexicon using word-formation to denote the new concept. We call it conceptual innovation, because this is n o t o n l y a n a l t e r a t i o n o f a g i v e n c o n c e p t , as is brought about by functional operations, b u t t h e a d d i t i o n o f a n e w i t e m t o t h e c o n ceptual system. The quantity of the types of lexical innovations has been very different in the history of the lexicon (see section 8), but let us first take a look at conceptual innovation. It is an operation which consists in assigning a whole concept to one or more motivating bases, resulting, for instance, in a scientific term or an expressive speech act. The denotative component of the added concept is not motivated by an existing concept, there is only a motivational link between the connotation of the motivated word and the motivating concept. Hence, the motivation is very weak. The given definition does not preclude that an analogous concept already exists in the language or that there is a synonym for the product of a conceptual innovation. In Lehmann (2004) terminological and expressive

58. Categories of word-formation

1031

innovations are mentioned as types of conceptual innovations. Chemical, medical, biological and other terms can often be recognised by certain suffixes: E. germanium, indium, mendelevium, cubanite/barracanite, gagarinite or R. germanij, indij, mendelevij, plutonij, kubanit, gagarinit are names for chemical elements with the suffixes -ium, -ite, or -ij, -it. In structuralism the suffix -eme was favoured (phoneme, morpheme, grammeme, etc.). Kanngießer (1987), who postulates a continuum of motivational compositionality for compound nouns, mentions the word Hausberufung ‘internal appointment as a professor at a university’ (← Haus ‘house’, Berufung ‘calling’) − a term from the German academic system − as an example for a compound that is not motivated compositionally. Scientific terms are usually not formed by changing a given concept. Instead, a word is sought that can be assigned to a definition, a word that can serve as the communicative carrier of this meaning. If Gagarin acts as the motivating word for gagarinit, then only because of the connotation connected with the name. If a new word is originally based on profiling, e.g., using a discoverer’s name as an eponym for the object discovered, cf. E. dahlia, named after the botanist Andres Dahl, and if it is demotivated historically, it functions synchronically like other conceptual innovations. You must have the concept of a specific plant in order to understand the word dahlia. The motivating base is no help to the denotative semantics (definition) of the motivated word. Even if you know or assume that dahlia is formed in accordance with an eponymic pattern. Arbitrariness is also a feature of expressive innovations as they occur in swear words, cursing, maledictions, and other pragmatic words. Rammelmeyer (1988) has described derivatives that cannot be labelled with any of the usual motivational relations, but are characterised by a strong expressiveness. Rammelmeyer distinguishes derivatives whose motivating words (i) are stylistically neutral, such as R. pereborščat’ ‘to exaggerate’ (← boršč ‘red-beet soup’), (ii) have an expressive connotation, such as R. vtreskat’sja ‘to fall in love’ (← treskat’ ‘to crack; to bust; to form cracks; (vulgar) to batter’), and (iii) are expressive themselves like R. vyebyvat’ ‘to leave, fuck off’ (← jebat’ ‘to fuck’). When fucking represents an expressive concept, especially when it is used as an infix in words like unfuckin(g)believable, fanfuckin(g)tastic, absofuckin(g)lutely, etc., there is no denotative component which could be motivated. Only its expressive content is motivated by expressive components of the verb to fuck. In both of these lexical domains, neologisms are entirely or strongly characterised by conceptual innovations. In addition, there are many further weakly or purely formally motivated words of this kind. However, these must be differentiated from other real or apparent derivatives that lack motivation, i.e. from demotivation and remotivation. Demotivation is present when the content of an originally motivated word loses its connection to the motivating word(s), e.g., E. hallmark ‘mark of quality’ (< ‘official stamp of purity in gold and silver articles’ < ‘mark from Goldsmiths’ Hall in London, site of the assay office’; cf. http:// www.etymonline.com/index.php, 23. 12. 2011). In order to distinguish conceptual innovations and demotivations, etymological studies have to be employed. After all, the reconstruction of motivations is a linguistic task and not a mental process.

8. Historical aspects Havránek, one of the most significant proponents of the Prague linguistic circle, emphasised intellectualisation as a line of development of standard language. By intellectualisa-

1032

Semantics and pragmatics in word-formation I: General aspects

tion of a standard language, the Prague linguists understood “its adaptation to the goal of making possible precise and rigorous, if necessary abstract, statements expressing the complexity and interaction of thoughts” (Havránek 1983: 147 [1932: 45]). The term is useful for the interpretation of some diachronic facts of word-formation. Russian word-formation, in areas such as trade, transportation, agriculture, law, and art in the period of intellectualisation, has been researched in a series of theses and term papers at the University of Hamburg. As has become obvious, conceptual innovations make up the smallest proportion of word-formation operations (for more information see Lehmann 1999: 232−238). The bulk of the data concerns recategorisation, modification, and profiling. Evidently, these functional operations serve primarily to make the syntax of content words more flexible (especially the operation of recategorisation) or to make implicit components of lexical concepts explicit, i.e. to give them a linguistic form (especially the operation of modification and profiling). These changes primarily serve the modus operandi of a language as well as the effectiveness of denomination processes thus expanding the possibilities to express particular contents by differentiated and effective means. Intellectualisation characterises the stage in the development of a language whose elementary means are already constituted and a system of lexical concepts has already been established, primarily by conceptual innovation. In the following stage of intellectualisation of a language like Russian on its way to becoming a standard language, central functional procedures are the derivationally based operations of recategorisation, modification, and profiling. For the 20th century, when intellectualisation is, in principle, achieved and a standard language established, the data show an increase in the use of the operation of uniting and an overwhelming majority of all types of abbreviations. Thus, for the lexicon of Russian and other Slavic languages, and it might be reasonably assumed for the lexicon of European standard languages in general, typological differences notwithstanding, we can suppose a development from lexical dynamics with an emphasis on conceptual innovations (and an abundance of synonymic word-formation in many languages especially during the middle ages), via a period of intellectualisation with alterations of lexical concepts by functional operations much more than with conceptual innovations, up to modern times with a predilection for changes of graphemic/ phonemic forms only and with an increasing number of conceptual innovations in scientific and technical varieties. Along with semantic extension and borrowing, word-formation is the key instrument used in expanding the vocabulary. The description of lexical expansion on the basis of the categories of word-formation and functional operations can show the development this expansion takes with regard to content.

9. References Aronoff, Mark 1984 Word formation and lexical semantics. Quaderni di Semantica 5(1): 45−50. Baeskow, Heike 2002 Abgeleitete Personenbezeichnungen im Deutschen und Englischen. Kontrastive Wortbildungsanalysen im Rahmen des Minimalistischen Programms und unter Berücksichtigung sprachhistorischer Aspekte. Berlin/New York: de Gruyter.

58. Categories of word-formation

1033

Bauer, Laurie 2002 What you can do with derivational morphology. In: Sabrina Bendjaballah, Wolfgang U. Dressler, Oskar E. Pfeiffer and Maria D. Voeikova (eds.), Morphology 2000, 37−48. Amsterdam/Philadelphia: Benjamins. Bergmann, Anka 2006 Binomina im Russischen als Kategorie der komplexen Benennung. Frankfurt/M.: Lang. Brekle, Herbert Ernst 1970 Generative Satzsemantik und transformationelle Syntax im System der englischen Nominalkomposition. München: Fink. Coseriu, Eugenio 1977 Inhaltliche Wortbildungslehre (am Beispiel des Typs “coupe-papier”). In: Herbert E. Brekle and Dieter Kastovsky (eds.), Perspektiven der Wortbildungsforschung. Beiträge zum Wuppertaler Wortbildungskolloquium vom 9.−10. Juli 1976 anläßlich des 70. Geburtstags von Hans Marchand am 1. Oktober 1977, 68−61. Bonn: Bouvier. Crystal, David 2003 The Cambridge Encyclopedia of the English Language. Cambridge: Cambridge University Press. Deltcheva-Kampf, Veronika 2000 Onomasiologisches Modell für eine kontrastiv-typologische Betrachtung des suffixalen und kompositionellen Wortbildungsbereichs (am Beispiel des Finnischen, Ungarischen und Russischen). Wiesbaden: Harrassowitz. Dokulil, Miloš 1962 Tvoření slov v češtině. Vol. 1: Teorie odvozování slov. Praha: Nakladatelství Československé Akademie Věd. Dokulil, Miloš 1968 Zur Theorie der Wortbildung. Wissenschaftliche Zeitschrift der Karl-Marx-Universität Leipzig. 17. Gesellschafts- und sprachwissenschaftliche Reihe 2/3: 203−211. Engel, Ulrich (ed.) 1999 Deutsch-polnische kontrastive Grammatik. Heidelberg: Groos. Fleischer, Wolfgang and Irmhild Barz 1995 Wortbildung der deutschen Gegenwartssprache. 2nd ed. Tübingen: Niemeyer. Grzegorczykowa, Renata, Roman Laskowski and Henryk Wróbel (eds.) 1998 Gramatyka współczesnego języka polskiego. Morfologia. Warszawa: Państwowe Wydawnictwo Naukowe. Havránek, Bohuslav 1932 Úkoly spisovného jazyka a jeho kultura. In: Bohuslav Havránek and Miloš Weingart (eds.), Spisovná čeština a jazyková kultura, 32−84. Praha: Melantrich. − English translation: The functional differentiation of the Standard language. In: Joseph Vachek and Libuše Dušková (eds.) 1983 Praguiana. Some Basic and Less Known Aspects of the Prague Linguistic School, 143−164. Amsterdam/Philadelphia: Benjamins. Kanngießer, Siegfried 1987 Kontingenzräume der Komposition. In: Brigitte Asbach-Schnitker and Johannes Roggenhofer (eds.), Neuere Forschungen zur Wortbildung und Historiographie der Linguistik. Festgabe für Herbert E. Brekle zum 50. Geburtstag, 3−30. Tübingen: Narr. Kukla, Julia 2013 Das Verb und sein Abstraktum im Russischen. München: Sagner. Kuryłowicz, Jerzy 1936 Dérivation lexicale et dérivation syntaxique: Contribution à la théorie des parties du discours. Bulletin de la Société Linguistique de Paris 37: 79−92. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Vol. 1. Stanford, CA: Stanford University Press.

1034

Semantics and pragmatics in word-formation I: General aspects

Łaziński, Marek 2006 O panach i paniach. Polskie rzeczowniki tytularne i ich asymetria rodzajowo-płciowa. Warszawa: Wydawnictwo Naukowe PWN. Lehmann, Volkmar 1999 Sprachliche Entwicklung als Expansion und Reduktion. In: Tanja Anstatt (ed.), Entwicklungen in slavischen Sprachen, 169−254. München: Sagner. Lehmann, Volkmar 2004 An den Grenzen der Motiviertheit: Zur funktionalen Beschreibung von Wortbildung und Polysemierung. In: Volkmar Lehmann and Ludger Udolph (eds.), Normen, Namen und Tendenzen. Festschrift für Karl Gutschmidt zum 65. Geburtstag, 65−85. München: Sagner. Lehmann, Volkmar 2005 Grammatičeskaja rekonstrukcija i leksikografija russkogo vida: Profilirovanie i drugie funkcional’nye operacii. In: Volkmar Lehmann (ed.), Glagol’nyj vid i leksikografija. Semantika i struktura slavjanskogo vida 4, 191−233. München: Sagner. Malkiel, Yakov 1978 [1966] Derivational categories. In: Joseph H. Greenberg (ed.), Word Structure, 126− 149. Stanford, CA: Stanford University Press. Motsch, Wolfgang 2004 Deutsche Wortbildung in Grundzügen. 2nd ed. Berlin/New York: de Gruyter. Ohnheiser, Ingeborg 1987 Wortbildung im Sprachvergleich. Russisch − Deutsch. Leipzig: Enzyklopädie. Ohnheiser, Ingeborg 1990 Neologismen in ihrem Verhältnis zu Wortarten und Wortbildungskategorien. Zeitschrift für Slawistik 35(6): 811−815. Ohnheiser, Ingeborg 1997 Dokulils Konzeption zur vergleichenden Wortbildung der slawischen Sprachen für den VI. Internationalen Slawistenkongreß 1963 − Versuch einer Bilanz 1998. Wiener Slavistisches Jahrbuch 43: 149−160. Ohnheiser, Ingeborg 2000 Wortbildungskategorie. In: Herbert Jelitte and Nina Schindler (eds.), Handbuch zu den modernen Theorien der russischen Wortbildung, 451−453. Frankfurt/M.: Lang. Raecke, Jochen 1999 Wortbildung. In: Helmut Jachnow (ed.), Handbuch der sprachwissenschaftlichen Russistik und ihrer Grenzdisziplinen, 150−181. Wiesbaden: Harrassowitz. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rammelmeyer, Mathias 1988 Emotion und Wortbildung: Untersuchungen zur Motivationsstruktur der expressiven Wortbildung in der russischen Umgangssprache. In: Bernd Harder and Hans Rothe (eds.), Gattungen in den slavischen Literaturen. Beiträge zu ihren Formen in der Geschichte, 185−208. Köln/Wien: Böhlau. Russkij associativnyj slovar’ online http://tesaurus.ru/dict/dict.php [last access 20 June 2013]. Švedova, Natal’ja Jur’evna (ed.) 1980 Russkaja grammatika. Vol. 1. Moskva: Nauka. Wierzbicka, Anna 1992 Personal names in expressive derivation. In: Anna Wierzbicka, Semantics, Culture, and Cognition. Universal Human Concepts in Culture-Specific Configurations, 225−307. New York: Oxford University Press.

Volkmar Lehmann, Hamburg (Germany)

59. Schemata and semantic roles in word-formation

1035

59. Schemata and semantic roles in word-formation 1. 2. 3. 4. 5.

Introduction Semantic role and schema in the history of word-formation research Semantic role and role schema in compound-analysis The properties of schemata, models and patterns and their interrelations The methodological and epistemological value of analyzing the use of schemata, models and patterns 6. Conclusion 7. References

Abstract In the 60’s of the twentieth century, theories of cognition and theories of sentence semantics converged and were increasingly applied to a hitherto neglected subject, i.e. wordformation research. This article reconstructs the historical and theoretical background of two then dominant descriptive categories, those of schema and semantic role, setting them off against related concepts such as pattern and model. The broad spectrum of possible uses of these categories in research on word-formation is illustrated. The specific properties of schemata, models and patterns and their interrelations show the concepts to be of central significance in the description of compounds in particular. The methodological value of schemata, models and patterns can be demonstrated ex negativo by analyzing mistakes, further stylistically unusual formations and variants and finally extreme manifestations of economy of expression.

1. Introduction Schemata based on semantic roles not only allow the abstracting of propositional relations in descriptive sentences but also the modeling of relations in word-formations as naming units, especially in compounds. The relationship of the constituents within a naming unit, i.e. part-whole or whole-part relationships constitute the dominant processing dimension, establishing the conceptual link to world knowledge, the formation of neologisms as well as the (re-)use and the interpretation of compounds. The schema-role conception, as explicated in the following, provided the basis for the project “Deutsche Wortbildung” [German Word-Formation], which was undertaken at the University of Innsbruck by the Innsbruck branch of the Institut für Deutsche Sprache, Mannheim [Institute for the German Language] (cf. Pümpel-Mader, Ortner and Müller-Bollhagen 2014). Only on this basis could word-formation be systematically described from a semantic point of view also in the field of compounding. After considering the history of this approach, the newly gained possibilities of typologization are exemplarily presented.

1036

Semantics and pragmatics in word-formation I: General aspects

2. Semantic role and schema in the history of word-formation research The concepts “semantic role” and “schema” were developed in theoretical linguistics and later adopted by research in word-formation in the early 70’s of the twentieth century. It was a time of the crossing of boundaries, of new approaches and experiments with much that was (conceptually) new. But the new did not merge into a unified theory. The theory of the time was marked by the appeal and dominance of Chomsky’s generative approach, according to which “die Grammatik zugleich als linguistisches Abbild des Sprachsystems und als Beschreibung der Sprecherkompetenz aufgefaßt wird” [the grammar is understood both as a linguistic representation of the language system and as description of the speaker’s competence] (Helbig 1986: 114). The Chomskyan revolution unsettled and challenged well-established research traditions, in particular the traditions following the Neogrammarians (Junggrammatiker) in continental Europe (in many cases still modeled on historical linguistics) and those inspired by structuralist linguistics, which achieved considerable success especially in word-formation research, though almost exclusively in the field of affixation. It was the life-time achievements of Wolfgang Fleischer and Johannes Erben that were paradigmatic for research in German word-formation, as was Hans Marchand’s reasearch in English word-formation. They set new standards, but at the same time retained their connections with the European tradition. It was Chomsky’s own criticism as well as the attempt at the development of Chomskyan theories by his critics that was at least as influential in the history of ideas as Chomsky’s original new approach (cf. Helbig 1986: 102−110). One of his critics was Fillmore, who introduced the concepts of “covert categories”, “semantic deep structure” and “deep (structure) case”: “The substantive modification to the theory of transformational grammar which I wish to propose amounts to a reintroduction of the ‘conceptual framework’ interpretation of case systems, but this time with a clear understanding of the difference between deep and surface structure” (Fillmore 1968: 21, cf. also 2, 5, 32, 88). With the introduction of the concept of deep case in syntax, a new descriptive level in terms of sentence semantics was adopted and is now unanimously recognized in any linguistic and psycholinguistic approach as one of many autonomous levels (i.e. as an independent module). There is, however, no consensus as to whether syntax or semantics/logic or utterance/text provides the function of guiding level (cf. Helbig 1986: 114− 116). The attribution of “deeper” structures to the linguistic surface structures, as paradigmatically defined by Chomsky and Fillmore, has become standard procedure in grammatical and/or (psycho-)linguistic description. Prior to Chomsky, European linguistics had left the “gnoseologisch-logischen Kategorien” [gnoseological-logical categories] (Dokulil 1964: 218) to the psychology of language. Yet, after Chomsky they had to form part of linguistic description. “Man unterscheidet zumindest zwischen den Ebenen der gedanklichen Planung [our emphasis], der grammatischen und phonologischen Enkodierung sowie der artikulatorischen Realisation von Sprechbewegungen” [A minimal distinction is made at least between the level of mental planning, of grammatical and phonological encoding as well as of the articulatory realization of speech movements] (Bosshardt 2003: 454). Since the revolution of deep semantics, traditional structuralist descriptions have been regarded as mere descriptions of surface structures: typologizations without a generativist

59. Schemata and semantic roles in word-formation

1037

foundation have been considered incomplete. The pre-Chomskyan rigid demarcation of the psychology of language has started to blur. The concepts and approaches derived from prior syntactic discussion have − at least tentatively − been adopted by wordformation research. For word-formation has always been considered as a syntax-related phenomenon (see article 45 on rules, patterns and schemata in word-formation). During the 1960’s, generativist impulses met with a tradition in word-formation research centred on models and patterns, while the central concepts of pattern and model had yet to be made theoretically explicit. They were used in an everyday or in an educated sense. Compare Fleischer for the significance of the concept model: “Die Zahl der Modelle ist begrenzt und überschaubar, die Zahl der nach diesen Modellen zu verbindenden Einheiten ist es nicht” [The number of models is limited and manageable, but the number of units to be combined in accordance with the models is not] (Fleischer 1978: 78). These models are also referred to as “Wortbildungsmuster” and “Wortbildungsparadigma (Funktionsstand)” [word-formation pattern and word-formation paradigm (functional word-formation class)] (Erben 2006 [1975]: 42) and “established patterns”/“types” (Marchand 1969: 9). (In contrast Hansen 1978 makes a clearer distinction between the terms word-formation pattern and word-formation type, cf. Kastovsky 2005: 108.) What was first defined by historical grammar was continued in structuralist linguistics, but could more easily be realized in studies on derivation than on compounding. Only in the case of types that were formally identifiable, i.e. by affixation, did word-formation research − especially its sense-related approach − venture into the field of semantics and systematically describe the semantic conditions within and the semantic interrelations among the distinct formal models (see article 5 on word-formation in inhaltbezogene Grammatik). This represented classic research on dominant paradigms/models with a focus on practice rather than theory. Its subject was the very restricted local phenomenon of isolated word-formation constructs. Due to its limitation to affixation, its material basis remained relatively manageable. In most contributions, the focus on affixation corresponded to the absolute dominance of superordinate, formal/morphological aspects. Compare for example the 3:1 ratio in the number of pages relating to prefixation/suffixation and to compounds in Marchand’s standard work (Marchand 1969: 85−156, 157− 289, 11−79; cf. also Henzen 1965: 98−108, 109−233, 36−97). The description according to (simple) morphological models/patterns formed the focus. In the 20th century the phenomenon of productivity was also more central in wordformation research than the phenomenon of creativity (as defined by Chomsky) was in syntax. This was largely due to the straightforwardness of the material and the simple morphological isolatability of the string of signs (cf., e.g., Fleischer 1969: 66−67; Fleischer and Barz 1992: 57−59). Yet, the “laws” governing productivity could not be discovered, and the few rules that were found did not allow a systematization within an algorithm-based theory that would have enabled a rule-based deduction. “Die DeterminantendesModellssindoffenbarzueinembeträchtlichenTeileinschränkend-negativ,alssogenannte ‘Restriktionsregeln’ zu fassen: keine Verbindung heimischer Basis mit nicht heimischem Suffix (neben verfügbar nicht *verfügabel, wohl aber neben disponibel auch disponierbar)” [Obviously, the determinants of the model are, to a great extent, to be regarded as negatively restrictive, as so-called ‘restriction rules’ such as: A native basis cannot be combined with a non-native suffix (verfügbar ‘available’ cannot be replaced by *verfügabel, but disponibel ‘disposable’ can be replaced by disponierbar)] (Fleischer

1038

Semantics and pragmatics in word-formation I: General aspects

1979: 76). Yet with respect to the total quantity of word-formation constructs, the attempt to describe word-formation as in principle regular fails, even when limited to the field of affixation. There are more exceptions than rules and, according to Erben, also “Affinitäten (Bindungsneigungen)” [affinities (binding tendencies)] in addition to “Einschränkungen [...] (Restriktionsregeln)” [restrictions [...] (restriction rules)] (Erben 2006 [1975]: 42). How can affinities, i.e. binding tendencies, but not strong regularities, be represented in a theory, such as that covering the phenomenon of strong or weak patterns that has always been recognized in word-formation? Prototype semantics and Wittgenstein’s use theory provide an answer to these issues that can only tentatively be used in word-formation research. Apparently their solutions are incompatible with the established generativist tradition: their models are indeed always formation models postulating quasi-regularity. Within this context, arbitrariness, qualitative leaps, and (system-)incompatible restrictions are generally not admissible, in contrast to cognitive science, which in the 90’s of the twentieth century added a connectionist perspective to its generativist perspective (e.g., Anderson 1996: preface, 2010: 374−375). Connectionism enables the conceptualization of parallel − but not homologous! − processing, besides and/or instead of sequential processing (cf. Kuhl 2001: 350; Keller 1990). In the connectionist approach part/whole dichotomies on multiple levels, in different modules, are to be considered parallel to the traditional part/whole dichotomies on one level (such as on the semantic level the relationships of the constituents to the schema of the semantic roles), in one dimension and in one module. In addition to the dichotomies, i.e. peripheral vs. central and local vs. global, this extension once again raises the urgent question regarding the guiding level. “Als Leitebene bezeichne ich den variablen Ausschnitt des Geschehens, auf den die bewußte Aufmerksamkeit des Handelnden zugespitzt ist. Komplexe Handlungen mit hohem Koordinationsbedarf beziehen einen beträchtlichen Teil ihrer Schwierigkeit daraus, daß man erst lernen muß, ihre Steuerung auf eine Leitebene zuzuspitzen” [What I refer to as the guiding level is the variable part of the event receiving most of the actor’s conscious attention. Complex processes with an urgent need for coordination derive a great deal of their difficulty from the fact that we first have to learn how to control them on a guiding level] (Knobloch 1994: 56). Another dichotomy, namely that of guiding and subordinate level (layer), arises, and with it the perspective of an almost chaotic interactivity, also reinforced by the fact that the constituents have to be subsymbolic elements, as shown by the analysis of errors and slips. Only by assuming such elements are we able to explain mistakes like the following: sah sie mit eisernem Blick an ‘looked at her with an iron (instead of eisigem ‘icy’) stare’ (taken from a school essay). The fact that the whole is in, and prior to, the part and that it is more than the sum of its parts, as previously formulated by Gestalt psychology, can reasonably be modeled from a connectionist perspective, whereas these modelings correspond to the multiple activation and multiple path approaches of cognitive science (cf. the modeling of intuition in Kuhl 2001: 626). The connectionist approach brings great benefits to compounding studies, such as the following insight: the constituents of compounds are peripheral, the models central − but sometimes it is the other way around. Referential semantics dominates constituent semantics and can even override it, e.g., in the case of metaphorical use, cf. Modezar vs. Zar ‘fashion king; lit. fashion tsar’ vs. ‘tsar’. The whole, i.e. the naming unit within the utterance in its context provides the central coordination.

59. Schemata and semantic roles in word-formation

1039

What proves even more important is that, through the connectionist models, the concept of network enters semantic research: in connectionism, apart from sequential-syntagmatic-algorithmic processes, spontaneous and chaotic processes in parallel-distributed networks become just as conceivable as incremental co-occurrence and qualitative leaps, e.g., in pattern recognition (cf. Kuhl 2001: 636). The conception of connectionist networks provided for the modeling of prototypicality, family resemblance as well as pattern and schema overlay as essential for word-formation. The clearer it becomes that each cognitive unit and each cognitive process is something highly complex, for which all cognitive abilities, all means and all operations have to be employed, the more urgently we need the concept of schema overlay. At first and long before something emerges as a lexical unit, these are tacted subsymbolic entities (e.g., semantic features), cf. mistakes like Gelähmten-Klo instead of Behinderten-Klo ‘toilet for the paralyzed’ instead of ‘for the disabled’ or angehortet instead of gehortet ‘hoarded’, i.e. a blending of gehortet ‘hoarded’ and angesammelt ‘gathered’: Wenn wir eine bestimmte Summe von Waren angehortet haben […] ‘As soon as we have hoarded a certain amount of goods […]’ (Lissmann in Österreichischer Rundfunk, Ö1, 23. 9. 2011). In view of the developments in the history of recent research, semantics has finally received due recognition in linguistic description. Contrary to the original Chomskyan intention, which was syntactically determined, semantics has achieved the status it deserves in developing from mere referential semantics to production and reception semantics. This has proved a long haul. Wittgenstein’s central concepts have had to be integrated to create a new semantics based on the ideas of “language game”, “use-theory of meaning”, the concept of family resemblance and the critique of essentialism, etc. In order to understand the formation of categories, a number of the basic Aristotelian concepts had either to be abandoned, and/or adapted, and/or extended. As it gradually established itself, prototype theory thus revolutionized the long tradition of psycholinguistic research on categorization as an “alternative au modèle classique de la catégorisation” [alternative to the traditional model of categorization] (Kleiber 1990: 150). Yet, prototype theory also went through phases of revision until it eventually arrived at the principle of family resemblance (cf. Kleiber 1990: 149−165). In word-formation research this idea still had to be radicalized in terms of level, layer and dimension: the “members of a category” on one level, on one layer and in one dimension “may be related to one another without all members having any properties in common that define the category” (Lakoff 1987: 12). This is the only way to describe and explain the syncretic chaos of model use and of the limits of model application. Moreover, a newly developed standard theory of semantics had to develop from an essentially monolithic semantics into a semantic theory that recognized subsymbolic processes in networks as constitutive, and it also had to develop from a semantics focusing on isolated units into a holistic semantics. The keywords here are frame, script and text/utterance. Text linguistics goes far beyond sentences as the largest unit of grammatical description (see article 123 on word-formation and text). From the 1970’s to 1995 these main currents in the history of ideas met with the still dominating practice of research that considered the formal aspect as a major criterion and only included semantic aspects in the area of (semi-)affixations (e.g., collectivity or augmentation), but not in the field of compounds. This was at the expense of the distortion of the descriptive proportions. Due to this approach, only little attention, effort and

1040

Semantics and pragmatics in word-formation I: General aspects

space was devoted to the description of the most frequent phenomenon of present-day language, i.e. compounding, whereas there was much focus on derivation. Also, neologisms, representing only 0.03 percent of words in a German text as Bosshardt (2003: 453) states, received disproportionally more space and attention in the field of (semi-) affixes and blends, but not compounds. This does not correspond to the factual situation because compounds are the most frequently used type of word-formation in German. Early on corpus-based classifications of semantic roles in compounds were used in English word-formation research, e.g., Adams (1973), Downing (1977), Hansen (1978), Warren (1978), and Shaw (1979; referring to English and German compounds). Yet research on German compounds has largely been restricted to exemplary demonstrations of the basic possibilities (e.g., Fanselow 1981: 155−202). Complete, systematic semantic analyses of substantial corpora were not undertaken. While in the description of derivations, semantics was indeed regarded as constitutive of subtypes (e.g., diminutives, augmentatives, etc.), the section on compounding usually just illustrated how diverse and different the basic possibilities of production and interpretation can be, thus implying that a systematic description is not possible. Although the problem deriving from the fact that semantics had received insufficient attention was recognized (cf. Rainer 1993: 131), the description of morphology and word-formation morphemes (including interfixes) and, at best, the description based on syntactic relations (e.g., Kürschner 1974) still took pride of place. Until today, the dominance of morphology in word-formation research has remained unchallenged (e.g., Dressler 2006: 43−44). Even in the early phases of the Innsbruck project “German Word-Formation”, each formal particularity and the most minute semantic detail in the derivation were registered and described as a distinct subtype provided they were repeatedly identified (at least three times in a corpus of more than 34,000 derivations). However, the three volumes dealing with derivations are also consistently “leistungsbezogene [...] Darstellung[en] der reihenhaft funktionierenden, systematisch zusammenspielenden Wortbildungsmittel” [function-related […] studies of serially functioning, systematically interacting morphemes in word-formation] (Kühnhold et al. 1978: 16; cf. also Kühnhold and Wellmann 1973; Wellmann 1975). This implies that semantic categories/functions − like sociative and locative prefix combinations, e.g., Ko-Direktor ‘co-director’ and Mitbegründer ‘cofounder’ or innerbetriebliche Mitbestimmung ‘shopfloor democracy; lit. co-determination within the firm’ and binnenstaatliches Abkommen ‘domestic agreement; lit. agreement within a country’ − are the superordinate categories combining forms of the same function. The Innsbruck project “German Word-Formation” however also aimed at putting an end to the underrepresentation of compounds. It was the first major “corpus-based and corpus-driven” project in German word-formation research (for the method of corpusanalysis cf. Biber 2010: 159−191). While the mainstream practice at that time predominantly generated examples derived from ideal language competence and, at best, analyzed individual examples, the aim of this long-term project was both comprehensively to generate and work on a substantial corpus − almost 75,000 compounds (analyzed in the years 1977−1991 without using computers). In the volumes “Deutsche Wortbildung IV and V” (Ortner et al. 1991; Pümpel-Mader et al. 1992), compounds were consistently analyzed as formal and semantic units, which admittedly fulfill certain form requirements (formal schemata) − first in reception and

59. Schemata and semantic roles in word-formation

1041

much later in production − but are constituted as much by schema-specific semantics as by the form. Only in the course and in consequence of this “corpus-driven research” (Biber 2010: 169) and by including semantic models in the structural-functional approach to description, did it become possible to understand word-formation holistically and comprehensively, without bias and furthermore − since the compound was to be described completely on the basis of a large corpus − to discern which structural units played what role in word-formation and what status these assumed. Semantic/thematic role, model, pattern and schema proved to be the central concepts. The last three, however, first had to be adopted, adapted and explicated as theoretical concepts by semantic theory. After the unsuccessful attempt to describe word-formation as an (exclusively) algorithmically deducible phenomenon, these concepts were intended to help understand the complexity of word-formation semantics. The term schema has had a long history of use in philosophy, most notably by Kant (cf. Kaulbach 1973), and in psychology by Bartlett et al. (cf. Flammer 1996). Piaget’s use of the concept schema was of special interest to constructivism: “Whatever is repeatable and generalizable in an action is what I have called a scheme […]” (Piaget 1970: 42). Through constructivism, the concept and the constructivist interpretation entered modern cognitive science. With the adoption of these ideas, attempts to develop a unified and comprehensive cognitive theory have become more promising. In cognitive psychology, the concept schema is central: “Schemata haben immer die Funktion, Invarianz und Ordnung in die konkrete Welt von singulären Ereignissen zu bringen” [Schemata always have the function of bringing invariance and order into the concrete world of singular events] (Flammer 1996: 1970). With the spread of cognitive psychology, actual-genetic and psycho-genetic aspects complemented historico-genetic aspects in linguistics. Also in the context of modern word-formation research a schema is a generalization, ideally one derived from specific cases (bottom-up, e.g., the cognitive approach to wordformation by Tuggy 2005: 325, 240−243). Within the framework of a dynamic usagebased model or construction morphology, schemata and sub-schemata are considered generalizing structures functioning on a morphological, phonological, syntactic, semantic and subsymbolic level, where there is a systematic relation between form and meaning (Langacker 2011: 293−296; Booij 2010: 51−93; for an overview on morphological constructions as schemas cf. Lampert and Lampert 2010: 37−41; see also article 10 on word-formation in cognitive grammar and article 12 on word-formation in construction grammar). In such approaches it is also possible to turn the much discussed dichotomy “analogy or schema?” from an either/or problem into a both/and issue (Booij 2010: 88− 93). Indeed, the existence of clear cases of “analogical word-formation based on an individual model word” (Booij 2010: 89) is not denied: e.g., Unser Stammbaum ist in Wirklichkeit ein Stammbusch. ‘Our family tree is in fact a family bush.’ (Spektrum der Wissenschaft 10/2007: 19) (see article 46 on word-formation and analogy and article 130 on word-formation and visuality). It must however also be emphasized that users have different abilities to “discover abstract patterns” and that analogical word-formation may also turn into “word-formation by means of a pre-established subschema” (e.g., watergate and other analogous formations with -gate; Booij 2010: 93, 90).

1042

Semantics and pragmatics in word-formation I: General aspects

3. Semantic role and role schema in compound-analysis The analysis of compounds in the “German Word-Formation” project was not only based on a traditional morphological, but also on a systematic description of the semantic conditions within the compounds. This approach relied on the description of the constituents and their relations which the paraphrase identified as especially relevant. Therefore basic factors are the semantic relation, the topicalization of the constituents and the schemata of semantic roles: (1) The semantic relation based on the fundamental predications ‘be’, ‘have’, ‘consist of’, ‘make’, ‘refer to’, ‘do’, ‘happen’ and ‘be named’: In the project, the meaning relation formed the most important criterion for the classification of compounds into basic types (Ortner et al. 1991; Pümpel-Mader et al. 1992; Ortner and Ortner 1984: 130−150). As shown in Table 59.1, thirty-one basic types of noun-noun and verb-noun compounds could be determined. For instance, the ‘have’ predication as the most abstract unit thus corresponds to two more concrete semantic relations, i.e. to inalienable possession (ornative) as well as to alienable possession (possessive), e.g., Henkelkorb ‘basket, that has handles’ and Gelddynastie ‘dynasty that has/ owns money’; see Figure 59.1. (For the central importance of the semantic relation in “conceptual combinations” as basis for representation and processing of compound words cf. Gagné and Spalding 2006: 145−168; see also article 63 on noun-noun compounds.) (2) The topicalization of the constituents: For the difference between Schaftstiefel ‘shaft boot’, i.e. ‘boot that has a shaft’ (relation of inalienable possession) and Stiefelschaft ‘boot shaft’, i.e. ‘shaft that is part of a boot’ (appurtenance relation) see Figure 59.1. (3) The concept of semantic role/deep case, as developed by Fillmore (1968) for syntactic analysis, and recently pursued, for instance, by authors in Kailuweit and Hummel (2004) and by Primus (2012): From the perspective of semantic roles, the “German Word-Formation” project identified 142 subtypes of noun compounds within the 31 basic types, e.g., the role pairs ‘characteristic trait − entity’ (Qualitätsstoff ‘quality fabric’), ‘whole − part’ (Mondoberfläche ‘moon surface’) or ‘possessor − possession’ (Vereinsvermögen ‘club funds’) within the ‘have’ pattern; see Figure 59.1 (cf. Ortner 1997: 31). The compilations in Table 59.1 and Figure 59.1 present a varied reflection of the differentiation into groups and, in some cases, subgroups (for the detailed classification of types and all 142 types of semantic-role relations cf. Ortner et al. 1991: 126−145).

59. Schemata and semantic roles in word-formation

1043

Tab. 59.1: Types of noun-noun and verb-noun compounds in German (after Ortner et al. 1991: 126−145) Category/Type of compound 1. additive copulative compounds 2. comitative concomitant compounds

Example Tier-Mensch ‘animal man’ ‘man animal’ Hosenbluse ‘trouser(-matching) blouse’

3. equative equative compounds 4. nominatory/appellative compounds of the type ‘[B] is called [A]’ 5. comparative comparative compounds 6. indicative compounds of the type ‘[B] amounts to [A]’ 7. substitutive compounds designating a service and its equivalent 8. specificative compounds specifying a type 9. mensurative compounds specifying a unit of measure 10. figurative compounds specifying a form/shape 11. constitutive compounds specifying a collective 12. substantial material compounds

Schockzustand ‘state of shock’ Njassa-Fluß ‘Njassa River’ Zitronenfalter ‘yellow butterfly; lit. lemon butterfly’ 1000-Mark-Gewinn ‘1000-mark windfall’ Transportgebühr ‘transport charge’ Kaffeesorte ‘coffee type’ Fleischportion ‘meat portion’ Milchpulver ‘milk powder’ Menschengruppe ‘group of people’ Holzhütte ‘wood(en) hut’

13. ornative/qualitative compounds naming an item after its part or characteristic trait 14. dimensional compounds naming an item after its dimension 15. possessive compounds naming an item after what it possesses or wears 16. existential compounds naming a place or period of time after a given element

Henkelkorb, Wertgegenstand ‘handle(d) basket’ ‘value object’ 10-m-Kabel ‘10-m cable’ Gelddynastie, Hosenmädchen ‘money(ed) dynasty’ ‘pant girl’ Arbeiterstadt, Friedenszeit ‘working-class town’ ‘peace time’

17. partitive/sociative compounds naming an item after its partitive or sociative relation to a superordinate item 18. possessory/benefactive compounds naming an item after its possessor/beneficiary

Mondoberfläche, Vereinsmitglied ‘moon surface’ ‘club member’ Vereinsvermögen, Lehrergehalt ‘club funds’,‘teacher salary’

1044

Semantics and pragmatics in word-formation I: General aspects

Tab. 59.1: (continued) Category/Type of compound

Example

19. referential object-relational compounds 20. competential compounds naming an item after its domain/function/theme 21. local compounds of local position and direction

Friedenssehnsucht ‘longing for peace’ Verkehrsministerium ‘ministry/department of transport’ Gebirgsdorf, Schulweg ‘mountain village’, ‘route to school’ Morgenkaffee, Zwei-Tages-Reise ‘morning coffee’, ‘two-day journey’

22. temporal compounds of temporal position and duration 23. conditional/occasional conditional compounds 24. causal causal compounds 25. consecutive/causative causative compounds

Nebelhorn ‘foghorn’ Feuerschaden ‘fire damage’ Todeskrankheit ‘lethal illness’

26. congruent correlative compounds 27. modal modal compounds 28. instrumental instrumental compounds 29. agentive/auctorial compounds naming an item after the agent or initiator

Kurswert ‘market value’ Etappenrennen ‘stage race’ Beilschlag ‘axe blow’ Polizei-Razzia, Picasso-Bild ‘police raid’, ‘Picasso painting’

30. nominative/commemorative naming compounds of the structure ‘[B] is named after [A]’

Shakespeare-Preis ‘Shakespeare Award’

31. actional compounds naming an item after the object/purpose of an action

Brotmaschine, Bohrmaschine ‘bread slicer’ ‘drill (machine)’

The respective schema (model) is constituted by the connection between the semantic roles as created by semantic relations based on specific predications. Due to the locally and structurally simple nature of word-formations as binary structures their determination by patterns and schemata becomes evident in a large corpus; among the 62,456 German noun compounds analyzed, strong and weak schemata can easily be detected (see Tables 59.2 and 59.3; Figures taken from Ortner et al. 1991: 126−141).

59. Schemata and semantic roles in word-formation

1045

‘ have’

‘B has A’

‘A has B’ = ‘B belongs to A’

Characterizing compounds

Appurtenance compounds

Relation of inalienable possession

ornative/qualitative

partitive/sociative

Compounds naming an item after a characteristic trait or part

Compounds naming an item after its partitive or sociative relation to a superordinate item

1. characteristic trait – entity Qualitätsstoff ‘quality fabric’

1. entity – characteristic trait Stoffqualität ‘fabric quality’

2. part – whole Schaftstiefel ‘shaft boot’

2. whole – part Stiefelschaft ‘boot shaft’ 3. collective – constituent Vereinsmitglied ‘club member’ 4. entity – associated entity Arztgattin ‘doctor’s wife’

Relation of alienable possession

possessive

possessory/benefactive

Compounds naming an item after what it possesses, wears or receives

Compounds naming an item after its possessor, wearer or recipient

1. possession – possessor Gelddynastie ‘money(ed) dynasty’

1. possessor – possession Vereinsvermögen ‘club funds’

2. worn – wearer Hosenmädchen ‘pant girl’

2. wearer (target) – worn Mädchenhose ‘girl pants’

3. benefit – recipient Zuschussbetrieb ‘subsidized firm’

3. recipient – benefit Lehrergehalt ‘teacher salary’

Fig. 59.1: Example of classification of German compounds expressing ‘have’ relation (after Ortner 1997: 31)

1046

Semantics and pragmatics in word-formation I: General aspects

Tab. 59.2: Strong role schemata in German (after Ortner et al. 1991: 126−141) Role schema

Example

Compounds*

purpose − instrument

Bohrmaschine ‘drill (machine)’

3,349

whole − part

Mondoberfläche ‘moon surface’

3,059

domain − institution responsible

Verkehrsministerium ‘department of transport’

2,273

material − product

Holzhütte ‘wood(en) hut’

2,203

object/goal − oriented entity

Friedenssehnsucht ‘longing for peace’

2,074

position − place

Gebirgsdorf ‘mountain village’

1,988

entity − characteristic trait

Haarfarbe ‘hair colour’

1,939

given element − place

Arbeiterstadt ‘working-class town’

1,903

topic − medium

Tierbuch ‘animal book’

1,873

characteristic trait − entity

Wertgegenstand ‘value object’ Glatteis ‘sheer ice’

1,820

* Numbers are total counts of 62,456 compounds Tab. 59.3: Weak role schemata in German (after Ortner et al. 1991: 126−141) Role schema

Example

Compounds*

counterpart − interaction

Frauenbekanntschaft ‘female acquaintance’

50

possession − possessor

Gelddynastie ‘money(ed) dynasty’

46

benefit − recipient

Zuschussbetrieb ‘subsidized firm’

45

unit − quantity

Fleischportion ‘portion of meat’

41

worn − wearer

Hosenmädchen ‘pant girl’

32

duration − action

Sechstagerennen ‘six day race’

15

equivalent − service

Lohnarbeit ‘wage labor’

13

* Numbers are total counts of 62,456 compounds

4. The properties of schemata, models and patterns and their interrelations The concept of schema allows a conceptualization that advances the analysis of wordformations by rendering complexity more understandable. According to Kölbl and Straub (2001: 520), schemata are “(mehr oder minder) abstrakte, relativ stark dekontextualisierte und generalisierte kognitive (Erwartungs-)Strukturen” [(more or less) abstract, rather strongly decontextualized and generalized cognitive structures (of anticipation)]. Piaget emphasizes that a schema has an invariant structure. Only if it is used repeatedly and is applicable to new objects can it be considered a schema (cf. Aebli 1980: 40). Invariant structure and stability are the most prominent features of schemata. There are further

59. Schemata and semantic roles in word-formation

1047

characteristics of individual schemata and models as units consisting of several “layered” schemata, as follows: (1) Multi-element schemata Schemata consist of diverse elements, both obligatory and facultative. For instance, the frequent binarity of word-formation schemata is only one manifestation on one level. (2) Multi-level character of the models As a linguistic surface phenomenon each concrete realization of a model consists of several overlapping schemata (= layering). These are the subject of linguistic description based on levels, as for instance in Motsch (1999: 3): phonological forms, inflectional morphological features, syntactic word categories, argument structure and semantic representation. Word-formation patterns are thus considered to be “Paare von semantischsyntaktischen und phonologisch-morphologischen Beschreibungen” [pairs of semanticsyntactic and phonological-morphological descriptions] (Motsch 1999: 1). (3) Relation It is the relation between the components (of the same level) that constitute the schema. The relations between the schemata of different levels constitute the model. Thus the model is the frame within which the schemata that constitute a model correlate: the model represents the sum of all the schemata. (4) Creation of a reference space Transferability and repeated use of a model thus creates a reference space consisting of the model and of that which it is applied and transferred to. The naming function holds all components and relations of the word-formation together. Just as the speech situation forms the frame of an utterance, which renders all elements and relations of the situation both interpretable and makes them parts of a whole, the naming function correlates all units, relations and levels of the word-formation construct in different contexts and connects them. (5) Strength The more often transfers and repeated uses occur, the stronger and more stable a schema is. The semantic roles create a particularly strong relationship and unit, so that they obtain the status and function of the guiding level, when, for example, the wordformation is resumed in a text. The unity of a schema is determined by the strength of the relation between its components. The strong role schemata manifest themselves in the establishment of a paradigmatic series, e.g., for the role schema ‘constituent − collective’ as in Insidergruppe ‘inside(r) group’, Insiderkreis ‘inside(r) circle’ or Hörerkreis ‘listening circle’, Hörerschaft ‘audience’ or, e.g., for the role schema ‘field of reference − first position’ as in Modestar ‘fashion model; lit. fashion star’, Tennischampion ‘tennis champion’, Literaturpapst ‘literary pundit; lit. literature pope’. Due to the process of lexicalization, strong role schemata also lead to the formation of systematic modifications (Ortner and Ortner 1984: 69−73) such as the semantic features ‘a lot of’ (Grätenfisch ‘bonefish’, i.e. ‘fish with a lot of bones’), ‘predominantly’ (Fichtenwald ‘spruce forest’, i.e. ‘forest predominantly containing spruce’), ‘typical of’ (Kinderstimme ‘child’s/children’s voice’, i.e. ‘a voice that is typical of a child/children’). The role sche-

1048

Semantics and pragmatics in word-formation I: General aspects

ma is just one schema among others: it is, similar to morphological schemata, a strong schema, but not one that works in isolation. The relevance to typologization corresponds to the strength of a type. Following the principle of performance linguistics, “nach dem der Gebrauch selbst zeichenbildend ist” [according to which usage itself is sign forming] (Feilke 1994: 225), the more relevant a factor is to typologization the stronger the pattern in which the factor takes effect. (6) Paradigmaticity of the parts and the whole, of the elements and the relations and the product of both The transferability of a schema presupposes the paradigmaticity of the elements constituting a structure and/or of the relations as well as of the whole. German word-formation is characterized by extensive paradigmaticity. Its models open up slots and enable total and partial analogies of all kinds, whereas the frequency of use and transfer of the constituents (to form a series) and the frequency of analogy relations are to be distinguished. The latter constitutes the strength of a type. Types are structurally stable in their core, but not at their periphery. Additional modifications interfere with, and modify type meaning (see above). The description and explanation of the limits of paradigmaticity is a main goal of word-formation research. The concepts pattern, model and schema are very helpful in this context. The meaning of the whole formation and of the constituents, as well as their complex interrelation determines limitations as shown in the formation of series: Fortsetzungsroman ‘serial novel’ is generally established, Fortsetzungsehe ‘serial marriage’ (Österreichischer Rundfunk 9. 7. 2005) is formed by analogy and only understood with sufficient context. (7) Insularity in networks and in reference space “Wenn man nach einer ‘Einheit’ des Gedächtnisses sucht, so ist es wohl am besten, wenn man das Gedächtnis aus Schemata aufgebaut betrachtet (Bartlett, 1931)” [When looking for a ‘unit’ of memory, it may be best to consider memory as being based on schemata (Bartlett, 1931)] (Dörner 2005: 150). In cognitive theories and theories of memory the schema is conceived of as a stable (not too large) processing unit within a “flow”, i.e. unit set in motion by activations within networks of dynamic knowledge. Not only the building blocks of a linguistic sign, but also the building blocks of the “cognitive reference points” (Rosch 1975) that linguistic signs relate to, stem from these networks. As postulated by revised prototype theory (cf. Kleiber 1990), they form the basis for the creation of constructs figuring as “the signified” in traditional theories of meaning. (Special res-verbum relationships, like designations of “natural” objects are specific, but frequent cases of the more general conditions determining the dimension of knowledge units/linguistic signs.) Each linguistic category is an island of stability in the flow of small and smallest subsymbolic entities. This is even observed in cases where mistakes occur. Gelähmten-Klo ‘toilet for the paralyzed’ instead of Behinderten-Klo ‘toilet for the disabled’ but not Rutsch-Klo ‘slide toilet’, etc. (the referential category of the first constituent is retained in addition to the schema of semantic roles; cf. Ortner, Ortner and Wellmann 2007: 118−119). Linguistic units are the results of cognitive processing and memorization of heard and/or read − linguistic and non-linguistic − input, i.e. material for referencing, for iconic processing, for constituting declarative knowledge, etc. (8) Constructs, not monoliths

59. Schemata and semantic roles in word-formation

1049

In mistakes and neologisms − but also “mere” metaphoric re-use − it becomes clear that word-formations, like all linguistic signs, are not monoliths hewn from one stone, but amalgamations, i.e. multiple constructs consisting of differently accentuable individual components of different formats and of different parameters (surface constituents being only one specific feature). Error analysis makes this particularly obvious: there is no use of words as such, even less a use of words in toto; at the same time, there is no use of constituents as such or in toto. Words are not monoliths ready to be reproduced, but constructs with a subsymbolic foundation (as phonemes are constructs of values in different parameters) varying according to the principle of analogy and/or the principle of family resemblance: the compoundedness of the issues is reflected in the malformation Abwässerungsgräben ‘draining ditches’ instead of Abwassergräben ‘drainage ditches’: [...] führte ein von tiefen Abwässerungsgräben gesäumter Feldweg nach links ‘a path lined with draining ditches led to the left’ (Dibdin 2000: 190). In this case, the partwhole Bewässerung- ‘watering’ prevails over the part-whole Abwasser- ‘drainage’, but the umlaut derives from other, obviously also co-present wholes (Entwässerungskanäle ‘drainage channels’, Bewässerungskanäle ‘watering channels’). (9) Deep layering Like syntax, word-formation is a multi-level, multi-layer and multi-dimensional phenomenon. The most important dimension is the naming function (and the closely related textual one); the second most important dimension is the formal, and the third most important is the “semantic depth structure” (Fillmore). Models are complex, layered constructs, not only hierarchically dependent schemata within schemata within schemata, but also − in parallel processing − schemata “above/below/beside/before …” schemata “above/below/beside/before …” schemata − and certainly not as a strict 1:1 homology. If we do not expect excessive modularity, but incremental conditions and conditions of parallel processing from the beginning of linguistic production and language use, we have to imagine the multi-layer concept as connectionistically used space. With this conception both mistakes and stylistically unusual word-formations can be explained by means of the same model. (10) Complex interrelationships It is obvious that not all factors − subsymbolic, lexical, morphological, syntactic − are equally effective in all contexts, that is, on all levels, in all layers and dimensions and that not all of them take effect with the same force at the same time, but the effective relative strengths and actual cooperative conditions have hitherto remained unexplored. We have to ask: which level is the guiding level, or what is the level par excellence, the only relevant one − the morphological, the syntactic or the semantic level? This question has long been asked, almost in the style of confessional disputes. Yet at the same time it has implicitly been accepted that the same conditions persist in all major patterns, and that the same forces assume control. With the concept of deep cases/semantic roles, the focus has shifted towards generative semantics, but due to the parallelist-connectionist extension, the conflict − syntactic or semantic origin − has become obsolete. The conditions are much more complicated and less stringent in every way, which is primarily reflected in the examination of model-pattern conditions.

1050

Semantics and pragmatics in word-formation I: General aspects

(11) Tension between concrete patterns, schemata and models; syncretism instead of/ beside regularity With the model providing the concepts of schema layering, the levels of the target, and finally the dimensions regarding the occurrence of a word-formation, it becomes possible to explain the thoroughly unsystematic application of rules in word-formation. They are limited, unsystematic and chaotic in such a way that the concepts of rule or regularity no longer seem applicable. This is also why we do not refer to rules, but to strong or weak series. Sometimes, word-formation occurs according to a paradigm/model based analogy, sometimes according to variation of a minimal unit (e.g., Rauhhaarlackel ‘coarse-haired churl’ as variation of Rauhhaardackel ‘coarse-haired dachshund’; cf. Ortner and Ortner 1984: 49−51). This frequently occurs according to the principle of family resemblance, but almost always guided by concrete instances of models which are used as patterns. A pattern emerges from the realization of a model which is perceived as marked according to the frequency of its occurrence or its prominent features. Analyzing a pattern entails considering the whole which exhibits the special features in dominant form which, in turn, determine its strong markedness (compare the concept of strength as explained above). In the case of compounds these features are frequently naming unit, form and constituent relation or constituent meaning. (12) Plasticity In different contexts, whatever is relevant to the text is foregrounded and other aspects are ignored, cf. Raubtieraugen funkelten in der Nacht ‘Predator eyes peered through the night’ − Der Mann blickte mich aus dunklen Raubtieraugen an ‘The man watched me with dark predator eyes’.

5. The methodological and epistemological value of analyzing the use of schemata, models and patterns Apart from the traditional approach of paraphrasing, i.e. of using the most explicit form to represent declarative knowledge, there exists an approach ex negativo. This approach identifies the conditions determining the range of possibilities by considering formations that are at the limit, i.e. mistakes, stylistically and economically excessive constructs and variants: (1) Mistakes Mistakes reveal the units processed by the brain (operative units and schemata). Mistakes also reflect the strength of patterns: Abwässerungsgräben ‘draining ditches’ instead of Abwassergräben ‘drainage ditches’. (2) Stylistically excessive constructs Semantically opaque and morphologically excessive compounds push the limits of patterns and models, cf. Ewigkeits-Zähne ‘teeth of eternity’ (Celan 1967: 57), Strohhaufen-Brombeer-Brennessel-Nullkosten-Nullenergie-Haus ‘straw-heap-blackberry-stingingnettle-zero-cost-zero-energy building’ (Doernach and Heid 1982: 1).

59. Schemata and semantic roles in word-formation

1051

(3) Economically excessive constructs These can be ellipses whose roles are not marked according to the statistically standard pattern, such as Kaltmiete ‘cold rent’ vs. Glatteis ‘slippery ice’ (not: ‘rent that is cold’, but: ‘ice that is slippery’). (4) Variants Variants are interpreted and based on the “Folie einer bestehenden Wortbildungskonstruktion und nicht ausgehend von der Bedeutung der Konstituenten” [foil of an existing word-formation construction and not the constituent meanings] (Ortner and Ortner 1984: 49), e.g., Klimahandel ‘climate trade’ after Klimawandel ‘climate change’. Formations of the types (1) to (4) at the very limit of the principles of word-formation show how complex the conditions and relationships are. This is why productive wordformation does not create model homologies or a 1:1 correspondence between all the schemata in one model. We frequently encounter paradigm constitution beyond semantic borders, for instance in the functional or naming dimension; e.g., the behavior of a constituent like voll- ‘full’ or halb- ‘half’ in Vollpreis ‘full price’ and Halbpreis ‘half price’ on the one hand, and Vollbruder ‘full brother’ and Halbbruder ‘half brother’ on the other hand: der dritte Halbbruder arbeitet in […] Hamids Vollbrüder […] sind noch erfolgreicher ‘the third half-brother works in […] Hamid’s full brothers […] are even more successful’ (Der Spiegel 2011, 29: 68). The components halb- ‘half’ and voll‘full’ constitute a graduative pattern, also found in word-formation types which are based on differences between word classes of the first constituents such as the verb-adjective difference in […] ist ein Schwelbrand ausgebrochen, der zum Vollbrand wurde ‘[…] a smoldering fire developed into a full fire’ (Österreichischer Rundfunk, Nachrichten, 15. 11. 2011). But they also transcend other major categories − derivation, compounding or synthetic compounding: Vollnarkose ‘full anaesthetic’, Vollholz ‘solid wood; lit. full wood’, Vollidiot ‘complete idiot; lit. full idiot’ (compounds); vollwertig ‘full(y)-fledged’, vollbusig ‘full-bosomed’ (synthetic compounds); one day, we might even come across vollbohren ‘to full(y) drill’, volljäten ‘to full(y) weed’, etc. As with other examples of the graduative pattern, the weak pattern would be the area of extension: Film-Film ‘filmfilm’ (i.e. ‘true film rather than series or part of a series’; announcement of true movies Sat-1 2010), Beinahe-Katastrophe ‘near catastrophe’, Pseudo-Krieg ‘pseudo war’, Alsob-Maxime ‘as-if maxim’ (Plank 1981: 75), Scheingrund ‘pretext; lit. purported reason’, Möchtegernsatiriker ‘would-be satirist’ (Die Zeit 1982, 16: 47), Analog-Käse ‘analogue cheese’, i.e. ‘not cheese (only a cheese substitute)’ (Der so genannte Analogkäse, ein aus pflanzlichen Fetten, Verdickungsmittel und Geschmacksverstärkern hergestellter Käseersatz ‘So-called analogue cheese, a cheese substitute made of vegetable fat, thickeners and flavor enhancers’; Tiroler Tageszeitung 22. 10. 2010: 19). Also, competing attributive formations with the same or similar function fall within the same naming unit paradigms, e.g., natur-identer Schnee ‘nature-identical snow’: Die Erfindung eines Generators, der [...] natur-identen Schnee produziert [...] Kunstschnee mit sehr geringer Dichte, der natürlichem, trockenen Schnee sehr ähnlich ist ‘The invention of a generator that […] produces nature-identical snow […] artificial snow with low density that is similar to natural, dry snow’ (Die Presse 19./20. 11. 2011: K 18). Apart from the ‘whole-pattern’ (with all its elements and their relations), dominant constituents are also provided by productive individual constituents. These are elements

1052

Semantics and pragmatics in word-formation I: General aspects

or the combination of elements within a schema within layers of the model and within dimensions of naming and/or processing. The concrete pattern with its dominant constituents becomes the driving force of productivity and the dominant constituents become powerful instances determining use. By assuming use of patterns according to the principle of family resemblance we can better explain regularities of use appearing everywhere − even if not very systematically organized. The Wittgensteinian metaphor of family resemblance refers to surface congruities and incongruities. A theoretical model has to reflect causes of and motives for these conditions but primarily has to be derived from a connectionist perspective and as the use of weak patterns. Transferability according to the principle of family resemblance begins with small and smallest subsymbolic units, in isolation and/or in connection. Or, vice versa, it starts from the connection, e.g., the relation between semantic roles of a schema or between the constitutive schemata in one model, etc. as illustrated by the formal model “adjective + noun” and the semantic relation between the constituents which are, semantically considered, not immediate constituents, e.g., Kaltmiete ‘cold rent’, i.e. ‘rent excluding heating bills’. Fleischer (1983: 256) explains the similar problem of noun-noun compounds, which are “Klammerformen, [...] bei denen die semantische Beziehung der UK [unmittelbaren Konstituenten] nur aufzudecken ist, wenn ein ‘Mittelstück’ ergänzt wird: Bier(glas)deckel” [clipped or elliptical formations, […] whose semantic relation between the IC (immediate constituents) can only be discovered by adding a ‘bridging element’: beer (mug) coaster]. Following Feilke’s (1994) performance-linguistic theorem, according to which only usage is sign-forming, we can indeed explain so-called elliptic formations differently, i.e. as realizations of patterns, of which no part has been deleted.

6. Conclusion The schema-based approach primarily focuses on the semantically founded autonomy of the specific word-formation patterns. This becomes evident in light of the recurrent frequency of patterns (productivity, word-formation series). With these patterns, specific, pattern-related features are also reproduced. This is exemplified by semantic modifications such as ‘a lot of’. In this context, we do not refer to universal, but concrete semantic patterns derived from language usage and only occurring prototypically in language usage. The concepts of semantic role, schema, pattern, model and analogy are elements of theorems, of description attempts whose integration on the basis of compound analysis was in fact the aim of the Innsbruck “German Word-Formation” project (cf. Ortner, Ortner and Wellmann 2007). These core concepts have indeed always been used in wordformation research, but they have neither been linked to nor integrated in a unified theory. This article refers to such an endeavour by relating to its conceptual and theoretical history, by explicating amply evidenced examples of use/usage, and by drawing the diverse strands together, thus providing both an integrated and therefore innovative point of view.

59. Schemata and semantic roles in word-formation

1053

Acknowledgement The article was supported by the University of Innsbruck. We would like to thank Svenja Grabner and Philip Herdina (University of Innsbruck) for translating and editing the text.

7. References Adams, Valerie 1973 An Introduction to Modern English Word-formation. London: Longman. Aebli, Hans 1980 Denken. Das Ordnen des Tuns. Vol. 1: Kognitive Aspekte der Handlungstheorie. Stuttgart: Klett-Cotta. Anderson, John R. 1996 Cognitive Psychology and Its Implications. 4th ed. New York: Freeman and Company. Anderson, John R. 2010 Cognitive Psychology and Its Implications. 7th ed. New York: Worth Publishing. Biber, Douglas 2010 Corpus-based and corpus-driven analyses of language variation and use. In: Bernd Heine and Heiko Narrog (eds.), The Oxford Handbook of Linguistic Analysis, 159−191. Oxford/New York: Oxford University Press. Booij, Geert 2010 Construction Morphology. Oxford: Oxford University Press. Bosshardt, Hans-Georg 2003 Morpho-syntaktische Planungs- und Kodierprozesse. In: Theo Herrmann and Joachim Grabowski (eds.), Sprachproduktion, 449−482. Göttingen: Hogrefe. Celan, Paul 1967 Atemwende. Frankfurt/M.: Suhrkamp. Dibdin, Michael 2000 Entführung auf Italienisch. Ein Aurelio-Zen-Roman. Aus dem Englischen von Ellen Schlootz. München: Goldmann. Dokulil, Miloš 1964 Zum wechselseitigen Verhältnis zwischen Wortbildung und Syntax. Travaux Linguistiques de Prague 1: 215−240. Doernach, Rudolf and Gerhard Heid 1982 Das Naturhaus. Frankfurt/M.: Krueger. Dörner, Dieter 2005 Gedächtnis. In: Astrid Schütz, Herbert Selg and Stefan Lautenbacher (eds.), Psychologie. Eine Einführung in ihre Grundlagen und Anwendungsfelder, 149−166. 3rd ed. Stuttgart: Kohlhammer. Downing, Pamela 1977 On the creation and use of English compound nouns. Language 53(4): 810−842. Dressler, Wolfgang U. 2006 Compound types. In: Gary Libben and Gonia Jarema (eds.), The Representation and Processing of Compound Words, 23−44. Oxford: Oxford University Press. Erben, Johannes 2006 [1975] Einführung in die deutsche Wortbildungslehre. 5th ed. Berlin: Schmidt. Fanselow, Gisbert 1981 Zur Syntax und Semantik der Nominalkomposition. Ein Versuch praktischer Anwendung der Montague-Grammatik auf die Wortbildung im Deutschen. Tübingen: Niemeyer.

1054

Semantics and pragmatics in word-formation I: General aspects

Feilke, Helmuth 1994 Common sense-Kompetenz. Überlegungen zu einer Theorie des ‘sympathetischen’ und ‘natürlichen’ Meinens und Verstehens. Frankfurt/M.: Suhrkamp. Fillmore, Charles J. 1968 The case for case. In: Emmon Bach and Robert T. Harms (eds.), Universals in Linguistic Theory, 1−88. New York: Holt, Rinehart and Winston. Flammer, August 1996 Schema. In: Wilhelm Arnold, Hans Jürgen Eysenck and Richard Meili (eds.), Lexikon der Psychologie. Vol. 3, 1970−1971. Augsburg: Bechtermünz. Fleischer, Wolfgang 1969 Wortbildung der deutschen Gegenwartssprache. 2nd ed. Leipzig: Bibliographisches Institut. Fleischer, Wolfgang 1978 Regeln der Wortbildung und der Wortverwendung. Deutsch als Fremdsprache 15: 78−85. Fleischer, Wolfgang 1979 Zum Charakter von Regeln und Modellen in der Wortbildung. Linguistische Studien (Berlin). Reihe A. Arbeitsberichte, Heft 62(3): 75−85. Fleischer, Wolfgang 1983 Wortbildung. In: Wolfgang Fleischer, Wolfdietrich Hartung, Joachim Schildt and Peter Suchsland (eds.), Kleine Enzyklopädie Deutsche Sprache, 237−273. Leipzig: Bibliographisches Institut. Fleischer, Wolfgang and Irmhild Barz 1992 Wortbildung der deutschen Gegenwartssprache. Unter Mitarbeit von Marianne Schröder. Tübingen: Niemeyer. Gagné, Christina L. and Thomas L. Spalding 2006 Conceptual combination: Implications for the mental lexicon. In: Gary Libben and Gonia Jarema (eds.), The Representation and Processing of Compound Words, 145−168. Oxford: Oxford University Press. Hansen, Klaus 1978 Probleme der semantischen Beschreibung von Komposita. In: Albrecht Neubert (ed.), Zur lexikalischen Semantik des Englischen. Linguistische Studien (Berlin). Reihe A. Arbeitsberichte, Heft 45: 24−61. Helbig, Gerhard 1986 Entwicklung der Sprachwissenschaft seit 1970. Leipzig: Bibliographisches Institut. Henzen, Walter 1965 Deutsche Wortbildung. 3rd ed. Tübingen: Niemeyer. Kailuweit, Rolf and Martin Hummel (eds.) 2004 Semantische Rollen. Tübingen: Narr. Kastovsky, Dieter 2005 Hans Marchand and the Marchandeans. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 99−124. Dordrecht: Springer. Kaulbach, Friedrich 1973 Schema, Bild und Modell nach den Voraussetzungen des Kantischen Denkens. In: Gerold Prauss (ed.), Kant. Zur Deutung seiner Theorie von Erkennen und Handeln, 105− 129. Köln: Kiepenheuer & Witsch. Keller, Jörg 1990 Konnektionismus − ein neues Paradigma zur Wissensrepräsentation? Linguistische Berichte 128: 298−331. Kleiber, Georges 1990 La Sémantique du prototype. Catégories et sens lexical. Paris: Press Universitaires de France.

59. Schemata and semantic roles in word-formation

1055

Knobloch, Clemens 1994 Sprache und Sprechtätigkeit. Sprachpsychologische Konzepte. Tübingen: Niemeyer. Kölbl, Carlos and Jürgen Straub 2001 Schema. In: Nicolas Pethes and Jens Ruchatz (eds.), Gedächtnis und Erinnerung. Ein interdisziplinäres Lexikon, 519−520. Reinbek bei Hamburg: Rowohlt. Kuhl, Julius 2001 Motivation und Persönlichkeit. Interaktionen psychischer Systeme. Göttingen: Hogrefe. Kühnhold, Ingeburg and Hans Wellmann 1973 Deutsche Wortbildung. Typen und Tendenzen in der deutschen Gegenwartssprache. Erster Hauptteil: Das Verb. Düsseldorf: Schwann. Kühnhold, Ingeburg, Oskar Putzer and Hans Wellmann, unter Mitwirkung von Anna Maria Fahrmaier, Artur Moser, Elgin Müller and Lorelies Ortner 1978 Deutsche Wortbildung. Typen und Tendenzen in der deutschen Gegenwartssprache. Dritter Hauptteil: Das Adjektiv. Düsseldorf: Schwann. Kürschner, Wilfried 1974 Zur syntaktischen Beschreibung deutscher Nominalkomposita. Auf der Grundlage generativer Transformationsgrammatiken. Tübingen: Niemeyer. Lakoff, George 1987 Women, Fire, and Dangerous Things. Chicago: The University of Chicago Press. Lampert, Martina and Günther Lampert 2010 Word-formation or word formation? The formation of complex words in cognitive linguistics. In: Alexander Onysko and Sascha Michel (eds.), Cognitive Perspectives on Word Formation, 29−73. Berlin/New York: de Gruyter. Langacker, Ronald W. 2011 A dynamic usage-based model. In: Adele E. Goldberg (ed.), Cognitive Linguistics. Critical Concepts in Linguistics, 276−331. London/New York: Routledge. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. Müchen: Beck. Motsch, Wolfgang 1999 Deutsche Wortbildung in Grundzügen. Berlin/New York: de Gruyter. Ortner, Hanspeter and Lorelies Ortner 1984 Zur Theorie und Praxis der Kompositaforschung. Mit einer ausführlichen Bibliographie. Tübingen: Narr. Ortner, Lorelies 1997 Zur angemessenen Berücksichtigung der Semantik im Bereich der deutschen Kompositaforschung. In: Rainer Wimmer and Franz-Josef Berens (eds.), Wortbildung und Phraseologie, 25−44. Tübingen: Narr. Ortner, Lorelies, Elgin Müller-Bollhagen, Hanspeter Ortner, Hans Wellmann, Maria Pümpel-Mader and Hildegard Gärtner 1991 Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Vierter Hauptteil: Substantivkomposita − Komposita und kompositionsähnliche Strukturen 1. Berlin/ New York: de Gruyter. Ortner, Lorelies, Hanspeter Ortner and Hans Wellmann 2007 Das Projekt “Deutsche Wortbildung”: Werkgeschichte und wissenschaftsgeschichtliche Position. In: Heidrun Kämper and Ludwig M. Eichinger (eds.), Sprach-Perspektiven. Germanistische Linguistik und das Institut für Deutsche Sprache, 91−131. Tübingen: Narr. Piaget, Jean 1970 Genetic Epistemology. Woodbridge Lectures Delivered at Columbia University in October of 1968. Translated by Eleanor Duckworth. New York/London: Columbia University Press.

1056

Semantics and pragmatics in word-formation I: General aspects

Plank, Frans 1981 Morphologische (Ir-)Regularitäten. Aspekte der Wortstrukturtheorie. Tübingen: Narr. Primus, Beatrice 2012 Semantische Rollen. Heidelberg: Winter. Pümpel-Mader, Maria, Elsbeth Gassner-Koch, Hans Wellmann, unter Mitarbeit von Lorelies Ortner 1992 Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Fünfter Hauptteil: Adjektivkomposita − Komposita und kompositionsähnliche Strukturen 2. Berlin/ New York: de Gruyter. Pümpel-Mader, Maria, Lorelies Ortner and Elgin Müller-Bollhagen 2014 Das Innsbrucker IDS-Projekt “Deutsche Wortbildung”: Ein Expeditionsbericht. In: Melanie Steinle and Franz Josef Berens (eds.), Ansichten und Einsichten. 50 Jahre Institut für Deutsche Sprache, 109−121. Mannheim: Institut für Deutsche Sprache. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rosch, Eleanor 1975 Cognitive reference points. Cognitive Psychology 7: 532−547. Shaw, James Howard 1979 Motivierte Komposita in der deutschen und englischen Gegenwartssprache. Tübingen: Narr. Tuggy, David 2005 Cognitive approach to word-formation. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 233−265. Dordrecht: Springer. Warren, Beatrice 1978 Semantic Patterns of Noun-Noun-Compounds. Göteborg: Acta Universitatis Gothoburgensis. Wellmann, Hans 1975 Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Zweiter Hauptteil: Das Substantiv. Düsseldorf: Schwann.

Hanspeter Ortner and Lorelies Ortner, Innsbruck (Austria)

60. Word-formation and argument structure 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

Introduction The structure of lexical items Phonetic form and its connections Basic conditions of semantic form Interdependencies of argument structure Interpreting categorization Head-complement combination Further configurations Argument inheritance Affixal adjuncts Remarks on conversion Argument-structure conditions on compounds Final synopsis References

60. Word-formation and argument structure

1057

Abstract After a brief survey of the structure of lexical items, which sets the boundary conditions for word-formation, the four components of lexical information − phonetic form, semantic form, argument structure, and categorization − are characterized with respect to their organization and interdependence. Particular attention is paid to the role of argument positions; their dependence on semantic variables, which determines their role in semantic or s-selection, their association with category features defining the c-selection and their control by the syntactic conditions of categorization is sketched with respect to the role played by the functor-argument connection in derivational morphology and compound formation. The basic effect of derivational affixes is illustrated with a number of canonical cases and their automatically emerging properties, including the peculiarities of argument inheritance, the difference between suffix-based head-complement and prefix-based head-adjunct structures and the different types of conversion or zero-derivation. Finally the role of argument structure in compound formation is considered with respect to the difference and exchange between complement-head and modifier-head combinations. A concluding survey summarizes and compares differences among the ingredients of argument structure in five respects.

1. Introduction Word-formation comprises the processes, rules and elements which are used in a given language to characterize or build up more complex (possibly new) lexical items. The argument structure (AS) of a lexical item represents the conditions according to which the item can or must combine with other linguistic elements to form complex expressions. Hence, argument structure plays a decisive role in the combinatorial processes of language. The expressions determined by these processes include complex lexical items as a particular, in fact central possibility, hence the question arises as to the role argument structure plays with respect to word-formation, compared to its role in combinatorial processes in general. On the background of these considerations the role of AS will be examined in three respects: first, as a complex property of lexical items that enter into a combinatorial operation (the initial AS), second, as a property of the resulting expression (the derived AS), third, the relation between the initial and the derived AS. These are general perspectives of which the third exhibits particular characteristics with respect to word-formation. To illustrate the aspects in question, consider examples like drink and drinkable. First, the initial AS of the verb drink has at least two, but presumably three positions, specifying subject, object, and event-reference; the initial AS of the suffix -able (an entry of a so-called bound morpheme) has one quite characteristic position which specifies a transitive verb. Second, the final AS of the derived adjective drinkable has one position, specifying the theme (or argument) to which the adjective can apply; likewise the final AS of the noun drink, usually considered as deriving from the homophonous verb drink, has one position, specifying the referent (or theme) to which the noun can refer. Third, the relation between the initial AS of the verb drink, in particular its object position, and the final AS of the adjective drinkable, and similarly the relation between the initial AS

1058

Semantics and pragmatics in word-formation I: General aspects

of the verb and the final AS of the noun drink, raise a number of problems: on the one hand, the same conceptual condition LIQUID is associated with the position in question in all three cases, on the other hand, the initial and final positions are subject to quite different grammatical conditions in the verb, the adjective, and the noun. These are typical problems of argument structure in word-formation, and they have been treated in different ways. The questions to be dealt with are: in what sense can initial and derived positions be identical, in what way and why can or must they differ, and how are they related? These problems are different with respect to different types of word-formation of which usually four kinds are distinguished: first, derivation by means of affixes that characteristically determine the resulting category, as in drinking, drinkable, drinker; second, combination with usually category-preserving prefixes, as in undrinkable, misbehave, transplant; third, conversion (or zero-derivation), such as the deverbal nouns drink, walk, rise or the denominal verbs like shelf, saddle, register; and fourth, compounding, such as wine drinker, drinking-water, water color painting. The criteria defining these types are not homogeneous and uncontroversial. If conversion is construed as zeroderivation, i.e. derivation by means of phonetically empty affixes (which must be recognized in other cases as well), and if prefix-combination can be analyzed as due to special types of affixes (or particles, for that matter), then the four types of word-formation reduce to two major domains: derivation and compounding, distinguished as combinations of lexical items with bound as opposed to free elements, where bound elements are morphemes that cannot appear as independent lexical items. None of the emerging types of word-formation is homogeneous. There are all sorts of boundary cases, as well as various types of ambivalent basic items such as like, trans, or under. Hence, what is more important than classifying cases of combination is the elucidation of the conditions that processes of word-formation and principles of argument structure might impose on one another. This endeavor in turn is closely related to the question of whether wordformation is to be conceived as an essentially syntactic or lexical phenomenon − a question that has been controversially debated especially (but not only) with respect to nominalization, among others in the syntactically oriented analysis of Lees (1960), Borer (2005) and Alexiadou (2007) vs. the lexicalist approach in Chomsky (1970) or Bierwisch (2009). In any case, some basic assumptions about the organization of lexical items must be made, not only with respect to the lexicalist view of word-formation.

2. The structure of lexical items Like linguistic expressions in general, lexical items are construed as pairs of sound and meaning. That is, they are assumed to associate a fairly abstract phonetic structure with an equally abstract condition on conceptual interpretation. The sound structure is called phonetic form (PF) and is characterized by arrays of phonetic features. The details of their interpretation via patterns of articulation and auditory distinctions need not concern us here. We only need to assume that PF representations are basically sequences of segments and syllables into which phonetic features are arranged. The structure of meaning is variably called logical form (LF) (Chomsky 1986), conceptual structure (CS) (Jackendoff 1990, 2002), semantic form (SF) (Bierwisch 1997), to mention just a few

60. Word-formation and argument structure

1059

pertinent proposals. The decisive point here is that meanings are considered as mental entities corresponding to external, physical, social, and other aspects of reality according to rather complex conditions, the details of which must be taken for granted and cannot concern us here. What needs to be considered, however, is the way in which the representation of meanings is built up from basic components or ingredients, i.e. semantic features or primes. In contrast to PF, representations of SF are not sequential in nature, but rely on functional connections and dependencies, leading to hierarchical configurations of semantic elements, about which more will be said below. So far, lexical items (and linguistic expressions in general) are characterized as pairs of PF and SF whose internal organization is subject to characteristic conditions and principles, such as constraints on syllable structure or the structure of possible concepts. Both PF and SF are systematically, but in abstract and complex ways, related to auditory/articulatory and conceptual interpretation, respectively. The connection between PF and SF of lexical items, however, is strictly arbitrary and purely conventional, as emphatically noted by Saussure (1916). Besides sound shape and meaning, linguistic expressions have a grammatical aspect which is crucial for the combinatorial operations in which they take part. Although this aspect is in some way connected to conditions of their SF, it constitutes an autonomous part of their organization which I will call the grammatical form GF of an expression. It consists of two components which concern rather different aspects of combination. The first component, the categorization (Cat), indicates the pertinence of an expression to a syntactic class like noun, adjective, or verb, but also to possible morphological categories like number, gender, or case (including conditions on their formal realization by inflection class, etc.). The second component of GF, called sub-categorization in Chomsky (1965) and much related work, was initially conceived as adding more detail to the classification, specifying additional conditions of morphosyntactic classification. Although distinctions of this sort play a crucial role in this component of GF, its actual character is quite different, as the present term argument structure AS correctly suggests. While there is general agreement that Cat is specified in terms of (essentially binary) syntactic and morphological features, such that it is a structured set of morphosyntactic features, the nature of AS is a matter of some debate. According to usual assumptions, AS specifies the selectional restrictions an expression imposes on required or possible partners for combination, where further semantic and categorial aspects must be distinguished, specified in terms of s-selection and c-selection, respectively. Thus s-selection is responsible, e.g., for the condition that the object of drink and the argument of drinkable meets the condition LIQUID, while c-selection requires the object of drink to be an accusative noun, a restriction that does not hold for the argument of drinkable. What must be made explicit is the basis and content of both types of selectional restriction. Obviously, AS has to do with semantic aspects of combination and with categorial, syntactic conditions on it. This not only requires access to the different participants involved in the SF of an item − such as the agent and the theme of the event in case of the verb drink −, it also implies a separate specification of its semantic and category requirements. More generally, AS turns out to be a sequence, or rather a hierarchy, of argument positions, selecting possible complements. Clarification of these matters is due to Grimshaw (1990), Williams (1981), Pesetsky (1982), among others. To sum up, linguistic expressions, including lexical items, are systematic structures of at least the following types of information:

1060 (1)

Semantics and pragmatics in word-formation I: General aspects a. b. c. d.

phonetic form PF: a sequence of segments, consisting of phonetic features semantic form SF: a non-linear configuration of semantic features/primes categorization Cat: a structured set of binary morphosyntactic features argument structure AS: a hierarchy of argument positions determining s-selection and c-selection

Further specification is needed with respect to the conditions and principles by which the different types of information are related. As already noted, the arbitrary relation between PF and SF is the basis on which the conventional nature of natural language rests. The arbitrariness is essentially limited by the combinatorial processes, where the PF-SF correspondence of complex expressions is motivated by the correspondence of their constituent parts. In a completely different way, SF relates to AS, which picks up positions of SF, attaching morphosyntactic features to them. Whether and in what way, finally, Cat relates to SF or AS is a matter of debate and has to be recognized as a core condition in word-formation. Before these relations are discussed in more detail, a general point is to be made with respect to the status of lexical items which raises intricate problems. On the one hand, lexical items are rightly assumed to be subject to the same organizational principles as linguistic expressions in general: the essentially linear nature of PF and the non-linear, functional structure of SF holds equally within and between lexical items. These fundamental conditions are essentially projected from lexical items to linguistic expressions in general by combinatorial processes operating in accordance with the conditions of AS. On the other hand, there is a crucial distinction between the basic, non-complex, lexical items of a language and the complex expressions (including all regular complex lexical items) based on them. The set of basic items in this sense is necessarily fixed and finite, in contrast to the set of derivable, complex expressions, which is extendible by means of combinatorial operations and principles. This means, by the way, that the frequently expressed assumption, that the set of words, but not the set of sentences of a language is finite could be valid only with respect to the set of basic lexical items. This observation creates, moreover, a non-trivial tension with respect to items like understand or undertake, which are complex because of their obvious component parts under, stand, and take, but (unlike drinkable) not derivable from these parts. Hence, at least three types of lexical items are to be distinguished: basic lexical items like drink, box, stand, regular complex items like drinkable, reading, and idiosyncratic complex items like understand, undertake. These distinctions suggest further differences since, e.g., understand is complex and idiosyncratic with an unpredictable SF, while understandable would be complex and regular with respect to understand. These considerations suggest that possibly further distinctions are needed in order to account for the distinct properties of free items such as drink or stand as opposed to bound morphemes like -able, -er, unand even more intricate cases like con-, de-, -ceive or -struct. Hence, basic lexical items are to be classified as free elements, like drink, stand, under, affixes like -able, un-, con-, de-, bound stems like ceive, struct, duct, and possibly further types. With these observations in mind, the assumption that the lexicon is “the repository of all idiosyncratic properties of particular lexical items” made in Chomsky (1965: 30) and much related work, is not easily projected onto the set of lexical elements of a language L. On the one hand, there is no way to collect all idiosyncrasies in one class of lexical items and, on the other hand, even the true basic items not only consist of

60. Word-formation and argument structure

1061

arbitrary configurations, but are subject to various principles of semantic and phonetic form. It should be added that the types of idiosyncrasy or partial regularity mentioned so far are noted because they relate to aspects of word-formation on which the present article concentrates; but further types of idiosyncratic, unpredictable properties must obviously be recognized, e.g., with respect to inflectional morphology, idiomatization, frozen metaphors, speech act peculiarities, phenomena of style and register, etc. Let me conclude this sketch of the different types of lexical information with a provisional illustration by means of the adjective drinkable, where λy indicates the adjective’s argument position, which corresponds to the object of the verb drink from which it derives. (2)

[drink-able ] [+V,+N] PF

Cat

λy

[POSS [e : [x TAKE-IN y ^ LIQUID y]]]

AS

SF

GF

While PF marks the concatenation of drink with the suffix able (which contributes the feature POSS (possibility) and the features in Cat, as we will see), there is more to say about SF below. For the moment the illustration should be construed as the conceptual condition according to which an event e is possible with a participant x taking in a liquid y − assuming that this is roughly the abstract structure of the property assigned by drinkable to its complement.

3. Phonetic form and its connections If one assumes that predictable features are supplied by general rules and principles, the PF information of lexical items consists of all and only the idiosyncratic features of their phonetic form. Thus the consonantal onset of [drink] needs only the specification of voiced, dental stop for [d] and liquid for [r], since the initial [d] of native items of English can only be followed by [r] or vowels. The rules and principles that account for facts of this sort are the matter of phonology. There are, however, rules and conditions according to which PF information − features, segments, or whole morphemes − can be directly related to features for tense, number, gender, case, etc. which belong to Cat or AS. The change from basic [i] to [æ] in drank, e.g., corresponds directly to the feature [+Past] in Cat, and the added segment [s] in [drinks] directly relates to the features [−Plural,−Participant] as a c-selection condition imposed on the subject of drink (by way of agreement). In other words, the condition third person singular of drinks, runs, is, etc., doesn’t concern the item’s categorization, but the c-selection it imposes. Features, segments, or affixes might furthermore directly be related to the structure of AS, inflectional passivization being a case in point as, e.g., in Latin bibitur where -ur comes with the deletion of the subject position from AS. The complex problems of these connections are the subject matter of inflectional morphology, which not only presupposes the organization of PF, but also accounts for direct connections between PF components and conditions in GF. Different approaches to the rules and principles of this intricate component

1062

Semantics and pragmatics in word-formation I: General aspects

are proposed, e.g., in the distributed morphology of Halle and Marantz (1993) or in the minimalist morphology of Wunderlich (1997a). What is important with respect to word-formation is the lack of a simple boundary between inflection and derivation. The type of direct connection between PF and GF features noted for inflection carries over to characteristic cases of derivation: inflectional alternations like drink, drinks, drank, or man, man’s, men have derivational parallels like believe, believer, belief or think, thinker, thought, where (quite different) conditions and changes in Cat and AS are directly connected to PF elements. It must be added that, besides these similarities concerning the PF-GF connection, there are notorious borderline cases, whose status between inflection and derivation is highly controversial (cf. article 14 on the delimitation of derivation and inflection). Intriguing cases in point are morphological causativization of verbs, or the comparative, which is traditionally considered as a characteristic category of adjective inflection although it characteristically extends the adjective’s AS, adding the position for the compared constituent. A related, but quite different aspect connecting morphologically relevant conditions of PF to Cat concerns features indicating inflection classes which cannot properly be part of PF, but clearly determine the PF realization of grammatical categories. Mainly nouns and verbs are candidates for this classification, which tends to be preserved under derivation, even if other properties of the items have nothing to do with the class. Thus, understand keeps to the inflection of stand, like forgive or misgive, give up, etc., keep to give. Hence this type of categorization is blind to SF and AS, it marks properties of PF with respect to morphological behavior. A similar categorization, which concerns PF only, obviously plays a role in derivation, distinguishing cases like refusal and recital from acceptance and assistance or director from leader. Hence morphology has rules and conditions by which features in PF are directly connected to GF information and it must, moreover, provide features that fix and classify exactly this kind of connection. One might call features identifying this widespread, but often ignored, identification of morphemes address features (usually showing up as [strong], [weak], [i-class] or just listing). The role of address features deserves particular attention in various respects of inflection and word-formation. To sum up, besides specifying the unpredictable features of its PF, a lexical item must indicate the (equally unpredictable) conditions by which it can relate features and segments of PF idiosyncratically to features in GF.

4. Basic conditions of semantic form The structure of SF, which must be assumed to organize the conceptual interpretation of linguistic expressions parallel to the way in which PF organizes their articulatory realization, raises problems and controversies far beyond its role in word-formation and the scope of the present survey. But as the role of AS in word-formation − the main topic of the present article − relies crucially on the structure of SF, some general assumptions about the nature of SF are indispensable. Four points might be made in this respect. First, SF consists of basic components − semantic features or primes − which are primitive elements with respect to SF, but may or may not be reducible to different components with respect to the conceptual structures and domains which interpret SF

60. Word-formation and argument structure

1063

intentionally and conceptually. Thus [LIQUID], [MALE], [BECOME], [DOG] or [SEE] might equally be candidates for basic SF elements, but with different conceptual simplicity or complexity outside SF. There is ongoing dispute about the primes of SF and semantic decomposition of basic lexical items, which need not concern us here, for detailed discussion see, e.g., Fodor (1981), Jackendoff (1990, 2002), Bierwisch (2011), Pustejovsky (1995). It is at this point sufficient to assume that SF is based on elementary components. Second, these components are categorized according to the way they combine with each other. Formally, a combinatorial system of this sort can be construed as a functorargument structure along the lines proposed by Ajdukiewicz (1935) and further developed in the theory of categorial grammar as surveyed among others in Moortgat (1989) or Steedman (2000), cf. article 8 on word-formation in categorial grammar. The relevant point here is that a functor Φ is of category )α/β* iff it takes an argument a of category α to build up a complex Φa of category β, schematically: (3)

a.

Φ

a

|

|

〈α/β 〉

α

b. [β [〈α /β〉Φ][α a ]]

c. [β [α a ][〈α/β〉Φ]]

β The structures in (3a), (b) and (c) are notational variants, indicating that neither the representational format nor the sequential ordering is relevant, but merely the functional relation between the functor Φ and its argument a, according to the fact that SF is based on conceptual dependencies rather than linear ordering of its components. The (recursively defined) functor-categories )α/β* (where α and β can themselves be functor categories) organize the complex structure of SF. They need, however, an initial repertoire of primitive categories which the complex categories can rely on. Following standard practice, I will assume two basic categories, viz. e (for entities or objects in the widest sense, written in italics in order to distinguish it from the e symbol for entities) and t (for states of affair or situations), both corresponding to fundamental conceptual categories. This yields complex categories like )e/t* for properties (such as LIQUID, PHYSICAL ), two-place functors like )e/)e/t** for relations (such as TAKE-IN, LOCATION), etc. (4)

x | e

ACT

| 〈 e/t 〉

CAUSE

| 〈 t/ 〈 t/t 〉 〉

y | e

SEE

| 〈 e/〈e/t 〉 〉

t

〈 e/t 〉

t 〈t/t 〉

t

z | e

1064

Semantics and pragmatics in word-formation I: General aspects

Third, instead of linear structure, SF has an essentially hierarchical organization with the arguments automatically depending on the pertinent functor, as illustrated in (4) for the (incomplete) SF of show, according to which x does something to let y see z. This structural hierarchy of SF has important consequences for the syntactic properties of lexical items, especially by way of the structure of AS, to be discussed below. Fourth, basic elements of SF are either conceptually fixed constants like ACT, CAUSE, SEE, LIQUID, etc., or variables like x, y, z, which indicate a kind of empty slot to be filled in on the basis of linguistic or merely situational context as, e.g., in I see it in contrast to I see, with see conceptually relating in both cases observer and target. Thus constants determine the substantial conditions that make up the meaning of an item, delimiting at once the range of admissible values for the variables it contains, thereby ultimately determining its s-selection, while variables are the formal means by which the argument positions create its combinatorial potential. In other words, variables connect or even integrate SF with the positions in AS.

5. Interdependencies of argument structure Thus AS is a kind of bridge or interface that semantically supports syntactic properties which ultimately have consequences for PF. There are at least two different views about the way in which argument positions are anchored in SF which might be termed the extrinsic and the intrinsic position. Both views consider AS as relating particular semantic conditions of an item to the syntactic constituent(s) it combines with. In spite of much similarity, they differ in a number of respects to be looked at. The intrinsic view assumes the positions in AS to be directly related to inherent variables in SF, such as x, y, and z in (4), where x is to be specified by the subject of show, and y and z by its indirect and direct object, respectively, as, e.g., in he showed him the book, but also in he showed me how to drive. In this way the semantic constants of the lexical item provide the syntactic relations (subject, object, etc.) with thematic substance like actor, causing the perception of the (visual) content. The extrinsic view takes this thematic content to be due to independent, separate relations, which connect syntactic constituents as participants to a lexically specified bare event e, that doesn’t determine arguments by itself. Thus, the verb show defines the conditions for a SHOW-event e, to which the thematic relations connect an agent, an experiencer, and a theme by means of pertinent two-place relations. Instead of (4), this yields (5) as the SF of show: (5)

[[SHOW e] o [AGENT x e] o [EXPERIENCER y e] o [THEME z e]]

With respect to the consequences of these views exemplified in (4) and (5), three remarks are indicted. First, the semantic content by which the argument positions are specified is a central topic of the extrinsic view. Intense discussions, at least since Fillmore’s (1968) case grammar, have resulted in a universal repertoire of thematic relations like (6), proposed, among others, in Parsons (1990), Grimshaw (1990) and adopted in Hale and Keyser (1993).

60. Word-formation and argument structure (6)

1065

agent, recipient, goal/source/location, theme

For the intrinsic view, initiated in Katz (1972) and developed in different ways, e.g., in Jackendoff (1987, 1990, 2002), Bierwisch (1997, 2006), Wunderlich (1997b) and many others, semantic conditions are imposed on argument positions as an integrated aspect of SF as a whole, as shown in (4). Although particular constants might be involved in this aspect − such as [ACT x] in causative verbs −, no independent repertoire must be stipulated. There are, of course, natural correspondences between the two views − thus [ACT x] in (4) is the obvious counterpart to [AGENT x e] in (5) −, but in general, the correspondence is less direct, as it is based on quite different perspectives. Second, the event specification that is decisive in (5) is not mentioned in (4). In fact for the intrinsic view, the event which an item like give, show or drink refers to is characterized by a configuration of more primitive elements, while the extrinsic view generally (but not necessarily) assumes simple constants like GIVE, DRINK, or SHOW, which are one-place predicates applying to an event variable e to which then participants are freely related by means of the constants in (6). This raises an additional problem though. Event types do not freely admit all kinds of participants: give allows the same participants as lend, or donate, but not as help, doubt, sleep or even rain. Hence additional constraints are necessary to determine the argument positions that go with a given event type − a problem that does not arise with the intrinsic view, where participants are inherently specified by the basic features. What must be added, however, is the condition that the whole SF of a lexical verb provides the characterization of an event e, which is the natural argument for tense and other specifications like location, purpose, etc. This, however, is a general condition presumably tied to the syntactic categorization of verbs. Third, the views under discussion are based on different (though not incompatible) notions about the nature and origin of the ranking that obtains between positions in AS. While the intrinsic view assumes the ranking of variables to follow from the inherent hierarchical structure of SF with functors systematically governing their arguments, as indicated in (4), the extrinsic view supposes the hierarchy to be a substantial aspect of the conceptual content in the set (6) of semantic relations: agent is higher than recipient, which is above theme, etc. with some controversial cases like the rank of goal. This leads to the next point to be noted with respect to the structure of AS. The ranking of variables in SF − whatever the right assumption about its origin might be − is a crucial factor in the assignment of AS positions to syntactic functions like subject, object, etc., and thus important for the combinatorial properties of lexical items. This aspect was captured by the notion of thematic grid, referring to the pattern by which semantically based thematic relations are syntactically realized. Assuming that participants of lower rank are selected earlier than higher ones, a ditransitive verb like show has the AS indicated in (7), where the variable z for the direct object is saturated first, and the highest ranking variable e is the place for the (top-most) clausal specifications. (7)

(z, y, x, e)

[[SHOW e] o [AGENT x e] o [RECIPIENT y e] o [THEME z e]]

An alternative way to specify the same hierarchy of argument-positions is indicated in (8), where the positions of AS are represented by lambda abstractors, i.e. operators that bind variables in SF. The SF in (8) is a notational variant of (4) with the colon “:” connecting the event variable e to the propositional structure. An important point to be

1066

Semantics and pragmatics in word-formation I: General aspects

noted here is the following: on the assumption that in principle a functor dominates its argument, the ranking of lambda operators becomes a direct consequence of the hierarchical ordering of variables in (8). To put it differently: higher variables are bound by higher operators. This is a simple and fairly general principle that covers a wide range of phenomena with respect to argument ranking. (8)

λz λy λx λe

[ee : [[ACT)e/t* xe]t [[CAUSE)t/)t/t** [ye [SEE)e/)e/t** ze])e/t* ]t ])t/t* ]t ]]

A further interesting consequence of this account is the fact that satisfying an argument position corresponds to the formal operation of lambda conversion. By this operation, an expression with a prefixed lambda operator λx […x…] combines with an argument a in two steps: first a is substituted for the variable x, second the operator λx is deleted. Formally, this operation is defined in (9): (9)

λ x [… x …] a ≡ [… a …]

Condition: a is of the same category as x.

In other words, a replaces x in the structure which λx is prefixed to, and the operator is dropped. More formally, the lambda-operator λx turns the expression it precedes into a functor, which requires an argument of the category which x belongs to. By this definition, the argument-positions of AS are connected to SF in precisely the sense required. They provide the semantic effect of syntactic combination: if a head and its complement are combined into a new syntactic object, the SF of the complement automatically becomes the value of the variable bound by the pertinent argument position. This standard construal of the head-complement combination is illustrated in (10) with the simplified SF of the clause Sue showed him the book, which results from combining the SF in (8) with the SF of Sue, him, and the book, abbreviated here as SUE, HE and DEF BOOK, respectively, while e BEFORE u represents the past tense, indicating that the event e precedes the utterance u: (10) [e

BEFORE

u o e : [[ACT

HE]

[[CAUSE [SHE [SEE [DEF

BOOK]]]]]]

This is not the whole story yet: positions in AS must also indicate the c-selection requirements which specify the morphosyntactic conditions to be met by the complements in question. Thus the AS of show in (8) must specify something like (11), which indicates the kind of morphological information (rather than the exact features) the complements must meet: (11)

λz [+Acc]

λy [+Dat ]

λx [+Nom]

λe

The Case features used in (11) are provisional, the assignment to their position is, moreover, predictable by general rules of English, and hence redundant. Therefore, the lexical entry of show in (8) must not contain them, although they do, of course, determine the c-selection of its complements. There are, however, idiosyncratic c-selectional conditions that must be lexically fixed, as cases like (12a) and (b) from Grimshaw (1979) illustrate. Different features must be matched by the complement of wonder and of think, although

60. Word-formation and argument structure

1067

they select the same type of theme. The difference in question must relate to the categorization of the complementizer who vs. that of the embedded clause. (12) a. Eve wondered who we met. (13) b. *Eve wondered that we met him.

a. b.

*Eve thought who we met. Eve thought that we met him.

Finally, AS must indicate whether the overt syntactic realization of an argument position is optional, i.e. whether the complement it selects must appear. It will be seen below that, again, optionality is largely predictable and hence not lexically fixed; but there are clear cases where it is a property of particular lexical items, as can be seen from cases like (14): under normal conditions know can appear without object, while consider without object is deviant. This contrast shows up not only with verbs, but also with cases like those in (16) and (17). (14) a. He would know it. b. He would know.

(15)

a. b.

He would consider it. *He would consider.

(16) a. the place below the window b. the place below

(17)

a. b.

the place under the window *the place under

The details of optionality might depend on further conditions. The crucial point is, that marked optionality is one of the factors by which AS determines the combinatorial properties of lexical items: First, positions in AS are based on variables in SF; second, the SF-based ranking of positions determines their assignment to syntactic functions; third, grammatical features attached to positions in AS determine the categorial properties to be met by c-selected complements; fourth, indicating the optionality of positions warrants non-optional positions to be necessarily filled; fifth, the highest (final) position in AS is obligatory, as otherwise the item would combine with arguments.

6. Interpreting categorization The categorization (Cat) of a lexical item (and of linguistic expressions in general) consists of a structured set of syntactic and morphological features like [+V,+Past, …] which control its combinatorial possibilities. Syntactic features indicate the lexical category to which an expression belongs, morphological features specify particular distinctions within this category. The role of these features is intricate, because they may appear in Cat as well as in the conditions attached to AS positions just discussed. This is an obvious consequence of c-selection implying that selecting features must match selected ones, the former belonging to AS, the latter to Cat. Thus a simple combination like BilI knows rests on the agreement between the 3rd person singular marked as selectional condition to be met by the verb’s subject, while Bill meets this condition by the features in its Cat. These are well known relations, treated in various ways, most strictly by the principle of feature checking in the minimalist program in Chomsky (1995). In any case, the double function of grammatical features is crucial and will be seen to determine also central aspects of word-formation.

1068

Semantics and pragmatics in word-formation I: General aspects

Another complication comes from the different origin one and the same specification can have. Thus, a morphological feature may be lexically inherent, like gender in brother vs. sister, or it can be introduced by rules of inflection, like tense in walked vs. walk. Various intermediate cases are possible in which idiosyncratic conditions are added to regular items, such as regular tense in thought vs. think or number in men vs. man. Depending on whether alternations like go − went, he − they, we − us are considered with regard to systematic categories of inflection like tense, number, and case or as characteristic instances of idiosyncrasy, suppletion appears as a lexical or inflectional phenomenon. It must be added that roughly the same ambivalence occurs with respect to syntactic category features and their role in derivation. Thus the relation between noun and adjective can be lexical (much vs. amount), but it is to large extent also the effect of systematic morphology, as in dark vs. darkness, bright vs. brightness, short vs. shortness, high vs. hight, or − with various degrees of idiosyncratic marking − broad vs. breadth, wide vs. width, etc. In any case, features in Cat can all be due to lexical or rule-conditions, and the question is: how? A rather different problem concerns the conceptual interpretation of morphosyntactic features − a notoriously controversial issue. To begin with, there is a remarkable tradition relating syntactic categories to ontological types, with verb for activity, process or state, noun for object or substance, adjective for property, and preposition for relation. But, of course, none of these constraints is even remotely valid: verbs frequently denote relations like approach or extend, or properties like smell or last, and nouns can cover all ontological types − concrete and abstract objects like bottle and number, processes like rain, properties like blue and relations like distance. Ontological constraints apply at best indirectly. The actual nature of syntactic features concerns the combinatorial properties of the categorized expressions. In this sense, Chomsky (1970) used the features [±Nominal] and [±Verbal] to specify the major categories N, V, A, P. This classification gave rise to various discussions among others in Jackendoff (1977), Wunderlich (1996), Bierwisch (2011) and various others, resulting in (18). The feature [+Referential] identifies nouns and verbs, the two fundamental categories, which can (by means of the functional elements D(determiner) for nouns and T(tense) for verbs) become referential expressions, while the feature [+StrongArg] indicates the fact that verbs and prepositions have obligatory positions in AS, where optional realization requires a special indication, in contrast to N and A that have facultative arguments throughout. (18) Referential StrongArg

V + +

N + −

A − −

P − +

It will be seen to be important that unrealized optional positions of all categories are not simply absent, but tacit and still available. A further important combinatorial effect of category features is the dependence of c-selection properties on an item’s categorization. A simple case in point is the different case assignment to the agent position in constructions like John walks vs. John’s walk: whether the alternation is construed as conversion or as ambiguity of walk between verb and noun, the case assignment varies with the categories. Turning to the conceptual interpretation of morphological categories, one may concentrate on features in Cat, ignoring c-selection features, as these remain − for reasons

60. Word-formation and argument structure

1069

already noted − semantically void: The 3rd person plural in agreement constructions like they came might be semantically relevant in they, but not in the agreement of came. The intriguing point is that the interpretation of morphological features even in Cat is subject to remarkable variation. On the one hand, a feature like [+Past] has a stable correspondence to the component [e BEFORE u] indicated in (10) as the SF of showed (or rather its past tense inflection -ed), while on the other hand the feature [+Acc] in her, for instance, has no semantic correspondence in the SF of a clause like I saw her come. As a matter of fact, impressive attempts to construct a semantic case theory in Hjelmslev (1935) and Jakobson (1936) have eventually been without conclusive success. Intermediate conditions hold, among others, for categories like number and gender which basically correspond to conditions like collection and sex, but are frequently suspended by idiosyncratic conditions, as in trousers (plural without collection) or kept without marking, as in brood (collection without plural). Without going into the intriguing details of inflectional categories according to which semantic components may be related to morphological features and their realization PF, it must be noted that idiosyncratic and systematic conditions are characteristically interrelated. The plural of oxen, for instance, is systematic in SF, but irregular in PF, the plural of glasses is regular in PF, but irregular in SF. Besides the address features mentioned above which mark ox, child, or woman for idiosyncratic plural, rule ordering, placing marked cases before regular or default ones, will play a crucial role. Revealing discussion of phenomena of this sort can be found in Halle and Marantz (1993) where SF-problems are left aside, however. But the phonetic as well as the semantic interpretation is crucially involved in closely related problems of derivational morphology. In any case, there may be morphological features in Cat without correspondence in SF, such as [+Plural] in scissors or jeans. Whatever correspondence between SF (and its conceptual interpretation) and morphological features in Cat can be established, it must be due to standard sound-meaning correspondence and subject to the general organization of lexical items.

7. Head-complement combination The role of argument structure in word-formation to be dealt with in this article rests on the assumption that word-formation follows essentially the general principles combining linguistic expressions, acknowledging however irregularities that are commonplace in expressions to be integrated into the system of lexical items. Idiosyncrasies of this sort will largely be left aside in the sequel, though, if there are no particular reasons to the contrary. Differences between word-formation and combinatorial operations in general should thus be due to elements entering word internal combination, especially derivational morphemes like those in darkness, refusal, scandalize, etc. In order to identify this difference, comparison with the standard head-complement combination indicated in (10) above will be useful. A very restricted case in point is a PP like in the book, as illustrated in (19) and (20), with P and N abbreviating the features of (18) and D stands for Determiner (where r indicates the referential variable integrated into the complement noun because of the definiteness-operator DEF (which comes close to the logician’s iota-operator ι):

1070

Semantics and pragmatics in word-formation I: General aspects

(19) a. [in] b. [the] c. [book]

[P] [D] [N]

(20) a. [the book] b. [in the book] PF

λx λy [y [LOC [INTERIOR x]]] λn [DEF r [n r]] λx [BOOK x] [D] [P]

λy

Cat

AS

[DEF r [BOOK r]] [y [LOC [INTERIOR [DEF r [BOOK r]]]]] SF

It might be noted in this connection, that lexical entries and complex expression alike exhibit the four components discussed above (including the empty AS of the book in (20a), as here no further complement can be selected), and that quite generally a linguistic expression X can be construed as a combinatorial rule that c-selects by its argument structure a complement-expression Y to combine with, in order to build up the complex expression Z. This combination has two sides: it determines the order of X and Y with respect to PF according to the role of the head X, and it integrates X and Y with respect to SF by turning the SF of Y into the value of the selecting variable in X thereby deleting the actual c-selecting position in AS by way of lambda-conversion as defined in (9). Finally, the c-selecting X becomes the head of Z by turning Cat and the remaining AS of X into the dominant GF regime of Z. For the present example in the book, all of this is realized in (20b), with one position left in AS, by means of which the PP can become a predicative or an adjunct (as in the typo in the book − an option to be discussed in section 9). This head-complement merging appears pretty complex. It is based, however, on quite general principles, triggered and controlled by the different components of lexical information, whose interrelation is the basis for the PF-SF correspondence, which the language faculty supports. It is important to note that this observation holds in much the same way for phenomena of word-formation, where especially derivational operations are largely to be conceived as determined by morphological heads combining with appropriately c-selected complements. An important step in the debate about lexical and syntactic aspects especially of nominalization was the proposal in Chomsky (1970) to consider minimal pairs like (21) and (22) as parallel lexical constructions rather than as one derived from the other by syntactic transformation, as assumed in earlier approaches. (21) a. John refused the offer.

b. John’s refusal of the offer

(22) a. They accepted the price.

b. Their acceptance of the price

Problems related to these types of facts have been dealt with in various ways in Jackendoff (1977), Grimshaw (1990), Borer (2005), Alexiadou (2007) and others. Looking more closely at cases like (21), one has to notice correspondences and differences between (a) and (b). First, it is trivial that the past tense verb refused and the noun refusal are both based on the verb refuse, which serves as complement to different affixal heads. That -ed and -al are the respective heads, is borne out by the category assignment, since refuse without the affix is neither past nor noun. Second, it is equally obvious that the meaning of refuse, whatever its details are, must provide a position for a refusing subject and one for the refused object, and − as any verb − a position for tense and event reference, respectively. It is furthermore obvious from the comparison of (21a) and (b), that the pertinent positions in AS have different, though systematically corresponding

60. Word-formation and argument structure

1071

conditions on Case-assignment and c-selection: nominative and accusative in (a) correspond to prenominal and postnominal genitive in (b), the former spelled out as possessive. (Rather similar observations would hold with respect to (22a) and (b).) A strictly local and maximally lexical account of these observations could be based on the following simplified representations, where the SF of refuse is provisionally indicated as negation of y’s accepting x. The Case-requirements in (23a) and (b) are default conditions of transitive verbs in English and need not be indicated for individual expressions. This holds also for the genitive associated with the corresponding AS positions in (23c), because the genitive must generally be assumed to be the default case for adnominal nouns and refusal differs from the verb only by the suffix -al. (23) a. [refuse] b. [refuse-d]

[V]

λx λy λe [Acc] [Nom]

[e : [NEG [y [ACCEPT x]]]]

[V,+Past]

λx λy λe [Acc] [Nom]

[e BEF u o e : [NEG [y [ACCEPT x]]]]

λx λy λe [Gen] [Gen]

[e : [NEG [y [ACCEPT x]]]]

c. [refuse-al] [N,…]

The relation between (23a) und (b) is a regular instance of verb-inflection, relying on the past tense morpheme and its semantic interpretation, mentioned above. This extension of (23a) is the automatic result, if something like (24) is the entry for the past tense morpheme, that merges according to the head-complement combination just discussed with (23a) as complement. (24) [-d]

[V,+Past]

λv [e [V]

BEF

u o v e]

In other words, the past tense inflection adds (in the default case) the suffix [-d] to the PF, the feature [+Past] to Cat, and [e BEFORE u] to the SF of the verb which it takes as complement, leaving its AS unchanged. This is, of course, not the whole story about past tense inflection, especially not with respect to past tense allomorphy (like think − thought, or can − could) and other tense categories (like refuse − refuses), and the important fusion with syntactically crucial agreement features, but it provides a sufficiently clear contrast for the derived nominal in (23c). Here the different categorization, which the affix -al brings in, leads to a derived AS that corresponds to, but clearly differs from, that of the verb with which the affix combines. To this effect, the nominalizing head must be something like (25), representing an affix, with an AS containing just one position, which c-selects a verb, whose SF directly replaces the SF of the derived nominal, providing nominal reference with respect to exactly that event, to which the verb would have temporal reference. (25) [-al]

[N]

λv [Val]

[v]

In other words, combining (25) as head with the verb refuse in (23a) leads to an expression with precisely the SF of (23c). For this assumption to go through, comments on four points must be made.

1072

Semantics and pragmatics in word-formation I: General aspects

First, address features: The c-selection indicated in (25) by [Val] must be conceived of as not merely abbreviating the features [+Referential,+StrongArg] for the category verb as introduced in (18) above, but as including the address features mentioned in section 3, such that (25) can select refuse, construe, deny, etc. but not delete, accept, etc. It is not obvious, to what degree the (quite incomplete and unsystematic) alternatives indicated in (26) are merely idiosyncratic, and what tacit subregularities might be involved. (26) a. b.

refuse explain

− refusal − explanation

recite derive

− recital − derivation

c. d.

conclude − conclusion decide expose − exposure depart develop − development resent

− decision − departure − resentment

e.

accept

− acceptance

assist

− assistance

f. g.

develop grow

− developing − growth

provide − providing

survive − survival compete − competition invent − invention close − closure govern − government perform − performance leave − leaving

Whatever might turn out to be the appropriate treatment of these differences, it must classify lexical items in such a way that c-selection has access to it, possibly as a separate component of Cat, which contains the characteristic subregularities concerning inflectional or derivational classes and their economic marking. (In what follows, subscripts will be used to indicate the distinctions in question.) Second, definiteness: a rather different phenomenon concerns the effect of prenominal genitives (and possessives) as in John’s refusal or their acceptance. This phenomenon holds in DP structure in general: the definiteness contrast in pairs like the novel of Faulkner vs. a novel of Faulkner disappears in Faulkner’s novel, where not only the proper name, but also the whole DP is definite. In the same way the DP John’s refusal would be definite with respect to John as well as refusal, as represented in the SF (27). Whether and in what way this effect results from general conditions of adnominal genitive morphology, must be left open here. As a matter of fact, more general questions with respect to adnominal DPs and their definiteness effect seem to be involved. (27) [DEF e [e : [NEG [ [DEF y [JOHN y] [ACCEPT x]]]]]] Third, of-phrases: this concerns the general assumption implied by the Case conditions indicated in the AS of (23c), where the Case requirements for both prenominal and postnominal complements are indicated by [+Gen] − very much in line with standard assumptions about the structure of John’s refusal of the offer and similar constructions. Whether the conditions associated with AS positions are represented in terms of Case features like [+Gen] or prepositions like [+of] depends on various assumptions, which must be left open here. They involve, among others, the alternation between [+Gen] and of, or [+Dat] and to (as in tell it Sue vs. tell it to Sue) and in general the indication of c-selection with respect to PPs as in believe in, get rid of, ask for, and other phrasal verbs.

60. Word-formation and argument structure

1073

Fourth, argument inheritance: this most important aspect of derivation, which has so far simply been taken for granted, will be discussed systematically in section 9.

8. Further configurations Before the details of this pervasive, but intricate phenomenon are taken up, the headcomplement combination as a central mechanism of derivational morphology will be illustrated by some further characteristic cases. A familiar instance is the so-called agent nominalization exemplified in (28c), which, like the event nominalization (28b), preserves the SF of the underlying verb practically unchanged. (28) a. Walter instructs the new group. b. Walter’s instruction of the new group c. Walter is the instructor of the new group. For the sake of illustration, (29a) serves as the lexical entry for instruct, whose conceptual interpretation (something like the activity of y to direct the behavior of x) might be compatible with the internal structure of INSTRUCT. With this proviso, the event nominalization, in (29b), which differs from that with -al in (25) merely by the “allomorph” -ion, again allows for (but does not require) a prenominal and a postnominal argument, with a definiteness effect for the event reference. In this respect, the agent nominal (29c) differs remarkably: even though the eventuality of the verb is still involved, it is somehow invisible. Its reference is not definite, but hovers between occasional and generic usage, depending on the interpretation. This (quite natural) indifference might very provisionally be represented by a variable e’. This variable must be supplied by the SF of the suffix -or and substituted for the variable which is bound by the argument position λe of the verb, whence instructor lacks the event reference. (29) a. [instruct]

[V]

λx λy λe [Acc] [Nom]

[e : [y

INSTRUCT

x]]

b. [instruct-ion]

[N]

λx λy λe [Gen] [Gen]

[e : [y

INSTRUCT

x]]

c. [instruct-or]

[N]

λx λy [Gen]

[e′ : [y

INSTRUCT

x]]

These different nominalizations result − with all the peculiarities described − from merging (29a) as a complement with the suffxes in (30a) and (b), respectively: (30) a. [-ion]

[N]

λv [Vion]

[v]

b. [-or]

[N]

λv [Vor]

[v e′]

c. [-er]

[N]

λv [Ver]

[v e′]

1074

Semantics and pragmatics in word-formation I: General aspects

There is no further comment needed with regard to (30a) (its entry differs from that of the suffix -al merely by its PF and the address features distinguishing the cases in (27), etc.). But (30b) involves a crucial point to be explained. The merging of -or with instruct will result in (30c) if and only if the variable v in (30b) covers not only the SF of the verb in question (as in all other cases of derivation discussed so far), such that the suffix preserves the meaning of the complement it combines with, but if it also neutralizes the event reference of the verb by substituting the indifference variable e′ for the variable e on which the event position in AS is based. In other words, the characteristic event reference of the event nominalization is withdrawn by the agent nominalization. This might look like a violation of the principles on which head-complement merging is based, but it is in fact a strict instance of lambda conversion as defined in (9) above: the argument position λe is deleted and e is replaced by the indifference variable e′. This is not only in line with the intuition that agent nominalization tacitly relies on the relevant kind of event without keeping it available for definite reference. It also observes the formal principles of functional combination by means of lambda operators, as will be seen in section 9. The deletion of the highest (innermost) position of AS in (29c) (which in (29b) provides the event reference) has an important, automatic consequence: the next highest ranking argument takes over the referential capacity, which the determiner can qualify, as in the/an instructor of the group. The lowest (i.e. first) position in the AS of instruct, which the underlying verb automatically assigns to the object as in instruct the group, is, as a further effect of the affix -or, turned into an adnominal genitive, e.g., in the instructor of the group, or alternatively the group’s instructor. These are mechanical consequences, if a verb that meets the address feature condition of Vor is selected by the affix -or. This holds in exactly the same way for the affix -er, which differs only in PF and in the address feature Ver that the complement verbs must meet. This condition might largely be predictable, since agent nominalization exhibits by far fewer variants than event nominalization as indicated in (26), and the choice between -er and -or might to some extent be determined by independently needed specifications like [±Native], as cases like instructor vs. teacher suggest. Completely independent of these address features, though, are intricate properties which derive quite regularly for the AS of teacher on the basis of the suffix -er in (30c) applied to the verb teach, which differs from instruct by licensing two objects, e.g., in she taught him French. If (31a) is provisionally assumed as the entry for teach, representing a kind of lexical causative to learn, where tentatively the indirect object is marked as a lexically optional argument (with otherwise standard Case requirements), then the properties to be assumed for teacher in (31b) follow automatically. Notice that the optionality of complements for nouns is not to be indicated lexically. (31) a. [teach]

[V]

b. [teach-er] [N]

λz (λy) λx λe [e : [ [ACT x] [CAUSE [y [LEARN z]]]]] [Acc] [Acc] [Nom] λz λy λx [e′ : [ [ACT x ] [CAUSE [y [LEARN z]]]]] [Gen] [Gen]

Notice that teacher differs from instructor in (29c) by admitting two adnominal complements (within the conditions of DP structure), due to the difference between teach vs.

60. Word-formation and argument structure

1075

instruct. The relevant data to be accounted for by (31b) are indicated in (32), and (33) shows similar relations for the two types of objects in conduct and conductor which emerge on the basis of the two types of objects of conduct as complement of (30b). (32) a. b. c. d.

Mary Mary Mary Mary

is is is is

the the the the

boys’ teacher of French. teacher of French. teacher of the boys. teacher.

(33) a. b. c. d. e.

Solti Solti Solti Solti Solti

conducts the philharmonic orchestra. conducts the Mahler symphony. is the conductor of the philharmonic orchestra. is the conductor of the Mahler symphony. is the conductor.

The effect of word-formation on argument structure as considered so far has two major, interdependent, aspects. First, due to the category features of the governing affixes which combine with the verbal complements, the derived expressions have a standard nominal categorization. Second, in accordance with this categorization, their argument structure has the regular properties of nominal items, i.e. their positions are optional, except for the highest (innermost) one which is, for general reasons, obligatory and unspecified for Case; moreover, the (maximally two) complement positions predictably require adnominal genitive, as indicated in (31b). These aspects have different, but analogous effects in other configurations, e.g., in deverbal adjectives like those in (34) and (35). Here, the adjective formation introduces the possibility operator, simultaneously passivizing the verb by withdrawing the argument position which the subject would have to spell out. (34) a. They can read the book. b. The book can be read. c. The book is readable.

(35) a. One can reduce the price. b. The price can be reduced. c. The price is reducible.

These conditions can be expressed as lexical properties of the suffix -able as in (36): its SF introduces the possibility operator POSS, which it applies to the SF of the verb it combines with, substituting the unspecific entity x′ as value for its subject position (thus creating the passive effect), and furthermore introducing the indifferent event operator e′ (already known from the agent nominalization (30) above) to suppress the verb’s event position. If (37) is assumed as a provisional entry for the verb read (with [READ y] short for UNDERSTANDING WRITTEN MATERIAL y), (36) could combine with it to yield (38a), if (36) is construed (as already mentioned) as an operation, which substitutes by the regular operation of lambda-conversion the SF of (37) for the v in (36), thereby dropping the λv of the suffix -able. Two important points are to be added, though. First, the SF of (36) consists of the operator POSS applied to [v x′ e′], which must jointly specify a situation type. For this reason, v must be a type with two open slots, which in this case means that the SF of the verb marks its two highest variables as open positions. Technically, this is achieved by taking the SF of V together with its two innermost positions of AS, as marked in (37), and substituting this whole complex for the variable v in (36). The result of this substitution is indicated in (38a). Formally, these are independently

1076

Semantics and pragmatics in word-formation I: General aspects

motivated assumptions, to be noted in section 9. Second, the configuration generated in (38a) automatically triggers substitution of x and e by x′ and e′ by means of regular lambda conversion, thereby converting (38a) into (38b), the appropriate representation for readable. (36) [-able] [A]

λv [POSS [v x′ e′]] [Vable]

(37) [read]

λy λx λe [e : [x [ READ y]]]

[V]

v (38) a. [rea d-able ] [A]

λy

[λx λe [POSS [e : [x [READ y]]] x′ e′] v

b. [read-able]

[A]

λy

[POSS [e′ : [x′ [READ y]]]]

The AS of the resulting adjective is − due to the operations triggered by the necessary lexical information of the suffix -able − automatically reduced to the preserved position that used to be the object position of the verb and is turned into the obligatory theme argument of the adjective, which is in line with the conditions imposed by its categorization. Other arguments of the underlying verb, such as the agent in the paper is readable for experts are not selected arguments of adjectives. A kind of converse effect triggered by category change shows up in nominalized adjectives (quality nouns) like cleverness, stupidity, or length, where the affix does not withdraw AS positions of the adjective it combines with, but adds one that comes close to the event-variable of verbs. The relation of pairs like those in (39) and (40) illustrates the point. (39) a. Paul is unusually small. b. Paul’s smallness is unusual. (40) a. The idea is surprisingly stupid. b. The stupidity of the idea is surprising. A suffix like -ness or -ity takes an adjective as complement, the SF of which will be extended with a state-variable s which supports an argument position λs. This position is analogous to the event position of verbs, as the paraphrase-like relation between his cleverness and his being clever indicates. These conditions are represented for the suffix -ity in (41), which combines with the entry of equal, sketched in (42a) to yield the noun in (42b). (41) [-ity]

[N]

λa λs [Aity]

[s : [a]]

(42) a. [equal] b. [equal-ity]

[A] [N]

λx [EQUAL x] λx λs [s : [EQUAL x]] [Gen]

60. Word-formation and argument structure

1077

The state-argument λs, inserted by the suffix -ity, becomes automatically the highest, obligatory position of the noun without a c-selection requirement. By this amendment, the adjective’s theme position λx, that originally held this position, is downgraded into an optional position of the noun that predictably requires the genitive. Again, lexical conditions of the suffix generate systematic consequences in the AS of its complement. This holds also in more complicated cases where the adjective actually imposes c-selection requirements on its complements as in (43). Here the relevant sense of familiar is associated with a c-selection condition on the respective complement. (43) a. Mary is familiar with the proposal. b. Mary’s familiarity with the proposal Although the c-selection conditions for prepositional complements are similar to Caseconditions, notably to genitive requirements realized as of-PP, it is unclear how these conditions are to be handled, given the fact that phonetic information is crucially involved. It should be noted, though, that the with-condition is not only lexically fixed, but also carried over to the derived nominal, as (43b) shows. A different question to be left aside concerns the relation between familiar with X and familiar to Y (which corresponds somehow to the phenomenon known as psych-movement, not to be dealt with here). It must be added that the state nominalization of adjectives, like the event nominalization of verbs, is subject to remarkable variation in the (determined) choice of the pertinent suffix, controlled by address features. Thus alongside (26) for verbs, there are a fair variety of alternatives for adjectives, indicated in (44), which includes not only different affixes, like -ness and -th, but also stem alternations like broad − breadth or long − length. This emphasizes the possibility of looking at affixes as rules, which not only add segments, but also change features in the PF of their complement. (44) a. b. c. d.

intelligent − intelligence, distant − distance, prudent − prudence clever − cleverness, ugly − ugliness, short − shortness wide − width, long − length, broad − breadth, warm − warmth familiar − familiarity, absurd − absurdity, odd − oddity, rare − rarity

These variations, which are a characteristic phenomenon in word-formation, raise at least two further problems. The first concerns the relation of arbitrary and systematic aspects between the affixes and their designated complements. What determines absurdity instead of ?absurdness, what allows for alternatives like pureness alongside with purity? The second problem concerns the fact, that the selectional conditions, abbreviated here as [Aity] or [Vion], must have a counterpart in the items to be selected. The source, character, representational format, and above all the (presumably partial) regularity of this information is largely unexplored. Although it is a characteristic phenomenon of word-formation, it only shows up in occasional hints at the differences to be respected.

9. Argument inheritance The basic assumption with respect to argument structure and word-formation was the notion that the actual phenomena derive from the standard lexical information associated

1078

Semantics and pragmatics in word-formation I: General aspects

with the items involved and general principles of head-complement merging. The lexical system includes the derivational affixes which are heads that turn out to function as rules selecting appropriate items as their complements; the principles of head-complement combination follow rules and principles of phonology, as far as PF is concerned, and they are subject to functional combination with respect to SF. This combination is controlled by the selectional conditions represented in AS, with the Cat features of the head determining the resulting categorization. This principle of functional combination requires the elucidation of a phenomenon that so far has been taken for granted without debate. The phenomenon in question is called argument inheritance and has been recognized in the literature primarily with respect to the event nominalization of verbs. It is naturally instantiated by the following cases. (45) a. That they observed a supernova is documented. b. Their observation of a supernova is documented. (46) a. That Sue derived these cases differently was surprising. b. Sue’s different derivations of these cases were surprising. The parallelism between the nominal and the verbal versions of (45) and (46) with respect to the participants in the event in question is obvious, including the event reference which carries the tense-information in the (a)-cases and the definiteness and number specification in the (b) cases. There are, however, different views of this correspondence in different approaches to derivation in general and to nominalization in particular. Three perspectives might be distinguished which can be called complete, partial, and noninheritance view. The non-inheritance view, advocated in Ehrich and Rapp (2000), assumes that if a derived nominal has argument positions, they are not derived from those of the underlying verb, but determined and motivated by the noun’s own semantic structure, hence not inherited from the base of derivation. Complements of a derived nominal are considered to be due to the noun, not inherited from the verb, which is taken to be confirmed by the adnominal conditions they are subject to: subject and object of the verb correspond to optional adnominal genitives. The partial inheritance view pursued, e.g., in Grimshaw (1990), Alexiadou (2007) and related work observes a systematic difference between AS-nominals with inherited argument structure and R-nominals which can refer to objects and pluralize, but don’t inherit argument positions of the underlying verb. Problems of delimitation are notorious for several reasons as cases like (47) illustrate, which suggest that argument inheritance and shift of interpretation are subject to different conditions. (47) a. The examination of the candidates lasted three hours. b. The examinations of the candidates were on the table.

(event) (result)

The complete inheritance view, adopted in Bierwisch (2009), Wunderlich (2012) and related work, takes head-complement combination to automatically proliferate open, unassigned argument positions of the combined constituents to the AS of the resulting combination subject to the Cat conditions of the combination. Thus if a verb like show combines with the pronoun it, the AS of the resulting VP preserves the argument positions of show except for the direct object position assigned to it. In the same sense, the

60. Word-formation and argument structure

1079

argument positions of a verb like derive carry over to the AS of the combination derivation, subject to the conditions of nominal AS, as (46b) shows. There is, however, an important difference in the preserved argument positions of derivation as opposed to show it, which is characteristic of the way in which argument positions function in wordformation. This difference has been ignored so far and needs to be noted explicitly. The point is this: argument positions inherited from the verb of a VP like show it are those of the head, while the positions inherited by a nominalization like derivation are those of the complement; the categorization in both cases is that of the head however. As a consequence of this difference, the positions of show it retain their original properties, which are those of the verbal AS, while the positions of derivation change to properties of a nominal AS. It is worthwhile to look at the ingredients of this operation by means of a standard example like (47) in order to make the peculiarities of this kind of argument proliferation clear. The head -ation of nouns like examination is one of the synonyms (or suppletives) of items like -al, -ion, discussed in (25) and (30a) above, from which it differs only by its PF and the corresponding address feature. The verb examine in (49) (with EXAMINE abbreviating something like look carefully with specific purpose) is strictly transitive with no optional positions in AS, but predictable Case-requirements for subject and object. (48) [-at-ion]

[N]

λv [v] [Vation]

(49) [examine]

[V]

λx

λy

λe

[e : [y [EXAMINE x]]]

(50) [examin-at-ion] [N]

λx

λy

λe

[e : [y [EXAMINE x]]]

On the condition that (49) meets the address feature of Vation , hence being a possible complement of (48), its SF, namely the whole configuration [e : [y EXAMINE x]], substitutes for the variable v in (48). It retains, however, its complete AS λx λy λe, while the saturated λv of the suffix is dropped by lambda-conversion. The result of these (apparently somewhat ad hoc) operations is (50), where in some way the AS and SF of examine are turned into the AS and SF of [examin-at-ion], while the categorization of examine is replaced by that of the suffix with the crucial consequence that the positions λx and λy become optional and automatically marked for genitive selection, leading to optional adnominal complements. So far, (50) and its ingredients give an apparently plausible, but somewhat tricky account of the relation between verbs and the properties of their potential event nominalization. This account is less ad-hoc than it appears, however, since its form strictly derives from the principle of functional composition which is inherent to the system of categorial grammar on which SF is based. Without going into too much technical detail, functional composition can be characterized as generalizing the combinatorial principles of categorial grammar. In addition to functional application introduced in (3) above, which combines a functor of category )α/β* with an argument of category α to get a complex of category β, a functor of category )α/β* might also combine with an argument of category )γ/α* (i.e. another functor) and yield a complex of category )γ/β*, which “inherits” the position γ from the category )γ/α* of the argument-functor. Schematically simplified, the two possibilities are given in (51), where (51b) shows the formal structure of argument

1080

Semantics and pragmatics in word-formation I: General aspects

inheritance: The position γ is preserved in )γ/β* such that the argument position γ of the constituent becomes an argument position of the resulting whole: (51) a. Functional application

b. Functional composition

β α

〈 γ/β 〉

〈α/β 〉

〈γ/α 〉 〈α/β 〉

As a result of this principle, argument positions of a linguistic expression carry over to the dominating complex expression taking their original c-selection conditions with them which now, however, are subject to the category condition of the new head which imposes the pertinent properties of verbs, nouns, or adjectives on the derived construction. This does not require arbitrary stipulations with respect to word-formation. One particular condition of this interaction is in need of clarification, though. The problem is illustrated by different derivations of admire as listed in (52) with indication of the respective AS. (52) a. b. c. d.

We We We We

expect expect expect expect

that he admires the result. his admiration of the result. him to be an admirer of the result. the result to be admirable.

λx λy λe λx λy λe λy λx λx

-at-ion -er -able

The full AS of the verb admire is inherited by the event nominalization admiration, the agent nominalization admirer cannot preserve the event position, and the adjective with -able inherits only the theme-position. Why? The difference is triggered by the lexical information in the SF of the different suffixes, repeated here as (53): (53) a. [-at-ion] b. [-er] c. [-able]

[N] [N] [A]

λv λv λv

[v] [v e′] [POSS [v x′ e′]]

(48) (30c) (36)

The crucial difference with regard to argument inheritance comes from the presence of the variables x′ and e′ in (53b) and (c), which fill pertinent AS positions of the complement verb. If these positions are filled and thereby neutralized, they are prevented from proliferation as open positions to the resulting AS. For this to come about, the relevant position(s) in the AS of the verb (the event position, and the subject position, respectively) must be filled by the variable(s) in the SF of the suffixes in question. In other words, the relevant argument positions of the verb are saturated by the variables provided by the SF of the suffixes. To this effect, the relevant positions of the complement verb’s AS must act as part of their SF: they absorb the variables in question by lambda conversion. For -able this is schematically represented in (38) above. Technically it boils down to functional composition in the sense of (51b): the suffix combines with the verb taking over all and only those verbal AS positions which the suffix does not saturate. The frame for this raising of AS positions into the verbal SF is determined by the SF principle (54), a type constraint that sets the limits for functional composition.

60. Word-formation and argument structure

1081

(54) The SF of a lexical entry is a configuration of type t. Intuitively, (54) claims that lexical meanings (and the combinations they enter) are propositional, representing the role of semantic configurations in situations or states of affairs. Technically, it means (among other things) that the variable v in the SF of event nominalizers like -ation, -ion, -al, etc. must be filled by a configuration of type t, the v of agent nominalizers like -er however by a configuration of type )e/t*, which together with the variable e′ is of the type t in (54), and the v of -able and similar suffixes needs a configuration of type )e/)e/t**, etc. This provides the mechanism of argument inheritance, to which then the conditions determined by Cat and the morphological rules for specifying c-selection systematically apply, turning, e.g., an obligatory accusative complement into an optional genitive adjunct. The mechanism of functional composition provides the basis for partial or complete inheritance of argument positions of complements. But it also accounts for derivational affixes that not only preserve, but also extend the argument positions of their complement. A case in point is the abstract nominalization of adjectives, which has already been considered, leading to nouns like smallness, stupidity, or length, which add the state reference to the adjective’s theme position in constructions like the rigidity of the plan. The entry (41) for the suffix -ity, repeated here as (55), adds the state reference, while the variable a takes up the SF of the adjective, inheriting its AS. A case like familiar with shows that more than the obligatory theme position might be inherited, as in Peter’s familiarity with the case. With the provisional and incomplete analysis of one reading of familiar, the result of the derivation is sketched in (57), assuming that prepositional objects are to be treated as particular instances of c-selection. (55) [-ity]

[N]

λa λs [Aity]

[s : [a] ]

(56) [familiar]

[A]

λy λx [with]

[x [KNOWS-WELL y]]

λy λx λs [with] [Gen]

[s : [x [KNOWS-WELL y]]]

(57) [familiar-ity] [N]

It must be noted that the c-selection [Genitive] for the theme position of the noun in (57) follows general rules of Case marking, whereas the preposition with, which is neither morphologically nor semantically determined, must be lexically fixed in the entry for the adjective in (56). Hence its preservation in (57) is a strong case for argument inheritance. What derivational suffixes like (55) contribute to their complement is restricted to the category shift and the state variable s, with the pertinent operator λs in AS. There are, however, derivational suffixes with more substantive semantic content. The operator POSS in -able is a small, but characteristic instance. Many verbalizing operations contribute more substantial conditions, even if they are phonetically empty, like the verb in shelf the books or bottle the fruits, where the activity of localizing is added to the noun by a zero suffix. Similar observations hold for the suffixes in verify, purify, falsify, intensify, exemplify, terrify, beautify or formalize, familiarize, verbalize, scandalize, computerize, etc. The semantic components are closely related to the conditions that characterize causativization in various configurations such as transitive and intransitive grow, melt,

1082

Semantics and pragmatics in word-formation I: General aspects

change, develop, etc., but also lexical pairs like give and have, show and see, cf. Wunderlich (2012) for an overview. The regular core of verbalizations with -ify is determined by something like (58), which combines with adjectives like (59) to yield (60). The suffix is subject to standard conditions of Cat, such that the c-selection features for the subject and object position are redundant. The status of the selectional condition [Aify] on the lexical complement (including the address feature) is somewhat complicated. On the one hand, the position is strictly obligatory, as it selects the lexical complement by which -ify qualifies as suffix. On the other hand, the classification of the lexical complement is open to lexical idiosyncrasy, as cases like exemplify, justify or terrify, dignify, show, for which there is no pertinent adjective. Hence, the condition [Aify] is incomplete. This sort of ambivalence is a usual phenomenon in word-formation, however, and does not blot the basic pattern. (58) [-ify]

[V]

λa λy [Aify]

(59) [pure]

[A]

λx

(60) [pur-ify]

[V]

λx

λe

[e : [[ACT y]

CAUSE [BECOME

[a]]]]

[CLEAN X] λy

λe

[e : [[ACT y]

CAUSE [BECOME [CLEAN

x]]]]

Like in other examples, the SF of pure is, of course, provisional and must be construed as denying disparate ingredients with respect to particular conditions. Its combination with the suffixal head -ify is a standard case of derivation, which nevertheless deserves two comments. First, the principle of functional composition derives a complex AS for the resulting verb which quite regularly combines the two verbal positions of the suffixal head with the inherited theme position of the complement adjective. Second, the inherited position gets, under the regime of the verbal head, the features of the suffix which turn it into the object position of the resulting verb, predictably associated with accusative case. Notice that the verb purify meets the c-selection and the address feature of -able, thus allowing the derivation of the adjective purifiable with inheritance of the theme position, now functioning as the new obligatory theme. This, in turn, is available for the address feature of -ity in (55), thus allowing the derivation of the noun in (61), where λx is inherited from the adjective, which inherited it from the object of the verb where it was inherited from the theme of the adjective pure. (61) [purifiabil-ity] [N] λx λs

[s : [POSS [e′ : [[ACT y′]

CAUSE [BECOME [CLEAN

x]]]]]]

Assuming that verbs like formalize, verbalize, mobilize, regularize, etc. are based on the suffix -ize in (62), which is essentially a synonym of -ify, the argument positions of familiar in (56) are taken over by the AS of the verb familiarize in constructions like Carla will familiarize the crew with the new situation. (62) [-ize] [V]

λa λy [Aize]

(63) [familiar-ize] [V] λz λx [with]

λy

λe [e : [[ACT y]

λe

[e : [[ACT y]

CAUSE [BECOME

[a]]]]

CAUSE [BECOME

[x [KNOWS-WELL z]]]]]

60. Word-formation and argument structure

1083

As in (60) for purify, the higher positions λx λy λe are standard verbal arguments with predictable c-selection conditions. The special position is the prepositional object with the inherited condition [with]. Altogether two of the four AS positions of familiarize are inherited from the complement, two are provided by the suffix.

10. Affixal adjuncts Besides the head-complement combination discussed so far, where the complement saturates an argument position of the head (possibly proliferating argument positions of the complement), a head can combine with an expression not as complement, but as adjunct, which crucially is not an argument of the head, but rather supplies an argument position which is to be assigned to the expression it combines with. Thus in (64) the institute is a complement of leave, while at noon is an adjunct, providing the theme position of at to be appropriately filled by Fred left. Likewise in (65), the story is complement of know in (65a), while which Mary knows is adjunct of the story (65b). (64) a. Fred left the institute. b. Fred left at noon. (65) a. Mary knows the story. b. the story which Mary knows Actually, possible adjuncts − adjectives, adverbials, prepositional phrases − are all expressions with a designated argument position, by means of which they may function as modifier, adding conditions to the interpretation of the modified head. This includes, in particular, the structure of relative clauses, the relative pronoun of which plays the role of an argument position with respect to the SF of the clause. The range of problems connected to modification goes far beyond the scope of this article. What needs to be noted in the present context is the tendency that in wordformation the distinction in question is related to the difference between the suffixal and prefixal realisation of affixes: In English, suffixes tend to be heads, prefixes are mainly adjuncts. The cases considered so far are crucially suffixal heads. An interesting, apparently simple example for prefixal adjuncts is the negative prefix un- with the systematic, but arbitrary partial synonyms in- and a-. (66) a. un-happy, un-even, un-bearable, un-successful, un-efficient, un-fair, un-kind b. in-finite, in-firm, in-expressible, im-pure, im-mortal, c. a-moral, a-septic, a-historical, a-typical There is quite a bit of arbitrariness with respect to the choice of accessible heads, and it is unclear whether the treatment in terms of address features is appropriate, while the restriction to adjectives seems to be appropriate, even in view of cases like unhappiness or untruth, which must be treated as nominalizations of unhappy and untrue, respectively. The basic entry for un- would thus be (67), where the lack of Cat features is assumed to indicate the status of un- as adjunct, which does not project category features, thus

1084

Semantics and pragmatics in word-formation I: General aspects

preserving the features of the head it combines with. (68) illustrates the application of un- to equal and familiar as sketched in (42) and (56), respectively: (67) [un-]

[]

(68) a. [un-equal] [A] b. [un-familiar] [A]

λa [Aun-]

[NEG a]

λx λy λx [with]

[NEG [EQUAL x]] [NEG [x [KNOWS-WELL y]]]

Notice that λx and λy are inherited from the adjoined head, preserving even the lexically fixed selectional condition with for the complement of familiar. In general, derived adjectives like practical, convincing, pretentious, or readable can equally be modifiable by un-, as shown for readable, repeated here as (69): (69) a. [read-able] b. [un-[read-able]]

[A] [A]

λy λy

[POSS [e′ : [x′ [READ y]]]] [NEG [POSS [e′ : [x′ [READ y]]]]]

Both adjectives can be nominalized, as discussed earlier, with the prefix un- in unreadability connected to the adjective, as un- does not normally combine with nouns. This yields appropriate analyses, as paraphrases like those in (70) illustrate. (70) a. an unreadable text ~ a text that is not possible to be read b. [the unreadability of the text] is obvious ~ it is obvious [that it is not possible to read the text] Examples like these suggest that affixal adjuncts are to be treated like other cases of functional combination: the semantic content of the affix − in this case NEG, a constant of type )t/t*, i.e. a propositional negation − applies regularly to the proposition to which it is prefixed, as determined by the entry in (67). This is not the whole story, though. One first has to note, that there is a closely related, but clearly different variant of the prefix un-, that shows up in entries like (71), where the prefix applies to a verb, which it modifies analogous to the way in which the adjectival prefix un- modifies the adjective to which it adjoins, but with different result. More precisely, un- with adjective negates a state, which un- with a verb negates as the result of a change. Thus uncover describes the removal of the coverage, which cover introduces. (71) un-cover, un-pack, un-couple, un-load, un-button, un-dress, un-lock, un-leash, unveil, un-earth, un-ravel, un-say, un-do Leaving aside item-specific idiosyncrasies, here the prefix un- in a way reverses a process by negating a condition P, the creation of which is the meaning of the un-modified verb. Thus uncover reverses the process described as cover, unload reverses load, etc. In other words, un- negates a situation in adjectives as well as verbs, but in verbs the negation does not apply to the event or process, but to its result. Somewhat simplified, the effect of un- is sketched in (72a) for adjectives and in (72b) for verbs, where CAUSE [BECOME p] is a kind of activity or process template that is familiar from various contexts, such as the different readings of adjectival vs. verbal open, clean, dry and others.

60. Word-formation and argument structure (72) a. b. y

CAUSE [BECOME

1085

Px ~ NEG [P x] [Q x]] ~ y CAUSE [BECOME [NEG [Q x]]]

There are different ways along which one might try to accommodate these observations. One is to redefine the element NEG, such that it is interpreted by an operation of inversion, instead of propositional negation, yielding contrariness rather than contradiction in case of states and reversal in case of processes or changes. This seems to be appropriate for elements like unclean, unsettle(d), but it cannot be extended to prefixes in general, including, e.g., mis-, over-, under-, re-, that yield problems similar to un-: misspell doesn’t mean a wrong act of spelling, but an act of wrong spelling, overestimate does not denote too much activity of estimation, but a too high value being assumed. More generally, the difference illustrated in (72) is one of scope, rather than interpretation of the functor. A solution in this direction is proposed in Pesetsky (1985) for these and a number of related phenomena in derivational morphology. Based on a framework, that distinguishes S-structure and Logical Form as levels of syntactic representation, Pesetsky assumes an operation QR of Quantifier Raising, that applies to affixes and relates structures corresponding to (73a) and (b) for Paul unloaded the car, where the suffix-initial (a) corresponds to the S-structure and (b) to the Logical Form. (73) a. [NEG [PAUL [CAUSE BECOME [LOADED b. [PAUL [CAUSE BECOME [NEG [LOADED

CAR]]]] CAR]]]]

The difference between these structures can be overcome in strictly semantic terms, however, if one assumes the prefix to apply not to the verb as a whole (negating the event), but only to the smallest possible part, skipping the CAUSE-BECOME template, negating just the result. Since the elements of SF are not linearly ordered, but only functionally dependent, (73) might be construed as attaching the operator NEG either from the left, as in (a), or from the right, as in (b), which means more correctly: either from the top or from bottom. For the adjectival prefix, the result would largely be identical, if there is only one component which the )t/t*-type functor can apply to (neutralizing largely the difference between contradictory and contrary negation, e.g., in dirty, not clean, and unclean). For adjectives with a complex SF and for verbs, however, there is a clear distinction between what might be called normal and minimal scope: not possible to be read crucially differs from possible not to be read, very much like not cause to be loaded differs from cause to be unloaded. The fact that the latter interpretation applies to unload, unearth, unravel, etc. might be due to a condition of narrow scope for verbal un-: instead of including the causative/inchoative template into the scope of NEG, it applies to the resulting state only, even if this state is not the SF of an independent adjective. With this proviso, the entry for un- in (67) can be supplemented with (74) for the verbal prefix, which applies to verbs like ravel to yield (75). The dotted part of the SF indicates the complex to be skipped according to the narrow scope condition of verbs: (74) [un-]

[]

(75) [un-ravel]

λv [Vun-]

[…

NEG

v]

[V] λy λx λe [e : [ACT x [CAUSE [BECOME [NEG [TWISTED y]]]]]]

1086

Semantics and pragmatics in word-formation I: General aspects

It is to be noted that shifted application of an operator as informally indicated in (74) is not included in the standard categorial system on which SF and the role of argument structure is based. Its systematic incorporation might be part of a more comprehensive amendment that will take care of properties of adjuncts more generally, as discussed, e.g., in Dowty (2003), Bierwisch (2003) and Maienborn (2003). The narrow scope condition seems to carry over from un- to several other verbal prefixes, as in overload, overstep, overtax, underestimate, understate, misspell, misread, etc., which describe the causation of a state that exceeds or fails to reach an expected norm or consist in an error. To illustrate the basic pattern by means of (one meaning of) the prefix over- and its combination with the verb state, the SF of the relevant reading of over might be indicated as surpassing a norm, which is relied on in stating a case or aim. A simplified sketch is given in (76). (76) [over-] [ ] λv [Vover-] (77) [state]

[… v >

NORM]

[V] λy λx λe [e : [ACT x] [CAUSE [BECOME [PRESENTED y]]]]

(78) [over-state] [V] λy λx λe [e : [ACT x] [CAUSE [BECOME [PRESENTED y] > NORM]]]] It is to be emphasized that this is a highly provisional sketch of a narrow section of the prefix over-, which in turn is only one restricted variant of over. It must furthermore be stressed that prefix combinations are particularly affected by the tendency towards idiosyncratic interpretation that pervades word-formation, as repeatedly noted. Cases like understand, undertake, overreach, overset are obvious examples of a wide variety of idiosyncratic, non-compositional instances of word-formation that extend in various directions.

11. Remarks on conversion The question of whether words like walk in expressions like they walk vs. their walk are instances of grammatical ambiguity or of derivational morphology by means of zeroderivation is a matter of ongoing debate which concerns not merely a problem of definition. The fact that walk and similarly occurrences of run, rise, jump, promise, excuse, rain, and many others can appear with the same meaning as verb and noun must of course be lexically registered; but the items in question also exhibit standard properties of nominalization, i.e. of derivational morphology, as shown by their corresponding, but different, argument structure and other aspects of combinatorial behavior, such as their surprising promise to stay vs. they surprisingly promised to stay. A plausible way to account for these observations is to assume that lexical entries like (79) for walk are categorized by the feature [+Referential], which according to the syntactic category features discussed in (18) classifies nouns and verbs, the characteristically different syntactic properties of which are automatically determined as soon as the second category feature [StrongArg] is specified as plus for the verbal variant, or minus for the nominal variant. In both cases λe provides the referential argument, with either

60. Word-formation and argument structure

1087

nominal or verbal properties, and λx is optional in the nominal case, requiring genitive by c-selection, and it is obligatory with nominative as c-selection in the verbal case. Although these are consequences determined by general conditions on lexical entries, they are also subject to standard conditions of derivationally specified argument structure. (79) [walk]

[+Referential]

λx λe

[e : [WALK x]]

Under this assumption, the initial non-specification of the feature [StrongArg] could be taken to lexically indicate possible conversion. Alternatively, one might consider one categorization − in the present case the verbal one − as basic, with the entry (80) instead of (79) and additionally a phonetically empty suffix (81), belonging to the remarkable family of event nominalizers. (80) [walk]

[+Ref, +StrongArg]

λx λe

[e : [WALK x]]

(81) [0̸]

[+Ref, −StrongArg]

λv [V0̸]

[v]

(81) is an apparently somewhat dubious suffix: it is phonetically and semantically empty, its only effect is to recategorize the verb as noun. The main contribution of the entry is to specify the items to which it may apply by means of the address feature V0̸. This is relevant, however, since verbs may have morphological properties that can’t carry over to nouns: the verb rise preserves its inflectional properties, just as fall or run, irrespective of the potential nominalization. There is, however, an additional reason for the apparently clumsy version based on (81) instead of the straight ambivalent entry (79), namely semantically non-neutral conversion creating verbs like clean, clear, dry, fat, open, round, tidy, etc. based on homonymous adjectives. In these cases the shift from adjectival to verbal usage is accompanied by systematic semantic differences as shown in (82). The pertinent verbal and adjectival version of clean is indicated in (83a) and (83b) (82) a. The table is clean. b. His shirt is dry. (83) a. [clean] b. [clean]

[A] [V]

− He got the table clean. − He got his shirt dry.

λx λx λy λe

− He cleaned the table. − He dried his shirt.

[CLEAN x] [e : [ACT y [CAUSE [BECOME [CLEAN x]]]]]

With respect to SF and argument structure, the verbal entry (83b) differs from the adjective entry (83a) by what has been called in section 10 the causative template (supporting corresponding positions in AS, including the event position λe, which accounts for the situational reference of the verb). Now, adjectives like (83a) are related to their verbalization in (83b) by the zero-affix (84), which applies in the same way as derivational heads in the cases considered so far: (84) [0̸]

[V]

λa λx (λy) λe [A0̸]

[e : [(ACT y [CAUSE) [BECOME [a x]]]]]

1088

Semantics and pragmatics in word-formation I: General aspects

It must be noted that the parentheses around the causative part of the template and the corresponding agent position in AS account for the intransitive, inchoative version in cases like the shop opens at ten, the lips round for u, etc. Besides the zero-suffix, there is, incidentally, a closely related overt suffix -en, that derives deadjectival verbs like shorten, widen, deepen, flatten, redden, blacken, harden, etc. crucially picking out the pertinent adjectives by the address feature Aen. An even more varied zero-derivation than that determined by (84) concerns cases of noun-verb conversion like shelf, box, bridge, saddle, shoulder, etc. This alternation, which is in some sense converse to the verb-noun conversion, is not merely a shift in categorization, but implies a crucial difference in SF which involves essentially the causative template of (84) extended by conditions that integrate the SF of the nouns box, shelf, bridge, saddle, etc. into the resulting situation of the verb by some locative or functional relation. The noun-verb conversion in examples like (85) vs. (86) differs remarkably from the verb-noun conversion of walk, run, or even promise or believe. (85) a. They shelved the dictionary. b. They shipped the furniture. c. They boxed the shoes.

(86) a. They put the dictionary on the shelf. b. They send the furniture by ship. c. They put the shoes in a box.

Simplified entries for the nominal and verbal version of box are sketched in (87), with specifying a (medium sized) container and IN the default relation a theme can have to this relatum, such that (88) would be the phonetically empty affix that turns the noun into the causative verb.

BOX

(87) a. [box] b. [box]

[N] λx [BOX x] [V] λz λy λe [e : [[ACT y] [CAUSE [BECOME [z

(88)

[V] λn λz λy λe [e : [[ACT y] [CAUSE [BECOME [z [N0̸]

[0̸]

LOC

x′ :

LOC

BOX

x′]]]]]

x′ : n x′]]]]]

Besides the non-trivial question, how specific the address feature N0̸ has to be that identifies the nouns to which (88) applies, an even more intriguing problem is whether, and with what content, (88) could be generalized to cover other nouns that surface as verbs. The condition indicated by LOC in (88) holds for containers like box, shelf, maybe house, but only marginally for vehicles like ship, cart, taxi, but clearly not for the different conditions for tools like hammer, drill, saw, harpoon, musical instruments like trumpet, flute, drum, harp, or quite different cases like bridge, book, hand, head, shoulder, saddle, crown, dress and many others, as extensively discussed in Clark and Clark (1979). This suggests two systematically different aspects to be distinguished with respect to this type of conversion (and in fact to word-formation quite generally). On the one hand, the verbal use of a basically nominal item like hand arises − just like the nominal use of a verb jump − through conversion by a zero affix roughly like (88), but with LOC replaced by a sufficiently abstract relation, the more specific instantiation of which might then depend on the different types of concepts represented by the noun undergoing verbalization. Consequently, the SF of the derived verb is to be construed as a frame for different modes of causation, depending on the underlying noun with, e.g., to crown or to dress someone being quite different types of process than to

60. Word-formation and argument structure

1089

bridge or to saw something. On the other hand, the actual SF of a verb corresponding to a PF-identical noun might belong to a lexical entry of its own with default grammatical properties and an SF that defines a situation with a process crucially involving the meaning of the noun, as in to mail, to smoke or to sluice, exploiting the (repeatedly mentioned) tendency of derivational morphology, to include idiosyncratic information into lexical items. These two aspects need not be mutually exclusive, since the first identifies the abstract pattern or skeleton, to which idiosyncratic conditions according to the second aspect can be added. This distinction corresponds closely to that between potential and actual words proposed in Halle (1973), where potential words are determined by morphemes and rules of morphology and related to the actual words by some kind of filter which is responsible for the lexical idiosyncrasies.

12. Argument-structure conditions on compounds A final look has to be taken at the role of argument structure in compound formation which is, in addition to derivational morphology, the second systematic possibility for creating lexical items which can enter into further syntactic combination. While derivation combines a stem with an affix, compound formation combines two stems or lexical items to build up a new lexical item. The delimitation of compound formation is not simple and clear cut, neither with respect to derivation, because of cases like overweight, crosswise, policeman, and countless others, where items like over, wise, or man oscillate between affix and stem, nor with respect to phrasal combinations, where cases like red light, white wine, or pizza eating have properties of words and phrases alike. Because of these and a wide range of other problems, the phenomena and questions concerning compound formation go far beyond the limits of the present article. The following remarks are therefore strictly concerned with issues related to the main topic of this article, namely the role of argument structure in compound formation. A basic orientation to this effect is given by the so-called head principle formulated in Jackendoff (2010), which simply identifies nominal compounds as the combination of two nouns [N N1 N2], one of which is the head which syntactically determines the categorization of the compound and semantically represents the general concept, the further determination of which is provided by the compound as a whole. In other words, in these so-called determinative compounds the determining element provides specifications which restrict in one way or another the meaning of the head. In English and other Germanic languages, the head of determinative compounds is usually N2, as in wine drinker, student discussion, topic choice, or hard cover. This restriction leaves aside combinations like theory of syntax, line of research, brother-in-law, or idiosyncratic cases like pickpocket, which are not noun-noun compounds. Two types of well known systematic exceptions are to be noted: so-called bahuvrihi compounds, such as block-head, green-beret or blue-helmet, rely on metonymic interpretation of the head (cf. article 61 on word-formation and metonymy), the so-called dvandva compounds, like poet-composer, panty-hose, are based on the identification of the denoted concepts and can be treated as a special type of determination of the head. On this background, two issues will be taken up: first, how far can compound formation be explained by the regular assignment of a complement to an argument position of

1090

Semantics and pragmatics in word-formation I: General aspects

the head, as in topic choice, or wine drinker, where N1 serves as object of N2? And second, how are the components of a compound connected, if N1 is not a complement of the head, as in snowman or wineglass? Before these problems are raised, the formal status of compound components is to be clarified. On closer inspection, the N1 of a compound like visitor control or helicopter attack does not exhibit, neither syntactically nor semantically, the standard conditions of phrasal constituents: neither as subject nor as object of N2 does N1 exhibit normal Case properties or one of the standard referential conditions as a term with definite, indefinite or universal reference. It is the kind of attack which is characterized by helicopter as agent or as patient. Whether the unspecific reference (assumed above for non-instantiated variables) would be appropriate in this connection as well, is to be left open. Various explicit and implicit proposals in this direction are discussed among others in Olsen (2012), Bücking (2010) and Jackendoff (2009). Taking for granted that one or the other notational variant will take care of the particular referential conditions and the suppressed c-selection requirements, as a provisional notation, arguments coming with compound constituents will be characterized by primed variables x′, y′, etc., such that wine in wineglass, wine drinker, wine export will be something like x′ [WINE x′]. Turning now to the problem of compound-internal argument structure, as a first step two types of noun-noun compounds are to be distinguished, indicated in (89). In (89a), the first noun meets the s- and c-selection conditions (as far as this is possible for a bare noun) of (one of) the available positions in the AS to which it is assigned. For the cases in (89b) the head does not provide a suitable argument position. (89) a. wine drinker, house cleaning, student assembly, director proposal b. wine bottle, doorknob, Sunday shirt, wool emblem In (90), the combination of the agent nominalization of the verb drink with the determining noun wine is illustrated. (90) a. [drink] b. [-er] c. [drink-er] d. [wine] e. [wine drink-er]

[V] [N] [N] [N] [N]

λy λx λe λv [Ver] λy λx λz λx

[e : [x [DRINK y]]] [v e′] [e′ : [x [DRINK y]]] [WINE z] [e′ : [x [DRINK y′ [WINE y′]]]]

Three more general remarks are to be made. First, the designated (highest) position in AS can never be assigned to complements, but has to provide the position by which the compound as a whole enters further combination or takes up referential function. Hence in the head noun (90c), only λy can be assigned to the complement. For the same reason, there is only one option available in cases like production quality, ball shape, life length, family father, water color, where, although the head is relational, i.e. has two positions in AS, the higher one is the designated position and cannot be assigned to complements. Second, if the head of a compound has an AS with more than one argument position beside the designated position, at least three possibilities are to be observed. (91a) exemplifies cases, where N1 is assigned to the object or theme position. This seems to be the prevailing option (and has occasionally been considered as the only one). The N1 of

60. Word-formation and argument structure

1091

cases like (91b) is construed as agent or subject, if nothing else intervenes. The most intricate cases are those in (91c), where N1 can be assigned to different (in fact inverse) roles, depending on contextual conditions: in defect survival the deficit might be what remains or what is overcome, just as the helicopters of the attack might be aggressive or defensive. (91) a. tax reduction, water supply, pasta eating, book reading, play rehearsal, repair promise, game finish (N1 = theme/object) b. bird singing, child fear, reader correction, student demonstration, driver, agreement, moon rotation (N1 = agent/subject) c. defect survival, helicopter attack, government critique, winner proposal, bank management, machine repair (N1 = object or subject) This shows that the assignment of argument positions in compounds corresponds to semantic and conceptual conditions rather to underlying syntactic structure as advocated in different ways, e.g., in Alexiadou (2007), Borer (2005), and various others. Third, these cases also indicate the preference (or even exclusiveness) of AS-determined relations in noun-noun compounds over other, plausible or at least conceptually possible, interpretations such as survival by means of a defect or even survival due to insurance for defect for defect survival; the same applies equally to other arbitrary cases that cannot rely on AS-positions like singing together with birds for bird singing or reduction through different tax rates for tax reduction. The paradigm case for this preference of AS-positions is (92) discussed in Selkirk (1982). (92) to eat pasta in a tree ===> pasta eating in a tree ===> *tree eating of pasta This does not mean, that N-N compounds like tree eating (interpreted as eating in trees) are excluded or anomalous, as other locative determinations like open air concert or Berlin tower and innumerable other cases indicate. This leads to the second question noted above: how are the components of compounds connected and interpreted otherwise? To answer this question, it is useful to summarize the phenomena discussed so far by the following generalizations: (93) Argument conditions on compound formation: a. The head of a compound determines its categorization, from which the properties of the AS of the compound as a whole derive. b. The designated AS position of the head serves also as the designated position of the compound and hence cannot be occupied by N1. c. If possible, N1 is assigned to one of the positions in the AS of the head, preferably that of the direct object or the subject. According to these conditions, the relation between the components of a compound is determined by the AS of the head and the SF of the head and complement without additional assumptions. It is easy to see, however, that a vast range of noun-noun compounds are not covered by these conditions, either because the head noun does not have the necessary positions in its AS, such as snowman or rainbow and non-relational nouns

1092

Semantics and pragmatics in word-formation I: General aspects

generally, or because the argument positions of the nouns do not fit the pertinent selectional conditions, like watercolor. It has long been observed, e.g., by Meyer (1993), Jackendoff (2010), Bücking (2010), Olsen (2012), that besides the relation determined by the argument conditions in (93), a set of other relations must be available. In this sense, Fanselow (1981) had proposed more than twenty types of compound formation, which Meyer (1993) was able to reduce to essentially two basic types, corresponding to what Jackendoff (2010) later on distinguished by means of argument schema (similar to (93)) and modifier schema (similar to (94)), which covers all other types of noun-noun compounds, emphasizing in particular the different ways in which they go beyond regular semantic compositionality by conditions that do not belong to the component parts. Thus a rainbow is a bend or bow and depends on rain, although in a (semantically unpredictable, but definitional, hence idiosyncratic) way: the origin of the prismatic light is physical, not linguistic. The basic relation involved in compound formation beyond the argument schema consists in the modifier specifying the head. More explicitly, three conditions are to be noted: (94) Modifier conditions on compound formation: a. If N1 is not a complement satisfying an argument position of the head, it becomes a modifier assigning its designated argument position to the head. b. Modification is generally realized by unification of the designated argument position of the modifier with that of the head. c. Unification of argument positions in compounds may involve completely context-dependent conceptual relations between the unified arguments. Unification of AS positions to realize (extensional) modification in phrasal construction was proposed by Higginbotham (1985) and adopted among others in Bierwisch (2003). Its essentials are exemplified in (95): the unification of x and y in (95c) and of x and z in (95e) in the two steps of modification is accompanied in both cases by logical conjunction connecting the semantic components. (95) a. b. c. d. e.

[water] [green] [green water] [in the bowl] [green water in the bowl]

[N] [A] [N] [P]

λx λy λx λz λx

[WATER x] [GREEN y] [[WATER x] o [GREEN x]] [z [IN [DEF w [BOWL w]]]] [WATER x] o [GREEN x] o [x [IN [DEF w [BOWL w]]]]

Along these lines, compounds like seashore, sunshine, street pavement, bed frame or even copulative compounds like poet composer, tractor-trailer can be analyzed as identifying the designated position of their constituents and to (roughly) conjoin their SF. More generally, however, there are two problems with this straightforward identification. One concerns the fact that although street pavement refers to regular instances of pavement and is subject to regular specification and quantification, the modifier street is merely a conceptual qualification without its own reference. The other point is, that most determinative compounds do not really identify the reference of the components: birthday cake does not refer to cakes that are conceptually birthday, and coffee pot is not some coffee and a pot, but a pot with the purpose of containing coffee. Many more

60. Word-formation and argument structure

1093

complex cases are easily adduced. Thus Jackendoff (2010) notes, e.g., police car, where N2 is used by N1, while in grease car N2 is equipped to be fuelled by N1. To account for the almost unlimited range of possibilities for relating the components of determinative compounds, Bücking (2010) and Olsen (2012) propose a compound template roughly like (96), which can be construed as a phonetically empty affix turning N1 into a modifying adjunct, comparable to prefixes like un-, over-, etc. in section 10. The specific property is its empty PF and, like modifying prefixes in general, the lack of features in Cat. (96) Compound template:

[0̸] [ ]

λP λQ λx [[P y′] o [y′ R x] o [Q x]]

P and Q are variables over the semantics of nouns, R is a variable over relations between conceptual instances y′ and referential variables x. The content of R is open to conditions of contextual interpretation. (97) illustrates the application of the template in forming the compound teapot, first instantiating R by IN, which abbreviates the purpose of x to be a container for y′, and then applying this modifier to the head noun pot. (97) a. b. c. d.

[tea] [pot] [tea-] [tea-pot]

[N] [N] [] [N]

λz λy λQ λx λx

[TEA z] [POT y] [[TEA y′] o [y′ [[TEA y′] o [y′

IN IN

x] o [Q x]] x] o [POT x]]

Because of the unrestricted interpretability of R relating the head to a determining concept, the template (96) accounts for arbitrarily farfetched compounds such as bottle flower for a flower that is taken out of (or to be put into) or has the shape of a bottle. Whether and in which way the template (96) is to be associated with restricting conditions, must be left open. Two points are obvious, though, and are independently to be taken for granted. First, there is a kind of general default principle dominating conceptual interpretation: (98) Minimal effort principle: Deviate from the simplest, most direct value for an open variable only on explicit demand. Even though the gist of this principle is obvious and perhaps uncontroversial, it is clearly in need of restriction and more explicit specification, not only with respect to what counts as explicit demand. With regard to (96), the relation of identity could be taken as a kind of minimal effort. The second point is less obvious. It concerns the distinction between modifier- and argument-structure compounds. It is an interesting, and in a way natural, consequence of the compound template that its unrestricted application would allow for the complete elimination of the systematic difference between different types of compounds, turning all complements into modifiers, as the following consideration shows. By means of appropriate interpretive values for R, each complement in an argument-structure compound can be interpreted as a modifier. Intuitively, a wine-drinker is a type of a drinker, hence the object of drinking modifies the head. More explicitly, the semantic relation of an argument to its head can by definition be taken as a possible value for R, relating the complement to the designated position in the head’s AS. (In fact, the Davidsonian con-

1094

Semantics and pragmatics in word-formation I: General aspects

cept of thematic roles is based on exactly this assumption.) The optional realization of this position belongs to the “weak” status of arguments of nouns which can always be dropped. In order to avoid their illegal reintroduction as modifiers, the minimal effort principle has to warrant the prevalence of proper arguments over back-door modifiers. Thus the conditions in (93), according to which an argument position can be assigned to N1 have preference over extending R to saturate head-modifier conditions. It should finally be noted that the observations made with respect to N-N compounds carry over to some extent to N-A compounds such as earthbound, mind-boggling, visitorfriendly with nominal complements and honey-sweet, life-long, knee-deep with nominal modifiers. For obvious reasons, N-A compounds with complements are by far more restricted than N-N compounds, since an adjectival head has fewer argument options than the head in N-N compounds. Therefore modifier compounds like widespread, darkblue, deep-frozen or far-fetched are fairly common, while complement compounds like user-convenient are restricted just as governed complements of adjectives. And potential compounds of the V-A type like ?rhyme-weary would rather be construed as N-A compounds. In any case, items of category N and A are ready to be the head of a compound, compared to elements of V and P which have strong argument positions, but are reluctant to head compounds (presumably a further aspect of the category feature [±StrongArg] considered in (18) above). As for prepositions, X-P compounds are generally excluded, unless combinations like far behind, high over, deep in, long before are construed as words rather than phrases. And for verbs, particle compounds like outline, underscore, cross-classify are the only candidates for compounding (besides a wide range of nonrighthand head items like turn out, muddle through, go ahead, etc.). As indicated in (99), Selkirk (1982) considers therefore only eight of the sixteen logically possible [X Y] compounds as actually attested: (99) a. N N schoolteacher sandstorm

A N highschool hotline

V N rattlesnake blink-light

P N afterthought uptown

b. N A worldwide homesick

A A dark-green long-lasting

V

A

P A underestimated above-noted

c. N V

A

V

V

P V undergo overstate

V

The interaction between the category membership of the head, complement or modifier of a compound, the assignment or the inheritance of an argument position (as in undergo, where the verb proliferates the object position of the preposition) raise questions beyond the present survey. The regular combinations and the gaps indicated in (99) depend on argument structure, but are only partially a matter thereof.

60. Word-formation and argument structure

1095

13. Final synopsis Word-formation is a central component of the combinatorial system of language. Like other combinatorial operations, word-formation consists essentially in the combination of two expressions X and Y to build up a complex expression Z. In the operations involved, the categorization and the argument structure of X and Y play a crucial role. Categorization is a set of morphosyntactic features, argument structure is a hierarchy of positions, each of which relies on a semantic variable associated with conditions on the categorization of expressions whose semantic structure is appropriate to specify the variable in question. On this background, four distinctions are important in order to account for characteristic phenomena of word-formation. (i)

Stem vs. affix. This asymmetry distinguishes free and dependent elements with respect to their combination in complex lexical items: two stems or a stem and an affix, but not two affixes, can be a stem. This does not exclude, however, a recursive stem-affix combination as in un-mis-under-stand-abil-ity, just as a recursive stemstem combination yields a complex compound. A stem-affix combination is usually contrasted as derivation with compound formation, although the borderline between stem and affix is not always clear-cut, as items like able, under, over or out show. (ii) Head vs. non-head. This distinction concerns the morphosyntactic categorization of X and Y, one of which projects its category features to the resulting complex Z, thereby functioning as its head. This projection includes features of inflection, such that inflectional morphology (not considered in this article) is necessarily a matter of the head. The non-head becomes its complement or adjunct, depending on the type of their combination. As the categorization of an expression controls its combinatorial potential – including especially its argument structure – the properties of Z in this respect are those of its head; the categorization of the non-head is not projected, but only checked by the head’s selectional conditions, in case it is to become its complement. The head may be a stem, as in passport-control or overstate but it may also be an affix as in drinkable or denial, where the suffix determines the categorization and hence the emerging argument structure. (iii) Functor vs. argument. This distinction concerns the semantic relation connecting the expressions involved, applying one as a functor to the other, thereby saturating an argument position of the functor. This is possible in two ways: in the headcomplement combination, the head is the functor, while in the head-adjunct combination the non-head is the functor, unifying its argument position with an argument position of the head. These different options hold for compounding as well as for derivation, where the head of a derivation entails the systematic possibility of turning argument positions of the complement into those of the resulting combination as in painter with the head -er and the argument positions of the verb paint by socalled argument inheritance. (iv) Right vs. left. This distinction concerns the linear ordering of head vs. non-head. It is subject to parameterization in particular grammars. English, for example, has a clear preference for the righthand-head parameter in word-formation, both for compounds and derivations, where suffixes are generally heads, while prefixes are mainly modifiers. However problematic cases exist like enlarge or debug.

1096

Semantics and pragmatics in word-formation I: General aspects

Distinction (iv) relates to the PF conditions of an expression, while (i) to (iii) deal with its semantic and grammatical properties as schematized in (2) above. As a further difference with respect to the combinatorial principles of word-formation, one might consider the distinction between word and phrase, which relates to different aspects of the resulting expression Z, especially to the fuzzy boundary between word-formation and syntax, i.e. combinatorial principles inside vs. outside the lexical system, discussed, e.g., by Di Sciullo and Williams (1987). There are various types of consequences of this distinction which could not be dealt with in this article, including stress and intonation or the different types of constraints on linear ordering. One aspect that had to be noted repeatedly, however, is the different degree to which lexical expressions are open to idiosyncratic properties, i.e. to conditions that hold for individual cases only. Although the lexical system of a language does not consist – as sometimes suggested – of all and only the idiosyncratic information of a language, two things are unquestionable: first, arbitrary phenomena must be lexically registered if they belong to the linguistic knowledge that makes up a language and, second, the inclination to accept idiosyncratic properties and interpretations of linguistic expressions is a characteristically lexical phenomenon which no language seems to get along without. This inherent tension between systematicity and arbitrariness, which does not allow for a simple separation of lexicon and syntax, has led, among other things, to the idea of construction grammar (cf. Fillmore, Kay and O’Connor 1988; Goldberg 1995; Goldberg and Jackendoff 2004 as well as article 12 on construction grammar), to mention just a few, according to which the combinatorial patterns of complex expressions are just as much lexical entries as stems, affixes, words, and even idiomatic phrases. Whatever the right place for idiosyncratic properties of complex expressions might be, they still presuppose systematic conditions, i.e. basic elements and principles of combination that allow for the arbitrary and underivable properties of pickpocket or kick the bucket. No proponent of construction grammar would fail to recognize that even idiosyncratic configurations involving argument structure are possible only on the basis of their general principles.

14. References Ajdukiewicz, Kazimierz 1935 Die syntaktische Konnexität. Studia Philosophica 1: 1−27. Alexiadou, Artemis 2007 Argument structure in nominals. In: Artemis Alexiadou, Liliane Haegemen and Melita Stavrou (eds.), Noun Phrases in the Generative Perspective, 477−546. Berlin/New York: de Gruyter. Bierwisch, Manfred 1997 Lexical information from a minimalist point of view. In: Chris Wilder, Hans Martin Gärtner and Manfred Bierwisch (eds.), The Role of Economy Principles in Linguistic Theory, 227−266. Berlin: Akademie Verlag. Bierwisch, Manfred 2003 Heads, complements, adjuncts: Projection and saturation, conceptual interpretation. In: Ewald Lang, Claudia Maienborn and Cathrine Fabricius-Hansen (eds.), Modifying Adjuncts, 113−159. Berlin/New York: de Gruyter. Bierwisch, Manfred 2006 Thematic roles − universal, particular, and idiosyncratic aspects. In: Ina Bornkessel, Matthias Schlesewsky, Bernard Comrie and Angela D. Friederici (eds.), Semantic Role Universals and Argument Linking, 89−126. Berlin/New York: de Gruyter.

60. Word-formation and argument structure

1097

Bierwisch, Manfred 2009 Nominalization − lexical and syntactic aspects. In: Anastasia Giannakidou and Monika Rathert (eds.), Quantification, Definiteness, and Nominalization, 281−320. Oxford: Oxford University Press. Bierwisch, Manfred 2011 Semantic features and primes. In: Claudia Maienborn, Klaus von Heusinger and Paul Portner (eds.), Semantics. An International Handbook of Natural Language Meaning. Vol. 1, 322−357, Berlin/New York: De Gruyter Mouton. Borer, Hagit 2005 In Name Only. Oxford: Oxford University Press. Bücking, Sebastian 2010 German nominal compounds as underspecified names for kinds. In: Susan Olsen (ed.), New Impulses in Word-Formation, 253−281. Hamburg: Buske. Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, Noam 1970 Remarks on nominalization. In: Roderick Jacobs and Peter Rosenbaum (eds.), Readings in English Transformational Grammar. Waltham, MA: Blaisdell. Chomsky, Noam 1986 Knowledge of Language. Its Nature, Origin, and Use. New York: Praeger. Chomsky, Noam 1995 The Minimalist Program. Cambridge, MA: MIT Press. Clark, Eve V. and Herbert H. Clark 1979 When nouns surface as verbs. Language 55: 767−811. DiSciullo, Anna-Maria and Edwin Williams 1987 On the Definition of Word. Cambridge, MA: MIT Press. Dowty, David 2003 The dual analysis of adjuncts/complements in categorial grammar. In: Ewald Lang, Claudia Maienborn and Cathrine Fabricius-Hansen (eds.), Modifying Adjuncts, 33−66. Berlin/New York: de Gruyter Ehrich, Veronika and Irene Rapp 2000 Sortale Bedeutung und Argumentstruktur. Zeitschrift für Sprachwissenschaft 19: 245− 303. Fanselow, Gisbert 1981 Zur Syntax und Semantik der Nominalkomposition. Tübingen: Niemeyer. Fillmore, Charles 1968 The case for case. In: Emmon Bach and Robert T. Harms (eds.), Universals in Linguistic Theory, 1−88. New York: Holt, Rinehart and Winston. Fillmore, Charles, Paul Kay and Mary Catherine O’Connor 1988 Regularity and idiomaticity in grammatical constructions. Language 64: 501−538. Fodor, Jerry A. 1981 Representations. Cambridge, MA: MIT Press. Goldberg, Adele E. 1995 Constructions. A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Goldberg, Adele and Ray S. Jackendoff 2004 The English resultative as a family of constructions. Language 80: 532−568. Grimshaw, Jane 1979 Complement selection and the lexicon. Linguistic Inquiry 10: 279−326. Grimshaw, Jane 1990 Argument Structure. Cambridge, MA: MIT Press.

1098

Semantics and pragmatics in word-formation I: General aspects

Hale, Kenneth and Samuel Jay Keyser 1993 On argument structure and the lexical expression of syntactic relations. In: Kenneth Hale and Samuel Jay Keyser (eds.), The View from Building 20, 53−109. Cambridge, MA: MIT Press. Halle, Morris 1973 Prolegomena to a theory of word formation. Linguistic Inquiry 4: 3−16. Halle, Morris and Alec Marantz 1993 Distributed morphology and the pieces of inflection. In: Kenneth Hale and Samuel Jay Keyser (eds.), The View from Building 20, 111−176. Cambridge, MA: MIT Press. Higginbotham, James 1985 On semantics. Linguistic Inquiry 16: 547−593. Hjelmslev, Louis 1935 La Categorie des Cas, I. Aarhus: Universitetsforlaget. Jackendoff, Ray 1990 Semantic Structures. Cambridge, MA: MIT Press. Jackendoff, Ray 2002 Foundations of Language. Oxford: Oxford University Press. Jackendoff, Ray 2009 Compounding in the parallel architecture and conceptual semantics. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 105−128. Oxford: Oxford University Press. Jackendoff, Ray 2010 Meaning and the Lexicon. Oxford: Oxford University Press. Jakobson, Roman 1936 Beitrag zur allgemeinen Kasuslehre. Travaux du Cercle Linguistique de Prague 6: 240− 288. Katz, Jerrold 1972 Semantic Theory. New York: Harper and Row. Lees, Robert 1960 The Grammar of English Nominalizations. Bloomington, IN: Indiana University Press. Maienborn, Claudia 2003 Event-internal modifiers: Semantic underspecification and conceptual interpretation. In: Ewald Lang, Claudia Maienborn and Cathrine Fabricius-Hansen (eds.), Modifying Adjuncts, 475−509. Berlin/New York: de Gruyter. Meier, Ralf 1993 Compound Comprehension in Isolation and in Context. Tübingen: Niemeyer. Moortgat, Michael 1989 Categorial Investigations. Dordrecht: Foris. Olsen, Susan 2012 Semantics of Compounds. In: Claudia Maienborn, Klaus von Heusinger and Paul Portner (eds.), Semantics. An International Handbook of Natural Language Meaning. Vol. 3, 2120−2150. Berlin/New York: De Gruyter Mouton. Parsons, Terence 1990 Events in Semantics of English. A study in Sub-Atomic Semantics. Cambridge, MA: MIT Press. Pesetsky, David 1982 Paths and categories. Ph.D. dissertation, Department of Linguistics, MIT. Pesetsky, David 1985 Morphology and logical form. Linguistic Inquiry 16: 193−246. Pustejovsky, James 1995 The Generative Lexicon. Cambridge, MA: MIT Press.

61. Word-formation and metonymy

1099

Saussure, Ferdinand de 1916 Cours de linguistique générale. Paris: Payot. Selkirk, Elisabeth 1982 The Syntax of Words. Cambridge, MA: MIT Press. Steedman, Mark 2000 The Syntactic Process. Cambridge, MA: MIT Press. Williams, Edwin 1981 Argument structure and morphology. Linguistic Review 1: 81−114. Wunderlich, Dieter 1996 Lexical Categories. Theoretical Linguistics 22: 1−49. Wunderlich, Dieter 1997a A minimalist model of inflectional morphology. In: Chris Wilder, Hans-Martin Gärtner and Manfred Bierwisch (eds.), The Role of Economy Principles in Linguistic Theory, 267−298. Berlin: Akademie Verlag. Wunderlich, Dieter 1997b Cause and the structure of verbs. Linguistic Inquiry 28: 27−68. Wunderlich, Dieter 2012 Operations on argument structure. In: Claudia Maienborn, Klaus von Heusinger and Paul Portner (eds.), Semantics. An International Handbook of Natural Language Meaning. Vol. 3, 2224−2259. Berlin/New York: De Gruyter Mouton.

Manfred Bierwisch, Berlin (Germany)

61. Word-formation and metonymy 1. 2. 3. 4. 5. 6. 7. 8.

Introduction Literal meaning Principles and problems of metonymy Dot-objects and conceptual shift Aspects of word-formation Metonymy and compound formation Conclusions References

Abstract Metonymy belongs to the mechanisms by which linguistic expressions obtain a transferred interpretation. As a prerequisite, the notion of literal meaning is distinguished from transferred meaning and the different phenomenon of ambiguity. Metonymy is described as referring to a concept by means of a characteristic property, such as an event by its place or a book by its author. A different type of interpretive variation for a given expression is distinguished as “conceptual shift”, based on so-called “dot-objects”, which rely on different sortal aspects of a lexical meaning, like newspaper for a particu-

1100

Semantics and pragmatics in word-formation I: General aspects

lar paper, the information it contains, or the institution producing it. Although morphological derivation systematically modifies the structure and interpretation of linguistic expressions, it usually does not affect the possibilities of transferred interpretation. Finally the asymmetry of compound constituents and the metonymy of a compound like blue helmets are dealt with. The role of fixed, idiosyncratic interpretation is noted as a characteristic property of word-formation, especially affecting transferred meaning. In sum, conceptual shift, metonymy, and metaphor are three quite different principles of non-literal interpretation.

1. Introduction Metonymy is a major type of semantic shift or transposition, which − alongside metaphor and a number of related operations − plays a pervasive role in interpretation of linguistic expressions. Metonymy replaces one conceptual entity by another one that is directly related to it, such as a piece of music replaced by its composer as in (1) or an event replaced by its location as in (2), in contrast to metaphor, a different type of semantic shift, which construes one concept on the basis of another one in terms of similarity or analogy, as in (3). (1)

a. Bach lived in Leipzig for almost thirty years. b. Bach was the most surprising part of the program.

(2)

a. Fukushima suffered from a heavy earthquake. b. Reducing nuclear energy was the answer to Fukushima.

(3)

a. He is the father of two young girls. b. He is the father of the new technology.

For intrinsic reasons, the boundaries between these different types of transposition are not clear cut; (2b) is actually a metonymy based on the location of the tsunami and its consequences, but at the same time a metaphorical interpretation of answer, construing a political decision as an act of communication. On closer inspection, the interpretation of many lexical items involves aspects of conceptual shift and differentiation in one way or the other. There is, in fact, no strict boundary between literal and transposed meanings for a large number of words, including even literal uses of standard expressions like drive, circle, line or head. For principled reasons, metonymy and related operations apply also to the products of word-formation and their conceptual interpretation. What is essential in this respect is not so much the classification of the various “tropes” that have been distinguished, but rather the elucidation of the conditions and mechanisms involved in conceptual shift in contrast or addition to literal meaning. The point to be noted in the present context is that these mechanisms are essentially the same for basic and complex expressions and carry over to products of word-formation − depending on the properties of the expressions they apply to. For the sake of illustration, compare the type of transfer between the interpretation as institution and as building shown for the basic noun bank in (4) with the same type of transfer between products of derivation as illustrated by Metropolitan Opera in (5):

61. Word-formation and metonymy

1101

(4)

a. The maneuvers of the bank ended in a total bankruptcy. b. The height of the bank exceeds that of all neighboring buildings.

(5)

a. The performance of the Metropolitan Opera was exciting. b. The parquet of the Metropolitan Opera will be renovated.

The shift between (4a) and (4b) leaves aside the ambiguity of bank as in riverbank, etc., to which the shift in question does not apply. Different possibilities of transfer are to be observed (besides characteristic parallels) in cases like church, school, or university, depending on specific, in fact idiosyncratic information, which does not belong to derivations like congregation or acceptance. These preliminary considerations indicate the interaction of different factors in the actual interpretation of linguistic expressions. What needs to be clarified is the way in which different conditions contribute to the phenomena of transposed interpretation, to which metonymy belongs. To this effect, certain basic distinctions are to be made that are sometimes difficult to verify.

2. Literal meaning Two aspects are to be noted in this respect: first the distinction between linguistic and extra-linguistic conditions on conceptual interpretation, second the distinction between literal and transferred interpretation. These are different, but not independent aspects, and they will be discussed in turn. To clarify first the distinction between linguistic and extra-linguistic aspects of conceptual interpretation, which is not always obvious, one might compare the different contributions depending on the word open in sentences like (6): (6)

a. The store will be open tomorrow. b. The store will open tomorrow. c. She will open the store tomorrow.

The three occurrences differ not only syntactically in uncontroversial ways − open as an adjective, an intransitive and a transitive verb, respectively −, but also by the contribution of the word open to the semantic structure of the three expressions, specifying a property, a change of state with this property as its eventual result, and an activity by which this change comes about. The linguistic information underlying these three closely related lexical items has been extensively discussed, (7) indicates one of the various versions of representing it, suggesting that open is at least three ways ambiguous, as it belongs to three distinct, though systematically related lexical items: (7)

a. [open] b. [open] c. [open] PF

[+V,+N] [+V,−N] [+V,−N] Cat

x x x

e y AS

e

[NOT CLOSED x] [[BECOME [NOT CLOSED x]] e [[y CAUSE [BECOME [NOT CLOSED x]]] e] SF

1102

Semantics and pragmatics in word-formation I: General aspects

Here, [open] abbreviates the phonetic form PF of open. PF is an array of phonetic features that specifies the linguistically fixed conditions determining the articulatory realization. Likewise, the structure called semantic form SF indicates the linguistically fixed conditions the alternative entries of open contribute to the interpretation on the conceptual side. More specifically, [NOT CLOSED x] in (7a) indicates that CLOSED, whose interpretation covers a large variety of conceptual possibilities to be taken up immediately, is not a property of the entity x. The familiar components BECOME and CAUSE of the intransitive and transitive variant of the verb specify the minimal conditions any change of state and its causation must provide. The causative component supplies a further variable y, to be specified by the agent of the causation. Finally the SF of open as a verb brings in a variable e indicating the event or situation, which the verb specifies and to which primarily (but not exclusively) the information of tense and mood applies. As already noted, PF and SF define the interfaces, in terms of which linguistic expressions interact with extra-linguistic, articulatory and conceptual conditions. The categorization Cat and the argument-structure AS on the other hand indicate the intra-linguistic, combinatorial conditions, which determine the organization of complex expressions. Cat consists of morphosyntactic features like [+V,+N] (for adjectives) and [+V,−N] (for verbs), which determine the role of an expression with regard to the elements it combines with, while the argument-structure AS determines the selection of complements an expression requires or admits. It is represented here by lambda-operators, which identify the argument by indicating the semantic variables, for which further specification is to be adduced by the complements in question. On the basis of entries like (7), the SF of clauses like (8) to (10) comes out as shown, where [DEF SHOP] and [EVE] are short for the SF of the shop and Eve, and [u BEFORE e] is a camouflage for the tense information of will, to be construed as “utterance before event”, i.e. e is in the future: (8) (9)

a. the shop will be open b. [[u BEFORE e] : [NOT CLOSED [DEF a. the shop will open b. [[u BEFORE e ] : [BECOME [NOT

SHOP]]

e]

CLOSED [DEF SHOP]]

(10) a. Eve will open the shop b. [[u BEFORE e] : [EVE CAUSE [BECOME [NOT

e]]

CLOSED [DEF SHOP]]]

e]]

These structures are oversimplified in several respects, but they reasonably represent the linguistically determined semantic relations. In this regard, it should be noted that in (8b) the presence of the event variable e and the tense information applying to it derives from the auxiliary will be, which combines with the predicative adjective open; hence the BECOME-component, which shows up in (9) and (10), does not appear in (8). The difference between (9) and (10), however, is due to the different argument structure of the variant of open, according to which the shop is subject in (9), and Eve is subject in (10), with corresponding consequences in SF. The crucial point is that all the semantic differences (and relations) in these cases are straightforward consequences of linguistic conditions, although in different ways. This holds also for the relation between (10) and the cases in (11), where the syntactic structure is identical, but the lexically based semantic difference is obvious.

61. Word-formation and metonymy

1103

(11) a. Eve will open the window. b. Eve will open the door. c. Eve will open the eyes. What is not obvious at all, however, is the conceptual identity of the resulting state of being open and the process by which it comes about in cases like (10) and (11). Before turning to this type of variation, it is worthwhile to acknowledge the invariant pattern, that open shares with a fair range of other lexical items, as exemplified in (12) by a small selection of typical cases. These items are all based on an adjective that lexically represents a property, to which the processes in question are related. (12)

Adjective a. dry clean tidy b. red black sharp c. false pure liquid

intransitive Verb dry clean redden blacken sharpen

liquefy

transitive Verb dry clean tidy redden blacken sharpen falsify purify liquefy

Cases like (12a) are usually called conversion, as their only non-semantic shift concerns the features in Cat (and their consequences), while (12b) and (12c) are instances of word-formation with the suffixes -en or -ify, the latter (loosely) bound to the SF-component CAUSE. Whether these items are related by rules of conversion and derivation or just by lexical patterns, is to be left open for the moment. Which relations are expressed in which way, is in any case a matter of idiosyncratic lexical determination, in view of cases like black/blacken, but pink/*pinken and many others. The variants these cases exemplify clearly affect the SF-structure and hence the conditions imposed on literal meaning. This does not carry over to open in cases like (11), where the different interpretation of [NOT CLOSED] with respect to the SF of shop, door, window, eye, etc. is not likely to derive from lexical idiosyncracies of open, but merely from the objects involved. If this is correct, then the conceptual distinction between open in open the door vs. open the room is neither a matter of ambiguity nor of transposed interpretation, although it exhibits a clear conceptual difference. This is not a trivial matter, as the denial of the condition CLOSED concerns the boundary of the object in open room, but the object as such in open door, insofar as object and boundary are different in room, but not in door. In other words, two conceptually distinct conditions are involved whether the object is a container or (a part of) its boundary. Thus these two variants of open are not a case of ambiguity with different representations in SF, comparable to, e.g., the two meanings of light (contrasting with dark or heavy). Further variants (still within the range of the condition [NOT CLOSED x]) are easily distinguished, as shown in (13): (13) a. She opened the door. b. She opened the bottle. c. She opened the book.

1104

Semantics and pragmatics in word-formation I: General aspects d. e. f. g. h.

She She She She She

opened opened opened opened opened

the bookstore. her eyes. her hand. her mind. the season.

Each of these cases refers to an event of opening, which conceptually differs from the others − not just in terms of vagueness, but by sufficiently specified factual or conceptual conditions, and (13) is by no means a complete list, since further cases can easily be adduced, such as opening a faucet, a fan, a pipeline, a meeting, a faire, a button, a computer, etc. Different and well defined actions are involved, leading in each case to accessibility of the object’s interior, with the differences based on the conceptual organization by which objects are identified and classified. Whatever details may contribute to the various kinds of being open − they are all variants that realize the condition [NOT CLOSED x] with the variation depending on the instantiation of x. And what is equally important: the related processes of change and causation that result in the situation in question are of the same conceptual diversity: to open the eyes has little to do with opening a bookstore, except that both events interpret the components [CAUSE [BECOME]]. The conditions sketched with respect to open are very general in nature and hold in comparable ways for linguistic expressions at large. To hint at just one further example − even the purely spatial interpretation of the preposition in requires remarkably different relations, if the PP in the water occurs as modifier of fish or boot or platform or salt or turbulence. On the background of these and innumerable other variants illustrating the space of literal interpretation, one has to recognize a fundamental asymmetry between linguistic invariance of meaning determined by SF on the one hand and conceptual specification according to different areas of encyclopedic knowledge on the other, as discussed, among others, in Bierwisch (2007). A comparable sort of asymmetry is to be observed, by the way, with respect to the phonetic form, specifying the interface of linguistic expressions with regard to the sound shape: the articulatory and perceptual properties interpreting PF are rich in terms of extra-linguistic, but still systematic information (like sex, age, mood or identity of the speaker), and the additional conditions are, of course, determined by extra-linguistic systems which knowledge has to cope with. Schematically, the linguistic interfaces and their interpretation can be sketched as follows: (14) A/P !#% PF *0 SF !#% C/I A/P abbreviates the various aspects of articulation and perception and C/I the different domains of conceptual knowledge and intentional organization of experience. Two important points are to be made on this background. First, properties and relations relevant in C/I are not exclusively determined by conditions of SF, but may be due to additional, extra-linguistic frameworks of knowledge and orientation. Searle (1979) emphasizes the principled distinction and dependency among the two aspects, which he calls “sentence meaning” and “context dependent literal meaning”. Second, even though the structures determined by SF and their interpretation in C/ I may depend on different principles, they are not in general incompatible − they might

61. Word-formation and metonymy

1105

even be projectable according to both types of principles (somewhat like black-andwhite drawings vs. colored paintings showing the same scenario). As a rough approximation, the principles according to which SF and C/I seem to be related can be indicated as follows: (15) a. Autonomy: Structures of SF and C/I are subject to independent conditions, based on at least partially different principles. b. Compatibility: SF and C/I are organized by units and dependencies that exhibit a partial correspondence. c. Intersection: This correspondence requires the identity of certain key elements. Intuitively, the key elements stipulated in (15c) organize the conceptual integration, as illustrated above by the conditions related to open due to the relevant type of object. These formulations are clearly in need of clarification, far beyond what the present limits allow for. Hence for the time being, provisional plausibility must be sufficient. It might be noted, however, that the compatibility condition is to a large extent the rationale for what is usually called S-selection, i.e. the semantic selection restriction, bound to the argument positions in AS, according to which, e.g., the interpretation of he doesn’t sleep requires he to be animate. The observations related to (13) and similar phenomena are decisive for the nature of literal interpretation, but they have − by definition − little to do with metonymy and other types of transposed interpretation. They provide, however, a useful perspective on the domain where metonymy and transferred interpretation actually occur.

3. Principles and problems of metonymy Transferred interpretation is not always easily distinguished from literal interpretation on the one hand and regular ambiguity on the other. The space for literal meaning is provided and delimited by the linguistic properties of a given expression, essentially its semantic form, determined by the lexical items and their morphosyntactic combination. Ambiguity originates from the literal interpretation of an expression with two (or more) alternative SF-representations, while transferred interpretation is the effect of an expression that has no literal interpretation and needs accommodation to yield an acceptable conceptual interpretation. Cases like (13) above illustrate the range of literal interpretation for transitive open, while (16) exemplifies the ambiguity of right, contrasting either with left or with wrong: (16) a. He is sitting on the right side as the left side was already occupied. b. He is sitting on the right side to have an excellent view. Although the two meanings of right are etymologically related and must at some stage have been connected by variants of literal interpretation, they are a sufficiently clear case of ambiguity with necessarily distinct representations in SF, specifying roughly direction vs. correctness. What is not obvious, though, is whether cases like right line and a right angle indicate further distinctions in SF of right or merely different conceptu-

1106

Semantics and pragmatics in word-formation I: General aspects

al conditions of application (which seems to hold for varieties like the right proposal vs. the right person). In other words, variation and ambiguity are not always completely separate, due to the fact that the correspondence between systematic differences in SF and the corresponding distinctions in C/I might be more or less fixed and obvious. Different, but comparable observations hold with respect to literal vs. transferred interpretation. Clear cases of indirect, metonymic interpretation are examples like (17), extensively discussed, among others, in Nunberg (1979) and Jackendoff (2010). (17) a. The ham-sandwich is sitting at table 20. b. You are parked just around the corner. c. Russell is on the top shelf next to Plato. Expressions like (17) do not by themselves provide a plausible literal interpretation, but on the background of appropriate situations, they are readily construed along the lines indicated in (18), where the parenthesized components supply the adaptation to be assumed in order to arrive at an acceptable literal interpretation: (18) a. (The person with) the ham-sandwich is sitting at table 20. b. You (r cars) are parked just around the corner. c. (The book by) Russell is on the top shelf next to (the sculpture of) Plato. The four instances of transferred interpretation can be made explicit as indicated in (18) by the parenthesized amendments, which are abridged in (19). These correspondences can then be construed as manifestations of a general pattern that can roughly be indicated as (20). (19) a. b. c. d.

the ham-sandwich → the person with the ham-sandwich you → the cars of yours Russell → the book by Russell Plato → the sculpture of Plato

(20) X → the Y of/with/by According to (20), the actual metonym, i.e. the replacing term X, is to be construed as a kind of modifier of the replaced element Y. In other words, metonymy does not substitute the metonym X for the (actually envisaged) element Y, but turns X into a marked modifier of the tacit head Y, which relates to the actually intended underlying concept. Hence in these cases, a syntactically organized head-modifier-combination is the linguistic frame in which the metonym is interpreted. (It is worth noticing that the same semantic relation is the core of what has been called bahuvrihi by Panini with regard to socalled exocentric compounds like redbreast, i.e. (bird with) red breast, cf. section 7.) The technical term for the adjustment, by which an apparently deficient expression becomes conceptually interpretable, is coercion. A couple of remarks seem to be indicated in this respect. First, the basic element, which the intended interpretation eventually aims at, is deleted from the actual expression in favor of one of its prominent attributes. This foregrounding of a characteristic attribute (the place, the possessor, the author, etc.) instead of the

61. Word-formation and metonymy

1107

intended concept is the actual force of the maneuver. Second, this foregrounding thrives on the attribute’s factual connection to its relatum, which is thereby automatically kept in focus, providing the actual referent. This means, that the relation between X and Y, and the whole state of affairs which coercion exploits, is based on contextual and encyclopedic knowledge. It depends on linguistic conditions just as far as the specificity of the tacit unit Y is subject to selection restrictions, such as the subject of sitting at table 20 in (17a), which requires a person. Third, and most importantly, the linguistic structure of the expression as a whole, especially its syntactic and semantic make up, is subject to the autonomous requirements, which metonymy and coercion have to comply with. Thus in cases like (17b), the actual you are parked accommodates the conceptually appropriate and grammatically correct your car is parked. This mutual autonomy of linguistic and conceptual organization is one aspect of the correspondence between SF and C/I in general, as noted in (15). With regard to metonymy, this means that the foregrounded unit X usually provides the key noted in (15c) by which linguistic and conceptual structure are interfaced. The linguistic side of metonymy, based on a particular case of coercion, can be characterized as a peculiar sort of lexical entry, the specific effect of which is to switch or expand the PF-SF-correspondence of an already established expression in the way exemplified in (18). Phonetically, the entry in question is empty, similar to the nullmorpheme involved in conversion in cases like [walk]Verb → [walk]Noun for walk and similar items. The relevant operation is easily observed in an obvious case like Russell in (18c). The basic linguistic properties of this item can be indicated by an entry like (21) with standard PF- and Cat components and an argument structure that provisionally marks the definite referentiality of proper names by means of the iota operator ιx. As to SF, the element [PERSON] in the conjunct [PERSON x] o [RUSSELL X] is a basic semantic feature, very much like [ANIMATE], [CAUSE], etc.; [RUSSELL] however is a hybrid element, called dossier in Bierwisch (2011), (or frame in a somewhat different context in Barsalou 1999), the interpretation of which in C/I includes all aspects by which the individual in question is identified and characterized, including visual, auditory, and episodic information. In other words, with respect to SF, a dossier like [RUSSELL] is a highly idiosyncratic feature, whose interpretation in C/I brings in complex, specific knowledge, which may differ for different speakers and includes in any case all relevant encyclopedic knowledge. (21)

[rusǝl] [+N, –V, +Male,…] x PF

Cat

AS

[[PERSON x]

[RUSSELL x]] SF

The metonym Russell in (17c), based on an item of this sort, must be construed, however, as an expression with different linguistic information − roughly of the sort indicated in (22), where [BOOK] abbreviates another (type of) dossier, this time characterizing a characteristic type of object, rather than a named individual, again relying on all sorts of encyclopedic conditions though, including those that are ambivalent for reasons to be noted immediately, because the conceptual interpretation of the feature [BY], which is a relational element like [CAUSE] or [BEFORE], is heavily determined by the entities it relates. In the present case, the natural relation in C/I is something like produced by or

1108

Semantics and pragmatics in word-formation I: General aspects

more specifically authored by, which brings in quite different aspects of producing a book. (22)

[rusǝl] [+N, –V,…] PF

Cat

y

[[BOOK y]

[y BY x [[PERSON x]

AS

[RUSSELL x]]]]

SF

For obvious reasons, the book y, to which the metonymic use of Russell refers, must be different from the person x who produced the book in question. But the author of y is of course not the producer of the copy of y’ referred to in (17c). Hence the way in which the relation [y BY x] is to be interpreted involves the complete range of conditions related to the fact that Russell is the author of a book, a copy of which ends up on the shelf. As already noted, the type of relation between the structures involved in metonymy can be considered as the result of an operation which a very specific kind of lexical entry determines. For the present example, it relates (21) to (22), which means it substitutes the SF of (22) for that of (21). In other words, without changing the PF-information, coercion enriches the SF of (21) by adding the condition that y refers to a book with specific relation to x. This operation can be represented by the structure in (23), which is in fact a lexical entry with empty PF and transparent categorization: (23)

0/

0/

P y

[[BOOK y]

[y BY P]]

(23) is an operator that substitutes the argument to which it applies for the variable P in its own SF, dropping the argument position λP from its AS and preserving the categorization of its argument. Technically, (23) can in fact be interpreted as an operator that applies to (21) and substitutes its SF (via lambda-conversion) for the variable P in (23), thereby turning it into (22). (The formal details of this operation, which introduce the component [BOOK y] into (21) and switch the AS from ιx to ιy, can be taken for granted.) More generally, though, (23) could be taken as a kind of metonym mold with [BOOK] as the possible value of an appropriate variable, that is to be filled in by whatever shows up as the prevailing feature of the component P to be taken up. Thus in (24a) to (24d) it is likely to be instantiated by [MUSIC], [PAINTING], [EVENT], and [ROOM], respectively. (24) a. b. c. d.

He doesn’t like Wagner. Chagall is dominating the exhibition. Pearl Harbor shocked the United States. This (presenting a key) is your lodging tonight.

The linguistic side of cases like those considered so far concerns nominal metonyms with adjustments of the type based on (23). This is a familiar type of transferred interpretation, but by no means the only one. While one would hardly consider prepositions like in, of, or from in constructions like believe in, think of, or protect from as subject to transferred interpretation, but rather treat them as lexically fixed attendants or even markers of argument positions of their respective heads, the interpretation of verbs like stretch, infect, or contaminate, and the like, in constructions such as he stretched the

61. Word-formation and metonymy

1109

analogy, the proposal was politically infected, a paper contaminated with dubious slogans, are clear cases of transferred interpretation. Although they are usually not classified as metonyms, they are based on the same kind of coercion, as cases like (25) may demonstrate. (25) a. Their proposal is rooted in their tradition. b. The computer needs to be debugged. c. He houdinied the cage easily. Comparable to canonical cases like (17), the metonymic item of constructions such as (25) shows up in place of an unrealized, more general term, the specifying modifier of which surfaces as the actual metonym. As shown in (26), this modifier may acquire the necessary syntactic categorization by derivational adjustment. (26) a. Their proposal has deep roots in their tradition. b. The computer needs to be (cleaned by removing) bugs. c. He easily (escaped like) Houdini the cage. The morphosyntactic adaptation, which relates (25) to (26), requires a slight modification of the pattern of coercion (20), noted in (27) by Y à la X, suggesting that X specifies − exactly as in (20) − the particular mode of the general condition Y: (27) X → Y à la X While the derivation of entrench and debug is independent of metonymy, the verbalization of the proper name Houdini is restricted to transferred interpretation. What all these cases of transferred interpretation share with proper ambiguity like bank, light, or ring, etc. is the usually clear delimitation of the derived interpretation, which might in turn be open to variation like literal meaning in general, as debug shows with program, computer or embassy as object, or stretch with license, comparison, or duration. However, as already noted, the distinction between flexible literal interpretation and transfer as a self-contained, separate option is not always clear. Consider for example the semantic component SEE. Assuming the process of visual perception to provide the domain for literal interpretation of see, glance, view and related items, would for a simple utterance of I see, which usually has little to do with visual perception, but merely signals cognitive awareness, lead to the following alternative: either the component SEE is interpreted not just by visual perception, but refers in addition (or instead) to its natural effects and consequences, or see would in many instances be turned into a metonym, which substitutes − according to (27) − vision for experience in general. This problem extends to a wide range of cases. Consider hear and its interpretation by auditory perception for instances of did you hear about it? which naturally include language and other aspects of communication. Similar questions arise with verbs of motion and expressions of shape like line, square, etc. Further problems are related to examples of the type (28a), which require a kind of coercion along the lines sketched in (28b). (28) a. Could you keep an eye on it? b. Could you keep a (target of the activity of the) eye on it?

1110

Semantics and pragmatics in word-formation I: General aspects

The upshot of these considerations is that metonymy and transferred interpretation in general is not a peripheral phenomenon, occasionally extending the range of literal meaning, but a ubiquitous and fundamental principle of semantic interpretation.

4. Dot-objects and conceptual shift Besides the general flexibility of conceptual interpretation and the possibility to extend it by coercion, there is a type of ambivalence that neither relies on mere vagueness nor on metonymy or other types of coercion. In cases like (29) a regular alternative is part of the literal interpretation without creating an ambiguity or transferred interpretation. (29) a. The book lies on the table. b. The book was heavily criticized in the press. c. The book on the table is very fascinating. In (29a), the book means a physical object, in (29b) a particular information structure. These are entities with ontologically incompatible types of properties (the press does not criticize particular copies), but instances like (29c) − and a wide range of normal conditions − rely on just this standard integration (parallel to that of the computer and its program, or even of mind and body). Pustejovsky (1995) and Jackendoff (2002) have proposed a particular notation to account for this dual nature of so-called “dot-objects” (a term derived from the proposed notation), integrating elements of SF that are at variance with logical conjunction. (30) [PHYSICAL

OBJECT

•

INFORMATION]

It is not clear, whether this is more than a notational convention that allows one to rely alternatively or even simultaneously on incompatible conditions. It is clear, however, that the intended effects cannot be restricted to separate lexical items, as cases like (31) demonstrate: (31) a. Somewhat later Mary left the bank. b. Two hours later Mary left the bank. c. Two years later Mary left the bank. Putting aside the proper ambiguity, by which bank can refer to a financial or a river bank, (31a) has a straight literal meaning, indicating presumably a change of Mary’s location, subject only to standard conditions of vagueness, leaving various aspects of the event unspecified. But Mary as well as bank represent particular types of dot-objects, and (31a) covers at least two quite different scenarios, which become obvious if the unspecific somewhat is replaced by definite quantities. Thus in (31b) the change concerns Mary’s localization, while in (31c) it plausibly has to do with Mary’s social relations, professional interest, etc. and hardly her physical location. Hence bank means a room or building in (31b), but a particular institution located there in (31c). Likewise, Mary is a physical person in (31b), but a participant in social relations in (31c). Corre-

61. Word-formation and metonymy

1111

spondingly different conditions hold for the interpretation of leave, leading to change of location in (31b) and shift of social or professional relations in (31c). This foregrounding of different situations does not depend, of course, on the SF of hour vs. year, but on the encyclopedic knowledge about monetary institutions and the conditions applying to them. Different time schedules just go with independently structured scenarios. Examples like these show, moreover, that there are not only different kinds of dotobjects − physical objects integrated with information as in book, physical objects integrated with mental reality as in Mary (and people in general), physical objects hosting social institutions, as in bank or academy, and several others − but also systematic relations between them across lexical items: In (31b), Mary and the bank are addressed with respect to their physical nature, while in (31c) the institutional aspect of the bank picks up the mental or social aspect of the person, due to the fact that social institutions are ultimately the result of interacting minds. Hence conceptual shift involves not only integration of heterogeneous aspects within dot-objects, but also systematic interaction between dot-structures across different items, including the connections and processes relating them. It follows from this observation, that the role of dot-objects belongs to a more comprehensive range of phenomena that might be called “conceptual shift” as discussed in Bierwisch (1983). What is at issue is the fact that different interpretations of an expression in C/I can involve distinctions (up to ontologically different classification of the entities involved) which depend on distinctions in SF that are nevertheless not mutually exclusive. In other words, the dot-notation indicates the conceptual unification of distinctions that otherwise give rise to ambiguity in the sense of separate representation in SF. This observation automatically involves the question as to whether and in what way conceptual shift affects the SF of verbs like leave in (31), or open discussed above with respect to cases like (11), or lose in (32) with the dot-object money as a physical object in (32a), as an abstract value in (32c), and presumably both in (32b), such that the change concerns location in (32a), abstract assignment in (32c), and different combinations of both in (32b). (32) a. Fred lost all his money, as he didn’t take care of his pocket. b. Fred lost all his money gambling at cards the whole night. c. Fred lost all his money in the recent financial crisis. This raises the question of whether the SF of the verbs in question involves dotted conditions, qualifying, e.g., the feature [BECOME] in the entries (7a) and (7b) for open, or in the lexical entries for lose, leave, etc. It seems however, that the relevant verbs are just not specified with respect to the alternatives in question. They adapt conceptually to different domains according to conditions given by the entities to which they apply, as noted above with respect to open. This is in line with the principles of interpretation proposed in (15), according to which the dot-objects would be the key-elements of a coherent interpretation. This conclusion is corroborated by the observation that in cases like He lost a lot or She had left it the alternative option does not come up, if the object of the verb is left unspecified. This does not mean, of course, that verbs don’t impose conditions on the interpretation of their arguments. A simple case like he didn’t drink it allows the object to be non-liquid only by way of coerced adjustment to the feature [LIQUID].

1112

Semantics and pragmatics in word-formation I: General aspects

Similar conclusions seem to hold for adjectives. Independently of proper ambiguity, as in right, left or light, there are adaptations to rather different conditions, as in (33). (33) a. b. c. d.

A strong horse was drawing the cart. He used a strong stick. They made a strong impression. This is a strong argument.

Even though in these (and many other) cases the varying interpretation of the adjective can be related to physical, mental, or informational conditions, there is little reason to reduce this to SF-differences, in contrast to cases like heavy smoker, where heavy is not the usual qualification of physical weight. A rather different problem with dot-objects is illustrated by cases like (34), which seem to suggest that more than just two types of entities are veiled in the representation of one object. Like the meaning of book, that of newspaper integrates a physical object as in (34a) with an information structure as in (34b), but also the set of copies representing the same information, as in (34c). But it furthermore integrates the institution producing it, as in (34e) − thereby differing from book, which cannot include the publisher, − and the different types of representing the information, as in (34d). (34) a. b. c. d. e.

The The The The The

newspaper newspaper newspaper newspaper newspaper

fell from the table. caused a veritable political crisis. I was asking for is sold out. has now an internet edition. dismissed several employees.

Further relevant distinctions might be included, as other cases suggest. This sort of “multiple dotting”, which might further be diversified by cases like opera, museum, parliament, court, and others, with different possible combinations as exemplified in (35), raises the question as to whether the distinctions to be recognized are a matter of the semantic form of the expressions in question and in which way they are to be captured. (35) a. b. c. d. e.

The opera originated in the Renaissance. The opera is famous for its wonderful overture. He went to the opera an hour ago. The opera needs a new director. The publisher promised to send the opera by air mail.

Thus (35a) to (35d) may derive from SF-structures that are comparable to those required for (34), differing mainly by the qualification of the information structure, while (35e) metonymically replaces opera for score of opera by coercion along the lines sketched in (20). The different combinations of concrete and abstract entities relied on in (29) and (31) to (35) give rise to the question how dot-objects and coercion relate to each other, both dealing with different domains of interpretation for the same expression. Do they in fact invoke different principles?

61. Word-formation and metonymy

1113

A fairly clear example illustrating this point is shown in (36), where in both cases the localization of a physical object related to an author is at issue. Likewise in (37) the presentation of physical objects, metonymically identified through their painter, contrasts with the behavior of the artist himself − two different sorts of physical objects. (36) a. Russell is on the top shelf next to Plato. b. Russell is sitting over there talking to his wife. (37) a. Chagall is dominating the exhibition. b. Chagall went to the exhibition a couple of times. The contrast between the transposed interpretation on the basis of coercion in the (a)cases and the completely literal interpretation in the (b)-cases is obvious: the objects created by a writer or painter are separate entities, not integrated in the interpretation of the person. The boundary between dot-objects and coercion is not always simple and evident. Where, for instance, the different referents of opera are a matter of direct interpretation and where they are a case of coercion, can only be decided by closer study of the SF and its interpretations for items like opera, theatre, concert hall and related entities. The different principles are clear enough, however: interpretation that goes beyond mere flexibility is either based on integrated alternatives of the dot-object sort, or derives by coercion, which has a fair range of flexibility.

5. Aspects of word-formation These observations are independent of word-formation, insofar as literal and transposed interpretation of linguistic expressions is based on their semantic form, whether or not word-formation is involved. Thus the metonymic coercion in (38b), where the famous composer replaces (office of) the famous composer, and the conceptual shift in (39), where the three variants are parallel to those in (29) for book, are independent of the fact that composer and publication are based on derivation. (38) a. The famous composer is waiting in the lounge. b. This (pointing to the door) is the famous composer. (39) a. The publication has a red cover. b. The publication is highly stimulating. c. The publication with the red cover is very interesting. Cases like these (and plenty of others) suggest that the results word-formation, i.e. lexical items constructed on the basis of other lexical elements, participate normally in general possibilities of direct or transferred interpretation with no separate, independent effect of derivation and compounding. Although this seems to hold for a wide range of cases, it is not a general principle, as certain types of phenomena show, some of which are now to be discussed. One kind of problem is connected to deverbal nominalizations of the type which (39) belongs to. The actual problem has to be discussed in three steps. The first step concerns

1114

Semantics and pragmatics in word-formation I: General aspects

the parallel properties to be observed in constructions like (40), where the verbal and the nominal realization have roughly the same conceptual interpretation, while relevant details of the morphosyntactic structure differ systematically: (40) a. Paul reconstructed the document only provisionally. b. Paul’s reconstruction of the document was only provisional. These observations can be reduced to a suffix -ion, which turns the verb reconstruct into the noun reconstruction with pertinent consequences for its argument structure, while it leaves the semantic form unchanged, just turning it into the SF of the noun. Formally, the suffix takes the verb as argument, substituting its SF for the variable bound by its only argument position. With apt simplification, this can be represented as in (41), which is formally a standard lexical item with PF, Cat, AS, and SF. The argument position λv must be associated with grammatical features abbreviated here as [+F], classifying the lexical items which the suffix can select as the argument it combines with. (41) [-ion] [+N,−V]

λv [v] [+F]

Following more general principles, this combination is realized as functional composition, a formal operation by which the resulting combination inherits the argument positions of its former argument, saturating them, however, according to the now dominating categorization. This means in the present case, that the subject and object position of the verb are realized as (optional) genitive attributes of the noun, and the event position of the verb becomes the referential position of the noun. To illustrate these conditions, the entry for reconstruct (itself a derivationally complex item) is abbreviated as (42a): its Cat is assumed to contain the feature [+F], such that the suffix -ion can apply to yield (42b), where the position λe is now the standard referential position of a noun, subject to quantification, modification, etc. as in this provisional reconstruction. (42) a. [ri:konstrʌkt] b. [ri:konstrʌkt-ion]

[+V,−N,+F] [−V,+N]

λy λx λe λy λx λe

[[x [[x

RECONSTRUCT RECONSTRUCT

y] e] y] e]

The nominal event reference leads to the second step concerning the meaning of derived nouns. While retaining the conceptually unmodified event reference of the verb seems to be appropriate in cases like (43a), this does not hold for the examples in (43b−d), where ontologically different types of entities are referred to: (43) a. b. c. d.

The The The The

reconstruction reconstruction reconstruction reconstruction

of the bridge was proposed two years ago. of the bridge took more than three years. consists completely of aluminum. was saluted with a great firework.

Event Process Result-Object Result-State

The properties assigned to the entity which the noun refers to in these cases are disparate in the same sense as the physical and informational properties of book discussed earlier. As the differences in (43) obviously depend on the nominalization by -ion, this operation appears not to be semantically vacuous. One might account for these differences by a

61. Word-formation and metonymy

1115

modified entry for the suffix whose SF provides different alternatives, referring to the event, process, resulting object or state of the underlying verb. Instead of (41), one would have to assume something like (44), where the derived noun inherits the AS of the verb except for the event reference, which is reinvested, so to speak, with the properties in question imposed on it (cf. Dölling forthc. for an analysis along these lines). λV [λe′ [V e′ • P e′]] [+F] where P varies over PROCESS, RESULT STATE,

(44) [-ion]

[+N,−V]

RESULT OBJECT

This proposal assumes that the possible alternatives for P are connected to the primary event reference much like the alternative options of dot-object discussed above. On this basis, (42b) would be replaced, e.g., by (45), to account for the reading of the noun in (43b). (45) [ri:konstrʌkt-ion]

[−V,+N]

λy λe′

[[x

RECONSTRUCT

y] e′ •

PROCESS

e′]

This implies the need to specify the value for P, i.e. PROCESS rather than OBJECT or STATE in the present case. Before this problem is taken up, the third step regarding the interpretation of derived nouns is to be adduced. It concerns the fact that English (like many other languages) has several suffixal means, including zero-derivation (or conversion), to formally realize this morphological relation. The main options, which include various idiosyncratic adjustments, like destroy − destruction, derive − derivation, etc. are listed and exemplified in (46). (46) a. b. c. d. e. f.

-ing: -(at)ion: -0̸: -al: -ment: -ance:

open publish promise propose accomplish perform

→ → → → → →

opening publication promise proposal accomplishment performance

grow construct walk refuse settle accept

→ → → → → →

growing construction walk refusal settlement acceptance

Putting aside the additional (and more systematic) role of -ing as a gerundive suffix, these are different formal ways of nominalization with essentially the same syntactic and semantic effect. In other words, the suffixes in (46) are different realizations of the same abstract morpheme nom (just as the abstract morpheme past is differently realized in walked, fell, came, went, etc.) Hence, the entry (44) must therefore be extended into a family of otherwise identical items, differing in their PF and in the correspondingly different selection features [+Fi] of their argument position λv, such that each of the items selects its aptly categorized verbs. Could these differences involve a different SF, contributing thereby to different interpretations of the sort illustrated in (43)? This would only be possible, if the categorization of the selected verbs could reflect shared semantic properties for each of the groups, such that each suffix is bound to specific options with respect to the variants PROCESS, RESULT OBJECT, etc. noted in (44). Even superficial inspection shows this to be impossible, and conditions of this sort have in fact never been stipulated for the different types of nominalization. The choice of suffixes is idiosyncratic; to be selected by one or the other suffix is no more systematic than to belong

1116

Semantics and pragmatics in word-formation I: General aspects

to one or the other inflection class. In short: Neither the PF of the suffix nor the class of verbs it selects contribute to its role in conceptual interpretation. Looking now into the actual interaction between derivation and different types of interpretation, one has to distinguish two sources for conditions involved in the cases exemplified in (39) and (43). The examples in (39) − especially the choice among the concrete and abstract aspect of the interesting publication with the red cover − rely on the alternatives of the dot-object type discussed above. They are only available, however, if as a first step the nominalization of publish by means of something like (44) has led to RESULT OBJECT as the value to be chosen for the variable P. These two interpretive decisions − reference to an object and specification of its aspects − are in a way orthogonal to each other, with the second depending on, but not determined by the first. Several questions concerning the origin and the determination of these options immediately arise. As far as they have been discussed, e.g., in Dölling (forthc.), some sort of coercion operation has been proposed. Technical problems aside, it still seems doubtful whether it is appropriate (or even coherently possible) to specify these distinctions in SF. Whether or not an event leads to a result object (with or without dot-object ontology), need not appear in the SF it interprets. It is involved as a fact in most cases of the interpretation of publish, but usually not of refuse (considering the event, not its content), and this obviously does not change in the SF of the publication and refusal. Similar considerations hold for a wide, in fact an indefinite, range of cases. Which possibilities for event reference and its pertinent aspects emerge, e.g., for experience in his experience was awful depends largely on the imagination of the speaker/hearer, which means they are undetermined (but not vacuous), if they are not fixed by their verbal expression. This is roughly the reasoning already proposed above with respect to open, where the conditions of the resulting state as well as the processes leading to it were claimed to derive from the nature of the object involved. More generally, if one is serious about the assumption of a self-contained conceptual structure C/I as a system of mental organization − and this assumption seems to be unavoidable − then it doesn’t make sense to stipulate the SF of a linguistic expression to be a complete duplication of the configurations that mentally represent complex states of affairs which interpret the SF in question. In other words, the structures and principles of C/I represent the way in which experience is organized in its own right, and the elements and relations in SF determine the way in which its conceptual interpretation is ultimately mapped on PF, but it does not reflect all elements, principles and laws, which C/I relies on. The domain where this disparity and the interdependence between linguistic and conceptual structure behind it is most extensively explored, is spatial structure and spatial orientation. For a survey of various aspects see, e.g., Bloom et al. (1996), where semantic and spatial representations are shown to be subject to rather different conditions of detail and efficiency. Under this perspective, the SF-representation of dotobjects might simply be a way to reconcile formal requirements with the fact that apparently disparate physical and mental aspects of the corresponding entities must not be construed as semantically incompatible. The common sense ontology of mind and body or of sound and meaning is just not that of classical logic and mechanical dualism. These remarks must not be read as a plea for escape from formal analysis, but as a hint at the fact that SF is necessarily more abstract than C/I and not subject to the same principles. For the sake of illustration, consider the phrase X is blue, which has very different interpretations, if X is a car, the sky, the ink in a pen, a person’s eye, a plum,

61. Word-formation and metonymy

1117

a canvas, or a flower, but it doesn’t make sense to assume a different SF-structure for is blue in each of these cases (where not delimitation of the hue is at issue, but the way in which the color applies to aspects or parts of an object). In a similar vein, different aspects of a situation − event, process, state, result, etc. − can be relied on in nominalization without identical distinctions represented in SF. Crucial dependencies of interpretation remain unchanged, however: publication or reconstruction can only exhibit the “dual ontology” if they are interpreted as objects, not as processes. This might still be considered as the interplay of word-formation and interpretation. These considerations carry over to other types of derivation, such as deadjectival nominalizations like redness, stupidity, or height, again based on a family of suffixes, as indicated in (47). (47) a. b. c. d.

-ity: -ness: -hood: -th:

absurd clever likely strong

→ → → →

absurdity cleverness likelihood strength

stupid tall false long

→ → → →

stupidity tallness falsehood length

The entries underlying these nominalizations have much in common with those applying to verbs, although with systematic differences. The point is the verbal event reference, picked up and proliferated in the nominalization by (44), while for adjectives event reference must first be introduced, in order to then rely on it, as, e.g., in Sue’s cleverness. Thus, among the nominalizers for adjectives, there must be an entry like this: (48) [-nis]

[+N,−V]

[λe [A e ]] λA [+Fʺ]

Assuming that the entry for the adjective open, given in (7a) above, contains the feature [+Fʺ] in its Cat, such that it can be selected for the argument position λA of (48), one gets (49) for openness, which inherits the (now optional) position λx of the former subject position of the adjective. (49) [open-nis]

[+N,−V]

λx λe

[[NOT

CLOSED

x] e]

In a way, an adjective nominalized by a suffix of this sort brings in the reference to a state e in which its subject x exhibits the property in question. That this is different from referring to the property, can be seen in cases like (50), where the color term red by its lexical structure allows direct reference to the color. (50) a. The red of the car is the same as that of the pullover. b. The redness of the car is due to the rust. That this subtle difference seems actually to be reflected in SF is apparently due to the particular structure of color terms, which leads to minimal pairs like (51), where (51a) creates a contrast for which apparently little conceptual support exists and which other adjectives do not allow, as (51b) indicates.

1118

Semantics and pragmatics in word-formation I: General aspects

(51) a. the car is red b. the car is short

the red of the car *the short of the car

the redness of the car the shortness of the car

Other differences originate with spatial dimensions and the particular conditions of antonymy on which they are organized, as shown in (52). The fact that adjectives for the positive extension like long, high, old, but not their counterparts short, low, young, allow measure phrases is rooted in their SF and carries over under nominalization. (52) a. The length of the car was surprising. The car was ten feet in length. b. The shortness of the car was surprising. *The car was six feet in shortness. Cases like (51) and (52) show, by the way, that the relation between SF and C/I noted earlier is not a one-sided affair, where SF merely eliminates subtleties of C/I. There are characteristic cases − in fact lots of them − where SF on the basis of its proper elements and inherent principles provides systematic distinctions which C/I does not bother very much to go along with. This, no doubt, contributes to the enormous power that linguistic structures bring to the overall mental machinery. In general, though, nominalized adjectives exhibit the usual interpretive flexibility, comparable to that of nominalized verbs, but restricted by their inherent structure: the properties which adjectives primarily refer to give rise mainly to dispositions or states, but appropriate conditions can give rise to material objects or time dependent processes as options among the types of sortal interpretation. (53) a. The likelihood of his arrival was taken for granted. b. Likelihood cannot be bought. c. The changing likelihood is now visible on the screen. (54) a. b. c. d.

The reality of the proposal is obvious. Reality is more important than imagination. Under this illumination the reality is colorful indeed. The reality expands even faster than predicted by the theory.

While changing likelihood in (53c) is a proper metonym to be construed as (representation of the) changing likelihood, the quasi-metonym (matter of) the reality in (54c) and (54d) may perhaps count as an instance of ambiguity which relies on the SF supporting things being real as a fixed alternative to the regular state of being real. Different and in part less subtle interaction between derivation and metonymy shows up with agent nominalization as in (55), where different morphological realization corresponds largely to the native or non-native origin of the nominalized verb. (55) a. -er: b. -or: c. -ant:

observe direct participate

→ → →

observer director participant

believe possess attend

→ → →

believer possessor attendant

The most frequent version is to be accounted for by a suffix, -er (or -or, for that matter), which converts a verb with the appropriate c-selectional conditions, indicated here by the feature [+F′], into a noun which takes the subject as the designated, nominal refer-

61. Word-formation and metonymy

1119

ence. To that effect, the basic event reference provided by the λe of the verb gets neutralized by a dummy variable e′ without any further change in the verb’s SF. Hence agent nominalization by (56) differs from event nominalization by (41) above merely by its PF and the dummy variable e′, whereby the subject (instead of the event) achieves the referential capacity: (56) [-er]

[+N,−V]

λV [V e′] [+F′]

For the sake of illustration, consider the noun painter, which derives from the verb paint, an ambiguous verbalization of the noun paint with the meaning ‘to cover with paint’ or alternatively ‘to create a representation by paint’, such that the entry (57) represents two alternatives, where [P-PAINT] abbreviates something like [COVER-BY-PAINT] or alternatively [CREATE-BY-PAINT]. One of these options shows up in example (58a) with the simplified SF in (58b), the other in (59a), with the simplified SF in (59b). (57) [peint] [+V,−N,…] λy λx λe [[x [P-PAINT y]] e] (58) a. Eve painted a flower. b. de [e BEFORE u] : [EVE [[CREATE-BY-PAINT] FLOWER] e] (59) a. Fred painted the door. b. de [e BEFORE u] : [FRED [[ COVER-BY-PAINT] [DEF DOOR]] e] Now, painter derives by combining (56) with (57) to get (60), which on the one hand preserves the ambiguity between (58) and (59) while at the same time it opens up additional options for the event variable e′, as it does not support the tense-bound event reference e in (58) and (59), nor does it lead to nominal event reference, as it would in Eve’s painting of a flower or in Fred’s painting of the door. Different possibilities to specify and to interpret the variable e’ in (60) are foregrounded in cases like (61). (60) [peint-er] [+N,−V] λy λx [[x [P-PAINT y]] e′] (61) a. The painter must not to be disturbed, he is at work. actual actor b. The painter of this scene lived in the 19th century. author c. The painter of a picture must always be taken into account. principled role Whether and how these different values of e′ are merely a matter of interpretation in C/I, depends among other things on the analysis of sentences like (61c), which is usually considered as a standard phenomenon of logical form. In any case, neither variants like (61) nor ambiguities like (58)−(59) are instances of metonymy or conceptual shift. Both of them can be found to play the standard role to be expected for agents according to their regular position in the SF of the underlying verbs. (62) exemplifies the canonical options of dot-objects. (62) a. b. c. d.

The painter of these pictures is sitting over there. The painter of these pictures had enormous influence. The painter you just met has been heavily criticized. An auction will yield an enormous price for this painter.

physical object information object • information exchange value

1120

Semantics and pragmatics in word-formation I: General aspects

Further conceptual shifts arise on the basis of additional scenarios of interpretation. More importantly, though, similar interpretive variants go with an agent nominalization like violinist, which is based on the suffix -ist combined with the noun violin, where the agent position must be indirectly supplied by the verb play (as in the synonymous violin player): (63) a. b. c. d.

The violinist left the hall a moment ago. This violinist is the label’s most surprising success. The violinist who is famous for the Bach sonatas arrived by plane. No musician sells better than this young violinist.

In other words, dot-object variation is available whatever the origin of the agent role may be. This includes not only instances of the fairly regular pattern (64a) using musical instruments and the implicit verb play, but also a range of different nominalizations, most systematically the parallel derivation of a view and its representatives as in (64c), and a number of other suffixes with similar effects, as indicated in (65). (64) a. -ist: piano → pianist, trombone → trombonist, harp → harpist b. -ist: machine → machinist, reserve → reservist, special → specialist c. -ist: realism ~ realist, purism ~ purist, idealism ~ idealist, nihilism ~ nihilist (65) a. -ian: music → musician, library → librarian, custody → custodian b. -er: pot → potter, trumpet → trumpeter c. -eur: massage → masseur, pose → poseur An almost trivial observation shows this to hold for cases of personal reference in general, including systematic nominalizations like American, Bulgarian, Canadian, etc. together with idiosyncratic cases like Cypriot, emperor, but including also basic items like king, child, poet, girl, etc., which are subject to the same range of context determined, conceptual shift, whether or not word-formation is involved. This carries over to proper, coercion-supported metonymy as shown in (66), where X appears instead of (signature of) X, (printed name of) X, or (portrait of) X, which is not covered by the primary dot-object interpretation of X: (66) a. b. c. d.

You will find the painter in the lower left corner of the picture. The publisher is given on the cover of the volume. The French sprinters will show up in red and blue on this list. The musicians at the entrance are left from a former concert.

This observation is corroborated in an interesting way by the natural constraints that apply to non-personal as opposed to personal agent nominalizations as in (67): (67) a. The piano player was late / ??out of order. b. The record player is out of order / ??late. There are at least two ways in which the derivation of record player might be construed in contrast to piano player: either the head nouns of record player and piano player are

61. Word-formation and metonymy

1121

supposed to derive as agent nouns from a uniform verb play with subsequent coercion of the non-personal interpretation of player in cases like (67b) or play is assumed to have one reading with a personal and one with an impersonal agent in the argument position of the subject, with record player based on the latter. In any case, the non-personal reading of player, however it is derived, does not allow for the conceptual shift available for personal nouns, a completely natural consequence, once the semantic condition of the agent is fixed (suppressing initial ambiguity or coercion). It might be added that this characteristically holds for a fair range of instrumental agent nominals with often no animate ambiguity: (68) a. send → sender serve → server print → printer transmit → transmitter compute → computer b. regulate → regulator ventilate → ventilator project → projector calculate → calculator

hang → hanger adapt → adapter receive → receptor connect → connector

As usual in derivation, unpredictable, particular, lexically fixed specifications can be attached to the derived items, as, e.g., in computer, adapter, projector, etc., without interfering with the general conditions of conceptual interpretation that interact with the pattern of derivation. This interaction is highly asymmetrical for systematic reasons, though: the properties of linguistic expressions, notably their SF (including possibly incorporated effects of coercion) provide the boundaries for their interpretation in C/I, which complies with the conditions sketched in (15) and exploits the possibilities available for dot-object semantics, such that the computer might be interpreted as a physical object, as its technical design, or as the constructional principle of the device, and even as the socio-cultural phenomenon with its various ramifications and political consequences. Idiosyncratic, lexicalized features may further be added for a wide range of cases like home trainer, land rover, screwdriver, etc., where compound formation comes in as a factor which participates in the transposed interpretation, raising problems which will be taken up briefly in the next section.

6. Metonymy and compound formation Compound formation consists in the combination of two lexical items into a new one with usually one of them as the head which determines the categorization and thereby the major syntactic properties of the whole. The wide variety of syntactic and semantic phenomena of this domain has been surveyed, among others, by Meier (1993), Jackendoff (2009), Olsen (2012); see also article 20 on composition. The focus of the present discussion will be on compounds with a nominal head as in (69), because these prevail with respect to the role of metonymy and related phenomena, although various other types of compounds must of course be recognized as well, as loosely indicated be the examples in (70). (69) a. b. c. d.

office hour, street corner, coastguard, screwdriver mainstream, bluebook, hard cover, software springboard, turntable, think-tank, playground up take, after thought, bypass, through-train

(noun noun) (adjective noun) (verb noun) (preposition noun)

1122

Semantics and pragmatics in word-formation I: General aspects

(70) a. dark green, grass green, blue-green, soft-green b. overact, underestimate, downplay

(adjective adjective) (preposition verb)

The semantic interpretation of compounds covers a wide variety of possibilities for relating the two component parts, even if one restricts the formal aspect further to noun-noun compounds with the second noun as head. On the background of the rich literature about these problems, Jackendoff (2009) recognizes two basic N-N compound schemata indicated in (71), where X and Y represent the SF of N1 and N2, respectively. The argument schema (71a) deals with the large group of cases, where the head N2 takes the complement N1 as argument according to the head’s argument structure, as in tree pruning, piano-player, book cover, or whistle-blower, while the modifier schema (71b) deals with the cases whose N1 modifies (and thereby specifies) the head N2, without filling an argument position of it, as in apple-tree, piano-music, church tower or phrase structure and a wide variety of other types. (71) a. Argument schema: [N1 N2] [Y [X]] b. Modifier schema: [N1 N2] [F [Y X]] where F is a variable over functions relating X and Y. Jackendoff considers fourteen basic types of functions F, such as PLACE, PURPOSE, MATERIAL, CONTENT, and combinations of them, all of which eventually turn N1 into the specification of a particular type or possibility of the head N2. Now the argument relation can reasonably be construed as a special case of these functions: if one element of the compound is appropriately construed as a regular argument of the other, it provides by this very relation a modification for it, the impact of which is determined by the SF of the head. In a compound like portrait-painting, for example, the first part provides the argument of the second, but it thereby also characterizes − like a regular modifier − a subclass of paintings; in oil-painting, however, the first element does not refer to the object, but to the material of painting, and in wall-painting it could indicate the place or technique of painting, rather than its argument. And under appropriate conditions, the same compound allows different types of combination for its components, such as test translation with at least three possibilities for combining the constituents semantically, as indicated in (72). (72) a. test translation ~ translation of the test b. test translation ~ translation for test c. test translation ~ translation as test

(head − argument) (head − modifier, purpose) (head − modifier, purpose)

Hence the two schemata in (71) can be collapsed into one overall condition, with the head providing the frame for modification, within which the complement modifies the head by serving as its argument or as one of various types of modifiers. This unification corresponds neatly to the observation that proper nominal arguments of nouns − as in John’s pruning of the tree or the painter of the portrait − are syntactically realized usually like attributes and semantically as conditions on the arguments in question. Without going into the non-trivial details of this aspect, one would get (73b) on the basis of the entry for painter sketched in (60) above, repeated here as (73a), with C-PAINT for creation by painting.

61. Word-formation and metonymy

1123

(73) a. [painter] [+N,−V] λy λx [[x [c-paint y]] e′] b. [portrait painter] [+N,−V] λx [[x [c-paint y] o [portrait y]] Turning next to the way in which compound formation interacts with transferred interpretation, viz. conceptual shift, coercion, and especially metonymy, it must first be noted, that the compound members are mutual determinants with respect to selection of ambiguity and conceptual variants. Thus, in cases like (74), the modifying N1 specifies a particular instance of the head N2 , which by this very specification selects the actual interpretation of N1 in case of ambiguity and thereby provides the general concept from which N1 selects a special instance. It is the modifying N1, which selects this general concept in case of ambiguity, such that bank in (74) is a monetary establishment with an entrance, a lobby, a manager, etc., and not a river bank. (74) a. bank entrance, bank location, bank lobby: b. bank manager, bank customer, bank raid: c. bank rate, bank book, bank account:

bank = building bank = institution bank = principle of organization

There is, however, a second selection, by which the head N2 may determine the conceptual selection in the dot-structure of N1, such that bank refers to a building in (74a), to an institution in (74b), and to the financial regulations and principles in (74c). The modifier-head relation amounts to a twofold asymmetry, so to speak: the sortal specification of N1 is selected by N2 which at the same time provides the general conditions for it. In pertinent configurations, these two conditions may create converse dependencies, with the sortal specification of the head N2 being selected by the modifier N1, which at the same time specifies its particular properties, as in (75), where ring, foundation, or frame are ontologically depending on the modifier, which also specifies their particular properties. (Due to the lacking lexical specificity of the components, the construal of cases like concrete foundation is ambiguous, but that does not prevent the dependencies obtaining in the interpretation in question.) (75) a. gold ring : physical object b. concrete foundation : physical object c. picture frame : physical object

espionage ring : social institution research foundation : social institution mind frame : mental structure

In general, then, the modifier N1 characterizes properties of the head N2, whether or not the sortal conditions of N1 are determined by N2 or vice versa. Turning now to the particular interaction of compounding and metonymy proper, it is useful to set aside for the moment compounds like the ham sandwich of example (18a), because there metonymy concerns the compound as a whole, which will be taken up shortly, and to look first at the characteristic head-modifier asymmetry internal to compounds. (76a) illustrates the point in question. Although the compound Mahler finale is well-formed, it is in need of conceptual adjustment, just as much as Wagner or Chagall must be adjusted in cases like (24). One of the possible amendments the conceptual background might provide is the standard coercion indicated in (76b), the more explicit version of which would be (76c) or even (76d). On this basis, Mahler in (76a) is meto-

1124

Semantics and pragmatics in word-formation I: General aspects

nymic for (symphony by) Mahler in exactly the same sense in which Wagner is metonymic for (music by) Wagner in (24). (76) a. b. c. d.

The The The The

[Mahler finale] was exciting. [[Mahler (symphony)] finale] was exciting. finale of the [Mahler symphony] was exciting. finale of the symphony by Mahler was exciting.

← ← ←

Now, in contrast to the interpretation of Mahler as a metonymic modifier, (76a) cannot normally be construed with finale as abbreviating the compound (symphony) finale, as spelled out in (77), ignoring possible delicacies of different stress patterns: (77) a. The [Mahler finale] was exciting. b. ??The [Mahler [(symphony) finale]] was exciting. c. ??The [symphony finale] by Mahler was exciting. There is no conceptual obstacle that would prevent symphony finale from being construed as the identifying notion of final movements, from which Mahler would have to select those by Mahler. What makes (77) implausible, is the conflict between the headmodifier asymmetry of compounds, and the omission of the characteristic part. According to (77), the qualifying component would be suppressed in favor of the backgroundproviding head. The general conclusion from these considerations is this: (78) Metonymy can apply to compounds as a whole or to their modifier. This generalization has various consequences. One is the contrast between (79a) and (79b), the latter being ambiguous between the metonymic reduction (79b) and the nonmetonymic (79c), which would refer to the Russells’ separate theatre-compartment. (79) a. the Russell shelf ← b. the Russell box ← c. the Russell box

the Russell (book) shelf the Russell (book) box

Likewise, a Leonardo frame can be metonymic for a [Leonardo (portrait)] frame, i.e. a frame for a portrait of Leonardo, but hardly for a Leonardo [(portrait) frame], as a portrait of Leonardo can contextually be motivated, but at best marginally an unspecified portrait frame. Similar cases are car union for [car (owner)] union, visitor rate for [visitor (admission)] rate, etc. Possible metonymic reduction of this sort carries over to recursive compound formation in general, as indicated in (80). (80) a. Russell box repair ← Russell (book) box repair b. Mahler finale performance ← Mahler (symphony) finale performance c. visitor rate increase ← visitor (registration) rate increase The interesting point is that − with appropriate background − metonymic interpretation of the compound as a whole is possible, in the sense already mentioned with regard to the ham sandwich of example (17). For the sake of illustration, one might imagine the

61. Word-formation and metonymy

1125

janitor of a theatre presenting a key and uttering one of the sentences in (81), where (81b) and (81c) are different metonyms of the explicit case (81a): (81) a. This is the Russell box key. b. This is the Russell key. c. This is the Russell box. While (81b) is the type already discussed, with Russell (box) determining the head key, (81c) is remarkably different, as here the modified head key is dropped, which turns it into an exocentric compound. Cases of this sort are natural under various conditions, as indicated in (82) and (83) for academic and clinical contexts, respectively. (82) a. The bachelor exam candidates are waiting upstairs. b. The bachelor candidates are waiting upstairs. c. The bachelor exams are waiting upstairs. (83) a. This is the knee operation patient. b. This is the knee patient. c. This is the knee operation. More importantly, the (context dependent) reduction in the (c)-cases is a direct link between proper metonymy and exocentric compounds, especially the so-called bahuvrihi-constructions. An exocentric compound is a construction that lacks a head expressing the concept which the compound’s potential referent belongs to, such that the concept in question must be supplied otherwise. Thus bachelor candidate in (82b) is endocentric, as a bachelor candidate is a candidate, while bachelor exam in (82c) is exocentric, as the intended interpretation relates to candidates which the suppressed head of the compound provides. Now, determining reference by means of a missing concept coerced by the context has earlier been characterized as the very principle of metonymy. As a consequence, exocentric compounds are a major type of metonyms, not only in ad hoc settings (such as in (81) to (83)), but also as lexicalized items with often highly idiosyncratic, fixed interpretations, as a few examples in (84) show. (84) a. green beret, blue jacket, red shirt, blue stocking, brass hat, red cap b. red skin, flatfoot, red head, long nose c. pickpocket, fly over, scarecrow, breakfast Lexicalized metonyms are frequently adjective-noun compounds with the bearer of the specified attributes providing the head, as the examples (84a) and (84b) show; other types are based on verb complement combination where the omitted agent of the verb supplies the head, as in cases like (84c). Similar to coercion in general, the supplied head determines the sortal properties of the metonym, especially the dot-object conditions wherever they are to be identified, as for example in The Blue Helmets were waiting in close vicinity and clearly responsible for the subsequent events. Lexical idiosyncrasy may suspend such regularities to different degrees − a fly over (for bridging road junction) has no tacit agent of fly −, but to the extent to which features of word-formation are preserved, the regularities are still in place.

1126

Semantics and pragmatics in word-formation I: General aspects

7. Conclusions To sum up, some general remarks with respect to metonymy and its relation to wordformation seem to be indicated. (i) Metonymy is a specific phenomenon of the human language capacity, belonging to the range of conditions and possibilities by which the relation of linguistic expressions to extra-linguistic mental structures comes about. More specifically, metonymy participates in the mapping between semantic form SF, i.e. the linguistically determined structure of meaning, based on lexical knowledge and the rules and principles of grammar, and the various domains of conceptual and intentional structure C/I, by which different aspects of mental and physical experience are organized. Although linguistic expressions must systematically refer to and in some respects also determine structures of different conceptual domains, SF and C/I cannot in general be isomorphic, as they are subject to different laws and principles. (ii) Assumptions about the correspondence between SF and C/I are still controversial, but a core of literal meaning providing the direct interpretation of basic elements, relations, and distinctions seems indispensable, although its range and content is not obvious. Of corresponding importance is the role and nature of transferred, non-literal interpretation, which is crucially determined by intra- and extra-linguistic context conditions and the way in which these provide the SF-C/I-correspondence. (iii) A central issue in this respect concerns the way in which language deals with sortal or ontological conditions and distinctions and their consequences to be recognized in the different systems of C/I, such as, e.g., a physical object, a complex of information, a social institution, or a group of people, covered (under different conditions) by the meaning of the same lexical item newspaper, or a book, its author, and his portrait covered by the name of the author. Taking the core of literal meaning as basis (no small commitment) and leaving aside other conditions and distinctions of the SF-C/I-correspondence, three types of transferred or non-literal interpretation can be distinguished, simplified as: a) Conceptual shift refers to alternative aspects of concepts with inherently mixed sortal conditions, such as human beings, social institutions, or newspaper. b) Metonymy replaces reference to a concept by a salient characteristic of it, such as the name Fukushima referring to the event for which the place is famous. c) Metaphor uses one concept to indicate another concept which is analogous to it in relevant respects or believed to be otherwise related to it, as John is bitter. Technically, conceptual shift is essentially based in SF on so-called dot-objects, i.e. concepts with physical, mental, social and other sortal aspects. Metonymy involves coercion of semantic conditions that supply the entity whose special characteristic is foregrounded. Metaphor goes beyond the principles pursued here in more detail, and it is sometimes claimed to include metonymy as a special case. Anyhow, (a) to (c) indicate clearly different ways to adjust the relation between linguistic structure and extra-linguistic mental organization, although the boundaries between them are not always clear-cut. (iv) Metonymy as well as conceptual shift rely on regular linguistic structure and its literal interpretation, extending it, if necessary. To this effect, conceptual shift unfolds,

61. Word-formation and metonymy

1127

so to speak, inherent alternatives of structures in SF, while metonymy closes semantic gaps by means of contextually triggered default options. In simplified terms − conceptual shift exploits inherent sortal flexibility, while metonymy adds contextually obvious bridges. (v) Both strategies apply to basic lexical items as well as to results of derivation and compounding, with combined effects, if indicated. A characteristic case in point is the role of exocentric compounds like red shirts or green berets, where the coerced semantic head bearer (of) provides the basis for the usual conceptual shift, unfolding the alternatives inherent in the supplied head person with .... Orthogonal to these productive possibilities, which are based on context dependent, although predictable interpretive mechanisms, idiosyncratic constraints and their lexical registration play a crucial role in metonymy, as in all areas of word-formation. Thus products of regular word-formation are often turned into fixed lexical items because of their idiosyncratic interpretation. Arbitrary selected cases like scarecrow, fly-past, or supercargo illustrate the point. Even completely transparent compounds like newspaper or typewriter must be recorded as separate lexical items, whose literal interpretation cannot be derived from their recognizable, regular structure. As a matter of fact, metonymy heavily contributes to the position of lexical information between grammar and idiosyncratic list.

8. References Barsalou, Lawrence W. 1999 Perceptual symbol systems. Behavioral and Brain Sciences 22: 577−660. Bierwisch, Manfred 1983 Semantische und konzeptuelle Repräsentation lexikalischer Einheiten. In: Rudolf Růžička and Wolfgang Motsch (eds.), Untersuchungen zur Semantik, 61−99. Berlin: Akademie Verlag. Bierwisch, Manfred 2007 Semantic form as interface. In: Andreas Späth (ed.), Interfaces and Interface Conditions, 1−32. Berlin/New York: Mouton de Gruyter. Bierwisch, Manfred 2011 Semantic features and primes. In: Claudia Maienborn, Klaus von Heusinger and Paul Portner (eds.), Semantics. An International Handbook of Natural Language Meaning. Vol. 1, 322−357. Berlin/Boston: De Gruyter Mouton. Bloom, Paul, Mary A. Peterson, Lynn Nadel and Merrill F. Garrett (eds.) 1996 Language and Space. Cambridge, MA: MIT Press. Dölling, Johannes forthc. Sortale Variation der Bedeutung bei ung-Nominalisierungen. In: Christian Fortmann, Wilhelm Geuder, Anja Lübbe and Irene Rapp (eds.), Situationsargumente im Nominalbereich. Berlin/Boston: De Gruyter Mouton. Jackendoff, Ray 2002 Foundations of Language. Oxford: Oxford University Press. Jackendoff, Ray 2009 Compounding in the parallel architecture and conceptual semantics. In: Rochelle Lieber and Pavol Stekauer (eds.), The Oxford Handbook of Compounding, 105−128. Oxford: Oxford University Press.

1128

Semantics and pragmatics in word-formation I: General aspects

Jackendoff, Ray 2010 Meaning and the Lexicon. Oxford: Oxford University Press. Meier, Ralf 1993 Compound Comprehension in Isolation and in Context. Tübingen: Niemeyer. Olsen, Susan 2012 Semantics of compounds. In: Claudia Maienborn, Klaus von Heusinger and Paul Portner (eds.), Semantics. An International Handbook of Natural Language Meaning. Vol. 3, 2120−2150. Berlin/Boston: De Gruyter Mouton. Nunberg, Geoffrey 1979 The non-uniqueness of semantic solutions: Polysemy. Linguistic and Philosophy 3: 143− 184. Pustejovsky, James 1995 The Generative Lexicon. Cambridge, MA: MIT Press. Searle, John R. 1979 Expression and Meaning. Cambridge: Cambridge University Press.

Manfred Bierwisch, Berlin (Germany)

62. The pragmatics of word-formation 1. 2. 3. 4. 5.

Introduction Demarcation of the area Pragmatic meaning History of research References

Abstract This contribution is an overview article intended to provide a definition, demarcation and description of the pragmatic perspective in word-formation as originated and developed in the relevant literature, and to present a full-fledged theoretical model − morphopragmatics − precisely covering the issue, and contrast its account of pragmatic meaning with that of other competing approaches.

1. Introduction In the literature on word-formation, pragmatic facts start to be accounted for only gradually and sparsely, first as a mere observation and description of extra features/meanings non-reducible to denotational semantics and only later within theories and theoretical models centered on morphology as a specific field interfacing with pragmatics. The present article will retrace the gradual emergence of a pragmatic consciousness relative

62. The pragmatics of word-formation

1129

to some morphological facts, which led to the identification and demarcation of the subfield of research indicated in the heading. In dealing with aspects of pragmatics and word-formation, we are confronted with major theoretical issues that pertain independently to each of the two disciplines. In word-formation, for example, a major issue is the characterization of meaning entailed in various input/output morphological operations, on which there is very little consensus. Meaning, in fact, has received conflicting formulations and divergent interpretations, also in the very phenomena that are at issue in the present article. On the other hand, the area of pragmatics is wide-ranging and its components are often controversially described and delimited. In the sub-area of its interface with morphology, for example, issues concerning pragmatic functions and effects of relevant word-formation mechanisms, prototypically represented by diminutives, will have to undergo discussion. The central point of the overall investigation carried out in this article is the characterization of pragmatic vs. semantic meaning in word-formation. This contribution, in fact, has two priorities, one is to provide a definition and description of the pragmatic perspective in word-formation, as it originated and was developed in the relevant literature, the other is to present a full-fledged theoretical model − morphopragmatics − precisely covering the issue, and to contrast its account of pragmatic meaning with that of other competing approaches. The article is structured as follows: In section 2, the sub-area under analysis, where pragmatics and word-formation meet, will be delimited precisely against competing areas of research. Section 3 deals with a preliminary delineation of pragmatic meanings and a sample list of the word-formation means that predictably can express it. Contextualized examples of their possible uses follow. Section 4 will be devoted to an outline of the history of research, including some pre-history, when studies on morphology identified pragmatic meanings ante litteram, at a time when pragmatics was not yet a stable discipline. Then, more recent literature will be reviewed and discussed, mainly with reference to some aspects inherent in the semantics/pragmatics antinomy and pertinent to the controversial interpretation of word-formation meaning. Subsection 4.3 will be reserved for the presentation and discussion of the model that represents the position being advocated here, contrasted with another important model, opposed in its premises and approach.

2. Demarcation of the area In order to approach the facts of pragmatics in word-formation, it is necessary to delineate and delimit a pertinent area of research, with its range of phenomena and distinguish it from neighbour areas. In terms of meanings involved in morphological operations, I would contrast it with the following established sub-areas (Dressler and Merlini Barbaresi 1994: 55): a) morphosemantics; b) lexical semantics of morphology; and also with: c) morphopragmatics; d) lexical pragmatics of morphology; e) sociopragmatics relative to morphological rules; f) general discourse pragmatics. A careful demarcation of the field will allow us to be consistent and clear in attributing phenomena to pragmatics. a) Morphosemantics. This deals with the semantic meaning of morphological rules, that is the denotational and connotational meaning change between the input and the output

1130

Semantics and pragmatics in word-formation I: General aspects

of a morphological rule. For the majority of cases, the wealth of studies in evaluative morphology are focused on the formal morphological properties of evaluative suffixes. When they explore the complexity of the suffix meaning, they give a semantic representation of it (among the most important of these are, e.g., Scalise 1984, Wierzbicka 1984, Bauer 1997, and more recently Grandi 2002). Rainer’s study (1990) on Italian diminutives is specifically focused on their morphosemantics, as is Schwarze’s (2001). If we accept the notion that describing the semantic meaning of a word-formation rule means disposing of all the pragmatic variables of the speech situations, in terms of participant interrelations, contextual place and time, speech acts, etc., then we should obtain a delineation of the field of morphosemantics straightforward enough to prevent any interference with a pragmatic account of word-formation. Yet, there are various blurry cases, having to do, for example, with contrasting interpretations of connotative meanings (see Klimaszewska 1983 and Volek 1987). Sometimes there is a very slim margin of difference between semantic connotation and pragmatic meaning. For example, an English word like doggy is connoted as a child’s word and, as such, picks up various fixed, positive features of a child-centered speech situation, including emotional overtones. When moved to adult use, though, the diminutive formation may index extra pragmatic meanings, e.g., tenderness and playfulness towards one’s pet or irony when, for example, doggy is used to refer to a giant dog. The connotation may also involve socio-cultural values and personal ideologies, e.g., a clipped word like commo, American for communist, is negatively connoted and its use indexes an opponent in political ideology, as, e.g., is Trudeau a commo? revisionist nonsense: a liberal, yes. a commo? No (from the web). Stable connotative features (developed over time and use) attached to words must be distinguished from pragmatic meanings/effects contextually and dynamically created by a morphological operation in the course of a speech event. b) Lexical semantics of morphology. This refers to the denotational and connotational semantics of a morphologically complex word. More precisely, it explores the meanings of morphemes and how they combine to form the meanings of complex words. See, for example, Lieber (2004), who specifically approaches the lexical semantics of derivation and compounding. Questions of affix polysemy, of synonymous and antonymous affixes, as in Lehrer (1987, 1995, 1998, 2003), are borderline between lexical semantics and morphosemantics (see also Bauer, Lieber and Plag 2013). c) Morphopragmatics. This refers to the idiosyncratic meaning of individual morphologically complex words. The majority of the early studies aiming at a pragmatic account of word-formation belong here. The pragmatic meaning of a morphological rule is often conflated with that of the individual lexical item. For example, in a word like bunny, a diminutive of Scottish dialectal bun, pet name for rabbit, the pragmatic meanings/effects (tenderness, playfulness) belong to the word itself and not to the word-formation operation. The field must be restricted and specified with respect to current general theories of lexical pragmatics within cognitive linguistics, which investigates the processes by which linguistically-specified (literal) word meanings are modified in use through the pragmatic operations of narrowing, loosening and metaphorical extension. See, for example, Blutner (1998) and Wilson and Carston (2007). Of course, within this type of study, the focus is on the word and not on word-formation mechanisms.

62. The pragmatics of word-formation

1131

d) Lexical pragmatics of morphology. This specifically refers to a theoretical model (Dressler and Merlini Barbaresi 1987, 1994; Dressler and Kiefer 1990) that deals with the general pragmatic meanings of morphological rules. With respect to the heading of this work, it covers a wider range of phenomena as it includes both word-formation and inflection (see section 4.3). e) Sociopragmatics relative to morphological rules. This has to do with word-formation rules and their interpreters, with contextual and indexical factors and variables of their use (for an exhaustive account of context see Auer 1996). Within our field, examples may be social preferences in the use of specific word-formation mechanisms, as regulated by genre, sex, age, diastratic and diatopic varieties, channel of communication, formal vs. informal situations. The area of studies may overlap with that of the pragmatics of morphology in various ways, e.g., when dealing with the use of polite or strategic diminutives in discourse (It. solo un minutino ‘just a sec’; Mex. Sp. ahorita ‘immediately’, or in mitigated requests, like It. Mi fai un piacerino? ‘can you do me a favour?’), where it may be a factor regulating interactant rapport. A notion of adaptability in terms of socialinteractive sense and of language adaptive to the process of communication is especially relevant. Various restrictions on the pragmatic use of diminutives, for example, are regulated by such concepts. Pragmatics as a theory of linguistic adaptability is fully explored in Verschueren (1999). A study centered on the sociopragmatics of diminutives is De Marco (1998). f) General discourse pragmatics. This area becomes relevant because the contribution of word-formation rules is often overshadowed in discourse by more general pragmatic operations. That is, in discourse, pragmatic meaning/effects may be contributed by various co-occurring operations at different levels, and some levels, e.g., syntax or lexis, may be more powerful in obtaining a pragmatic interpretation. For example, it may be difficult to discern the contribution of the word-formation mechanism to the sarcasm of a request, as in It. (angry tone) Ti dispiacerebbe darmi un aiutino, invece di startene spaparanzato sul divano? ‘Would you mind giving me a help-DIM, instead of staying sprawled out on the sofa?’. The syntax (modalizing conditional), the extra polite formula, the lexis of ‘stay sprawled out’ all contribute, perhaps more strongly than the diminutive, to the pragmatic effect of sarcasm. The overwhelming competition of general syntactic and lexical pragmatics has been the reason for the state of neglect of individual morphological items, even after pragmatics had already become established as a discipline.

3. Pragmatic meaning We start by raising two preliminary questions: a) what is meant by pragmatic meaning and b) what are the word-formation means capable of expressing it?

3.1. Definition A definition of pragmatic vs. semantic meaning needs discussion. This is where approaches and especially theoretical premises diverge in the relevant literature. By using

1132

Semantics and pragmatics in word-formation I: General aspects

the term “pragmatic meaning”, I intend to refer more precisely to that part of wordformation meaning that can receive a characterization via a pragmatic perspective. I interpret this meaning as obtained in the course of a speech event by the application of the word-formation rule. Using Fradin’s (2003) (diminutive) meaning trichotomy, it is the type of meaning that belongs to the sphere of the “interlocutor’s pole” (vs. referent’s and locutor’s). Dal’s (1997) notion of the diminutive (in -ette) as a marqueur d’appropriation is also pertinent. It indicates a type of proximity between locutor and object established by the application of the rule.

3.1.1. Emotiveness In the earlier literature, pragmatic meanings were dealt with under the label of emotiveness, raised to the status of a unifying explanatory principle. Volek (1987, 1990), for example, is an authoritative representative of this approach. She identifies emotiveness as part of the structural meaning of Russian suffixes (diminutives and augmentatives), but she attributes both the emotive and the notional components to the stable semantic configuration of the suffixes. A morphosemantic approach of this type, however, can hardly accommodate the paradoxical fact that the same diminutive may express totally contradictory emotions (tenderness and contempt) and perform opposite functions (mitigation and intensification). The notion that all pragmatic meanings would derive from the semantics of the suffixes, i.e. from the single abstract meanings ‘small’ and ‘big’, through contextualised inferences, is at the basis of various other important studies (e.g., Ettinger 1974a, 1974b; Wierzbicka 1984). Even Schneider (2003), in spite of his pragmatic program, actually conceives of “attitudinal meaning”, identified as affection and emotion, as a feature of the diminutive suffix’s semantics (nice/sweet+small). Dressler and Merlini Barbaresi (1994) do not negate the importance of emotions (see Kerbrat-Orecchioni 2000, Weigand 2004, and Bazzanella 2004 for a more recent account of emotions) but they attribute to them the value of regulative rather than constitutive factors. They conceive of pragmatic meaning as separate from morphosemantic meaning and claim that a morphological rule can contribute pragmatic meaning if it contains a pragmatic variable which is necessary for the description of its meaning. Their approach is in line with their more general theoretical claim that pragmatics is a superordinate of semantics (more in section 4.3).

3.1.2. Prototypes and primitives Jurafsky (1996), in his cognitive model, offers a different description. In his radial polysemous category, he assumes that the semantic meaning of ‘child’ is the central prototype from which all the meanings/functions of diminutives are derived, both semantic and pragmatic, in a continuum reconciling synchrony and diachrony (more in section 4.3). Close to a conception of semantic primitives is also Grandi’s (2002) configuration of the meaning of evaluatives (diminutives and augmentatives) as represented by modulations along the scales of the fundamental values of quantity (small and big) and of quality (good and bad), from which all meanings/functions are derived. In his approach, as

62. The pragmatics of word-formation

1133

well, there are no actual borders between synchrony and diachrony. Dressler and Merlini Barbaresi’s extended work (1987, 1989, 1994) is where the various meanings and functions of evaluatives are modelled in detail and exemplified in discourse. Successively, the authors’ basic claim of the priority and autonomy of pragmatic over semantic meaning has been corroborated by results in studies on first-language acquisition (see Savickienė and Dressler 2007). There are positive indications from various languages that children learn some pragmatic meanings (e.g., those involved in cajoling) earlier than semantic meaning. According to Dressler and Merlini Barbaresi (1994), emotionality in child-centered speech situations, but also the ludic character of playfulness among intimates, familiarity and informality in general, sympathy and empathy and also understatement, euphemism, false modesty, irony and sarcasm, are all factors that favour a pragmatic use of diminutives and other morphological rules (there will be more discussion on this point in the pertinent literature surveyed below).

3.2. Morphological rules and mechanisms By explicitly referring to diminutives as prototypical representatives of evaluatives, I have in fact anticipated the choice of the word-formation phenomena that I consider as eligible for a pragmatic investigation. Diminutive formations are almost universally represented and remain the morphological mechanism that best exemplifies the variety of relevant meanings. But other morphological means are also exploited (see Schneider 2013). In my listing, I make random choices from various languages, with a practical preference for Italian, my morphologically rich native language, and, when possible, English, for more general understanding. First of all, in order to delimit the range of phenomena, we have to make a decision on whether to include only grammatical word-formation rules or be more comprehensive and also include extra-grammatical word-formation devices (cf. Dressler and Merlini Barbaresi 1994; Doleschal and Thornton 2000). A safer choice would be to refer to stable patterns within grammar, which are morphotactically more transparent and analysable, but extra-grammatical phenomena also deserve mention. Pragmatic meanings are especially conveyed during informal interactional discourse (in-group, homiletic or intimate), in which participants’ attitudes, emotions and beliefs are foregrounded. Such a type of discourse also creates a favourable locus for the expressive/pragmatic use of slangy formations and other extra-grammatical devices (Merlini Barbaresi 2001). Among the mechanisms that Zwicky and Pullum (1987) assign to what they call “expressive morphology” vs. “plain morphology”, the majority are of the extra-grammatical type. Some will be included in the following sample list: a) evaluative/alterative affixes (diminutives, augmentatives, pejoratives) (examples from It(alian), G(erman, E(nglish), Sp(anish), Fr(ench), ante-suffixal interfixes included, as in various Romance languages, cf. article 31 on interfixes in Romance, (e.g., It. -ol- in ludic top-ol-one vs. serious top-one ‘sewer rat-AUG’) b) reduplicatives (e.g., E. Lizzy-wizzy, teensy-weensy, It. (occhi) neri neri ‘very dark (eyes)’, Fr. joujou ‘toy’, Zizou for Zidane, sousoupe; see also Yiddish shm-reduplica-

1134

Semantics and pragmatics in word-formation I: General aspects

tives) as, for example, the ironical/derisive He’s just a baby! Baby-shmaby. He’s already 5 years old! c) clipped forms, blends or portmanteaus (e.g., It. bibe(ron); E. digiteria ← digital cafeteria, Fr. intell-o ← intellectuel, Sp. telebobela ‘silly soap opera’ ← telenovela ‘soap opera’ and boba ‘silly’) d) patterns in discourse involving word-formation operations in speech acts (e.g., diminutive in attenuated requests and mitigated orders; It. intensifier -issimo in rebuttals to assertions or questions) e) the use of some formatives, prefixoids and suffixoids derived from negatively connoted words (e.g., It. -poli < Tangent-o-poli in vall-ett-o-poli ‘irregular recruitment of TV starlets’; E. -gate < Watergate in Enron-gate). Analytic forms with reduced, unstressed little (or Fr. petit, ti in Québec) may express similar meanings, as sarcasm in E. I can’t stand your little tricks or tenderness in Fr. tiJean. Also metaphorical compounds (e.g., E. giant-killer, baby-dolphin, baby-trees, picobrain, uber-brain, It. literary pietre-bambine, i.e. pietr-uzze ‘stone-DIM-PL’ (in D’Arrigo’s novel “Horcynus Orca”), pargoletta mano ‘child-DIM hand’, metaphor for ‘little hand’) (in Carducci’s poem “Pianto Antico”) are often pragmatically exploited, e.g., tenderness in baby-dolphin, pargoletta mano, exaggeration and irony in pico-brain, uberbrain. I have not included here particles (e.g., Japanese), even if cliticized or incorporated, because they are actually external to word-formation mechanisms, as they do not contribute to the creation of the word. For the same reason, we also avoid expletive insertions (e.g., E. every-bloody-body), in spite of their being indices of attitudinal meanings. A few contextualized examples will suffice to show the effects of such processes in discourse (from the web and other research books): 1. Playful irony: It. È anche un tant-in-ello schifoso, è un po’ schifos-etto. ‘It is also a so much-DIM1-DIM2 disgusting, a bit disgusting-DIM’ (‘It is rather revolting, pretty disgusting’). 2. Emotion, tenderness: Viennese G. (to a child) Nein, rühr das Wass-erl nicht an ‘no, don’t touch the water-DIM’. 3. Derogatory irony: E. He’s got a wife and a couple of wif-ie-s ‘girl friends’. 4. Euphemism: G. Er hat ein Gläs-chen über den Durst getrunken ‘He has just drunk one little glass too many’ (‘he is drunk’). 5. False modesty: It. Avrei anch’io una mia teori-etta ‘I’d have a little theory of my own’. 6. Playful irony: E. We’ve had Watergate, Iran-gate and Zipper-gate. Now our football has … Lancaster Gate. 7. Emotion, affection: It. Non fate gli sciocch-er-elli ‘Don’t be silly-Interfix-DIM’ (and German and English equivalents: Seid’s keine solchen Dumm-erl-n and Don’t be a silly-billy). 8. Empathy: It. Ah, è già ora del tuo whisk-ino ‘Oh, it’s already time for your whiskyDIM’ (‘your beloved whisky’). 9. Emotion, anger: E. Do me a teensy weensy little favour! Get out. 10. Hedged request: Sp. Espera un minut-ito solamente ‘Just wait a minute-DIM’ (‘Can you wait a little, please’). Cf. Fr. Voulez vous attendre pour un petit moment? 11. Pleading: Sp. Deme un pedac-ito de pan! ‘Give me a little piece of bread’.

62. The pragmatics of word-formation

1135

12. Contradicting rebuttal: It. A: Ma, non è la stessa! B: la stess-issima! A: But it’s not the same! B: ‘the same-ELATIVE’ (‘The same, the very same!’).

4. History of research 4.1. Pre-history We will begin our survey with a cursory mention of the scholars in modern times that first focused on morphological phenomena (mainly diminutives) and identified pragmatic meanings, at a time when pragmatics had not yet been established as a discipline. Two important precursors to a pragmatic approach in the study of diminutives in Romance languages were Leo Spitzer and Amado Alonso (cf. article 88 on the semantics and pragmatics of Romance evaluative suffixes). To Spitzer (1921: 201−202) we owe various important observations on the ludic, emotional character of diminutives and, more importantly, the fundamental conception of “sentence diminutive”, i.e. a diminutivized word which extends its meaning scope to the entire sentence. Such a meaning is clearly nonsemantic (cf. Dressler and Merlini Barbaresi’s 1994 concept of “landing site”). Alonso (1935/51) is a pioneer in the pragmatic analysis of diminutives; he has produced the richest and most pragmatically-oriented study. He downgrades the denotative meaning of smallness, in favour of emotional values whose meanings and effects depend on the context, the participants’ attitudes and the type of speech act (again ante litteram). Following Spitzer (1933), he also characterizes diminutives as sentence diminutives. He left a rich inheritance, especially for his accurate survey of emotions, which was the source for various “emotionalists”, as, for example, Gaarder (1966) on Mexican diminutives and augmentatives and later Volek (1987), and many others. Sieberer (1950) also claims that the essence of diminutive meaning is not denotational diminution but emotionality. Hasselrot’s (1957) study deals, on the other hand, with the formal properties of diminutives in Romance languages. In more recent times, Ettinger (1974a, b) has produced very ample and accurate accounts of the formal properties and denotational semantics of Romance and German diminutives and augmentatives. He has been a precious source for later studies on the subject. But his investigation does not take pragmatic aspects into consideration.

4.2. More recent studies Grabiaś (1981), in his study of Polish diminutives and augmentatives, bases his representation on expressiveness, but his account of pragmatic aspects is too general, in spite of important intuitions about speaker’s evaluation and emotional meanings. Klimaszewska (1983), in her contrastive study on Dutch, German and Polish diminutives, consciously adopts a pragmatic perspective (1983: 2) but her way of elaborating on pragmatic meanings leads her to co-identify them with expressive connotations (1983: 6, 27, 30). Nieuwenhuis’s study (1985) is an unpublished doctoral dissertation, but widely cited among scholars. He provides a wealth of cross-linguistic data and many interesting observations

1136

Semantics and pragmatics in word-formation I: General aspects

in his attempt at finding universals in diminutive formation. But he is not specifically focused on pragmatics, although a large number of his examples would deserve a pragmatic analysis. Volek (1987), in what she calls a “pragmatic analysis” (1987: 149−175) of Russian emotive signs (diminutives), makes various relevant observations. Most importantly, she distinguishes between the “emotive attitude” towards the phenomenon named in the base vs. phenomena not named in the base, but rather inherent in the addressee or the speech situation, i.e. in adult-child interaction, jocular or intimate familiar environments, or specific speech acts, like requests. She interprets the first type of attitude as due to a connotational meaning feature of the diminutive word base, and classifies the second type as aspects statically pertaining to the speech situation, with a prejudice, in my view, to a notion of pragmatic meaning as dynamically obtained in the course of the speech event via the application of the morphological rule. The component of the speaker attitude is interestingly developed and modelled in later studies, for example in Dal (cf. her notion of marqueur d’appropriation, 1997) and Fradin (2003) (cf. his notion of interlocutor’s pole) mentioned in section 2.1 above. Wierzbicka (1984, 1991, 1999, 2009) carries out many important analyses of the illocutionary meanings of morphological devices (Russian hypocoristics, Polish diminutives, Italian intensifiers) and on emotion in general. But, in her theoretical configuration, she doubts the rational existence of clear borders between referential/denotational and pragmatic/attitudinal meanings, as, in her words, “all meanings conveyed in natural languages are inherently subjective and anthropocentric” (1991: 17), and she preferably handles pragmatic meanings within her general semantic framework (see a more recent elaboration in Prieto forthc.). Many other studies, mainly on language-specific analyses (Stefanescu 1992; Rainer 1993; Mutz 2000; Gracia and Turon 2000; Laalo 2001; Cantero 2002 among many others) or on more general relevant issues (e.g., Grandi 2002) would deserve comments, but the space left must be devoted to studies precisely focused on the pragmatics of morphology.

4.3. Morphopragmatics The term defines a sub-discipline dealing with phenomena that pertain to morphology and pragmatics at the same time. A theory of morphopragmatics, modelling the sub-field where these phenomena interact, was pioneered and elaborated in successive steps by Dressler and Merlini Barbaresi (1987, 1989, 1992) and expanded into a full-fledged model in 1994 that was further elaborated in 1999 and 2001. Other applications are Dressler and Kiefer (1990) on German and Hungarian excessives, Kilani-Schoch and Dressler (1999) on the French -o suffix, Crocco Galéas (1992) on Italian ethnics, Merlini Barbaresi (2001) on the English -y/ie suffix (see also Merlini Barbaresi forthc.). The model has been paralleled and contrasted with important competitors, mainly Jurafsky (1996), by various researchers worldwide, from Europe to Jordan, Africa, China, Korea (cf. the recent African Morphopragmatics Conference, Manchester 2011). Dressler and Merlini Barbaresi’s and Jurafsky’s studies, although based on opposite theoretical premises and methodologies, are, to date (and to my knowledge), the most complete and influential studies centered on the complex meaning configuration of diminutives (as the

62. The pragmatics of word-formation

1137

central representatives of the category of evaluatives, cf. Grandi and Körtvélyessy forthc.). They both deserve special mention, because Dressler and Merlini Barbaresi’s offers a new perspective on semantic vs. pragmatic meaning in morphology, with the evidence of a myriad of contextualized examples, also contributing to language-specific research, whereas Jurafsky’s major assets are his universalist cognitive approach: he explores 60 languages, and identifies a synchronic complex network of meanings, all diachronically emanated from a central meaning ‘small’, of which each language instantiates at least a portion coherently connected with the centre.

4.3.1. The tenets of the theory Morphopragmatics describes grammatical morphological phenomena (within both wordformation and inflection) capable of systematically contributing autonomous pragmatic meanings to discourse. The model marginalizes extra-grammatical phenomena. It is definable as the set of general pragmatic meanings/effects regularly obtained by grammatical morphological rules. The privileged objects of a morphopragmatic description are evaluative suffixes, such as diminutives, augmentatives, on which the authors base their main argumentation, but also elatives (It. -issimo), reduplicatives, excessives (as German das Aller-schlecht-este EXC-bad-SUPERL ‘the very worst of all’), and, within inflection, personal pronouns of address and Japanese honorifics, which basically interact with the same factors relevant for evaluatives. The morphological operation may be totally responsible for the added utterance meanings, with the word base being either neutral (dogg-y) or contributory (E. dear-ie, It. piccol-ino ’small-DIM’) or even contrary (It. gross-ino ‘big-DIM’) to the effect pursued. Diminutives and augmentatives are evaluative in the sense that they express an evaluation ‘as to value’ (not ‘as to fact’), according to the evaluator’s attitude, beliefs and standards (cf. also Grandi’s argumentation on the evaluative character of diminutives, 2002: 50 ff.). Evaluations may also involve emotion, as is often the case, but not necessarily. Dressler and Merlini Barbaresi (1994) characterize the sub-field of morphopragmatics as the end-point of a diachronic and synchronic process of grammaticalization of pragmatic phenomena. Specifically, they configure a level of morphologized pragmatics, which is meant to cover the area of the general pragmatic meanings of morphological rules. A main goal of this theory is to demonstrate their autonomy in conveying pragmatic meaning. Essential to their definition of morphopragmatics is, therefore, a clear separation between semantic and pragmatic meanings. The authors disprove the “minimalist” hypothesis, which assumes invariant morphosemantic denotation and connotations and then tries to derive all morphopragmatic uses directly from them, with no intermediate morphopragmatic invariants. In that account, morphopragmatic effects of the morphological rule are all derived via general pragmatic strategies, quite independently of morphology, i.e. via the same strategies also relevant for the lexical, syntactic and discourse levels. Dressler and Merlini Barbaresi clearly want to falsify the reduction of morphopragmatics to being a mere result of general pragmatics applied to morphosemantic meaning. They advocate, instead, the thesis of a “maximalist” approach, whereby the denotative meaning is attributed to morphosemantics and the remainder of the meaning components to morphopragmatics. Also connotations, pragmatically originated and lexicalized over time as stable features, are to be accounted for within semantics. The au-

1138

Semantics and pragmatics in word-formation I: General aspects

thors’ contributions (see 2001), in fact, primarily intend to defend the priority of pragmatics over semantics, in opposition to major proposals of semantically-based meanings for diminutives and augmentatives. In order to do so, in addition to the basic semantic meaning ‘small’, with its allosemes ‘unimportant’ and ‘young’ for diminutives (reference to Italian diminutives) and ‘big’ for augmentatives, they propose for both an invariant, non-semantic, still more basic pragmatic feature ‘fictive’, which naturally inheres in and conforms to the fuzziness of subjective evaluations. Fictiveness is conceptualized as a departure from conventionally accepted standards of meaning. It generates a frame of personalized values where such standards glide according to the speaker’s evaluation. Evidence of the pragmatic nature of many of the meanings/effects obtained by diminutives is the fact that many such effects can also be obtained by augmentatives, which share with diminutives the pragmatic feature ‘fictive’, but certainly not the semantic meaning ‘small’. See, for example, It. Sono dei bei sold-ini/sold-oni ‘That’s a pretty penny; lit. They are some money-DIM/AUG’, where both suffixes have an upgrading function, or Mangi come un maial-ino/one ‘You eat as a little piggy/huge pig; lit. You eat like a pig-DIM/AUG’, where the remark does not sound offensive with either suffix, because, here, both suffixes function as attenuators (for a general account of attenuation, see Caffi 2001; Kerbrat-Orecchioni 2004). In diminutives, fictiveness is further specified as a feature ‘non-serious’, which is responsible for the majority of their meanings in discourse (e.g., imprecision, attenuation, but also irony, meiosis, etc.) and which, in general, indexes a speaker’s lowered responsibility and entails lower distance between speaker and addressee. This feature also predicts child/pet/family-centered speech situations to be the privileged locus for diminutives (with emotions and affects involved), while, at the same time, disfavoring (but not excluding) serious, formal interactions. The pragmatic meaning conveyed by diminutive and augmentative operations extends from the suffixed word to the entire utterance and is also capable of modifying the illocutionary strength of the speech act, i.e. the suffixed word is just one of the possible “landing sites” in the sentence.

4.3.2. A competing theory Jurafsky (1996: 563) challenges Dressler and Merlini Barbaresi’s core argument relative to the feature ‘non-serious’ (see Dressler and Merlini Barbaresi 2001 for a general critical rebuttal), which he (following Wierzbicka 1984) proposes to replace with ‘child’ and its meaning ‘small’, as a semantic prototype that he postulates to be at the centre of a universal radial category of diminutive meanings. ‘Child’ is historically prior and motivates all the other senses. The radial structure consists of the central meaning prototype and its extensions, represented by a network of nodes and links. In Jurafsky’s words (1996: 542): “nodes represent prototypes of senses, while links represent metaphorical extensions, image-schematic transfer, transfers to different domains, or inferences”. Links are created by four types of cognitive operations: inference, metaphor, generalization and lambda-abstraction. As a synchronic object, the radial category motivates the sense relations of a polysemous category, but as a diachronic object, it gives account of the meaning changes from the more physical, central sense of ‘child’ to the more general, abstract and qualitative meanings (e.g., pragmatic) of the edge. His model would have been wider-reaching if he had devoted the same type of analysis to augmentatives, not

62. The pragmatics of word-formation

1139

only as the antonymic counterpart of diminutives but especially because the two types of evaluatives share a significant number of pragmatic meanings and functions (cf. Mutz forthc., where this lacuna is filled in). Jurafsky’s model has the great merit of motivating the complexity of the diminutive meaning on a quasi-universal basis and of providing a coherent synchronic and diachronic picture of it. His static representation acquires dynamicity at a cognitive level. Morphopragmatics is dynamical thanks to the feature ‘fictive’, inherent in evaluatives, capable of immediately generating pragmatic meaning, given a favorable set of situational circumstances. Both models, although theoretically diverging, bring new, consistent perspectives to the study of evaluative meanings and are landmarks in the study of semantic vs. pragmatic meaning in word-formation.

5. References Alonso, Amado 1935/51 Noción, emoción, acción y fantasía en los diminutivos. In: Amado Alonso (ed.), Estudios lingüísticos. Temas españoles, 161−189. Madrid: Gredos. Auer, Peter 1996 Context and contextualization. In: Jef Verschueren, Jan-Ola Östman, Jan Blommaert and Chris Bulcaen (eds.), Handbook of Pragmatics, 86−101. Amsterdam/Philadelphia: Benjamins. Bauer, Laurie 1997 Evaluative morphology: In search of universals. Studies in Language 21(3): 533−575. Bauer, Laurie, Rochelle Lieber and Ingo Plag 2013 The Oxford Reference Guide to English Morphology. Oxford: Oxford University Press. Bazzanella, Carla 2004 Emotions, language and context. In: Edda Weigand (ed.), Emotion in Dialogic Interaction. Advances in the Complex, 55−72. Amsterdam/Philadelphia: Benjamins. Blutner, Reinhard 1998 Lexical pragmatics. Journal of Semantics 15(2): 115−162. Caffi, Claudia 2001 La Mitigazione. Münster: Lit Verlag. Cantero, Monica 2002 La morfopragmática del español. München: LINCOM Europa. Crocco Galéas, Grazia 1992 Gli etnici italiani. Studio di morfologia naturale. Padova: Unipress. Dal, Georgette 1997 Grammaire du suffixe -et(te). Paris: Didier Érudition. De Marco, Anna 1998 Sociopragmatica dei diminutivi in italiano. Arcavacata di Rende: Università degli Studi della Calabria, centro editoriale e Librario. Doleschal, Ursula and Anna Maria Thornton (eds.) 2000 Extragrammatical and Marginal Morphology. München: LINCOM Europa. Dressler, Wolfgang U. and Ferenc Kiefer 1990 Austro-Hungarian morphopragmatics. In: Wolfgang U. Dressler, Hans C. Luschützky, Oskar E. Pfeiffer and John R. Rennison (eds.), Contemporary Morphology, 69−77. Berlin: Mouton de Gruyter. Dressler, Wolfgang U. and Lavinia Merlini Barbaresi 1987 Elements of morphopragmatics. In: Jef Verschueren (ed.), Levels of Linguistic Adaptation, 33−51. Amsterdam/Philadelphia: Benjamins.

1140

Semantics and pragmatics in word-formation I: General aspects

Dressler, Wolfgang U. and Lavinia Merlini Barbaresi 1989 Grammaticalizzazione morfopragmatica: Teoria e tipologia, con particolare riguardo ai diminutivi nell’italiano, spagnolo e inglese. Quaderni Dipartimento di Linguistica [Bergamo] 5: 233−255. Dressler, Wolfgang U. and Lavinia Merlini Barbaresi 1994 Morphopragmatics. Diminutives and intensifiers in Italian, German and other languages. Berlin: Mouton de Gruyter. Dressler, Wolfgang U. and Lavinia Merlini Barbaresi 1999 Morphopragmatics. In: Jef Verschueren, Jan Blommaert and Jan-Ola Östman (eds.), Handbook of Pragmatics, 1−14. Amsterdam/Philadelphia: Benjamins. Dressler, Wolfgang U. and Lavinia Merlini Barbaresi 2001 Morphopragmatics of diminutives and augmentatives: On the priority of pragmatics over semantics. In: István Kenesei and Robert M. Harnish (eds.), Perspectives on Semantics, Pragmatics, and Discourse, 43−58. Amsterdam/Philadelphia: Benjamins. Ettinger, Stefan 1974a Diminutiv- und Augmentativbildung. Regeln und Restriktionen. Morphologische und semantische Probleme der Distribution und der Restriktion bei der Substantivmodifikation im Italienischen, Portugiesischen, Spanischen und Rumänischen. Tübingen: Narr. Ettinger, Stefan 1974b Form und Funktion in der Wortbildung. Die Diminutiv- und Augmentativmodifkation im Lateinischen, Deutschen und Romanischen. Ein kritischer Forschungsbericht 1900− 1970. Tübingen: Narr. Fradin, Bernard 2003 Problemi semantici in morfologia derivazionale. In: Anna Maria Thornton and Maria Grossmann (eds.), La formazione delle parole. Atti del 37 Congresso Internazionale di Studi della Società di Linguistica Italiana. L’Aquila, 25−27 settembre 2003, 1−30. Roma: Bulzoni. Gaarder, A. Bruce 1966 Los llamados diminutivos y aumentativos en el espaňol de México. Publications of the Modern Language Association 81: 585−595. Grabiaś, Stanisław 1981 O ekspresywności języka. Lublin: Wydawnictwo Lubelskie. Gràcia, Lluïsa and Lídia Turon 2000 On appreciative suffixes. Acta Linguistica Hungarica 47(1−4): 231−247. Grandi, Nicola 2002 Morfologie in contatto. Le costruzioni valutative nelle lingue del Mediterraneo. Milano: Franco Angeli. Grandi, Nicola and Livia Körtvélyessy (eds.) forthc. Edinburgh Handbook of Evaluative Morphology. Edinburgh: Edinburgh University Press. Hasselrot, Bengt 1957 Etude sur la formation diminutive dans le langues romanes. Uppsala: Uppsala Universitets Ǻrsskrift. Jurafsky, Daniel 1996 Universal tendencies in the semantics of the diminutive. Language 72(3): 533−578. Kerbrat-Orecchioni, Catherine 2000 Quelle place pour les émotions dans la linguistique du XXe siècle? In: Christian Plantin, Marianne Doury and Véronique Traverso (eds.), Les émotions dans les interactions, 33− 74. Lyon: Presses Universitaires de Lyon. Kerbrat-Orecchioni, Catherine 2004 L’adjectif petit comme procédé d’atténuation en français. Travaux et Documents [Université Paris 8-Saint-Denis] 24: 153−175.

62. The pragmatics of word-formation

1141

Kilani-Schoch, Marianne and Wolfgang U. Dressler 1999 Morphopragmatique interactionelle: Les formations en -o du français branché. In: Livia Tonelli and Wolfgang U. Dressler (eds.), Natural Morphology − Perspectives for the Nineties, 31−52. Padova: Unipress. Klimaszewska, Zofia 1983 Diminutive und augmentative Ausdrucksmöglichkeiten des Niederländischen, Deutschen und Polnischen. Eine konfrontative Darstellung. Wrocław: Ossolineum. Laalo, Klaus 2001 Diminutives in Finnish child-directed and child speech: Morphopragmatic and morphophonemic aspects. Psychology of Language and Communication 5(2): 72−80. Lehrer, Adrienne 1987 A note on the semantics of -ist and -ism. American Speech 63: 181−185. Lehrer, Adrienne 1995 Prefixes in English word formation. Folia Linguistica 29: 133−148. Lehrer, Adrienne 1998 Scapes, holics and thongs: The semantics of English combining forms. American Speech 73: 3−28. Lehrer, Adrienne 2003 Polysemy in derivational affixes. In: Brigitte Nerlich, Zazie Todd, Vimala Herman and David D. Clarke (eds.), Flexible Patterns of Meaning in Mind and Language, 217−232. Berlin/New York: Mouton de Gruyter. Lieber, Rochelle 2004 Morphology and Lexical Semantics. Cambridge: Cambridge University Press. Merlini Barbaresi, Lavinia 2001 The pragmatics of the “diminutive” -y/ie suffix in English. In: Chris Schane-Wolles, John Rennison and Friedrich Neubarth (eds.), Naturally!, 315−326. Torino: Rosenberg & Sellier. Merlini Barbaresi, Lavinia forthc. A pragmatic approach to evaluative morphology. In: Nicola Grandi and Livia Körtvélyessy (eds.), Edinburgh Handbook of Evaluative Morphology. Edinburgh: Edinburgh University Press. Mutz, Katrin 2000 Die italienischen Modifikationssuffixe. Frankfurt/M.: Lang. Mutz, Katrin forthc. Evaluative morphology in a diachronic perspective. In: Nicola Grandi and Livia Körtvélyessy (eds.), Edinburgh Handbook of Evaluative Morphology. Edinburgh: Edinburgh University Press. Nieuwenhuis, Paul 1985 Diminutives. Ph.D. dissertation, University of Edinburgh. Prieto, Victor forthc. The semantics of evaluative morphology. In: Nicola Grandi and Livia Körtvélyessy (eds.), Edinburgh Handbook of Evaluative Morphology. Edinburgh: Edinburgh University Press. Rainer, Franz 1990 Appunti sui diminutivi italiani in -etto e -ino. In: Monica Berretta, Piera Molinelli and Ada Valentini (eds.), Parallela 4. Morfologia/Morphologie, 217−218. Tübingen: Narr. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Savickienė, Ineta and Wolfgang U. Dressler (eds.) 2007 The Acquisition of Diminutives. A Cross-Linguistic Perspective. Amsterdam/Philadelphia: Benjamins.

1142

Semantics and pragmatics in word-formation I: General aspects

Scalise, Sergio 1984 Generative Morphology. Dordrecht: Foris. Schneider, Klaus P. 2003 Diminutives in English. Tübingen: Niemeyer. Schneider, Klaus P. 2013 The truth about diminutives, and how we can find it: Some theoretical and methodological considerations. SKASE Journal of Theoretical Linguistics 10(1): 137−151. Schwarze, Christoph 2001 Variation und Entwicklung im Lexikon. Konstanz: Fachbereich Sprachwissenschaft der Universität Konstanz. Sieberer, Anton 1950 Das Wesen des Deminutivs. Die Sprache 2: 85−121. Spitzer, Leo 1921 Das Suffix -one im Romanischen. In: Ernst Gamillscheg and Leo Spitzer (eds.), Beiträge zur Romanischen Wortbildungslehre, 183−205. Genève: Olschki. Spitzer, Leo 1933 Review of Amado Alonso, Para la lingüística de nuestro diminutivo. Humanidades 21: 35−41. Stefanescu, Ioana 1992 On diminutive suffixes. Folia Linguistica 26: 339−356. Verschueren, Jef 1999 Understanding Pragmatics. London/New York: Arnold. Volek, Bronislava 1987 Emotive Signs in Language and Semantic Functioning of Derived Nouns in Russian. Amsterdam/Philadelphia: Benjamins. Volek, Bronislava 1990 Emotive semantics and semiotics. Grazer Linguistische Studien 33/34: 327−347. Weigand, Edda (ed.) 2004 Emotion in Dialogic Interaction. Advances in the Complex. Amsterdam/Philadelphia: Benjamins. Wierzbicka, Anna 1984 Diminutives and depreciatives: Semantic representation for derivational categories. Quaderni di Semantica 1: 123−130. Wierzbicka, Anna 1991 Cross-cultural Pragmatics. The semantics of human interaction. Berlin: Mouton de Gruyter. Wierzbicka, Anna 1999 Emotions across Languages and Cultures. Diversity and Universals. Cambridge: Cambridge University Press. Wierzbicka, Anna 2009 Language and metalanguage: Key issues in emotion research. Emotion Review 1(1): 3−14. Wilson, Deirdre and Robyn Carston 2007 A unitary approach to lexical pragmatics: Relevance, inference and ad hoc concepts. In: Noel Burton-Roberts (ed.), Advances in Pragmatics, 230−259. London: Palgrave. Zwicky, Arnold M. and Geoffrey K. Pullum 1987 Plain morphology and expressive morphology. In: Jon Aske (ed.), Proceedings of the Thirteenth Annual Meeting (February 14−16, 1987). General session and parasession on grammar and cognition, 330−340. Berkeley: Berkeley Linguistic Society.

Lavinia Merlini Barbaresi, Pisa (Italy)

VII. Semantics and pragmatics in wordformation II: Special cases 63. Noun-noun compounds 1. 2. 3. 4. 5. 6. 7.

Introduction Relation-based theories of conceptual combination Empirical support for the RICE theory Applying the RICE theory to established compounds Meaning construction and opaque compounds Conclusion References

Abstract Noun-noun compounds are theoretically interesting due to the variability in how the meanings of their constituents are related to each other in the meaning of the whole compound. This article presents theoretical and empirical evidence that the processing of established noun-noun compounds is affected by meaning construction processes that are common to the processing of novel compounds.

1. Introduction The processing of noun-noun compounds is affected by semantic, conceptual, and pragmatic information. In this article, we provide an overview of these effects from a psycholinguistic perspective and discuss how research on novel compounds can provide valuable insight into the processing of established compounds (i.e. compounds that are part of the language). In particular, we examine the involvement of a meaning construction process during the interpretation of both novel and established compounds. We focus on interpretation of noun-noun compounds rather than on production because the vast majority of psycholinguistic work has focused on this aspect of compound processing. Determining how language-users construct the meaning of a compound is not straightforward. Taft (2003: 127) noted that although some of a compound’s meaning can be derived from the constituents, “other information is needed as well”. We propose that the “other information” consists of the conceptual knowledge which is used to construct a relational structure during the interpretation of a compound. The constituent nouns are not merely linked to each other. Instead, in the case of endocentric compounds, there is a relation that denotes the manner in which the head noun should be modified. This relationship between the constituents of a compound has been referred to as “variable R” (Allen 1978), and there have been several attempts to characterize the specific link that exists between the constituents (e.g., Downing 1977; Finin 1980; Lees 1960; Levi 1978; Li 1971; Warren 1978).

1144

VII. Semantics and pragmatics in word-formation II: Special cases

The impact of this relation is particularly obvious and profound for noun-noun compounds. Although the majority of English adjective-noun compounds can be paraphrased as “noun IS adjective”, noun-noun compounds offer a much wider range of possible relations (for discussions of predicating and non-predicating modifiers see Downing 1977; Levi 1978; Murphy 1990). For example, apple pie is a pie made with apples, whereas apple tree is a tree that produces apples. The relation that links the constituents of a compound heavily influences the meaning of the compound; a newspaper truck could be a truck that delivers newspapers or a truck that is made of newspaper.

2. Relation-based theories of conceptual combination Most research on relations has been conducted using novel compounds. Novel compounds do not yet have a representation at either the lexical or conceptual level and, thus, the meaning of these items must be computed. Therefore, we begin by discussing the relevant findings in the conceptual combination literature. Conceptual combination refers to the process by which two or more concepts are combined to form a new concept. Recognizing that compound words (in the language system) correspond to combined concepts (in the conceptual system) provides a new perspective and raises some important questions about how compounds are represented and accessed. Given that every compound starts out as novel (both in the language and to the individual language user), it seems natural to ask whether the processes that are involved in interpreting novel compounds are maintained once the compound becomes familiar. That is, what knowledge do people use to determine the meaning of novel compounds, and does that information remain influential even after the compound has entered the language? The competition-among-relations-in-nominals (CARIN) theory of conceptual combination is based on the premise that relational information plays a central role in the processing of novel noun-noun compounds (Gagné and Shoben 1997; Spalding and Gagné 2008). Information about how objects, people, and so on interact in the world is used to select a relation that links the constituent concepts during the formation of a new concept. According to this theory, the availability of relational information varies from constituent to constituent. For example, language users know that when chocolate is used as a modifier, the compound usually can be paraphrased using a MADE OF relation, but that other relations such as FOR and ABOUT are also possible. Consequently, some relations are more readily available than others and this difference in availability influences the time required to interpret a compound. Another key prediction of this theory is that relations compete for selection. Consequently, it should be easier to interpret a compound that requires a relation that is highly available (i.e. a relation that is a strong competitor) than to interpret a compound that requires a relation that is less available (i.e. a relation that is a weak competitor). Moreover, changing the availability of a relation (via prior context, for example) should also affect ease of comprehension. According to the CARIN theory, relation selection is most heavily influenced by relational information pertaining to the modifier concept. The relational-interpretation-competitive-evaluation (RICE) theory (Spalding et al. 2010) is derived from the CARIN theory. Like CARIN, the RICE theory posits that multiple relational structures are constructed and evaluated. These structures compete

63. Noun-noun compounds

1145

with one another and ease of interpretation depends on how quickly a single relation structure can be identified as the most likely candidate. However, the RICE theory is more comprehensive than the earlier theory in that it more clearly identifies the roles played by the modifier and head noun, and discusses two aspects of the relation selection process: suggestion and evaluation. In addition, the RICE theory removes CARIN’s claim that relation availability is associated with the modifier concept, but not with the head-noun concept. According to RICE, the interpretation of a compound is obtained through a “suggestevaluate-elaborate” process. The search for suitable relations is triggered after the constituents have been assigned to their respective morphosyntactic roles and relation availability is associated with concepts in a particular role. As in the CARIN theory, the relations are initially suggested with a strength proportional to the relations’ availability for the modifier. Next, the appropriateness of the suggested relations is evaluated. During this evaluation stage, relational information about the head plays a role. An important aspect of the evaluation process concerns the ability of the constituents to function as arguments for a given relation structure. In particular, a constituent must fit the entailments required to fulfill a particular function within a particular relation. Consequently, the constituent snow can function as a modifier in the MADE OF relation because it is a material. Likewise, the constituent planet can function as the head noun of a LOCATIVE relation because planets can be in a physical location. Furthermore, although each constituent must fit the restrictions of its respective roles, the restrictions are co-determined. To illustrate, the MADE OF relation requires that the modifier be a material, but not just any material. It must be a material that is appropriate for the head noun. Snow sculpture satisfies these restrictions, but snow hospital does not. This aspect of the evaluation process relies on world knowledge. Gagné (2002; Spalding and Gagné 2008) uses the example of mountain planet to illustrate the use of this type of knowledge. The interpretation “planet LOCATED mountain” is rejected because, to name just one pragmatic restriction, planets are too large to be located in the mountains. Levi (1978) refers to this type of knowledge as extralinguistic knowledge and she outlines several semantic and pragmatic considerations that are used to determine the contextually most plausible reference for a given compound (for further discussion concerning the use of extralinguistic information see Downing 1977; Finin 1980; Levi 1978; Meyer 1993; Štekauer 2005, 2006, 2009). Finally, once a suggested relation has been selected and evaluated as appropriate, there is an elaboration stage in which the full meaning of the compound is developed.

3. Empirical support for the RICE theory The RICE theory makes two main predictions. First, relations compete for selection and, consequently, the relative availability of the to-be-selected relation affects the ease of comprehension such that the more available the relation, the faster it will be to select that relation. Furthermore, the RICE theory posits that the modifier and head noun are differentially involved in the suggestion and evaluation of relations. Second, the RICE theory distinguishes between the assignment of a constituent to a particular morphosyntactic role (e.g., to either the modifier or head noun role) and the selection of a relational

1146

VII. Semantics and pragmatics in word-formation II: Special cases

structure. According to this theory, relational information about the constituents is accessed in the context of the constituents’ morphosyntactic roles. Both predictions have received empirical support.

3.1. Relation availability influences ease of processing Initial evidence for the influence of relation availability on compound processing was provided by Gagné and Shoben (1997). To estimate the competitiveness of various relations for particular constituents, they created a set of potential novel compounds by crossing 91 modifiers and 91 head nouns. They determined whether each pairing had a sensible literal interpretation and classified these 3,239 sensible modifier-noun phrases in terms of Levi’s (1978) categories. For example, plastic bee was classified as using the noun MADE OF modifier relation. Next, they computed the frequency with which each modifier and head noun appeared with the various relations. For example, of the 38 compounds in which plastic was used as a modifier, 28 used the noun MADE OF modifier relation, 7 used the noun ABOUT modifier relation, 2 used the noun DERIVED FROM modifier relation, and 1 used the modifier CAUSES noun relation. The relation distributions were used to select items that would fit one of three experimental conditions. HH compounds were defined as compounds for which the underlying relation was among the set of highly competitive relations for both the modifier and head, HL compounds were compounds for which the underlying relation was among the set of highly competitive relations for the modifier only, and LH compounds were compounds for which the underlying relation was among the set of highly competitive relations for the head noun only. During the experiment, each item was presented on a computer screen and participants performed a sense-nonsense judgment task (i.e. they indicated, by pressing a key, whether the item had a sensible literal interpretation). Nonsense filler items (e.g., scarf soda) were included to balance the number of sense and nonsense responses in the experiment. Responses to the HH and HL phrases were faster than responses to the LH phrases, indicating that it was easier to determine that the phrase had a sensible interpretation when the underlying relation was among the set of highly-competitive relations for the modifier than when it was a less competitive relation. Responses in the HH and HL conditions did not differ, indicating that the availability of the relation for the head noun constituent did not strongly affect response time in this task. Further evidence for the influence of relation availability comes from experiments using a priming paradigm. Gagné (2001, 2002) showed that it takes less time to make a sense-nonsense judgment about a novel compound such as oil treatment (treatment USES oil) when it is preceded by a compound using the same modifier and the same relation (oil moisturizer) than by a compound using a different relation (e.g., oil accident). This finding suggests that processing the prime compound alters the competitiveness of the relation used to interpret the prime compound such that the relation becomes more able to successfully compete with other relations. Consequently, the relation is easier to select if it is subsequently needed to interpret the target compound. On the other hand, the required relation is more difficult to select for the target item if that relation is not the same as the one used in the prime. This relation priming effect was not observed when

63. Noun-noun compounds

1147

the head noun was in common between the prime and target (Gagné 2001, 2002), except when the target phrase had more than one equally plausible relation (Gagné and Shoben 2002). The finding that head-based relational information did not affect the ease of making sense-nonsense decisions was initially taken to mean that relational information about the head noun is not used during conceptual combination. However, more recent research has found head-based relational effects when the judgment task is verification (in which participants indicate whether a definition is appropriate for the compound) rather than a sense-nonsense decision (Spalding et al. 2010). The two tasks differentially engage the suggestion and evaluation stages. Relative to the sense-nonsense task, the verification task more heavily engages the relation evaluation stage of the combination process because the task is to indicate whether the presented relation is appropriate. In contrast, the sense-nonsense task more heavily engages the relation-suggestion stage, as no particular relation is presented to the participant. Across a series of experiments, Spalding et al. (2010) found that modifier-based relational effects were more robust than the head-based relational effects in the sense-nonsense task (similar to the results of Gagné 2001, 2002; Gagné and Shoben 1997). However, head-based relational information strongly impacted performance in the verification task. Indeed, the head-based relational effects were more robust than modifier-based relational effects in the verification task. These results obtained when relational availability was manipulated using the same design and materials as Gagné and Shoben (1997) and when relational availability was manipulated by relation priming (as in Gagné 2001, 2002). Thus, for example, relations that are stronger competitors for a modifier lead to faster sense-nonsense decisions than relations that are not strong competitors, but the availability of the relation for the head has relatively little effect. When the decision task is verification, however, the relation availability for the head has a strong effect. Similarly, when the target and prime share a modifier, repeating the relation leads to faster sense-nonsense decisions than not repeating the relation, but the effect on verification decisions is modest. When the prime and target share a head, however, repeating the relation does not strongly affect the ease of sense-nonsense decisions, but it does strongly affect the ease of verification decisions. Spalding et al. (2010) results are consistent with the RICE theory’s main claim that relational information associated with the constituents of compounds affects processing of novel compounds. The results show that modifier-based and head-based relational information both make important, but different, contributions to relational interpretation of novel compounds. Furthermore, they are consistent with the RICE theory’s claims that the suggestion of relations and the evaluation of relations are two separate stages and that modifier-based relational information is most heavily involved in the suggestion of relations, whereas head-based relational information is more heavily involved in the evaluation of relations.

1148

VII. Semantics and pragmatics in word-formation II: Special cases

3.2. Access of relational information occurs after morphosyntactic assignment Gagné et al. (2009) examined whether morphosyntactic assignment and relation selection are two distinct psychological processes. They used a priming paradigm and manipulated whether the prime used the same relation as the target. In experiment 1, the shared constituent moved from the modifier position in the prime to the head noun position in the target. For example, the target research mouse (mouse FOR research) was preceded by either mouse trap (trap FOR mouse) or mouse whisker (whisker HAS mouse). In experiment 2, the shared constituent moved from the head noun position in the prime to the modifier position in the target. For example, summer car (car FOR summer) and metal car (car MADE OF metal) served as primes for the target car port (port FOR car). In experiment 3, the design included primes in which the constituent was used in the same position for both the prime and target, which allowed us to determine whether there is any benefit to repeating a constituent in the same morphosyntactic position. A sense-nonsense task was used in all experiments. The results were consistent with the predictions of the RICE theory. The ease of interpreting compounds depended on the ease of mapping constituents to particular morphosyntactic roles, and on the ease of selecting an appropriate relational structure. In terms of the role of morphosyntactic information, processing of the target was faster when the shared constituent was in the same position in both the prime and the target than when the shared constituent was in a different position, and this benefit occurred regardless of whether the prime used the same relation as the target. To illustrate, both fur blanket and fur trader aided the processing of fur gloves relative to the processing of brown fur. This finding indicates that the constituents are assigned morphosyntactic roles without reference to any particular relational structure. In addition, because both the same-position and different-position targets used the same constituent (i.e. fur) the effect is not solely due to repetition priming. In contrast, the use of relational information was contingent on the morphosyntactic role of the shared constituent. In all three experiments, repeating the relation with the constituent in a different morphosyntactic role did not speed processing of the target. Experiment 3 confirmed that the same-relation prime only benefited the processing of the target combination when the shared constituent remained in the same position.

4. Applying the RICE theory to established compounds The results presented in section 3 show that relational information plays an important role in the construction of a novel compound’s meaning. Conceptual combination is obligatory for novel compounds because the meaning of such compounds is not yet part of the lexical or conceptual system. Whether conceptual combination (or meaning construction) is also involved in the processing of established compounds is a particularly relevant issue because, although many existing theories of complex-word processing (such as Libben 1998, 2006; Sandra 1994; Schreuder and Baayen 1995; Taft 1994; Zwitserlood 1994) posit the existence of a conceptual level of representation that is linked to an access (word-form) representation of that word, they do not yet contain a

63. Noun-noun compounds

1149

mechanism for how the meaning representation of the compound is formed, and relational information has not yet been incorporated into these views. In this respect, the RICE theory provides insight into the nature of compound processing. The RICE theory is based on the premise that the language system will always attempt to compute a meaning whenever meaning properties of the constituents are available, and that meaning construction occurs not only for novel compounds, but also for established compounds. This assumption aligns with Schreuder and Baayen’s (1995, 1997) statement that the primary function of morphological processing is to compute meaning. In the current section, we argue that research on conceptual combination, in general, and the RICE theory, in particular, is relevant for understanding the processing of established compounds (i.e. compounds that are part of the language). We discuss both theoretical and empirical reasons for positing that some meaning construction process occurs for established compounds and then discuss recent research showing specifically that relational effects occur in the processing of established compounds.

4.1. Theoretical grounds for positing meaning construction The psycholinguistic literature on complex word processing has recently emphasized computation and the role of semantic/conceptual information. According to Libben’s (2010: 324) principle of maximization of opportunity “the human mind in general and the human lexical processing system in particular is organized to maximize the opportunity for meaning creation”. Similarly, Kuperman, Bertram and Baayen (2010) argue for a multiple-route model in which the language processing system attempts to use all information that is available to it. In addition, some researchers suggest that some of the effects for complex words that were originally assumed to originate from the lexical level might actually derive from the level of syntactic and/or semantic assignment (e.g., Hyönä, Vainio and Laine 2002; Lehtonen et al. 2006). The RICE theory, with its emphasis on meaning computation and on the retrieval of all available sources of information, is compatible with these recent trends in the literature because it is based on the theoretical assumption that if conceptual information about the constituents is available to the system, then the system will attempt to use that information to construct a unified representation (i.e. it constructs meaning).

4.2. Empirical grounds for positing meaning construction Meaning construction in the processing of established compounds depends on the lexical and semantic/conceptual representations of the constituents being available. In addition, those representations must then be integrated into a meaning for the compound.

4.2.1. Availability of the constituents’ lexical representations A necessary pre-condition for meaning construction is that the constituent representations of a compound are available during compound processing. There are several streams of

1150

VII. Semantics and pragmatics in word-formation II: Special cases

research indicating that they are available. First, the processing of a compound word is influenced not only by the frequency of the whole word but also by the frequency of the constituents. Compounds with high frequency constituents require less time to process than do compounds with lower frequency constituents (e.g., Andrews, Miller and Rayner 2004; Bien, Levelt and Baayen 2005; Dunabeitia, Perea and Carreiras 2007; Juhasz 2007; Juhasz et al. 2003; Shoolman and Andrews 2003). Also, people with aphasia are less likely to produce errors in naming compounds with high frequency constituents than in naming compounds with lower frequency constituents (e.g., Blanken 2000). Second, priming studies have revealed that exposure to one of the constituents facilitates the subsequent processing of compound words (Duñabeitia et al. 2009; Forster and Davis 1984; Jarema et al. 1999; Libben et al. 2003; Monsell 1985; Shoolman and Andrews 2003; Zwitserlood 1994). For example, responses to bookshop were faster when preceded by either book or shop than when the previous trial was an unrelated word, such as house (Shoolman and Andrews 2003). This benefit has been found for both semantically opaque and transparent compounds in English (Libben et al. 2003) and in French and Bulgarian (Jarema et al. 1999). The reverse is also true; it was easier to process a word (e.g., black) following recent exposure to a compound (e.g., blackbird or blackmail) in which that word was used as a constituent (Masson and MacLeod 1992; Sandra 1990). Finally, studies using a picture naming task have revealed that naming latencies were faster when words presented earlier in the experiment were morphologically related to the picture and that this facilitation occurred regardless of the position of the shared morpheme (Zwitserlood, Bolte and Dohmes 2002). For example, both Blumentopf ‘flowerpot’ and Topfblume ‘pot plant’ speeded the naming of a picture of a flower (Blume). Third, further evidence that a compound’s constituents become available during compound processing comes from the analysis of the types of errors produced by people with aphasia. Errors to compound words include the omission or substitution of one of the constituents (Badecker and Caramazza 2001; Blanken 2000; Hittmair-Delazer et al. 1994; Jarema, Perlak and Semenza 2010; Mäkisalo, Niemi and Laine 1999; Semanza, Luzzatti and Carabelli 1997). For example, producing waterhorse rather than seahorse, or producing ash vase rather than ashtray (Badecker and Caramazza 2001). Even when an error has been made, however, the compositional structure often remains intact; that is, the correct constituents are typically kept in the correct position (Hittmair-Delazer et al. 1994; Delazer and Semenza 1998).

4.2.2. Availability of the constituents’ semantic representations The results discussed in section 4.2.1 suggest that orthographic (or word-form) and lexical representations become available during the processing of transparent compounds, as well as of opaque compounds. However, are the semantic representations of the constituents also available? Libben, Derwing and de Almeida (1999) used ambiguous novel compounds that could be parsed in two ways. For example, clamprod could be parsed as clam + prod or as clamp + rod. Libben et al. (2003) found that the presentation of the ambiguous compound aided the recall of semantic associates of all four possible constituents (i.e. sea, hold, push, and stick, which, in order, are related to clam, clamp,

63. Noun-noun compounds

1151

prod, and rod ), which indicates that during the processing of clamprod, the lexical and semantic representations of all constituents were accessed. In addition, they found that the parsing choice for ambiguous compounds is driven by the relative semantic plausibility of the parses. This latter finding is consistent with Koester, Holle and Gunter’s (2007) finding that German three-word compounds with plausible constituents were easier to interpret than were compounds with less plausible constituents. A number of studies have also found semantic priming effects between transparent compounds and their constituents. For example, Zwitserlood (1994) found a semantic priming effect for transparent Dutch compounds and their constituents (e.g., kerkorgel ‘church organ’ as a prime for muziek ‘music’), as did Sandra (1990). However, evidence for semantic priming with opaque compounds is much more mixed. Opaque compounds can be fully opaque (e.g., neither hog nor wash is related to hogwash, which means ‘nonsense’), or partially opaque (e.g., berry is related to strawberry, but straw is not). Zwitserlood (1994) found semantic priming with partially opaque Dutch compounds (e.g., drankorgel, which means ‘drunkard’ as a prime for muziek ‘music’), but Sandra (1990) did not. Zwitserlood did not observe semantic priming for completely opaque compounds. Isel, Gunter and Friederici (2003) investigated semantic priming and compound transparency in a cross-modal priming paradigm. In their study, participants heard the compound (e.g., Gasthaus ‘guesthouse’) and then made a lexical decision judgment on a word that was semantically related to one of the compound’s constituents (e.g., Besuch ‘visit’). The left constituent yielded a semantic priming effect only when the right constituent was transparent (i.e. only for OT and TT items), which indicates that the activation of the left constituent of German compounds is affected by the transparency of the right constituent. However, when prosody information was removed from the signal (by recording the constituents separately and then splicing the recordings together to form the auditory signal for the compound), then the left constituent yielded semantic priming regardless of the head noun’s transparency. Thus, in this condition Isel, Gunter and Friederici (2003) found semantic priming even for completely opaque compounds. Finally, Shillcock (1990) found semantic priming even with pseudo-compounds (e.g., carpet) which serve as a kind of purely opaque condition because they do not actually include morphological constituents. We will return to the issue of semantic opacity in section 5.

4.2.3. Constituent integration and semantic composition Much research suggests that the processing of complex words requires the integration of those words’ constituents (e.g., Schreuder and Baayen 1995; Caramazza, Laudanna and Romani 1988; Fiorentino and Poeppel 2007; Koester, Gunter and Wagner 2007; Juhasz, Inhoff and Rayner 2005). For example, Inhoff, Radach and Heller (2000) examined the effect of inter-word spacing on eye-fixations for German compounds and found that spacing facilitated access to constituent word forms, but hampered the creation of a unified compound meaning, as indicated by longer final fixation times for spaced compounds. Based on this finding, they suggested that two processes are involved in the interpretation of compounds, one that involves accessing the constituent word forms (and is aided by the presence of a space) and one that involves integrating the constitu-

1152

VII. Semantics and pragmatics in word-formation II: Special cases

ents (and is hindered by the presence of a space). Juhasz, Inhoff and Rayner (2005) found that these results extend to English compounds. For the most part, constituent integration has been described in terms of co-activation (e.g., Libben 1998; Schreuder and Baayen 1995, 1997; Taft 2003; Taft and Kougious 2004; Zwitserlood 1994; Zwitserlood, Bolwiender and Drews 2005); that is, the activation of the lexical representations of a compound’s constituents resulting in increased activation of the lexical representation of the compound due to facilitatory links between these representations. However, we posit that an important aspect of integration is not only the co-activation of constituents, but also composition of their conceptual representations (Gagné and Spalding 2009; see also Schreuder and Baayen 1995, 1997). Several streams of evidence strongly suggest that compound processing involves semantic composition beyond just co-activation. Kounios et al. (2003) performed memory studies in which participants were shown word pairs and asked either to integrate the words into a single meaning unit, or to remember the words as two associated items. Different patterns of brain activity were observed when participants combined word pairs to form a single conceptual unit than when they processed the words as two concepts, which suggests that combining words consists of something beyond just forming an association between the constituents. Koester, Gunter and Wagner (2007) compared the size of a negative shift in the ERP signal (which is thought to represent semantic composition) that occurred when participants listened to novel three-constituent German compounds. The negative shift was larger for compounds that were judged by participants (in a post-hoc classification) to be more difficult to integrate and, hence, required a greater degree of semantic composition, than for items that were judged to be easier to interpret. In addition, this shift was larger for transparent compounds, which can be processed using a semantic composition process, than for opaque compounds, which do not rely on semantic composition. Further support for semantic composition comes from studies that indicate that multiple interpretations are constructed and evaluated during compound processing. Gagné, Spalding and Gorrie (2005) found that as the dominance of a particular meaning for a novel compound (as measured by the percentage of people preferring that meaning) increased, the time required to select the best definition for the compound decreased. Also, as noted in section 4.2.2, semantic plausibility of the parses affects the parsing choice (Libben, Derwing and Almeida 1999) and semantic plausibility of the constituents affects the time required to interpret the compound (Koester, Holle and Gunter 2009). The computed meaning can intrude on the conventional meaning. Libben (1998) presents examples in which a person with mixed aphasia produced the literal meaning of opaque compounds. For example, the person paraphrased blueprint as ‘a print that is blue’ and bellybutton as ‘a button in your stomach’. This phenomenon is not restricted to aphasics. In a study by Gagné, Spalding and Gorrie (2005), participants read a sentence consistent with the established meaning (e.g., The thread that a silk worm produces is often used by Kim to make beautiful scarfs) or with an innovative meaning (e.g., Kim decided it would be fun to make a silk worm out of the fabric she had bought) of a compound such as silk worm. Immediately after viewing this sentence, the participants viewed the compound with either the established meaning (e.g., ‘a worm that produces silk’) or the innovative meaning (e.g., ‘a worm made of silk’). They were told that the definition did not have to be the best definition, but that they should indicate “yes” if the definition was plausible. When the sentence used the established meaning, the estab-

63. Noun-noun compounds

1153

lished definition was judged plausible 89 % of the time. However, when the sentence used the innovative meaning, the established definition was judged plausible only 64 % of the time. In terms of response time, participants took longer to indicate that the established definition was plausible when the sentence supported the innovative meaning than when it supported the conventional meaning. These findings suggest that competition with the innovative meaning constructed in the previous sentence decreased the availability of the established meaning.

4.3. Influence of relation availability In the preceding sections, we discussed substantial evidence indicating that constituent representations are available and that semantic composition might be occurring during the processing of established compounds. These findings strongly suggest that the processing of established compounds involves meaning construction of some sort. Importantly, there is also direct evidence that processing established compounds involves integrating the modifier and head noun constituents into a relational structure. In particular, the relation priming effect that was observed for novel compounds (see section 3.1) also occurs for established compounds. For example, Gagné and Spalding (2009, see also Gagné and Spalding 2004) found that response time to an established compound (e.g., snowball ) was faster when the compound was preceded by a compound using the same relation (e.g., snowfort, MADE OF) than by a compound using a different relation (e.g., snowshovel, FOR). Furthermore, consistent with the RICE theory’s emphasis on the competition among relational interpretations, recent evidence indicates that the relation priming effect observed for lexicalized compounds arises due to slower processing following a different relation prime rather than faster processing following a same relation prime (Spalding and Gagné 2011).

5. Meaning construction and opaque compounds Thus far, we have mostly discussed the issue of meaning construction in the context of semantically transparent compounds which have a relational structure. However, semantically opaque compounds such as jailbird or hogwash by definition have no such structure because at least one of the constituents does not contribute to the meaning of the compound. Although the final interpretation is not determined via semantic composition of the constituents, there has been some preliminary evidence that the meaning construction process occurs during the interpretation of such compounds. The language processing system is unable to determine a priori whether a compound is transparent or opaque and, as discussed in section 4.2.1, there is clear evidence that the lexical representations of the constituents become available during processing. In Libben’s (1998) theory, activation of lexical representations results in either the activation (in the case of transparent constituents) or the inhibition (in the case of opaque constituents) of the corresponding semantic representations (see also Libben et al. 2003; Zwitserlood 1994). Unlike these approaches, the RICE theory posits that the activation of the semantic representation corresponding to an activated lexical representation occurs

1154

VII. Semantics and pragmatics in word-formation II: Special cases

regardless of whether the constituent is transparent. Rather than explaining the role of semantic transparency in terms of facilitation or inhibition among the lexical and semantic levels, we ascribe the influence of semantic transparency to differences in the nature of the processing that is afforded by transparent and opaque compounds. In particular, the influence of the semantic representations can be obscured due to conflict resulting from the computed meanings based on these representations. If the lexicalized meaning and the computed meaning (or meanings, if more than one is possible) become available within the same time frame, then the system must evaluate which meaning is most likely. Whether the computed meaning is maintained depends on an evaluation process that evaluates the various meanings available to the system and selects the meaning that is most appropriate for the current context (Spalding et al. 2010). In the case of opaque compounds, the constructed meaning conflicts with the conventional meaning, which introduces processing costs because this conflict must be resolved as the system attempts to settle on one meaning. On average, there is less competition between the lexicalized meaning and computed meaning for transparent compounds than for opaque compounds because there is more consistency among these meanings for transparent compounds. Consistent with this claim, Ji, Gagné and Spalding (2011; Ji 2008) found that experimental manipulations that aided morphological decomposition (and, thereby, aided semantic composition) slowed the processing of opaque compounds, but did not slow the processing of transparent compounds. Additional support for the claim that there is a processing cost associated with semantic composition for opaque compounds comes from the observation that the frequency of the first constituent differentially influenced response time for transparent and opaque compounds; the processing of transparent compounds was helped by having a high frequency first constituent, but opaque compounds were hindered by having a high frequency first constituent. That is, the more available the constituent representation, the more difficult it was to process an opaque compound. Taken together, these results are consistent with the view that meaning construction occurs for both opaque and transparent compounds, but for opaque compounds this process produces a meaning that conflicts with the conventional meaning.

6. Conclusion Relative to adjective-noun compounds, noun-noun compounds are especially variable in terms of the relational structures that underlie their meaning. In this article, we have proposed that research on conceptual combination is useful for studying noun-noun compounds. In our discussion, we have focused on the RICE theory, which is a relationbased theory of conceptual combination. The empirical evidence suggests that meaning construction occurs during the processing of both novel and established compounds, including opaque compounds, and that relational information plays an important role in this meaning construction process.

63. Noun-noun compounds

1155

7. References Allen, Margaret R. 1978 Morphological investigations. Ph.D. dissertation, University of Connecticut. Andrews, Sally, Brett Miller and Keith Rayner 2004 Eye movements and morphological segmentation of compound words: There is a mouse in mousetrap. European Journal of Cognitive Psychology 16: 285−311. Badecker, William and Alfonso Caramazza 2001 Morphology and aphasia. In: Andrew Spencer and Arnold M. Zwicky (eds.), The Handbook of Morphology, 390−405. Malden, MA: Blackwell. Bien, Heidrun, Willem J. M. Levelt and R. Harald Baayen 2005 Frequency effects in compound production. Proceedings of the National Academy of Sciences of the United States of America 102: 17876−17881. Blanken, Gerhard 2000 The production of nominal compounds in aphasia. Brain and Language 74: 84−102. Caramazza, Alfonso, Alessandro Laudanna and Cristina Romani 1988 Lexical access and inflectional morphology. Cognition 28: 297−332. Delazer, Margarete and Carlo Semenza 1998 The processing of compound words: A study in aphasia. Brain and Language 61: 54−62. Downing, Pamela 1977 On the creation and use of English compound nouns. Language 53: 810−842. Duñabeitia, Jon A., Itziar Laka, Manuel Perea and Manuel Carreiras 2009 Is Milkman a superhero like Batman? Constituent morphological priming in compound words. European Journal of Cognitive Psychology 21: 615−640. Duñabeitia, Jon A., Manuel Perea and Manuel Carreiras 2007 The role of the frequency of constituents in compound words: Evidence from Basque and Spanish. Psychonomic Bulletin and Review 14: 1171−1176. Finin, Timothy W. 1980 The semantic interpretation of compound nominals. Ph.D. dissertation, University of Illinois. Fiorentino, Robert and David Poeppel 2007 Compound words and structure in the lexicon. Language and Cognitive Processes 22: 953−1000. Forster, Kenneth I. and Chris Davis 1984 Repetition priming and frequency attenuation in lexical access. Journal of Experimental Psychology: Learning, Memory, and Cognition 10: 680−698. Gagné, Christina L. 2001 Relation and lexical priming during the interpretation of noun-noun combinations. Journal of Experimental Psychology: Learning, Memory, and Cognition 27: 236−254. Gagné, Christina L. 2002 Lexical and relational influences on the processing of novel compounds. Brain and Language 81: 723−735. Gagné, Christina L. and Edward J. Shoben 1997 Influence of thematic relations on the comprehension of modifier-noun combinations. Journal of Experimental Psychology: Learning, Memory, and Cognition 23: 71−87. Gagné, Christina L. and Edward J. Shoben 2002 Priming relations in ambiguous noun-noun combinations. Memory and Cognition 30: 637−646. Gagné, Christina L. and Thomas L. Spalding 2004 Effect of relation availability on the interpretation and access of familiar noun-noun compounds. Brain and Language 90: 478−486.

1156

VII. Semantics and pragmatics in word-formation II: Special cases

Gagné, Christina L. and Thomas L. Spalding 2009 Constituent integration during the processing of compound words: Does it involve the use of relational structures? Journal of Memory and Language 60: 20−35. Gagné, Christina L., Thomas L. Spalding, Lauren Figueredo and Allison C. Mullaly 2009 Does snow man prime plastic snow? The effect of constituent position in using relational information during the interpretation of modifier-noun phrases. The Mental Lexicon 4: 41−76. Gagné, Christina L., Thomas L. Spalding and Melissa C. Gorrie 2005 Sentential context and the interpretation of familiar open-compounds and novel modifier-noun phrases. Language and Speech 48: 203−221. Hittmair-Delazer, Margarete, Barbara Andree, Carlo Semenza, Ria De Bleser and Thomas Benke 1994 Naming by German compounds. Journal of Neurolinguistics 8: 27−41. Hyönä, Jukka, Seppo Vainio and Matti Laine 2002 A morphological effect obtains for isolated words but not for words in sentence context. European Journal of Cognitive Psychology 14: 417−433. Inhoff, Albrecht W., Ralph Radach and Dieter Heller 2000 Complex compounds in German: Interword spaces facilitate segmentation but hinder assignment of meaning. Journal of Memory and Language 42: 23−50. Isel, Frederic, Thomas C. Gunter and Angela D. Friederici 2003 Prosody-assisted head-driven access to spoken German compounds. Journal of Experimental Psychology: Learning, Memory, and Cognition 29: 277−288. Jarema, Gonia, Céline Busson, Rossitza Nikolova, Kyrana Tsapkini and Gary Libben 1999 Processing compounds: A cross-linguistic study. Brain and Language 68: 362−369. Jarema, Gonia, Danuta Perlak and Carlo Semenza 2010 The processing of compounds in bilingual aphasia: A multiple-case study. Aphasiology 24: 126−140. Ji, Hongbo 2008 The influence of morphological complexity on word processing. Ph.D. dissertation, Department of Psychology, University of Alberta. Ji, Hongbo, Christina L. Gagné and Thomas L. Spalding 2011 Benefits and costs of lexical decomposition and semantic integration during the processing of transparent and opaque English compounds. Journal of Memory and Language 65: 406−430. Juhasz, Barbara J. 2007 The influence of semantic transparency on eye movements during English compound word recognition. In: Roger P. G. Van Gompel, Martin H. Fischer, Wayne S. Murray and Robin L. Hill (eds.), Eye Movements. A Window on Mind and Brain, 373−389. Oxford: Elsevier. Juhasz, Barbara J., Albrecht W. Inhoff and Keith Rayner 2005 The role of interword spaces in the processing of English compound words. Language and Cognitive Processes 20: 291−316. Juhasz, Barbara J., Matthew S. Starr, Albrecht W. Inhoff and Lars Placke 2003 The effects of morphology on the processing of compound words: Evidence from naming, lexical decisions and eye fixations. British Journal of Psychology 94: 223− 244. Koester, Dirk, Thomas C. Gunter and Susanne Wagner 2007 The morphosyntactic decomposition and semantic composition of German compound words investigated by ERPs. Brain and Language 203: 64−79. Koester, Dirk, Henning Holle and Thomas C. Gunter 2009 Electrophysiological evidence for incremental lexical-semantic integration in auditory compound comprehension. Neuropsychologia 47: 1854−1864.

63. Noun-noun compounds

1157

Kounios, John, Peter Bachman, Daniel Casasanto, Murray Grossman, Roderick W. Smith and Wei Yang 2003 Novel concepts mediate word retrieval from human episodic associative memory: Evidence from event-related potentials. Neuroscience Letters 345: 157−160. Kuperman, Victor, Raymond Bertram and R. Harald Baayen 2010 Processing trade-offs in the reading of Dutch derived words. Journal of Memory and Language 62: 83−97. Lees, Robert B. 1960 The Grammar of English Nominalizations. Bloomington, IN: Indiana University. Lehtonen, Minna, Victor A. Vorobyev, Kenneth Hugdahl, Terhi Tuokkola and Matti Laine 2006 Neural correlates of morphological decomposition in a morphologically rich language: An FMRI study. Brain and Language 98: 182−193. Levi, Judith N. 1978 The Syntax and Semantics of Complex Nominals. New York: Academic Press. Li, Charles N. 1971 Semantics and the structure of compounds in Chinese. Ph.D. dissertation, University of California. Libben, Gary 1998 Semantic transparency in the processing of compounds: Consequences for representation, processing, and impairment. Brain and Language 61: 30−44. Libben, Gary 2006 Why study compound processing: An overview of the issues. In: Gary Libben and Gonia Jarema (eds.), The Representation and Processing of Compound Words, 1−21. New York: Oxford University Press. Libben, Gary 2010 Compound words, semantic transparency, and morphological transcendence. In: Susan Olsen (ed.), New Impulses in Word-Formation, 317−330. Hamburg: Buske. Libben, Gary, Bruce L. Derwing and Roberto G. Almeida 1999 Ambiguous novel compounds and models of morphological parsing. Brain and Language 68: 378−386. Libben, Gary, Martha Gibson, Yeo Yoon and Dominiek Sandra 2003 Compound fracture: The role of semantic transparency and morphological headedness. Brain and Language 84: 50−64. Masson, Michael and Colin MacLeod 1992 Re-Enacting the route to interpretation: Enhanced perceptual identification without prior perception. Journal of Experimental Psychology: General 121: 145−176. Mäkisalo, Jukka, Jussi Niemi and Matti Laine 1999 Finnish compound structure: Experiments with a morphologically impaired patient. Brain and Language 68: 249−253. Meyer, Ralf 1993 Compound Comprehension in Isolation and in Context. The Contribution of Conceptual and Discourse Knowledge to the Comprehension of German Novel Noun-Noun Compounds. Tübingen: Niemeyer. Murphy, Gregory L. 1990 Noun phrase interpretation and conceptual combination. Journal of Memory and Language 29: 259−288. Monsell, Stephen 1985 Repetition and the lexicon. In: Andrew W. Ellis (ed.), Progress in the Psychology of Language. Vol. 2, 147−195. Hillsdale, NJ: Erlbaum. Sandra, Dominiek 1990 On the representation and processing of compound words: Automatic access to constituent morphemes does not occur. The Quarterly Journal of Experimental Psychology 42A: 529−567.

1158

VII. Semantics and pragmatics in word-formation II: Special cases

Sandra, Dominiek 1994 The morphology of the mental lexicon: Internal word structure viewed from a psycholinguistic perspective. Language and Cognitive Processes 9: 227−269. Schreuder, Robert and R. Harald Baayen 1995 Modeling morphological processing. In: Laurie B. Feldman (ed.), Morphological Aspects of Language Processing, 131−156. Hillsdale, NJ: Erlbaum. Schreuder, Robert and R. Harald Baayen 1997 How complex simple words can be. Journal of Memory and Language 37: 118−139. Semenza, Carlo, Claudio Luzzatti and Simona Carabelli 1997 Morphological representation of compound nouns: A study on Italian aphasic patients. Journal of Neurolinguistics 10: 33−43. Shillcock, Richard 1990 Lexical Hypotheses in Continuous Speech. Cambridge, MA: MIT Press. Shoolman, Natalie and Sally Andrews 2003 Racehorses, reindeer, and sparrows: Using masked priming to investigate morphological influences on compound word identification. In: Sachiko Kinoshita and Stephen J. Lupker (eds.), Masked Priming. The State of the Art, 241−278. New York: Psychology Press. Spalding, Thomas L. and Christina L. Gagné 2008 CARIN theory reanalysis reanalyzed: A comment on Maguire, Devereux, Costello, and Cater (2007). Journal of Experimental Psychology: Learning, Memory, and Cognition 34: 1573−1578. Spalding, Thomas L. and Christina L. Gagné 2011 Relation priming in established compounds: Facilitation? Memory and Cognition 39: 1472−1486. Spalding, Thomas L., Christina L. Gagné, Allison C. Mullaly and Hongbo Ji 2010 Relation-based interpretation of noun-noun phrases: A new theoretical approach. In: Susan Olsen (ed.), New Impulses in Word-Formation, 283−315. Hamburg: Buske. Štekauer, Pavol 2005 Meaning Predictability in Word Formation. Novel, Context-Free Naming Units. Amsterdam/Philadelphia: Benjamins. Štekauer, Pavol 2006 On the meaning predictability of novel context-free converted naming units. Linguistics 44: 489−539. Štekauer, Pavol 2009 Meaning predictability of novel context-free compounds. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 272−297. Oxford: Oxford University Press. Taft, Marcus 1994 Interactive-activation as a framework for understanding morphological processing. Language and Cognitive Processes 9: 271−294. Taft, Marcus 2003 Morphological representation as a correlation between form and meaning. In: Egbert M. H. Assink and Dominiek Sandra (eds.), Reading Complex Words. Cross-Language Studies, 113−137. Amsterdam: Kluwer. Taft, Marcus and Paul Kougious 2004 The processing of morpheme-like units in monomorphemic words. Brain and Language 90: 9−16. Warren, Beatrice 1978 Semantic Patterns of Noun-Noun Compounds. Gothenburg: Acta Universitatis Gothoburgensis.

64. Gender marking

1159

Zwitserlood, Pienie 1994 The role of semantic transparency in the processing and representation of Dutch compounds. Language and Cognitive Processes 9: 341−368. Zwitserlood, Pienie, Jens Bölte and Petra Dohmes 2002 Where and how morphologically complex words interplay with naming pictures. Brain and Language 81: 358−367. Zwitserlood, Pienie, Agnes Bolwiender and Etta Drews 2005 Priming morphologically complex verbs by sentence contexts: Effects of semantic transparency and ambiguity. Language and Cognitive Processes 20: 395−415.

Christina L. Gagné and Thomas L. Spalding, Edmonton (Canada)

64. Gender marking 1. 2. 3. 4. 5. 6.

Defining the problem Characterization of the category Composition Derivation Problems related to female marking References

Abstract Gender marking has been in the focus of linguistic research since the 1960s, and especially since the feminist critique of language of the 1970s. The article presents the central theoretical problems linked to gender marking, as well as providing an overview of the language-internal and cross-linguistic variability of the category.

1. Defining the problem Gender marking is a way of explicitly signalling that a linguistic expression refers to a male or female being (person or animal). This can be achieved by various linguistic means, e.g., attributive adjectives as in the phrase male nurse, female kangaroo, appositions such as in French madame le premier ministre ‘madam prime minister’ and, last but not least, by word-formation: compounding, as in Turkish erkek öğretmen man teacher ‘male teacher’, German Papageienweibchen ‘parrot female’ or affixation as in Italian attrice ‘actress’. The crucial point shared by all these examples is that the semantic feature of gender be signalled by a recurrent and identifiable marker, since not all genderspecific nouns denoting persons carry such a marker. Thus, e.g., the Turkish kinship terms oğul ‘son’ or kız ‘daughter’ are gender-specific by means of their semantics, not

1160

VII. Semantics and pragmatics in word-formation II: Special cases

their form or grammatical features (since there is no grammatical gender in Turkish), and therefore do not fall under gender marking as defined here. On the other hand, the same kinship terms in Italian figlio ‘son’ and figlia ‘daughter’ signal their gender-specificity by the endings -o and -a, respectively, and so are included in the definition. In the present article we will be concerned only with gender marking by word-formation. We will mainly deal with the domain of nouns denoting persons and only touch upon nouns denoting animals. This focus on the former is due to the fact that the topic of gender marking in personal nouns is much more prominent in both the literature and public awareness. In particular, since the feminist critique of language from the 1970s onward, gender marking has continuously received attention and triggered substantial research within grammatical theory, studies of language use and psycholinguistics. Speaking of gender, we first have to clarify some terminological issues. In their seminal handbook Gender Across Languages Hellinger and Bußmann (2001−03 Vol. 1: 5−11) distinguish four distinct “categories of gender” relevant to linguistics: social, referential, lexical and grammatical gender. Lexical gender is defined as the lexical specification of nouns “as carrying the semantic property [female] or [male], which may in turn relate to the extra-linguistic category of referential gender (or “sex of referent”)” (ibid.: 7), as, e.g., in kinship terms, such as mother or father. Social gender refers to the semantic bias of an otherwise unspecified noun towards one or the other gender, e.g., nurse denoting stereotypically female persons and surgeon male ones. Referential gender, in turn, is defined as relating “linguistic expressions to the non-linguistic reality” (ibid.: 8). In such a way, a term which is lexically specified for female gender, such as Mädchen ‘girl’ in German Mädchen für alles ‘maid of all work’, may in metaphorical use refer to a specific man (ibid.: 8). Similarly, a lexically unspecified noun as Russian vrač ‘physician’ may refer to a specific woman, indicated by the hybrid agreement of the predicate, as in vrač skazal-a ‘the physician:m said-f ’. Note that vrač usually has masculine grammatical gender, but here may trigger feminine agreement for semantic reasons. The latter example leads us immediately to the intricate question of how lexical and referential gender, and thus also gender marking by word-formation devices, are related to grammatical gender. Grammatical gender is defined as a classificatory feature of all nouns of a language that is obligatorily signalled by agreement (cf. Corbett 1991: 4). Moreover (following Hellinger and Bussmann 2001−03 Vol. 1: 5−6 in contradistinction to Corbett 1991: 5), grammatical gender will here be distinguished from noun class by the correspondence between class membership and the lexical specification of nouns as male-specific and female-specific. Thus, the term grammatical gender will be reserved for languages which have a comparatively small number of noun classes (usually 2 to 4) that are semantically related to maleness and femaleness (as in many Indo-European languages), whereas the term noun class will be applied to languages where classification relies on other semantic principles, e.g., humanness, and where we usually observe a greater number of classes (up to 20, such as in the Bantu languages). Grammatical gender is an inflectional category, but is closely related to gender marking in the sense adopted here, and sometimes even substitutes for genuinely word-formational means as in the case of gender conversion (cf. section 4). From a typological point of view, gender marking is much more common in gender languages than in languages lacking grammatical gender, particularly as far as gender marking by derivation is concerned. The very existence in a given language of the

64. Gender marking

1161

grammatical category of gender seems to force the concept of a lexical and referential gender dichotomy on nouns denoting persons, whereas gender marking in languages without grammatical gender is facultative and context dependent. E.g., ethnonyms as American usually exist in a masculine and a feminine form in gender languages, cf. German Amerikaner m, Amerikanerin f, whereas in languages which lack grammatical gender (or where gender is expressed only in pronouns, as in the case of English), there is just one expression as in Turkish Amerikalı. This indicates that grammatical gender is itself a marker of lexical and referential gender. However, the semantics of grammatical gender presents an intricate problem that has given rise to significant controversy and has a turbulent history within grammaticography (cf., e.g., Doleschal 2002; Roca 2005: 21−25). The most influential position until today is Roman Jakobson’s, stipulating that the neuter gender signals asexuality, the feminine gender the sexedness of the denoted object and the masculine gender signals neither. In Jakobson’s (1971: 184) words: “The masculine is a twice unmarked gender. Contrary to the neuter, it signals neither the asexual character of the entity named, nor, in contradistinction to the feminine, does it carry any specification of the sex”. Given this view, it is not surprising that gender marking is often treated as “formation of the feminine” (as observed by Thornton 2004: 241). This point of view is also due to the fact that nouns denoting females are often morphologically dependent on semantically parallel nouns denoting males, e.g., German Lehrer-in f ← Lehrer m ‘teacher’, where the whole masculine agent noun is the base of the feminine correlate and the feminizing suffix bears merely the meaning ‘female’. I.e. the feminizing suffix -in is attached to a masculine base which is itself a free form. The fact that the opposite type, the formation of nouns denoting males from feminine ones, is very rare is another reason why gender marking is often seen as “female marking”. One of the controversial questions related to gender marking is whether the base from which the gender-marked noun is derived is itself gender-specific or not − in other words, is the base gender-indefinite, as argued by Jakobson (and many linguists after him, e.g., Kalverkämper 1979, Krongauz 1996 and still Roca 2005−06)? The behaviour of masculine personal nouns is ambiguous in this respect: on the one hand, they are mostly used to refer to male persons when they are used in the singular and in a referential NP, e.g., Italian Il professore arriva ‘the professor (a man) is arriving’. On the other hand, they are widely used in a gender-indefinite sense in non-referential NPs such as Non vedo nessun professore ‘I do not see any professor’ and especially in the plural i professori della facoltà di lettere ‘the professors of the faculty of Humanities’ (cf. Doleschal 1995). All such phenomena are subsumed under the term “generic masculine”. The generic masculine can be treated in two ways: 1) In the Jakobsonian, structuralist tradition the unmarkedness of the masculine is related to a gender-indefinite semantics, the gender-specific interpretation is brought about by the context: if there is a female noun, the opposition [male]:[female] becomes salient. 2) In the feminist tradition the masculine gender is seen as signalling male gender, since psycholinguistic evidence supports this interpretation as the more plausible and basic one, cf. Gygax et al. (2012). Under this premise, the generic masculine can be explained as a kind of synecdoche where the prototypical case (male) stands for the whole category (human). This interpretation is also in line with cognitivist and naturalist theories of language. The present description follows the latter approach, because abstraction is more easily explained than its oppo-

1162

VII. Semantics and pragmatics in word-formation II: Special cases

site. Thus, masculine personal nouns are understood as inherently ‘male’, gender-indefinite uses being derived from this basic meaning by means of abstraction. It is noteworthy that generic human terms tend to be used and interpreted as denoting male beings in non-gender languages as well. In such languages most nouns denoting human beings have no lexical gender and − for the lack of grammatical gender − no inherent classification that would lead to gender-specific interpretations. In such languages social gender comes into play, so that some personal nouns are typical for male and female domains, respectively, and so are associated more strongly with one or the other gender. However, in order to make the interpretation as male or female explicit, such nouns have to be combined with overt gender markers, e.g., kadın doktor woman doctor ‘female doctor’ in Turkish. Interestingly, gender marking is much more common with reference to women in these languages as well, i.e. gender marking is used for feminization, not masculinization (cf. Braun 2000: 319, 2001: 287−295, and the other relevant articles in Hellinger and Bußmann 2001−03). In order to denote men, bare nouns denoting persons are sufficient (provided that social gender does not indicate the contrary). Thus, the general unmarkedness of the masculine grammatical gender and its subsequent “gender-ambiguity” (Nissen 2002) is paralleled by the unmarkedness of male referential gender in non-gender languages and leads to a similar kind of distribution of personal nouns: The general (“human”) term is identified with maleness, femaleness is the exception that has to be made explicit, e.g., Indonesian dokter ‘doctor’ (both genders and male) vs. dokter perempuan doctor woman ‘female doctor’ (Kuntjara 2001). This identification of humans in general with men has also been found in psycholinguistic research (Braun et al. 1998) and seems to be widespread in semantic change in the sense that nouns with the original meaning ‘human being’ (as Turkish adam, German Mann) undergo semantic narrowing and acquire the meaning ‘male human being’. It is this asymmetry in gender marking − both in gender languages and languages without grammatical gender − that is pointed out by the feminist critique of language. Indeed, the opposition between the referential genders [male] and [female] is an equipollent one with a distinct hyperonym [human]. On the morphological level, however, it is very often instantiated as a privative opposition, where [male] is the unmarked member and thus expressions associated with this semantic value also function as hyperonyms (cf. Waugh 1982).

2. Characterization of the category As has been mentioned, gender marking can be achieved by lexical, syntactic and wordformational means and it is not always easy to delimit word-formation, some phenomena being transitional. On the semantic side, gender marking means that a specific word-formation rule has the word-formational meaning ‘male X’ or ‘female X’. This word-formational meaning can be signalled by a noun or pronoun with male or female lexical gender, e.g., man, woman, he, she as in English chairman, chairwoman, he-dog, she-dog or by affixes, e.g., Italian casaling-o m ‘housewife-m:M’ (male housewife ← casaling-a f ‘housewifef:F’) or German Sänger-in f ‘singer-F’ (← Sänger m) and, last but not least, by gender

64. Gender marking

1163

agreement as in Italian l-a cantante ‘the-f singer’, il cantante ‘the-m singer’. In the case of derivation, it is usually the affix which carries the lexical gender information. However, there are also cases of zero-derivation, where the lexical feature [female] or [male] remains covert. The latter is possible only in languages with grammatical gender, since the change in lexical meaning is in this case signalled by the agreeing elements (as well as, in some cases, by the inflectional endings of the noun). In the following, lexical maleness or femaleness will be indicated by majuscular M, F, whereas minuscular m, f stand for masculine and feminine grammatical gender. Examples in the following sections are taken from the literature cited, especially from the articles in Hellinger and Bußmann (2001−03).

3. Composition One of the word-formational possibilities of marking the gender-semantics of a lexeme is compounding. In this case, a lexically gender-specific morpheme is involved, as man in chairman or lady in lady doctor. Two types of compounds can be distinguished: In cases such as chairman, a gender-specific morpheme is the head of the compound and makes it a noun denoting persons, while the modifier is not a noun denoting persons and may belong to any word class. The semantic relationship between the two elements can be any beside ‘X is also a Y’. In addition to the lexemes for ‘man’ and ‘woman’, ‘mister’ and ‘lady’ as well as kinship terms are candidates for this kind of gender marking, e.g., German Kindergartentante nursery aunt ‘nursery school teacher’. In languages with productive compounding, this type is a very common means of creating personal nouns denoting professions, functions, titles, etc. such as Finnish lakimies law man ‘lawyer’, Turkish bilim adamı science man ‘(male) scientist’, Danish sportskvinde ‘sportswoman’, Swedish statsman ‘statesman’, which tend to get lexicalized. Lexicalization may blur gender-specificity, which is, however, typical only for nouns with male lexical elements: the German noun Hintermann ‘person behind somebody (in a row)’ or ‘person behind something’, e.g., is hardly understood in a gender-specific way. Nevertheless, the relation to a gender-specific interpretation is never completely lost. This phenomenon criticized by the feminist critique of language (pace Hellinger and Bußmann 2001−03), has raised awareness in some languages and led to political measures of language planning. As a consequence, compounds with the morpheme {man} as head noun have been either substituted by gender neutral terms, as, e.g., English chairman by chairperson, or parallel female forms have been introduced, cf. German Obfrau ‘chairwoman’, Dutch cameravrouw ‘camerawoman’, Turkish bilim kadını science woman ‘female scientist’. Gender marking by compounding also occurs in another variation, where the lexically gender-specific morpheme combines with another noun denoting persons. Following Braun (2000: 67−71) the word-formational semantics of this type of compound is the conjunction of two predicates X and Y, meaning a ‘Y that is also an X’ or an ‘X that is also a Y’. Let us illustrate this with Braun’s examples from Turkish: in the compounds in question, the gender-specific noun can be either in the position of head (being the rightmost element) as in futbolcu kadın lit. ‘football-player woman’ or in the position of modifier as in kadın futbolcu lit. ‘woman football-player’. In the first case, the lexeme ‘woman’ in the head position highlights the gender aspect, denoting a ‘woman who is a

1164

VII. Semantics and pragmatics in word-formation II: Special cases

football player’, whereas in the second case the principal information is that we are dealing with a ‘football player who is a woman’. Finnish shows the same possibilities of gender marking (cf. Engelberg 2002). Another case in point is Italian, where the head noun is on the left: donna magistrato lit. ‘woman judge’ corresponds to English ‘magistrate-woman’ and magistrato donna lit. ‘judge woman’ to English ‘female magistrate’. In Italian, only the head noun is inflected and agreed with, e.g., l-e donn-e magistrat-o lit. ‘the-f.pl women-(f ).pl judge-(m)sg’. The same is true for French (cf. Schafroth 2003). In English and Icelandic, however, we find only the structure “gender specific modifierpersonal head noun” as in English lady doctor and Icelandic kvenprestur ‘woman priest’. This kind of compounding is a convenient form of gender marking in languages which either lack grammatical gender completely (such as Finnish) or in which gender is reflected merely in the pronoun (as in English) (cf. the data in Manzelli 2006). Although further research is needed, it appears that the compounds with the wordformational meaning ‘X is also a Y’ or ‘Y is also an X’ do not often become lexicalized. This indicates the transitional status of such constructions between compounding and apposition, which is also reflected in the literature (e.g., Rainer 1993: 247−249; article 38 on noun-noun compounds in French). Compounding is also used to mark the gender of nouns for animals. These compounds always mean an ‘X that is also a Y’. Usually, the gender-specific morphemes involved in these cases are different from the ones in nouns denoting persons. In English we find compounds with the animal noun in the position of head and a gender-specific element as modifier, e.g., a personal pronoun as in he-dog, she-dog, or a gender-specific (human) name: tomcat, jenny-wren. German displays a compounding type with a gender-specific morpheme as head and the animal noun as modifier: Papageienmännchen ‘parrot male’, Papageienweibchen ‘parrot female’, as does Italian, where the head noun is to the left: papagallo maschio parrot male ‘male parrot’, papagallo femmina parrot female ‘female parrot’. Both types are widespread in the languages of the world as documented by Manzelli (2006: 81−82).

4. Derivation Gender marking by derivation occurs in many languages, but is common and natural in gender languages (where it is sometimes treated as a part of inflection, cf. Dressler 1989, Rainer 1993: 41; see also Jobin 2004: 151−171 for an in-depth discussion and, from another viewpoint, Spencer 2002). Derivational gender marking involves the following types: affixation, conversion and transposition (Manzelli 2006 reports a case of modification (apophony) which is, however, exceptional). From the literature focused on gender marking, we can conclude that gender marking usually occurs as suffixation. The derivation of gender marked nouns can occur independently or in the correlation [male][female], where a female noun is derived from a male one or vice versa. Gender marking by derivation concerns common nouns as well as names. It is productive among nouns denoting persons and nouns denoting animals. This word-formational category was called motio substantivorum in the Latin grammatical tradition, and the term is still used as a convenient label in some languages, such as German (Movierung, Motion), Italian (mozione), Bosnian, Croatian, Serbian (mocija)

64. Gender marking

1165

and others. Originally the ability of adjectives as bon-us m − bon-a f − bon-um n ‘good’ as well as nouns as amic-us m − amic-a f ‘friend’ to mark (grammatical) gender by simply changing the inflectional endings was conceived of as “mobility”, and the words in question termed mobilia (cf. Schad 2007: 249), whereas the term motio is attested at least from the 16th century onwards. In modern linguistics, the terminological equivalents of motio refer to the derivation of a gender-specific noun denoting a person or an animal from a complementary term denoting the other gender, automatically causing a change of grammatical gender, e.g., Czech ministr ‘minister m(:M)’ → ministr-yně ‘ministerF.f ’. Thus, the kind of derivational gender marking called motio is by definition confined to gender languages. The occurrence of derivational gender marking in languages without grammatical gender, by contrast, is rare and often unproductive. The close interrelation between derivational gender marking and grammatical gender supports the conclusion that grammatical gender is in itself an indicator of lexical gender or may at least be used as such. This view is corroborated by the fact that gender changing affixes are specified for one grammatical gender in the lexicon, e.g., German -in for the feminine (Lehrer-in ‘teacher-F.f ’ from Lehr-er teach-AGENT.m:M ‘teacher’).

4.1. Zero-derivation In gender languages gender marking can occur as an instance of zero-derivation, i.e. without the addition of a gender-marking affix, so that lexical gender is signalled only by grammatical gender and (possibly) the inflectional endings of the noun. We distinguish between transposition and gender conversion. Transposition occurs when a part of the inflectional paradigms (i.e. the masculine and the feminine) of an adjective or participle is substantivized. Either can be transponed by itself, if the need to name a male or female person arises, so that there is no principled derivational relationship between masculine and feminine here. However, they can enter into a correlation where the masculine is the unmarked member. If the adjective in question distinguishes masculine and feminine inflectional paradigms, we speak of differential gender. Gender-marked nouns can be derived by substantivizing one gender paradigm, as in the case of Russian rul-ev-oj steer-ADJECTIVEm:M ‘steersman’, which has male social and lexical gender (and no feminine counterpart), gornič-n-aja room-ADJECTIVE-f:F ‘chamber-maid’ with female social and lexical gender (and no masculine counterpart). Besides such cases, we find symmetrical pairs such as russk-ij Russian-m:M and russk-aja Russian-f:F, where the two terms are in an equipollent relationship, i.e. as with kinship terms the masculine form cannot be used to denote a woman. This is not so for Russian bol’n-oj m ‘sick man, sick person’, where the masculine noun is the unmarked member and can thus be used generically. The feminine bol’n-aja f ‘sick woman’ is therefore seen to be derived from the masculine equivalent (cf. Protčenko 1985). The relationship between the two terms is thus asymmetrical. Differential gender is widespread in European languages, cf. German Angestellt-er m/Angestellt-e f, French employ-é m/employ-ée f, Italian impiegat-o m/impiegata f, Spanish emplead-o m/emplead-a f ‘(male)/female employee’, etc. A special case of transposition can be found in languages where not all adjectives and participles have distinct gender paradigms, e.g., Italian features (as does Latin) a

1166

VII. Semantics and pragmatics in word-formation II: Special cases

class of adjectives and participles which do not show distinct forms of grammatical gender, e.g., cant-ant-e sing-PARTICIPLE-m/f.sg. Thus il/la cantante m/f ‘male/female singer’ is a word that was formed on the basis of a transponed adjective. This example has inherited both genders and is therefore a noun of common gender. Again, this is not always the case, e.g., the older lexeme studente m ‘student’ was substantivized only in the masculine gender. With regard to gender marking as defined at the outset, the only marker here is the gender feature, which is not signalled by the noun itself, but still overtly by agreement. Common gender is also found in original nouns, such as Russian sud’ja m/f ‘judge’. Originally masculine or feminine nouns can acquire a second gender feature, as is the case with Italian il presidente m ‘the president’, which can also be used in a feminine version la presidente f (although this is not accepted unanimously, see Villani 2012). Such processes occur in languages that feature common gender in the strict sense (one lexeme/two agreement patterns), such as Italian and Russian. However, although this is a case of gender marking, it clearly does not belong to word-formation and so will not be pursued further here. In gender conversion, gender-marked nouns are formed by assigning an inflectional class that is prototypically bound to a certain grammatical gender to a stem. Usually gender conversion occurs as motio, i.e. a gender-marked lexeme is derived from another, signalling the opposite lexical gender. The change in lexical and grammatical gender is brought about without any overt word-formational suffix, merely by changing the inflectional class as in Hebrew saxkan-it f ‘actress’ ← saxkan m ‘actor’, Moroccan Arabic katib-a f ‘female secretary’ ← katib m ‘(male) secretary’, Italian figli-a f ‘daughter’ ← figli-o m ‘son’, infermier-a f ‘woman nurse’ ← infermier-e m ‘male nurse’. Note that the suffixes -it, -a, -o, -e are inflectional endings of distinct inflectional classes and not derivational suffixes. In the Hebrew and Arabic cases they signal both gender and number, in the Italian case -o, and -a (besides signalling the singular) are associated with masculine and feminine by default, while -e is ambiguous as to gender (Thornton 2004). This type of gender marking is the only one reported for Hebrew and Moroccan Arabic, where it is so regular that it appears to be on the verge of inflection. It is also very productive in Italian and Spanish, and although most often the feminine is derived from the masculine noun, we also find opposite cases such as Italian casaling-o m ‘male housewife’ ← casaling-a f ‘housewife’, Spanish prostitut-o m ‘male prostitute’ ← prostitut-a f ‘prostitute’. Gender conversion is also used to derive gender-marked nouns for animals, cf. Hebrew sus-ah f ‘mare’ ← sus m ‘stallion, horse’, Italian pinguin-a f ‘female penguin’ ← pinguin-o m ‘penguin’. Note that we distinguish this kind of gender conversion from hybridization (Corbett 1991: 183) as in Spanish l-a médico lit. ‘the-f physician:m’, Russian vrač skazal-a lit. ‘doctor:m said-f’, where a masculine noun is used with a feminine modifier or predicate but remains indeclinable. Hybridization is close to common gender, but usually brings about a restriction of the paradigm either to only the singular (Spanish) or to only one or two cases in the singular (Russian). We shall therefore not include it in our discussion of word-formation.

64. Gender marking

1167

4.2. Affixation Gender marking by affixation is arguably the most varied and best-studied means of gender marking. The most widespread type is motio, but gender-marked nouns can also be derived independently of a correlative noun, in languages both with and without grammatical gender, depending on the semantics of the respective affix. For instance, the Finnish suffix -kko/kkö derives relational nouns with female lexical gender, e.g., sisä-kkö ‘parlour-maid’ (from sisä ‘interior’), the Italian suffix -trice derives femininefemale agent nouns from verbs (also independently of its masculine counterpart -tore), e.g., ricama-trice embroider-AGENT.F.f ‘embroideress’, and the Russian patronymic suffix -ovna derives feminine-female patronyms from male names, e.g., Ivan-ovna IvanPATRONYM.F.f ‘Johnsdaughter’. And, of course, in gender languages almost all masculine nouns denoting persons are derived independently of any possibly occurring feminine counterparts by corresponding affixes from verbs, adjectives, nouns, etc., as, e.g., German Lehr-er teach-AGENT.m ‘(male) teacher’. However, these masculine suffixes are usually not seen as gender-marking, since the majority of nouns derived by them can be used gender-indefinitely in the sense of Jakobson’s analysis. In linguistic accounts of word-formation, they are usually treated as deriving “nouns denoting persons”. Exceptions are derivations such as German Witw-er m ‘widower’ ← Witwe f ‘widow’ which cannot take on the gender-indefinite reading. Gender marking by affixation is rare in non-gender languages, but nevertheless occurs. The bases for gender marking are nouns denoting persons, which are usually unspecified for lexical gender, e.g., Turkish sahib-e ‘female owner’ ← sahip ‘owner’, but there are also cases where correlative pairs of male-female are formed, e.g., the Hungarian feminizing suffix -né which is attached to names and means ‘wife of ’. Often the suffixes in question are borrowed from prestigious languages, e.g., Arabic for Turkish. Since gender-marking by affixation occurs most often as motio in gender languages, we will now concentrate on this phenomenon. As usually practised in the literature on gender marking, we will also subsume those cases which may be analyzed as independently derived pairs of nouns, but by their correlation show an asymmetrical semantic relationship. For motio by affixation a lexically specified gender affix is attached to a base with the opposite gender value, normally a feminine-female suffix to a masculine-male base, replacing the semantic feature [male] by the feature [female] or, in a strictly Jakobsonian approach, adding the same feature to a gender-indefinite meaning. Even under such a premise, however, generic masculine nouns must have the potential to refer to male beings only (a point that Roca 2005−06 fails to demonstrate). Nissen (2002: 256−258) points out rightly that not all masculine nouns are semantically male to the same degree, i.e. lexical gender can be more or less strong. The degree of maleness may also vary across languages, depending on the regularity of the male-female correlation among other factors. The gender marking affix can be added concatenatively to the whole (simplex or complex) base, e.g., Polish szef-owa f ← szef m ‘boss’, nauczy-ciel-ka f ← nauczy-ciel m ‘teach-er’, or replace another affix as in Russian ukrain-ka f ← ukrain-ec m ‘Ukrainian’. Affixation may also cause morphonological changes in the base: Czech filolož-ka f ← filolog m ‘philologist’, German Ärzt-in f ← Arzt m ‘physician’. Correlates may also show stem-suppletion as in Italian possedi-trice f ← posses-sore m ‘possessor’.

1168

VII. Semantics and pragmatics in word-formation II: Special cases

Nouns denoting animals can be gender-marked with the help of the same affixes or may have specialized ones such as German -erich m, which is used to derive lexically male nouns from feminine(-female) ones, e.g., Gäns-erich m ‘male goose’ ← Gans f ‘goose’. In Russian the affix -ixa f is the only productive one for deriving nouns denoting female animals (slon-ixa f ‘female elephant’), but may also derive nouns denoting persons, if only in colloquial style and producing a derogatory connotation. Languages vary as to the number of gender marking affixes: German, e.g., displays one productive female suffix -in f and four unproductive ones (Doleschal 1992), whereas Serbo-Croatian had four productive and eight unproductive female ones (Ćorić 1982). Accordingly, affixes may be restricted to a certain base-type, as French -euse f is to derivations with -eur m, or they may combine with various bases, as German -in can be attached to almost any masculine noun denoting persons (see Doleschal 1992: 36−38 for systematic restrictions). On the other hand there may be competing affixes for one and the same base-type as in the case of Bosnian -ica f and -ka f, deriving doktor-ica f as well as doktor-ka f from doktor m ‘doctor’, sometimes leading to stylistic or connotational differences between the derivations (Šehović 2003). Languages also vary as to the productivity and acceptability of motio. In Czech, e.g., “a feminine counterpart may be formed practically from any masculine form” (Čmejrková 2003: 41; Schwarz 1999: 122), whereas in Polish this is definitely not the case (Miemietz 1993).

5. Problems related to female marking In most languages reported on in the literature, female-marked nouns are valued less than their male counterparts. This fact has various consequences: The use of female-marked derivations is avoided, especially in official style. As a consequence, the formation of female nouns is avoided, too, especially if a prestigious role is designated, excluding, e.g., Russian *predsedatel’-nica f ← predsedatel’ m ‘chairman’, thus causing idiosyncratic exceptions to otherwise productive word-formation rules. Another consequence is hybridization, as in Spanish la médico ‘the-f physician:m’, which “looks” more masculine than the regular derivation médic-a ‘physician-F.f’. At the same time, female marking is usually unrestricted in colloquial language. This may lead to the classification of single formations (e.g., Russian načal’-nica ‘woman boss’) or certain suffixes (Russian -ša, -ixa) as colloquial. In turn, the use of such formations to denote women may be perceived as derogatory, but occasionally also as endearing, since less formal. Another variation is the possible semantic differentiation between the masculine and corresponding feminine term, cf. Mozdzierz (1999: 169 and 175) for Russian: “Some feminine agentives differ from their masculine counterparts in that their meanings are far narrower […]. While feminine sekretarša is limited to a low-to-mid-level secretary in an office, the masculine sekretar’ may refer either to an administrative or academic secretary, leader, or supervisor.” And: “the feminine agentive učitel’nica occurs with […] modifiers […] which imply lack of experience. By contrast, masculine učitel’ refers to women who are outstanding in their profession”. Lastly, some female-marked nouns may convey the meaning of “marital counterpart” (ibid.: 169). Such problems of attitude and language use can be successfully eliminated by consistent language politics, as in the case of English and German (cf. Romaine 2001; Elmiger

64. Gender marking

1169

2008: 354), because we are not dealing with morphological restrictions here (Miemietz 1993: 18; Mozdzierz 1999: 177; Haase 2010: 71) − even if they are sometimes presented as such.

6. References Braun, Friederike 2000 Geschlecht im Türkischen. Untersuchungen zum sprachlichen Umgang mit einer sozialen Kategorie. Wiesbaden: Harrassowitz. Braun, Friederike 2001 The communication of gender in Turkish. In: Marlis Hellinger and Hadumod Bußmann (eds.), Gender Across Languages. Vol. 1, 283−310. Amsterdam/Philadelphia: Benjamins. Braun, Friederike, Anja Gottburgsen, Sabine Sczesny and Dagmar Stahlberg 1998 Können Geophysiker Frauen sein? Generische Personenbezeichnungen im Deutschen. Zeitschrift für germanistische Linguistik 26: 265−283. Čmejrková, Světla 2003 Communicating gender in Czech. In: Marlis Hellinger and Hadumod Bußmann (eds.), Gender Across Languages. Vol. 3, 27−57. Amsterdam/Philadelphia: Benjamins. Corbett, Greville G. 1991 Gender. Cambridge: Cambridge University Press. Ćorić, Božo 1982 Mocioni sufiksi u srpskohrvatskom jeziku. Beograd: Filološki fakultet. Doleschal, Ursula 1992 Movierung im Deutschen. Eine Darstellung der Bildung und Verwendung weiblicher Personenbezeichnungen. München: LINCOM Europa. Doleschal, Ursula 1995 Referring to women. In: Richard A. Geiger (ed.), Reference in Multidisciplinary Perspective. Philosophical Object, Cognitive Subject, Intersubjective Process, 277−298. Hildesheim: Olms. Doleschal, Ursula 2002 Das generische Maskulinum im Deutschen: Ein historischer Spaziergang durch die deutsche Grammatikschreibung von der Renaissance bis zur Postmoderne. Linguistik online 11(2): 39−70. http://www.linguistik-online.de/11_02/ [last access 24 Nov 2014]. Dressler, Wolfgang U. 1989 Prototypical differences between inflection and derivation. Zeitschrift für Sprachwissenschaft und Kommunikationsforschung 42: 3−10. Elmiger, Daniel 2008 La féminisation de la langue en français et en allemand. Querelle entre spécialistes et réception par le grand public. Paris: Honoré Champion. Engelberg, Mila 2002 The communication of gender in Finnish. In: Marlis Hellinger and Hadumod Bußmann (eds.), Gender Across Languages. Vol. 2, 109−132. Amsterdam/Philadelphia: Benjamins. Gygax, Pascal, Ute Gabriel, Arik Lévy, Eva Pool, Marjorie Grivel and Elena Pedrazzini 2012 The masculine form and its competing interpretations in French: When linking grammatically masculine role names to female referents is difficult. Journal of Cognitive Psychology 24: 395−408. Haase, Peter 2010 Feminisierung im spanischen Sprachraum. Berufs-, Amts- und Funktionsbezeichnungen: El juez, la juez, la jueza? Hamburg: Kovač.

1170

VII. Semantics and pragmatics in word-formation II: Special cases

Hellinger, Marlis and Hadumod Bußmann (eds.) 2001−03 Gender Across Languages. 3 Vol. Amsterdam/Philadelphia: Benjamins. Jakobson, Roman 1971 The gender pattern of Russian. In: Selected Writings. Vol. 2: Word and Language, 185− 86. Den Haag: Mouton. Jobin, Bettina 2004 Genus im Wandel. Studien zu Animatizität anhand von Personenbezeichnungen im heutigen Deutsch mit Kontrastierungen zum Schwedischen. Stockholm: Almqvist & Wiksell. Kalverkämper, Hartwig 1979 Die Frauen und die Sprache. Linguistische Berichte 62: 55−71. Krongauz, Maksim A. 1996 Sexus, ili problema pola v russkom jazyke. In: Vladimir Toporov (ed.), Rusistika. Slavistika. Indoevropeistika, 510−525. Moskva: Indrik. Kuntjara, Esther 2001 Gender in Javanese Indonesian. In: Marlis Hellinger and Hadumod Bußmann (eds.), Gender Across Languages. Vol. 1, 199−228. Amsterdam/Philadelphia: Benjamins. Manzelli, Gianguido 2006 Il genere nelle lingue senza genere. In: Silvia Luraghi and Anna Olita (eds.), Linguaggio e genere. Grammatica e usi, 72−88. Roma: Carocci. Miemietz, Bärbel. 1993 Motivation zur Motion. Zur Bezeichnung von Frauen durch Feminina und Maskulina im Polnischen. Frankfurt/M.: Lang. Mozdzierz, Barbara M. 1999 The rule of feminization in Russian. In: Margaret Mills (ed.), Slavic Gender Linguistics, 165−182. Amsterdam/Philadelphia: Benjamins. Nissen, Uwe-Kjœr 2002 Gender in Spanish: Tradition and innovation. In: Marlis Hellinger and Hadumod Bußmann (eds.), Gender Across Languages. Vol. 2, 251−279. Amsterdam/Philadelphia: Benjamins Protčenko, Ivan F. 1985 Leksika i slovoobrazovanie russkogo jazyka sovetskoj ėpoxi. 2nd ed. Moskva: Nauka. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Roca, Ignacio M. 2005−06 La gramática y la biología en el genero del español. Revista Española de Lingüística 35(1): 17−44 and 35(2): 397−432. Romaine, Suzanne 2001 A corpus-based View of Gender in British and American English. In: Marlis Hellinger and Hadumod Bußmann (eds.), Gender Across Languages. Vol. 1, 153−175. Amsterdam/Philadelphia: Benjamins. Schad, Samantha 2007 A Lexicon of Latin Grammatical Terminology. Pisa: Serra. Schafroth, Elmar 2003 Gender in French. Structural Properties, Incongruences and Asymmetries. In: Marlis Hellinger and Hadumod Bußmann (eds.), Gender Across Languages. Vol. 3, 87−117. Amsterdam/Philadelphia: Benjamins. Schwarz, Jana 1999 Die Kategorie der WEIBLICHKEIT im Tschechischen. Die Verwendung femininer und maskuliner Personenbenennungen für Frauen im Vergleich zum Deutschen. Praha: Univerzita Karlova.

65. Singulatives

1171

Šehović, Amela 2003 Upotreba mocionih sufiksa (u nomina agentis et professionis) u savremenom razgovornom bosanskom jeziku. Pismo 1(1): 73−92. Spencer, Andrew 2002 Gender as an Inflectional Category. Journal of Linguistics 38: 279−312. Thornton, Anna M. 2004 Mozione. In: Maria Grossmann and Franz Rainer (eds.), La formazione delle parole in italiano, 218−227. Tübingen: Niemeyer. Villani, Paola 2012 Le donne al parlamento: Genere e linguaggio politico. In: Anna M. Thornton and Miriam Voghera (eds.), Per Tullio de Mauro. Studi offerti dalle allieve in occasione del suo 80° compleanno, 317−339. Roma: Aracne. Waugh, Linda 1982 Marked and unmarked: A choice of unequals in semiotic structure. Semiotica 38(3/4): 299−318.

Ursula Doleschal, Klagenfurt (Austria)

65. Singulatives 1. 2. 3. 4. 5.

Introduction Delineating the category Unitizing Singulatives and typology References

Abstract Singulatives are derived unit nouns. Their varieties revolve around a core notion of a map between nominalizing morphology and unitizing value, varying across languages in semantic latitude and relation to grammatical number, as illustrated by several examples. The meaning of unitizing is elucidated, in relation to packaging and individualizing. A final discussion places singulatives in the context of the typology of strategies for dividing reference.

1. Introduction An expression meaning ‘many x’ is not necessarily based, in form or in meaning, on a corresponding expression meaning ‘one x’. If a unit-denoting noun is morphologically derived from a more basic non-unit-denoting noun, the derived noun, its grammatical

1172

VII. Semantics and pragmatics in word-formation II: Special cases

category, the individualizing marker, or the individualizing derivation, are often called singulative. The creation of this term is attributed to Johann Caspar Zeuss (see Cuzzolin 1998), who in his Grammatica Celtica (1853: 299) adopted it to describe the derived singular of those nouns that in the Brittonic branch of Celtic express the plural with a bare stem and the singular with a suffixed stem: (1)

‘trees’ gwydd gveyth gwez

‘tree’ gwydd-en gveth-en gwez-enn (Pedersen 1913: 70)

Welsh Cornish Breton

What is distinctive about singulatives is that the derivative nature of the singular form corresponds to a derivative conceptualization: the unmarked plural form is also conceptually basic, and describes an entity as a discrete aggregate. Derivatively, the singulative describes the parts of this aggregate as individual atomic entities, denumerable and identifiable. Like its opposite collective, the term singulative is used in different senses, all revolving around a key notion of a unit-denoting noun morphologically derived from a more basic non-unit-denoting noun.

2. Delineating the category 2.1. Morphological and semantic dimensions The concept of singulative captures the intersection of two fundamental properties: morphologically, the derivation of singular from non-singular word forms; and semantically, the derivation of unit-denoting expressions from bases which denote aggregates of such units. I will call this semantic function “unitizing”. The term packaging refers instead to the segmentation of a non-atomic domain into units, as in Dutch bier ‘beer’ − bier-tje ‘a beer’. In languages where the two conceptualizations correlate with different grammatical properties, packaging turns mass nouns into count nouns. In the most restrictive use of the term, singulatives are the output of derivational morphology with a unitizing value which turns a description of many units into a description of one unit. In a broader sense, singulatives are the output of morphology with a packaging value where the input is not an aggregate of distinct units. The two functions are clearly distinct notionally, but are often brought about by the same morphological markers, with an overlap that varies substantially across languages. In addition, the same morphology may derive individual-denoting nouns not just from a plural- or mass-denoting base, but from a base that is not a noun. Particularly common is the use of the same morphology in packaging and evaluative function (cf. Jurafsky 1996: 555). Singulatives, then, are not just unit nouns, but unit nouns whose singular form is derived from a non-singular form by morphology with this specific function. By contrast, cow, for instance, is not morphologically derived from cattle, although semantically it stands to the collective in almost the same relation as the Arabic singulative baqar-a

65. Singulatives

1173

stands to the collective baqar ‘cattle, livestock’. The phrase piece of furniture is in a sense the singulative of furniture, but it is not a word, much less one derived from the collective. Using the term in this purely notional sense would lump together singulatives proper with all other linguistic strategies for identifying a unit out of a collection: by distinct lexical items, as in person − people; by different interpretations of the same lexical item, as in the count vs. mass readings of hair; or by syntactic means, as in periphrases like a piece of furniture or in classifier constructions. The adjective singulative describes the unitizing value of a linguistic expression or operation; singulative nouns, or just singulatives, can more specifically refer to individual-denoting nouns derived by unitizing morphology. Across languages, singulatives vary along several dimensions. They can represent a class of singular nouns, or they can occur in both number values. Unitizing morphology can be more or less regular, its application more or less productive, and the choice of markers more or less predictable from the form or meaning of the base. The markers involved may or may not have other semantic functions beside unitizing. Finally, a collective-singulative opposition may pervade the whole nominal lexicon, or concern only semantically determined classes of nouns, or a few lexemes. In general, the stronger the mutual implication between morphological exponents and unitizing interpretation, the greater justification there is for viewing singulatives as a category in the grammar of a language − as opposed to a description of the form or meaning of certain nouns.

2.2. Singulatives as a category It is customary to recognize a category of singulative nouns, opposed to underived collectives, in the Brittonic branch of Celtic (Welsh, Breton, Cornish). A morphologically uniform class of unitizing singulatives is also prominent in Arabic (including Maltese), traditionally identified as a specific value of nominal derivation under the name of ’ism l waħda, or nomen unitatis. Singulatives as derived singulars are also a salient feature of the number system of Nilo-Saharan languages (Dimmendaal 2000). The term can also denote nouns overtly marked as singular, opposed not just to overt collectives but also to an unmarked “general” form compatible with singular or plural reading, most typically in Cushitic languages. More broadly, many languages have derivational morphemes that can be called singulatives for their unitizing or packaging value, as in Tariana (Aikhenvald 2006: 171), Itelmen (Georg and Volodin 1999: 105), or Burushaski (Berger 1998: 39). The term can also indicate the singular interpretation of bare forms unmarked for number, but this no longer refers to a relation between morphology and semantics. Gender reassignment from masculine to feminine has a singulative function in Chadic languages, where a noun with generic or collective value is reinterpreted as individualdenoting (Newman 1990: 134); a similar function is performed by reassignment to class 5 in Swahili (Contini-Morava 1999). Welsh illustrates a language in which singulatives constitute a category in the grammatical system. This language has an inflectional number category with a straightforward singular-plural opposition, operative on agreement controllers and targets (cf. Corbett 2000). Most nouns have a basic singular form, from which the plural is derived by a variety of morphological means. There is however a class of nouns where the marked-

1174

VII. Semantics and pragmatics in word-formation II: Special cases

ness relation is inverted: as shown in (1), the plural is basic and the singular is suffixed. While these nouns are a “relatively small number” (King 1993: 67), they are too many to be regarded as lexical exceptions. What characterizes them as a local regularity, rather than a list of irregularities, is the systematic match of a uniform exponent with a constant interpretation: each singulative noun is suffixed by -yn (masculine) or -en (feminine), where the choice is determined by the gender of the base noun, and denotes an identifiable unit contrasting with the collective aggregate denoted by the base; for instance, King (1993: 67) suggests ‘leaf − foliage / leaves’ as a translation for dail-en − dail. Accordingly, membership in this set of nouns has a clear semantic motivation. Most collectives denote animal and vegetal living beings typically perceived as aggregates: plants (ysgall, ysgall-en ‘thistles’, coed, coed-en ‘trees’), trees (derw, derw-en ‘oaks’) and other particulate aspects of vegetation (cnau, cneu-en ‘nuts’, gwiail, gwial-en ‘twigs’); insects (clêr, cler-en ‘flies’), animals (adar, singulative ader-yn ‘birds’), but also children (plant, plent-yn). Other concepts include gravel (graean, greyen-yn), stars (sêr, ser-en) or lightnings (lluched, lluched-en), and a remarkably small number of artifacts, like clothes (dillad, dilled-yn) and bricks (brics, brics-en; the base is the English bricks, plural). Other bases are not bare stems: llygod ‘mice’ and pysgod ‘fish’ display the same ending -od which forms suffixal plurals for other names of animals, like cath ‘cat’ − cath-od ‘cats’. It seems plausible, however, that this stem-final -od goes back historically to a formant distinct from the plural suffix (cf. Cuzzolin 1998). What matters is that the base for singulative suffixation is not necessarily a bare stem. Finally, there are cases like blod-yn ‘flower’ − blod-au ‘flowers’: here both forms are suffixal, the singulative ending alternating with the plural one.

2.3. Singulatives as singular forms In some languages, singulatives are integrated in the system of number exponence, where they define a particular class of singulars. The clearest and best-studied example comes from Nilo-Saharan languages (Dimmendaal 2000), some of which express a singularplural syntactic opposition via a three-way partition of nouns: basic singulars, suffixed for plural; basic plurals, suffixed for singular; and nouns suffixed in both number values. The three patterns are exemplified by Baale, belonging to the Surmic family: (2) ‘crocodile’ ‘wagtail’ ‘buffalo’

singular kiɲáŋ dʊrsa-jí kʊ´ʊ´wá-n (Baale:

plural kiɲaŋ-ɛ´ dʊrsa kʊ´ʊ´wá-i(t̪ ) Dimmendaal 2000: 224−228)

Forms like dʊrsa-jí are singulative in the formal sense, as they fill the singular cell in the paradigm with a form derived from the form filling the plural cell (forms like kiɲaŋɛ´, which illustrate the opposite derivation, are sometimes called plurative). But this way of forming a singular by adding a marker to the plural is also a semantically unitizing operation, which marks the conceptual and perceptual priority of plural over singular referents for notions like ‘louse’, ‘ant’, ‘bird’, ‘feather’, ‘leaf ’, as well as some occurring

65. Singulatives

1175

in natural sets like ‘finger’ or ‘eye’; basic plurals include human-denoting terms like ‘child’, and intrinsically plural-referring names of populations, like suri-jí − suri ‘Suri person − Suri people’ (Dimmendaal 2000: 220−226). Deriving a description for individual humans from a collective description of a people is a widespread phenomenon which can give rise to collective-singulative oppositions even in languages that otherwise lack morphologized singulatives. A good example is provided by Russian human-denoting nouns like anglič-anin − anglič-ane ‘Englishman − Englishmen’, or dvor-ânin − dvorâne ‘nobleman − noblemen’ (from dvor ‘court’), where the singular ending -ânin derives historically by suffixing a singulative -in ending to a collective form in -ân. It is important to distinguish these cases, in which nouns marked as singulative are the singular in a number opposition, from superficially similar cases where nouns marked as singulative contrast not only with an overtly marked plural, but also with a bare form compatible with a singular or plural interpretation (Corbett 2000: 16−18). The bestknown example is provided by languages in the Cushitic branch of Afroasiatic where we find oppositions like the following (the superscript circle indicates an accented vowel in Afar): (3) ‘gazelle’ ‘guest’

general (sg/pl) híddi’ keesúmma’

plural singulative hiddí-ile’ − − kéesúmm-itʃ a (Oromo: Andrzejewski 1960: 66)

‘beehive’ ‘onion’

guruf båsal

guruf-wa guruf-ta − basål-tu (Afar: Hayward 1998: 627)

‘stone’

kina

−

kin-čo (Sidamo: Moreno 1940: 80)

The gaps are important: Hayward reports that in Afar only some nouns like ‘beehive’ have all three forms, while for Oromo Andrzejewski noted that use of the singulative or of the plural is very rare outside of a few nouns, adding that he only recorded a few human-denoting nouns (plus ‘young bull’) with a singulative, and only one with both a singulative and a plural (‘priest’). If we also consider that the singulative derivation may bring about semantic specialization (Oromo nama ‘person’, singulative nám-itʃa ‘man’; Afar daro ‘grain’, singulative daro-yta ‘loaf of grain’; Hayward 1998: 627), and that it may express not only singularity but definiteness (Andrzejewski 1960: 74) or specificity (Dimmendaal 2000: 238), the conclusion is clear that we are not just dealing with a way to mark the singular value in an inflectional number opposition.

2.4. Singulatives and plural In the Nilo-Saharan and Russian examples we have just considered, a singulative noun fills the singular cell in the paradigm, while the plural is expressed by a base collective form. But this coincidence between the two types of opposition, collective − singulative and plural − singular, is a property of certain number systems or of certain lexemes; it

1176

VII. Semantics and pragmatics in word-formation II: Special cases

is not a property of singulatives per se. Just as in English the collective cattle contrasts not just with the singular cow, but with the whole paradigm cow − cows, morphological singulatives can in principle be singular or plural, if this opposition is operative in the language. Maltese is a well-known illustration (Mifsud 1996: 37): (4) ‘fly’

collective dubbiin

singulative sg dubbiina

singulative pl dubbiniit

Maltese

The corresponding phenomenon in Classical Arabic involves the feminine suffix -a(t), plural -aa(t); the final consonant is dropped in pre-pausal position and the vowel has various realizations in the modern dialects, as in the following examples from Damascus (Cowell 1964: 297, 369): (5) ‘fly’

collective dəbbaan

singulative sg dəbbaan-e

singulative pl dəbbaan-aat

Arabic of Damascus

The singulative thus defines a formally and semantically regular opposition between singular and plural. The non-suffixed collective form also serves routinely to denote pluralities of what the singulative describes as a unit, but it is not “the” plural of the same noun; for the morphological system, it is another noun, and it can itself have alternative plurals − not arising by suffixation, like in the singulative, but by the rearrangement of the stem CV template which is the main expression of morphological oppositions in Semitic: (6) ‘fly’

collective1 dubbiin dəbbaan

collective2 dbiiben dababiin

singulative sg dubbiina dəbbaan-e

singulative pl dubbiniit dəbbaan-aat

Maltese Arabic of Damascus

Whether there is one or more collective for a given singulative, and what semantic distinctions may be expressed by different collectives beyond non-individuated plurality (often, ‘several types’ or ‘a great many’), is ultimately a matter of historical accident. But the word-formation component makes such alternative plurals possible, if not productively so. The same applies to a second type of singulative derivation in Arabic, which derives unit nouns from collective designations for human ethnics (and for the demons called jinn) by means of the ending -iiy. This is the marker of a pattern of relational adjectives traditionally termed nisba ‘relationship’: ‘arab ‘Arabs’ − ‘arab-iiy ‘Arab’, badw ‘bedouins’ − badaw-iiy ‘bedouin’. Here too, some singulatives are input to regular affixal pluralization, and some collective bases have a non-affixal plural: (7)

collective1 ‘Greek’ yuunaan ‘Turk’ turk

collective2 − atraak

singulative sg yuunaan-iiy turk-iiy

singulative pl yuunaan-iiyy-uun Arabic −

Several such plurals may be possible: Wehr (1976) lists ‘uruub, a‘rub, ‘urbaan, a‘raab for ‘arab ‘Arabs’.

65. Singulatives

1177

The collective-singulative opposition also cross-classifies with number in Celtic, most prominently in Breton. While singulatives in Welsh have one of two suffixes, corresponding to the gender of the base, modern Breton has generalized the feminine -enn; this means that the singulative derivation determines a gender value, effectively forming a new noun. The interaction with number confirms this interpretation. First, singulatives typically have a plural; in fact, in the wealth of plural formations of Breton, the pluralization pattern illustrated by sili-enn − sili-enn-ou ‘eel − eels’ stands out for its regularity. In addition, beside canonical collective-singulative pairs like del ‘leaves, foliage’ − delenn ‘leaf ’ or per ‘pears’ − per-enn ‘pear’, unit nouns can be built on suffixed plurals, like pesk-ed-enn from pesk-ed, sg. pesk ‘fish’, or trid-i-enn from trid-i, sg. tred ‘starling’. The disappearance of a bare singular form can lead to a replacive pattern X-enn (sg.) − X-[pl]; in the dialect described by McKenna (1998: 223) -enn alternates with the collective plural suffix -ad, as in gouri-ad ‘roots’ − gouri-enn ‘root’. This can give rise to series like ster ‘star’ [obsolete] − ster-ed − ster-ed-enn − ster-ed-enn-ou, where the last form, the plural of the singulative, signals a degree of individuation that makes it appropriate for entities that can be pointed to one by one, like stars on epaulettes or on a label, more than for the particulate appearance of a starry sky.

2.5. Form and function Even in languages where singulatives have a uniform morphology, the relation between markers and singulative function is rarely one-to-one. For example, Welsh has only two singulative suffixes, deterministically selected by the gender of the base. We can legitimately speak of a unified morphological marking for a unified semantic function; yet the relation is not one-to-one, because, as is often the case, the same suffixes also have a diminutive value (Cuzzolin 1998: 139 cites gron-ynn-yn, diminutive of gron-yn ‘(wheat) grain’, singulative of grawn ‘wheat’). In Breton, the connection between unitizing function and a dedicated morphology is weaker. The singulative suffix is always -enn, but, first, a few collectives are unitized by an alternative archaic construction involving the prefixes pen(n)- or pez- (Trépos 1956: 124, 236); second, -enn can also have a diminutive value; third, -enn is not limited to unitizing collectives, but acts more generally as an individualizing nominalizer. In addition to plurals, -enn can modify singular nouns, like glav-enn ‘raindrop’ from glao ‘rain’, or dour-enn ‘liquid substance, secretion’ from dour ‘water’; it attaches even to nouns describing single discrete entities, like kalon-enn ‘heart-shaped object, core’ from kalon ‘heart’, or karreg-enn ‘rock, boulder’, from karreg ‘rock’ (cf. Trépos 1956: 268−278; Favereau 1997). Crucially, these derivations do not have evaluative function. Finally, -enn can form individual-referring nouns from non-nominal bases: cf. glas-enn ‘lawn, green’ from glas ‘green/blue’, koantenn ‘beautiful woman’ from koant ‘beautiful’, and drailh-enn ‘fragment’ from drailh-a ‘to break’ (Favereau 1997; Trépos 1956: 270); already Zeuss (1853: 301) mentioned the derivation of goulou-en, glossed as ‘candela’, from goulou, glossed as ‘lux’. These uses of -enn are far from regular or productive, and it should be emphasized that often a form in -enn is functionally the only singular, like stered-enn ‘star’ from the obsolete ster. Similar remarks apply to Arabic. Here, the derivation of singulatives involves the deterministic choice of one of two suffixes, -iiy and -a, targets a semantically and formal-

1178

VII. Semantics and pragmatics in word-formation II: Special cases

ly coherent set of collectives, and creates count nouns interpreted as the corresponding units. While this creates a morphologically and semantically well-defined category in the grammar of Arabic, it is a particular use of morphological resources with broader semantic range. The ending -iiy derives relational adjectives in general, not just from collectives. As for -a, which creates feminine nouns, beside unitizing it derives nouns denoting portions of a mass (laħm ‘meat’ − laħm-a ‘piece of meat’); it produces deverbal nominalizations denoting bounded events (sariq ‘stealing’ − sariq-a ‘theft’), identified as a distinct function by the traditional label ’ism l marra (nomen vicis); and it derives singular but not unitized abstract nouns like waqaaħ-a ‘impudence’ from waqaaħ ‘impudent’. In this function, -a can actually combine with the relational affix, to form a complex ending -iyy-a, very productive in contemporary usage (as in ‘arab-iyy-a ‘Arabic civilization’). In fact, this value of -a can bring about an interpretation that is the very opposite of the singulative: maarr ‘pedestrian’ − maarr-a ‘[collectivity of] pedestrians’ (Holes 2004: 153). When the unitizing function is less directly related to a morphological class, it can still make sense to speak of singulative derivations. Wierzbicka (1988: 518) discusses under this rubric Russian formations like goroš-ina ‘pea’ or trav-in-ka ‘blade of grass’, from goroh ‘peas’ and trava ‘grass’. These correspond to canonical singulatives which turn nouns denoting particulate substances into nouns denoting the perceptual atoms of these substances. However, this type of lexical correspondences does not approach the extension of singulative affixation in the vocabulary of Arabic or Breton. In addition, it represents a particular use of a suffix which can have a range of other semantic functions (Townsend 1980: 190). First, -ina can express packaging from a mass, as exemplified by lëd ‘ice’ − l’d-ina ‘ice floe’ or krov’ ‘blood’ − krov-inka ‘drop of blood’, although this is lexically quite restricted. Secondly, the suffix can derive an entity-denoting noun from a verbal stem (razval-it’sâ ‘to collapse’ − razval-ina ‘wreck [person]’, pl. razvaliny ‘ruins’), or turn a general noun into a description for individual events or objects (konec ‘end’ − konč-ina ‘demise’; verh ‘top, summit’ − verš-ina ‘peak’). In a few cases, it also derives a mass noun, naming the edible meat of animals from the animal kind name (baran ‘lamb’ − baran-ina ‘lamb meat’). The label “singulative” in such cases identifies a particular individualizing function of nominalizing morphology. Correspondingly, in Tariana, Aikhenvald (2006: 171) describes both as “singulative” and “individualizing nominalizer” the suffix -seɾi, which has a plural -seni and attaches to suffixed collectives as well as to non-nominal bases: mawaɾi-ne-seɾi ‘one of the snake people’ (snake-PL-SGLT), nu-phune-seɾi ‘my follower, my enemy’ (1sg-accompany-SGLT).

3. Unitizing 3.1. Mass, unity, and individuation To say that singulatives derive a reading ‘one x’ from collectives interpreted as ‘many x’, or that the Welsh dail-en ‘leaf ’ denotes a unit in the denotation of the collective dail ‘foliage, leaves’, effectively restates what one already knows about the meaning of these words, or what is implied by their translation. Intuitively, singulatives describe as selfstanding entities what collectives describe as a mass-like multitude. This cannot just

65. Singulatives

1179

mean that singulatives turn a mass input into a count output. First, it would miss the key fact that the core cases of singulatives individuate not just parts of a mass, but units in a collection; this is what the terminological opposition of “packaging” and “unitizing” is meant to capture, if not to explain. Second, the concept of “mass” is problematic because it conflates a semantic notion of non-atomicity (cumulativity, divisibility) with a cluster of grammatical properties, which vary across languages and do not always align perfectly with each other. But the two are notoriously distinct, as made especially clear by grammatically mass terms which denote discrete units, like furniture (cf. Rothstein 2010). For singulatives, more important than a linguistic mass/count distinction, and cognitively prior to it (Soja, Carey and Spelke 1991), is the conceptual distinction between properties that define a standard of unity and properties that do not. A standard of unity is a way to be ‘one’. But being ‘one’ can only be defined relative to a concept (a point made by Frege 1884). Count nouns like leaf, stalk, branch, tree, wood define criteria of unity, or “built-in modes, however arbitrary, of dividing their reference” (Quine 1960: 91). But singulatives are not just count nouns; they denote entities conceptualized as fully individuated members of a set, each identifiable as distinct from others. Many nouns are grammatically count but lack this property, like units of measure, or property nominalizations with a count syntax (as in a particular intensity). This means that singulatives typically encapsulate a standard of identity (Acquaviva 2008: ch. 4): they conceptualize their referents not as interchangeable tokens but as identifiable individuals, which can be semantically distributed over, and enumerated by unit numerals.

3.2. Outline of a semantic typology The core singulative-collective oppositions involve concepts for entities that tend to be experienced in aggregates made up of atomic units but not numerically quantifiable. When a singulative unitizes a collective, both terms identify the same individual concept, which defines a uniform and stable granularity of the reference domain (contrast pairs like band − player, which define different criteria of unity). But what is primary is the indefinitely-numbered aggregate, formed by indistinguishable tokens of such an individual concept. This conceptualization is mass-like because it does not allow identifying reference to a particular token, not because it blurs the boundaries of discrete elements. A singulative derivation changes this conceptualization and allows identifying reference to individual tokens. Singulatives based on human group denominations represent a slightly different category, where the standard of unity is the cognitively salient notion of a human being, while the collective identifies a group. Another variety is represented by singulatives which describe the units, not of a collection of indefinite cardinality, but of a closed small set of mutually cohesive elements forming a natural complex: typically body parts like teeth, feet, breasts, or paired garments; Trépos (1956: 124, 277) reports the Breton series glin ‘knee’ − daoulin ‘pair of knees’ (a formal dual, denoting a natural set) − penn-daoulin ‘knee’ (a prefixal singulative). Related to this category are compounds with a morpheme leath meaning ‘half ’ to denote a single member of a natural pair: cf. Irish leath bhliain ‘half year’, but leath-shúil ‘one eye’ (as in fear leathshúile ‘one-eyed man’).

1180

VII. Semantics and pragmatics in word-formation II: Special cases

As we have seen, morphemes with a unitizing function can also have a packaging value, naming a bounded object from the name for a substance: cf. Arabic zujaaj-a ‘piece of glass, bottle, flask, drinking glass’ from zujaaj ‘glass’; for Welsh, Cuzzolin (1998: 129) discusses unusual but attested cases like cos-yn ‘piece of cheese’ from caws ‘cheese’. Unlike unitizing, packaging changes the granularity of the reference domain. But if the relation to the base is different, the output is the same as in unitizing: nouns denoting individual entities, as opposed to substances, property nominalizations, abstract units of measure, or indistinguishable members of an aggregate. The fact that location in space and time enhances identifiability probably explains the “concretizing” function of Breton singulatives like lod-enn ‘part (object extended in space)’ from lod ‘part (subdivision)’ (Trépos 1956: 268). What counts, however, is the identifiability of the referent, not concreteness per se; singulatives may refer to entities that can be told from each other but are abstract, like the Breton kred-enn ‘belief ’ (as in ‘the beliefs of the church’) from kred ‘faith’ (cf. Acquaviva 2008: 245). This individualizing function places singulatives on a par with “actualizing” nominalizations that are neither packaging nor unitizing, as the Russian pâtër-ka, from pât’ ‘five’, which may describe a bus, a school mark, or a banknote (cf. German Fünfer, English fiver).

4. Singulatives and typology Unlike other unitizing morphemes like classifiers, singulatives are formants of lexical words. Their functions range from a strictly unitizing value to a broader range of individualizing nominalizations; in the former case they attach to nouns (collectively interpreted) to derive nouns (singularly interpreted), otherwise they attach to a wider range of bases, and unitizing is one among several semantic functions that derive individualdenoting nouns. Their exponence is mostly suffixal. Prefixes with singulative value seem to be better characterized as compounding stems, in cases like the Breton pen(n)- or pez- (cf. section 2.5 above), from the nouns for ‘head’ or ‘piece’, or the Irish leath ‘half ’ (cf. Corbett 2000: 163 for prefixal “quantity markers” in Oceanic languages). As McKenna (1988: 223) notes, pen(n)- may occasionally attach to a noun that is formally a derived singulative but functionally a simple singular; cf. pen-salad-en ‘lettuce’. Apart from these cases, suffixes with a singulative value normally occur stem-finally closing off a nominalization, and before inflectional morphology which controls agreement, like the regular plural -ou of Breton or the fused inflectional endings of Russian. As a result, plural morphology can appear inside singulative affixes, when it lexically marks a collective reading on a stem, or outside, as a marker of contextually relevant plural inflection. Languages with a rich enough morphology can express both at the same time, like the Breton ster-ed-enn-ou. Breton also shows that a singulative noun may be input to verbal derivation, as in sili-enn-a ‘to slip between the hands’, formed from sili-enn ‘eel’ (Trépos 1956: 121). In European languages, the relation of singulative affixes with gender varies. Some are category-preserving with respect to the gender of the base, in the terms of Stump (1993), but affixes like the Breton -enn or the Arabic -a, which derive nouns with a fixed gender, are category-assigning. Diminutives that may otherwise inherit the gender of their base determine a fixed value when used as individualizing nominalizers; cf.

65. Singulatives

1181

Italian crema ‘cream’ (fem.) − crem-ina (fem.) ‘little cream’, but crem-ino ‘cream praline’ (masc.). Likewise, in Irish, -ín inherits the gender of the base when it has a diminutive value, but imposes the masculine when it derives a new nominal lexeme: fear-ín ‘little man’ is masculine, bean-ín ‘little woman’ is feminine, but paidr-ín ‘rosary’ is masculine even though derived from the feminine paidir ‘prayer’ (Ó Siadhail 1984; this explains why even cail-ín ‘girl’ is masculine). As lexeme formants, singulative markers do not take part in agreement or concord. This seems to hold even when they coincide with the exponent of the singular number, as in Nilo-Saharan languages (there are singulative adjectives in Turkana, but they are a small closed set; cf. Dimmendaal 2000: 218). This of course does not apply when a singulative interpretation is achieved by reassignment to a different gender or noun class. Two other properties of singulative morphology follow naturally from its lexemeforming character: it only applies to a semantically motivated (often small) subset of the nominal lexicon; and its exponents are few and deterministically selected for each base, often having an alternative non-unitizing value. It might appear that these properties logically belong together, distinguishing singulatives from numeral classifiers as lexical vs. grammatical unitizing devices. However, Seifart (2005, 2009) has called attention to the mixed character of unitizing morphology in the Amazonian language Miraña. Here, unitizing follows the lexical pattern of singulatives in being expressed by affixes on nouns, which are compatible with inflectional pluralization. However, the affixes are many and classify most of the nominal lexicon, like classifiers; above all, they co-occur on nouns and numerals: (8)

maːkíní-Ɂo-βa ɯ ´ hi-Ɂó-ːnɛ three-SNGLT-PL banana-SNGLT:OBLONG-PL ‘three bananas’ (Miraña: Seifart 2005: 5)

A satisfactory typology of unitizing, which would essentially contribute to a theory of how languages express the division of reference, is still a desideratum. Before this goal can be reached, more research is needed into the syntax and semantics of singulatives in particular languages, especially in quantified constructions.

5. References Acquaviva, Paolo 2008 Lexical Plurals. A Morphosemantic Approach. Oxford: Oxford University Press. Aikhenvald, Alexandra 2006 A Grammar of Tariana, from Northwest Amazonia. Cambridge: Cambridge University Press. Andrzejewski, Bogumil W. 1960 The categories of number in noun forms in the Borana dialect of Galla. Africa: Journal of the International African Institute 30(1): 62−75. Berger, Hermann 1998 Die Burushaski-Sprache von Hunza und Nager. Wiesbaden: Harrassowitz. Contini-Morava, Ellen 1999 Noun class and number in Swahili. In: Ellen Contini-Morava and Yishai Tobin (eds.), Between Grammar and the Lexicon, 3−29. Amsterdam/Philadelphia: Benjamins.

1182

VII. Semantics and pragmatics in word-formation II: Special cases

Corbett, Greville 2000 Number. Cambridge: Cambridge University Press. Cowell, Mark 1964 A Reference Grammar of Syrian Arabic. Washington, D.C.: Georgetown University Press. Cuzzolin, Pierluigi 1998 Sull’origine del singolativo in celtico, con particolare riferimento al medio gallese. Archivio Glottologico Italiano 84: 121−149. Dimmendaal, Gerrit 2000 Number marking and noun categorization. Anthropological Linguistics 42(2): 214−261. Favereau, Francis 1997 Dictionnaire du breton contemporain. Morlaix: Skol Vreizh. Frege, Gottlob 1884 Die Grundlagen der Arithmetik. Breslau: Köbner. Georg, Stepan and Alexander P. Volodin 1999 Die Itelmenische Sprache. Grammatik und Texte. Wiesbaden: Harrassowitz. Hayward, Richard 1998 Qafar (East Cushitic). In: Andrew Spencer and Arnold M. Zwicky (eds.), The Handbook of Morphology, 624−647. Oxford: Blackwell. Holes, Clive 2004 Modern Arabic. Structures, Functions, and Varieties. Washington, D.C.: Georgetown University Press. Jurafsky, Daniel 1996 Universal tendencies in the semantics of the diminutive. Language 72(3): 533−578. King, Gareth 1993 Modern Welsh. A Comprehensive Grammar. London: Routledge. McKenna, Malachy 1988 A Handbook of Spoken Breton. Tübingen: Niemeyer. Mifsud, Manuel 1996 The collective in Maltese. Rivista di Linguistica 8(1): 29−51. Moreno, Martino 1940 Manuale di Sidamo. Milano: Mondadori. Newman, Paul 1990 Nominal and Verbal Plurality in Chadic. Berlin/New York: Mouton de Gruyter. Ó Siadhail, Micheál 1984 A note on gender and pronoun substitution in modern Irish dialects. Ériu 35: 173−177. Pedersen, Holger 1910 Vergleichende Grammatik der Keltischen Sprachen. Göttingen: Vandenhoek & Ruprecht. Quine, Willard van Orman 1960 Word and Object. Cambridge, MA: MIT Press. Rothstein, Susan 2010 Counting and the mass-count distinction. Journal of Semantics 27(3): 343−397. Seifart, Frank 2005 The Structure and Use of Shape-Based Noun Classes in Miraña (North West Amazon). Nijmegen: Max Planck Institute for Psycholinguistics. Seifart, Frank 2009 Towards a typology of unitization: Miraña noun classes compared to numeral classifiers and singulatives. Ms., Leipzig: Max Planck Institute for Evolutionary Anthropology. Soja, Nancy, Susan Carey and Elizabeth Spelke 1991 Ontological categories guide young children’s inductions of word meaning: Object terms and substance terms. Cognition 38: 179−211. Stump, Gregory 1993 How peculiar is evaluative morphology? Journal of Linguistics 29: 1−36.

66. Collectives

1183

Townsend, Charles 1980 Russian Word Formation. Columbus, OH: Slavica. Trépos, Pierre 1956 Le pluriel breton. Rennes: Imprimeries Réunies. Wehr, Hans 1976 A Dictionary of Modern Written Arabic. Ed. by J. Milton Cowan. 3rd ed. Ithaca, NY: Spoken Language Services. Wierzbicka, Anna 1988 The Semantics of Grammar. Amsterdam/Philadelphia: Benjamins. Zeuss, Johann Caspar 1853 Grammatica Celtica. Leipzig: Weidmann.

Paolo Acquaviva, Dublin (Ireland)

66. Collectives 1. 2. 3. 4.

The semantics of collectives Collectives as a word-formation category Conclusion References

Abstract Collective nouns belong to a class of nouns which, just like the nominal categories mass noun, count noun, abstract noun, concrete noun, animate or inanimate noun, is characterized by very general semantic characteristics of nominal lexical items which may also have syntactic consequences such as number agreement and selection restrictions on adjectival modifiers. The aim of this article is to describe the most important word-formation patterns involved in creating collectives, illustrated by examples from several European − both Indo-European and non-Indo-European − languages. The focus will be on the relationship between the specific semantics of collectives and the particular means of word-formation in this domain, on cross-linguistic tendencies as well as some language-particular means.

1. The semantics of collectives The term collective is employed for a special semantic class of nouns obligatorily designating a plurality of entities, although similar phenomena can be observed for adjectives and verbs, e.g., adjectives referring to a sum of properties as in bitter-sweet, and verb compounds combining two types of actions such as German fräs-bohr-en ‘mill-drillINF’.

1184

VII. Semantics and pragmatics in word-formation II: Special cases

Apart from their specific semantic content, lexical items are characterized by lexicosemantic properties such as countability (Krifka 1989) or “Seinsart” (Rijkhoff 2004: 59; also see Meisterfeld 1998). From a typological perspective, obligatory number marking seems to be linked to the prevalence and prototypicality of count nouns designating one entity in the singular. Rijkhoff (2004: 50−56) calls these nouns “singular object nouns”, whereas in languages like Mandarin Chinese with numeral classifiers the prototypical nouns are transnumeral and resemble mass-nouns. Within the noun systems of the European languages collectives are therefore a marked category distinct from the majority of nouns, since in Indo-European languages (and other European languages) object nouns with an obligatory singular/plural distinction prevail (Corbett 2000: 88). Because the European languages (comprising both Indo-European and other language families) are mostly based on singular object nouns, I will describe collectives as they occur in these systems. In the European languages analysed here, collectives cross-cut the other very general noun categories, i.e. the non-count/count distinction, the distinction between animate and inanimate nouns and between concrete and abstract nouns. Collectives are above all defined as nouns inherently designating a plurality of entities (concrete or abstract entities, less frequently kinds), which explains their semantic closeness to the inflectional plural and which distinguishes them from mass nouns designating undifferentiated substances. On a morphosyntactic level, they can be count nouns which may inflect for plural and which then designate a plurality of sets. If they cannot inflect for plural they are either singularia tantum or pluralia tantum. Collectives have to be distinguished from abstract nouns such as purchase, which may refer to a plurality of entities on the utterance level as a result of type coercion (see Borillo 1997), whose intension, however, is not specified for the reference to a plurality of elements. As mentioned above, collectives can be found in any of the other general noun categories, they can be concrete or (less frequently) abstract (cf. furniture vs. humanities), they can designate human, animate or inanimate entities (cf. people, cattle and outfit). They can be semantically very specific like chapter (of a cathedral) or very general like set. They can refer to entities conceived of as homogeneous units as in the case of cattle, or obligatorily comprising a set of heterogeneous entities as in the case of outfit. Furthermore collectives can be count as in the case of family or non-count as in the case of cattle. Some collectives refer to a unique totality as in the case of fauna, or the French, and are thus close to proper names (also see Leisi 1971: 33). The second important property corresponds to the presence or absence of a unifying frame (Kuhn 1982: 56; Seiler 1986: 43). This aspect is closely related to countability. The unifying element can be a common function (hiking gear), a specific configuration (bunch of flowers), spatial contiguity (forest), social relationships (family) and others. Leisi (1971) calls these collectives “group collectives”. The frame can show up in the selection of adjectives, which may either modify the collectivity or the single entities (as in big crowd vs. big people) and other phenomena (cf. Mihatsch 2000). In many cases these collectives can be inflected for plural and then refer to several collectivities as in family/families. Some collectives, usually non-count collectives such as cattle are not specified for a frame. They designate classes of similar entities which share common properties. They usually cannot be inflected for number and are, morphologically speaking, either singularia or pluralia tantum. If they show plural marking it reflects the plurality of the elements (Mihatsch 2006: 128−137). Leisi (1971: 32−34) calls these nouns “generic nouns”. Unlike group collectives such as herd or crowd the relationship between the single elements composing them and

66. Collectives

1185

the collective corresponds to hyponymy rather than meronymy (Mihatsch 2006: 103−104, 123−127). All elements obligatorily fall under the generic noun, while in the case of the group collectives the member category is defined independently and does not semantically depend on the collective. For instance, not all human beings are part of a crowd, but all human beings are people. It is not entirely clear whether generic nouns are collectives since they are logically transnumerals referring to a class, thus, in principle they may refer to an individual (Leisi 1971: 32): (1)

I have eaten an apple. → I have eaten fruit.

Therefore some linguists do not classify these nouns as collectives, for instance, Bosque (1999: 53 f.) considers them mass nouns, Joosten (2010) calls them aggregate nouns, which he distinguishes from collectives in a strict sense. However, on utterance level, for instance, in the case of Welsh and Arabic transnumerals, these nouns usually refer to a plurality of entities (Kuhn 1982: 62, 66; also see Corbett 2000: 13, footnote 4). Nevertheless, the distinction between these two types is semantically important. The distinction also becomes apparent in the different positions they occupy in binominal constructions (also see Kuhn 1982: 57−59; Seiler 1986: 45−47). Group collectives appear in the first position, generic nouns as well as plural nouns in the second position as in a herd of cattle, a herd of cows. Breton and Welsh as well as Arabic have singulative morphemes isolating one entity from a generic noun (see article 65 on singulatives), for instance, Welsh plu ‘feather(s)’ (transnumeral) − plu-en feather-SING ‘a feather’ (singulative) (Kuhn 1982: 65 f.). Other languages have singulative derivational affixes, as in the case of Russian rjabin-inka ‘rowan berry-SING’ from rjabina ‘rowan berries’ (Nagórko 2009: 791). Singulative meaning can also be expressed by binominals such as head of cattle or compounds such as club member. Both types of collectives can arise as a result of word-formation processes as will be shown below, however, many of them are also lexical roots. Furthermore, many collectives, both roots and morphologically complex nouns, can be traced back to − usually metonymic − semantic changes, often from locative nouns to a collectivity of entities found in one place, and from abstract nouns, especially action nouns and property nouns acquiring collective readings, either via the reference to a plurality of entities resulting from an action or a plurality of entities possessing a common property. According to Kuhn (1982: 79) morphologically marked collectives in opposition to the corresponding unmarked non-collectives are always group collectives, while unmarked collectives opposed to marked singulatives or otherwise derived singulars tend to be generic nouns. Therefore one might expect more group collectives than generic nouns as a direct outcome of word-formation processes. However, there is a continuum between these two types of collectives. When the unifying frame fades out and the similarity of the members increases, group collectives may become generic nouns. This process is typical of concrete collectives above the basic level (see Mihatsch 2006: 123−127, 2007). Collectives above the basic level are cognitively more primitive than count hyperonyms, since they maintain the basic level imagery by combining several basic level concepts. This explains the relative frequency of collectives, but also the prevailing inflection for plural above basic level (Markman 1985: 39; Mihatsch 2006: 144−145, 2007). For instance, a typical path leads from group collectives meaning ‘outfit’ to generic nouns meaning ‘clothing’. Some of these generic nouns can even become singular object nouns, as in

1186

VII. Semantics and pragmatics in word-formation II: Special cases

the case of garment. The emerging generic or singular object nouns are perceptually more stable than the relational group collectives, since they are based on inherent rather than contingent properties, a typical case of lexicalization (Mihatsch 2006: 16−23, 123− 127), while in a few cases generic nouns such as French vaisselle ‘tableware’ can acquire a unifying frame, in this case ‘dirty dishes’ (Mihatsch 2006: 127). Due to the plural reference of collective nouns, there are strong synchronic and diachronic links between the inflectional plural and collectives. Plural morphemes, like other inflectional categories, may be an outcome of grammaticalization processes. They often emerge from collective affixes (Corbett 2000: 119, 266). This close relation is not a coincidence. Number marking on nouns is considered a case of inherent inflection as opposed to contextual inflection (Booij 1996: 2). Inherent inflection resembles derivational processes since it semantically interacts very strongly semantically with the stem (cf. Booij 1996: 2− 3), whereas contextual inflection such as agreement and case does not usually affect the semantics of the stem (Booij 1996: 11). This is also correlated with the processing and storage of the two types of inflection. Inflected plural forms are more likely to be stored as units than contextually inflected forms such as case (cf. Baayen, Burani and Schreuder 1997). This tendency explains the emergence of pluralia tantum, i.e. the fossilization of nouns inflected for plural as collective nouns (Booij 1996: 3) such as clothes, often in semantic domains showing locally unmarked plurals (Tiersma 1982). In some cases old plural forms are thus preserved in the lexicon, then often become pluralized again, in analogy to count nouns (see Baldinger 1950: 126, 173 for French examples). For instance, many Romance collectives go back to Latin neuter plural forms, which in turn can adopt a Romance plural morpheme, as in the case of French entrailles from Lat. intralia (DHLF, s. v. entrailles). In Italian the Latin neuter plural has even become the basis of a well established lexical pattern of group collectives (cf. Ojeda 1995), such as dita ‘all fingers of a hand’ vs. the inflected plural diti ‘fingers’. Acquaviva (2008: 123−124) considers these forms inherent plural nouns based on a now unproductive derivation. Some traces of the reanalysis of the Latin neuter plural as a collective can also be observed in other Romance languages. Acquaviva (2008) discusses cases of lexical plurals in Italian, Arabic, Irish and Breton and argues for derivational plural uses of productive inflectional plural morphemes in some cases, for instance in Breton (Acquaviva 2008: 241, 263−264). Some languages such as Maltese, in turn, seem to have a specific collective plural inflection (Gil 1995 and Corbett 1996, 2000 cited in Acquaviva 2008: 72). The transitions between inflection and the lexicon are accompanied by subtle semantic changes, since there is an important difference between the lexical category of collectives stored as such in the mental lexicon, and the inflectional plural at the utterance level. The latter can be interpreted distributively or collectively. In the case of the distributive reading, each single entity is conceptualized separately, a process underlying counting, which is a cognitively more complex operation than conceptualizing a collectivity visually representable in the case of concrete nouns (see Frazier, Pacht and Rayner 1999: 100−101). Thus the distributive plural tends to be semantically much more marked than the collective reading (cf. Gil 1995: 324−325, and Link 1998: 23, 35−36). In the course of the fossilization of nouns inflected for plural, only the cognitively more primitive collective reading is lexicalized. This can also be observed for some irregular plurals, which tend to have collective readings, as in the case of French yeux ‘both eyes’ as opposed to the regular plural œils for the distributive or collective plural of the metaphoric meaning designating different kinds of openings (Grevisse and Goosse 1993: 797).

66. Collectives

1187

2. Collectives as a word-formation category The semantic complexity of collectives also shows up in the different word-formation processes that may either reflect properties of the entities composing the collectives or the unifying aspects of the frame. As will be shown below, collectives as a word-formation category seem to be widespread, but not overly productive when compared to other categories such as nominalization. In the following, the focus will not be on the morphosyntactic and morphonological processes, but rather on the correlation between the semantics and the morphology of collectives. The aim of this overview is an illustration, not an exhaustive study, based on examples from a selection of European languages which show the most important tendencies of collectives as an outcome of word-formation processes.

2.1. Endocentric compounds and the grammaticalization of derivational affixes Endocentric collective compounds can be observed in languages with a generally productive system of compounding as in German or English. Here we typically find compounds consisting of a head designating the frame and a modifier specifying functions or members of the collective as a whole as in hiking gear, stamp collection and many others. If the modifier designates elements as in stamp collection, the modifying noun is usually not inflected for plural or, for instance in German, is ambiguous due to linking elements such as -en- in Dozent-en-schaft lecturer-LINK-SUFFIX ‘lecturer staff ’ (Wellmann 1969: 69). These compounds, at least with respect to their word-formation semantics, tend to be group collectives. In languages with less productive compounding, nominal syntagms show equivalent patterns, for instance, in Romance binominals containing prepositions as in French collection de timbres collection of stamps ‘stamp collection’. Semantically, the head tends to be more general than the modifier, thus many heads occur in larger series of compounds. Such series may eventually, via reanalysis, lead to the grammaticalization of these nouns, which then become semantically even more general and may eventually lose their autonomy and thus become semi-suffixes and in some cases even full-fledged derivational affixes. Wischer (2008) discusses arguments in favour and against a classification of these diachronic processes as grammaticalization processes. Beside many common properties, especially in diachrony, she points out one important difference: While inflectional morphemes originate in free word forms in determined syntactic contexts, derivational affixes based on nouns arise within compounds (Wischer 2008: 136, 145). The emerging semi-suffixes and the more strongly grammaticalized derivational suffixes are relatively well-studied for German (for an overview see Erben 2003: 2531 and article 104 on grammaticalization in German word-formation) and English (Trips 2009). Semiaffixes, and to a higher degree, derivational affixes, are semantically more general than their nominal sources, however, they still show certain semantic restrictions. In German (Drosdowski 1995: 492 f.; Wellmann 1969: 187) and English (Trips 2009) we find several relatively weakly grammaticalized collective morphemes, which still have fully nominal equivalents, that, however, differ in most cases semantically from the semi-suffix.

1188

VII. Semantics and pragmatics in word-formation II: Special cases

German -zeug originally had the meaning ‘equipment of a troop of people going in company’ in a non-sedentary context, then acquired the more general meaning ‘equipment, provisions, tools’ as in Badezeug ‘swim gear’, Strickzeug ‘knitting gear’ (Rosenkranz 1968: 230−235; Wellmann 1969: 200−201). Similarly German -werk, whose simplex meaning was originally ‘work’, has been attested as a collective morpheme since the 14th century (Erben 2006: 147; Rosenkranz 1968: 225). As a consequence of its grammaticalization, it is not only used for artefacts, as in Schuhwerk ‘footgear’ or abstract artefacts as in Regelwerk ‘set of rules’, but also for natural objects such as Astwerk ‘branches (of a tree or bush)’ (cf. Wellmann 1969: 188−190). Other semi-suffixes for inanimate concrete and/or abstract collections are -kram (Rosenkranz (1968: 242), -gut (Wellmann 1969: 210−211), -material (Wellmann 1969: 215, 219), -ware(n), and -wesen (Wellmann 1969: 220; also see Seiler 1986: 163). The collective nouns Leute ‘people (human beings)’ and, to a lesser extent, Volk ‘people (nation)’ (Drosdowski 1995: 492− 493) are employed to derive collectives of human beings as in the group collective Eheleute ‘married couple’, but also nouns such as Bergleute mountain-people ‘miners’ (Wellmann 1969: 42−43), also see Middle Bavarian Mannaleit men-people ‘totality of men’. English has a similar array of semi-grammaticalized collectives, e.g., wear, gear and stuff, in some cases with a slightly pejorative meaning as in foodstuff, -ware for artefacts as in tableware or kitchenware and, less productive today, folk, as in womenfolk. Other morphemes have become derivational affixes (Wischer 2008: 141). This is the case of German -schaft, originally ‘make-up, texture’ (Erben 2006: 147), which used to derive denominal abstract nouns designating properties or roles as in Vaterschaft ‘fatherhood’, from which collectives of humans such as Lehrerschaft ‘teaching staff ’ are derived, usually either designating the whole class or the staff of one particular institution. Some derivations also refer to localities as in Grafschaft ‘county’ (Wellmann 1969: 71, 162−166). German -tum as well as English -dom go back to a full noun meaning ‘judgment’, both suffixes are used to derive status nouns, as in clerkdom, but are now also − to a limited extent − collectives such as professordom and may derive spatial compounds such as kingdom (Plag 2003: 88; Wellmann 1969: 171−172). German -heit (cf. Gothic haidus ‘manner’) as in Menschheit man-SUFF ‘humanity’ was already used to derive deadjectival abstract nouns in the 8th century (Erben 2006: 145); collectives may arise as a secondary semantic change of derived nouns (Wellmann 1969: 174−176). Similarly, English -hood and -ship are used to derive abstract nouns such as childhood and hardship and, less frequently, collectives as in neighborhood and membership (cf. Plag 2003: 88; see Trips 2009 for a detailed analysis). This path of grammaticalization does not exist in Romance languages for collectives, only marginally for some heads of more or less lexicalized syntagms such as French gens as in gens d’affaires ‘business people’, also see gendarme ‘policeman’ from gens d’armes ‘armed men’, and I have not come across it in the literature on other nonGermanic European languages.

2.2. Co-compounds Co-compounds typically, but not exclusively, produce collectives (Wälchli 2005; see also article 40 on co-compounds). Collective co-compounds consist of two nouns, neither of

66. Collectives

1189

which is the head and which are either exhaustive members of a couple as in Mordvin t’et’a.t-ava.t father.PL-mother.PL ‘parents’ or typical members which are representative of a larger set of heterogeneous entities (Wälchli 2005: 5−6). In the case of concrete nouns the roots are usually basic-level nouns, for instance in Mordvin ponks.t-panar.t trousers.PL+shirt.PL ‘clothing, clothes’ (Wälchli 2005: 139), which can refer to the outfit of a person or to clothing as a category, which is thus both a group collective and a generic noun. Lezgian shows an analogous pattern, for instance buba-dide father-mother ‘parents’ or xeb.mal ‘sheep-cattle’, designating domestic animals (Haspelmath 1993: 108). As a means of compounding in the strict sense, this process becomes more frequent as we go from continental Eurasia toward the east (Wälchli 2005: 191, 236), with the exception of Basque, for instance anai-arrebak ‘brothers and sisters’ (Hualde 2003: 352). In Europe, co-compounds can also be found, apart from Lezgian and Mordvin, in a very limited number in Baltic languages, where we find relics such as Lithuanian kójosrankẽlės ‘feet and hands’ (Larsson 2002: 221). In other Indo-European languages there are some isolated cases of lexicalized syntagms based on coordination, such as English brothers and sisters, corresponding to the rather academic noun sibling. Speakers seem to prefer such syntagms to less transparent collectives or hyperonymic count nouns (Aitchison 2003: 96). Another special kind of compounding producing collective nouns can be found in Lezgian, which creates echo-compounds based on reduplication, with the onset of the first syllable of the second member being replaced by /m/ (Haspelmath 1993: 109). Semantically, these compounds correspond to associative plurals and mean ‘N and similar things’ as in sik’-mik’ ‘fox and other wild animals’. Reduplication does not seem to play an important role in producing collectives in European languages, where reduplication is generally very marginal. However, typological overviews show that reduplication is a well-known means of plural inflection in some languages such as Indonesian (Stolz 2008: 88).

2.3. Derivational affixes and metonymic change All in all, specialized collective derivational affixes do not seem to be very productive in European languages today, many of them primarily serve to derive abstract nouns such as action nouns or property nouns, only secondarily collectives, i.e. often via semantic change of individual derivational products. Many specialized collective affixes are unproductive and only persist in lexicalized derivational products (see Borillo 1997 for French; Grossmann 2004: 244 for Italian; Rainer 1993: 206−207 for Spanish; Rosenkranz 1968 for German dialects; Nagórko 2009 for Slavic languages). In all languages considered here collective derivational affixes are suffixes, with one exception in German, which will be described below. In many cases the lexical root designates members, the derivational affix serves to derive the collective as in French feuillage ‘leaves’ from feuille ‘leaf ’. These collectives tend to be group collectives with a frame determined by the utterance context, for instance in the case of Spanish herramental ‘set of tools used for a specific purpose’ (see MOL, s. v. herramental ), or feuillage for the leaves of a tree (PR, s. v. feuillage).

1190

VII. Semantics and pragmatics in word-formation II: Special cases

Unlike compounds and semi-affixes, derivational affixes are semantically very general. However, they still show certain semantic restrictions or at least preferences. Below I will present some of the most important domains. An ancient semantic domain of collective derivational affixes designates plants growing in one place. Classical Latin, Ancient Greek and Sanskrit had productive suffixes for collectives of plants, for instance Latin -(ē)tum for plants growing in one place and the place itself as in palmētum ‘palm grove’ (Lühr 2008: 64−65; Gaide 1989: 221−224), further suffixes in this domain are -īna and -ārium. Old High German had an equivalent collective suffix -ahi (Gaide 1989: 223−224), which is preserved in toponyms such as Lärchach ‘larch grove’. The Romance languages still have more or less productive suffixes in this domain. Many, but not all of them, go back to the equivalent Latin suffixes specialized in plant collectives with their spatial extension such as Italian -eto/-eta as in pineta ‘pine grove’ (Dardano and Trifone 1997: 536; Grossmann 2004: 251 for an overview of Italian suffixes in this domain). In Slavic languages we also find locative collectives such as Czech dubina ‘oak grove’, which are today mostly lexicalized und not any longer productive (Nagórko 2009: 791). Today a series of borrowed Graeco-Latin learned suffixes also derive plant collectives. These suffixes, however, designate classes or categories within biological taxonomies, i.e. kinds, such as Italian -acee for plant families. Interestingly, the plural inherent in the suffixes above the level of the genus reflects the heterogeneity of the subcategories typical for higher generalization levels (Mihatsch 2006: 166). In the domain of human collectives there are essentially two groups of derivational suffixes. In colloquial language human collectives tend to be pejorative, since the reference to a collectivity may show a disregard for the individual and may thus serve to express contempt. Some languages show a great variety of suffixes in this domain, for instance, Spanish where we find -ada, -aje, -alla, -ancia and some others (Rainer 1993: 207). The lexical root usually designates the members, sometimes even collectives as in Italian gentaglia people-SUFF ‘riffraff ’, sometimes properties of the members (see Dardano and Trifone 1997: 530, 536). When employed with inanimates these suffixes are less likely to be pejorative (see Grossmann 2004: 245−247 for Italian). Similar pejorative tendencies in the domain of human collectives can be observed in Slavic languages, cf. Polish -(s)tw(o) as in krzyżactwo ‘order of crusaders’ from krzyżak ‘crusader’ (Nagórko 2009: 790). Other typical not necessarily colloquial nouns referring to humans specify social relations or professions. Collective suffixes are often derived from suffixes originally producing abstract nouns as in Latin -ia, -tās (Gaide 1989: 225); plausibly via reanalysis triggered by series of individual lexical items having undergone parallel semantic changes. Collectives may also arise as an outcome of (very productive) metonymic changes of individual derivations. The Pan-Slavic collective suffix -stv-o is often used to derive collections of professionals, such as Russian učitel’-stvo ‘teaching staff, the teaching profession’ from učitel’ ‘teacher’ (Sussex and Cubberley 2006: 437). These human collectives tend to be group collectives, either linked to one specific situation or place or to an international totality as in Spanish -ado in campesinado ‘peasantry’ (Rainer 1993: 207). Many of these suffixes are learned borrowings such as English -ate as in electorate, Russian -at as in dekanat ‘dean’s staff ’ and others (Nagórko 2009: 791), going back to the Latin collective suffix -ātus. In many cases these expressions also have an abstract or a spatial meaning, such as ‘office of a dean’ (Nagórko 2009: 791).

66. Collectives

1191

In the domain of artefacts, the roots often designate members or common materials as in Italian cristallame ‘glassware’ (Dardano and Trifone 1997: 536). Many of these suffixes also derive nouns designating containers. Apart from the more frequent animate and inanimate concrete collectives there are special, often learned, suffixes for deriving abstract collectives, such as Italian -ario as in ideario ‘ideas (of a particular community or tendency)’ (Dardano and Trifone 1997: 535). Since, semantically, plural inflection and the derivation of collectives are very close (see article 14 on the delimitation of derivation and inflection), it is not surprising that many collective derivational suffixes contain fossilized plural morphemes and thus often produce pluralia tantum. Many suffixes originally taken from Latin go back to neuter plural forms as in Spanish -aria, -alia and -ilia (Morreale 1973: 124). Some collective affixes adopt vernacular plural morphemes as in German -alien as in Personalien ‘personal data’ (Wellmann 1969: 45; also see Baldinger 1950: 173 for French examples). While the derivation of collectives seems to be exclusively suffixal, German has a particular pattern with the prefix Ge-, early on also used for deriving collectives, cf. also Gothic gaskōhi ‘pair of shoes’ from skōhs ‘shoes’ (Gaide 1989: 221; cf. also Wellmann 1969: 155, 158; Drosdowski 1995: 501). In some cases it is difficult to decide whether the collective meaning goes back to the result of a semantic change affecting an individual derived noun or to a specialized affix semantically derived from another function. Both processes are typical of action and property nouns as well as place and container nouns acquiring collective readings as in German Garderobe ‘clothes, outfit’ from ‘wardrobe, closet’. This is a case of regular metonymy which can be observed in simplex and derived nouns in all European languages considered, even in Antiquity (Lühr 2008: 59−62, 237−240). As in the case of the grammaticalization of derivational affixes, plausibly a critical number of individual reinterpreted nouns is necessary for the emergence of a new affix function via analogy, as with Latin -tiō and -tūra, which first produced action nouns, but then also collectives. This type of change seems to be a particular tendency of post-Imperial Latin (Gaide 1989: 226), probably following a general tendency of concretization in the lexicon. Latin -men was first used to derive action nouns, but also nouns designating results and instruments as in aequamen ‘instrument for levelling’, and, already in Latin, collectives as in examen ‘swarm’, the function it still shows in Romance derivations (Pharies 2002: 395− 396). This process can also be seen in many modern languages (for Italian see Grossmann 2004: 244−249, for French see Grevisse and Goosse 1993: 208−219, for Spanish see Rainer 1993: 206−207, for German see Wellmann 1969: 178−179, for English see Plag 2003: 87−90, for Slavic languages see Nagórko 2009: 788−790, Sussex and Cubberley 2006: 433). Nominal roots as in French feuillage leaf-SUFFIX ‘leaves’ provide evidence for the new collective suffix function. The same can be said for some cases of conversion from verbs or adjectives to nouns primarily producing action nouns and property nouns, but which may secondarily produce individual nouns or collectives, as in the case of English the young, the old, which necessarily refer to a plurality. In a few cases, however, collective nouns or, via reanalysis, collective affixes, may acquire an abstract meaning (Gaide 1989: 228; Seiler 1986: 163), as in Polish obywatel ‘citizen’ → obywatelstwo ‘Bourgoisie’ − while today obywatelstwo has acquired an abstract meaning, ‘citizenship’ (Nagórko 2009: 788).

1192

VII. Semantics and pragmatics in word-formation II: Special cases

The cross-linguistic tendency of collective affixes to be derived from affixes for abstract nouns, place nouns or container nouns reflects a general semantic tendency of abstract nouns to become concrete nouns via metonymy in the course of their entrenchment in the lexicon, i.e. lexicalization (Mihatsch 2006: 16−23).

3. Conclusion Although collectives are a relevant and stable category expressed by different kinds of word-formation processes in European languages, collectives seem to be a rather peripheral semantic domain of word-formation and often parasitic on other domains, in particular abstract nouns. As to the semantics of word-formation processes leading to collective nouns, group collectives seem to prevail, while generic nouns seem rather to arise as later semantic changes affecting individual lexical items.

4. References Acquaviva, Paolo 2008 Lexical Plurals. A Morphosemantic Approach. Oxford: Oxford University Press. Aitchison, Jean 2003 Words in the Mind. An Introduction to the Mental Lexicon. Oxford: Blackwell. Baayen, Harald, Christina Burani and Robert Schreuder 1997 Effects of semantic markedness in the processing of regular nominal singulars and plurals in Italian. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1996, 13−33. Dordrecht: Kluwer. Baldinger, Kurt 1950 Kollektivsuffixe und Kollektivbegriff. Berlin: Akademie Verlag. Booij, Geert 1996 Inherent versus contextual inflection and the split morphology hypothesis. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1995, 1−16. Dordrecht: Kluwer. Borillo, Andrée 1997 Statut et mode d’interprétation des noms collectifs. In: Claude Guimier (ed.), Co-texte et calcul du sens, 105−121. Caen: Presses Universitaires de Caen. Bosque Muñoz, Ignacio 1999 El nombre común. In: Ignacio Bosque Muñoz and Violeta Demonte (eds.), Gramática descriptiva de la lengua española. Vol. 1, 3−75. Madrid: Espasa Calpe. Corbett, Greville G. 1996 Minor number and Plurality Split. Rivista di Liguistica 8(1): 101−122. Corbett, Greville G. 2000 Number. Cambridge: Cambridge University Press. Dardano, Maurizio and Pietro Trifone 1997 La nuova grammatica della lingua italiana. Bologna: Zanichelli. DHLF = Rey, Alain (ed.) 1998 Dictionnaire historique de la langue française. 3 Vol. 2nd ed. Paris: Dictionnaires Le Robert. Drosdowski, Günther (ed.) 1995 Duden. Grammatik der deutschen Gegenwartssprache. 5th ed. Mannheim: Dudenverlag.

66. Collectives

1193

Erben, Johannes 2003 Hauptaspekte der Entwicklung der Wortbildung in der Geschichte der deutschen Sprache. In: Werner Besch, Anne Betten, Oskar Reichmann and Stefan Sonderegger (eds.), Sprachgeschichte. Ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung. 2nd ed. Vol. 3, 2525−2538. Berlin/New York: de Gruyter. Erben, Johannes 2006 Einführung in die deutsche Wortbildungslehre. 5th ed. Berlin: Schmidt. Frazier, Lyn, Jeremy M. Pacht and Keith Rayner 1999 Taking on semantic commitments, II: Collective versus distributive readings. Cognition 70: 87−104. Gaide, Françoise 1989 Les dérivés synchroniques dénominaux collectifs en latin. Revue de philologie 63(2): 221−228. Gil, David 1995 Universal quantifiers and distributivity. In: Emmon Bach, Eloise Jelinek, Angelika Kratzer and Barbara H. Partee (eds.), Quantification in Natural Languages. Vol. 1, 321−362. Dordrecht: Kluwer. Grevisse Maurice and André Goosse 1993 Le bon usage. Grammaire française. 13th ed. Paris/Louvain-La-Neuve: Duculot. Grossmann, Maria 2004 Nomi collettivi. In: Maria Grossmann and Franz Rainer (eds.), La formazione delle parole in italiano, 244−252. Tübingen: Niemeyer. Haspelmath, Martin 1993 A Grammar of Lezgian. Berlin: de Gruyter. Hualde, José Ignacio 2003 Compounds. In: José Ignacio Hualde and Jon Ortiz de Urbina (eds.), Grammar of Basque, 351−362. Berlin/New York: Mouton de Gruyter. Joosten, Frank 2010 Collective nouns, aggregate nouns, and superordinates: When ‘part of ’ and ‘kind of ’ meet. Lingvisticæ Investigationes 33(1): 25−49. Krifka, Manfred 1989 Nominalreferenz und Zeitkonstitution. Zur Semantik von Massentermen, Pluraltermen und Aspektklassen. München: Fink. Kuhn, Wilfried 1982 Formale Verfahren der Technik KOLLEKTION. In: Hansjakob Seiler and Franz Josef Stachowiak (eds.), Apprehension. Das sprachliche Erfassen von Gegenständen. Teil 2: Die Techniken und ihr Zusammenhang in Einzelsprachen, 55−83. Tübingen: Narr. Larsson, Jenny Helena 2002 Nominal compounds in the Baltic languages. Transactions of the Philological Society 100(2): 203−231. Leisi, Ernst 1971 Der Wortinhalt. Seine Struktur im Deutschen und im Englischen. 4th ed. Heidelberg: Quelle & Meyer. Link, Godehard 1998 Ten years of research on plurals − where do we stand? In: Fritz Hamm and Erhard Hinrichs (eds.), Plurality and Quantification, 19−54. Dordrecht: Kluwer. Lühr, Rosemarie 2008 Nominale Wortbildung des Indogermanischen in Grundzügen. Vol. 1. Hamburg: Kovač. Markman, Ellen M. 1985 Why superordinate category terms can be mass nouns. Cognition 19: 31−53.

1194

VII. Semantics and pragmatics in word-formation II: Special cases

Meisterfeld, Reinhard 1998 Numerus und Nominalaspekt. Eine Studie zur romanischen Apprehension. Tübingen: Niemeyer. Mihatsch, Wiltrud 2000 Wieso ist ein Kollektivum ein Kollektivum? Zentrum und Peripherie einer Kategorie am Beispiel des Spanischen. Philologie im Netz 13: 39−72. http://www.fu-berlin.de/ phin/phin13/p13t3.htm [last access 8 Oct 2014]. Mihatsch, Wiltrud 2006 Kognitive Grundlagen lexikalischer Hierarchien untersucht am Beispiel des Französischen und Spanischen. Tübingen: Niemeyer. Mihatsch, Wiltrud 2007 Taxonomic and meronomic superordinates with nominal coding. In: Dietmar Zaefferer and Andrea Schalley (eds.), Ontolinguistics. How ontological status shapes the linguistic coding of concepts, 359−378. Berlin: Mouton de Gruyter. MOL = Moliner, María 1998 Diccionario de uso del español. 2 Vol. 2nd ed. Madrid: Gredos. Morreale, Margherita 1973 Aspectos gramaticales y estilísticos del número (Segunda parte). Boletín de la Real Academia Española 53: 99−206. Nagórko, Alicja 2009 Diminutiva/Augmentativa und Kollektiva. In: Sebastian Kempgen, Peter Kosta, Tilman Berger and Karl Gutschmidt (eds.), The Slavic Languages. An International Handbook of their Structure, their History and their Investigation. Vol. 1, 782−792. Berlin/New York: de Gruyter. Ojeda, Almerindo E. 1995 The semantics of the Italian double plural. Journal of Semantics 12: 213−237. Pharies, David 2002 Diccionario etimológico de los sufijos españoles y de otros elementos finales. Madrid: Gredos. Plag, Ingo 2003 Word-formation in English. Cambridge: Cambridge University Press. PR = Rey-Debove, Josette and Alain Rey (eds.) 1993 Le Nouveau Petit Robert. Dictionnaire alphabétique et analogique de la langue française. Paris: Dictionnaires Le Robert. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rijkhoff, Jan 2004 The Noun Phrase. Oxford: Oxford University Press. Rosenkranz, Heinz 1968 Komposita auf -ding, -sache, -werk, -(ge)zeug, -geschirr und -kram im thüringischen Sprachraum. Beiträge zur Geschichte der deutschen Sprache und Literatur 90: 212− 248. Seiler, Hansjakob 1986 Apprehension. Language, Object, and Order. Part III: The Universal Dimension of Apprehension. Tübingen: Narr. Stolz, Thomas 2008 Grammatikalisierung ex nihilo. Totale Reduplikation − ein potentielles Universale und sein Verhältnis zur Grammatikalisierung. In: Thomas Stolz (ed.), Grammatikalisierung und grammatische Kategorien, 83−109. Bochum: Brockmeyer. Sussex, Roland and Paul Cubberley 2006 The Slavic Languages. Cambridge: Cambridge University Press.

67. Action nouns

1195

Tiersma, Peter Meijes 1982 Local and general markedness. Language 58: 832−849. Trips, Carola 2009 Lexical Semantics and Diachronic Morphology. The Development of -hood, -dom and -ship in the history of English. Berlin/New York: de Gruyter. Wälchli, Bernhard 2005 Co-compounds and Natural Coordination. Oxford: Oxford University Press. Wellmann, Hans 1969 Kollektiva und Sammelwörter im Deutschen. Ph.D. dissertation, Universität Bonn. Wischer, Ilse 2008 Zum kategorialen Status von Derivationsaffixen im Rahmen von Grammatikalisierungsprozessen. In: Thomas Stolz (ed.), Grammatikalisierung und grammatische Kategorien, 135−146. Bochum: Brockmeyer.

Wiltrud Mihatsch, Bochum (Germany)

67. Action nouns 1. 2. 3. 4. 5. 6.

Introduction Action nouns, event nouns and other verb-based nominals Internal syntax of action-noun constructions Verbal and nominal categories in action nouns Inflection vs. derivation vs. transposition References

Abstract Action nouns constitute a cross-linguistically robustly attested phenomenon. The present article provides an overview of the central problems related to their status as a mixed category, situated in-between prototypical verbs and prototypical nouns, such as the verbal and nominal properties of action nouns and internal syntax of action noun constructions, the status of action nouns as inflected, derived or transposed words and the different theoretical approaches to their formation. It combines a large-scale typological perspective on action nouns in the languages of Europe with a closer look at some particular phenomena in particular languages.

1. Introduction Action nouns (AN, often referred to as action nominals, nomina actionis), such as destruction, conquest, singing or reading, constitute a cross-linguistically robustly attested

1196

VII. Semantics and pragmatics in word-formation II: Special cases

phenomenon. In my global sample (Koptjevskaja-Tamm 2005), action nouns occur in 126 out of the 168 languages, i.e. in 75 % of the investigated languages. The same overall distribution of action nouns (76.4 %) is found in the balanced fifty-five language sample of the world’s languages in Štekauer, Valera and Körtvélyessy (2012: 293−297). Importantly for the present publication, the languages of Europe constitute one of the strongholds for this phenomenon.

2. Action nouns, event nouns and other verb-based nominals Action nouns are either nouns or they at least occur in typical nominal positions and show inflectional properties and/or combinability with adpositions typical of nouns. They are, however, in some reasonably productive way formed from verbs, either derivationally or inflectionally, and refer, broadly speaking, to situations, rather than just to actions, as the name might imply (Comrie and Thompson 1985; Koptjevskaja-Tamm 1993, 2003, 2005). Formation of action nouns is, most probably, the best-known kind of nominalization and is often understood as the nominalization proper. However, nominalization covers a much broader range of phenomena, both processes and operations, whereby a word which is not a noun (but also not necessarily a verb) is turned into a noun (not necessarily denoting a situation), e.g., red → redness, or teach → teacher (cf. Comrie and Thompson 1985). Nominalization itself belongs to word-class-changing operations, or transposition (in the broad sense, as in Haspelmath and Sims 2010: 253−261; or in a more restricted sense, as in Spencer 2013, section 5). What is meant by “situations” in the definition of action nouns, is an interesting issue. There is a long tradition of distinguishing among several recurrent meanings typically expressed by action nouns, more or less following in Vendler’s (1967, 1970) footsteps. The discussion normally takes into account constructions that action nouns head and that may also contain expressions referring to the participants in the situation designated by them (action noun, or action nominal constructions, ANCs). These, in turn, most often serve as arguments (complements) to various predicates. Vendler’s (1970) semantic taxonomy distinguished among three major meanings of such complements − events, facts and propositions. For instance, the verbs assert and be unlikely combine with ANCs referring to PROPOSITIONS (The collapse of the country’s economy is unlikely), the verbs know, regret and lead take ANCs referring to FACTS (as in The enemy’s destruction of the city led to a complete collapse of the country’s economy), while the verbs hear and go on take ANCs referring to EVENTS (The singing of the Marseillaise went on for hours). Action nouns, as understood above, largely correspond to what Grimshaw (1990) defines as complex-event nominals, but the literature does not show complete consensus on the use of each of these terms. Grimshaw argues that complex-event nominals preserve the event structure associated with the underlying verb, i.e. both its lexical aspectual properties akin to the Vendlerian categories of activity, accomplishment etc., and its obligatory (internal) argument structure. They cannot pluralize or co-occur with indefinite articles, but may combine with duration modifiers (e.g., frequent, constant), cf. the complex-event nominal building in example (1).

67. Action nouns (1)

1197

a. The constant building *(of new houses) (by private persons) requires large numbers of timbers. b. *a building of new houses (by private persons) c. *numerous buildings of new houses (by private persons)

Grimshaw (1990) opposes complex-event nominals to simple-event nominals (traditionally called act nominals, or nomina acti). Simple-event nominals denote events, but without preserving the event structure of the underlying verb, i.e. without specifying its lexical aspectual properties and without preserving its arguments. The simple-event nominal development in (2) has no arguments and co-occurs with an indefinite article. Many simple-event nominals are not even derived from verbs, e.g., journey, concert or party. Grimshaw’s distinction between complex-event nominals and simple-event nominals, although quite popular in the generativist literature, is not uncontroversial (see Sleeman and Brito 2010 for some problems with Grimshaw’s diagnostics and definitions). (2)

The country is witnessing a new development.

Some languages have verbal nouns which refer to the way, or manner of performing the action denoted by the base verb − mode, or manner nominals (nomina modi), distinct from action nouns, e.g., in Amharic akkiyahed ‘the way of going’ vs. mohed ‘to go, going’, or in Turkish yürü-yüş ‘the way of walking’. In many accounts, “action nouns” would include act nominals and mode nominals, as well as stative nominalizations (such as belief, love, preoccupation) (cf. Melloni 2010 and also article 68 on action nouns in Romance). The status of the latter is otherwise debatable: Rozwadowska (1997), Spencer and Zaretskaya (1998) and Fábregas and Marín (2012) describe nominalizations from certain stative verbs as sharing certain properties with, but also being quite different from the typical complex-event nominals/action nouns. Situation-denoting nominals are further opposed to verb-based nominals with more concrete meanings, such as naming one of the arguments or various circumstances concomitant to the situation designated by the erstwhile verbs. Examples include result nominalizations proper (see article 71 on result nouns), e.g., agreement (as in The agreement they signed was submitted in four copies), agent nouns, e.g., singer, employer, and instrument nouns, e.g., sharpener (see article 74 on agent and instrument nouns), object nouns, e.g., employee (see article 52 on semantic restrictions on word-formation: the English suffix -ee, and article 75 on patient nouns), locative nominals, e.g., refinery (see article 76 on place nouns) or West Greenlandic puisinniarvik ‘place for hunting seal’ (Fortescue 1984: 319). Note that the term result nominals in the literature following Grimshaw’s influential study is often used as a cover label for all these more concrete verb-based nominals. In the European languages, the by far most frequent word-formation process used to form action nouns is suffixation, but conversion and syntactic nominalization are also well attested. Most languages normally have several word-formation types of action nouns, and most of these word-formation types are not tightly bound to just one meaning. One and the same word may often have several readings, both situation-related, such as FACT/PROPOSITION/EVENT (i.e. complex event), ACT (i.e. simple event) and MANNER, on

1198

VII. Semantics and pragmatics in word-formation II: Special cases

the one hand, and more concrete readings, such as RESULT, PLACE, TIME, AGENT, which complicates the exact demarcation of the category of action nouns in particular languages and across languages. Such multiple readings and polysemy of the related word-formation devices have attracted considerable attention in theoretical literature (see article 68 on action nouns in Romance). Another problem for the delineation of action nouns is found in languages where verbal derivates combine properties of action nouns with those of typical non-finite verb forms, such as infinitives, participles and converbs. In Altaic languages, many verbbased forms that conform to all the criteria for action nouns, are also used in relativization, i.e. as typical participles, which finds parallels in many languages of the world. Finno-Ugric languages confront researchers with the challenge of distinguishing between action nouns and infinitives, since the latter often show a subset of nominal inflectional and combinatorial properties. However, typically, the whole paradigm is significantly reduced compared to the normal nominal paradigm; in addition, the inflected forms may have very specialized meanings not always easily deducible from the “normal” case meanings and should rather count as converbs, i.e. non-finite verbal forms used in adverbial subordination (for the details see Koptjevskaja-Tamm 1993: 33−45; Ylikoski 2003; Serdobol’skaja et al. 2012). These are examples of a more general process whereby individual inflected forms of action nouns, often used in particular functions, may gradually detach themselves from the original paradigm and become non-finite verb forms on their own. Thus, whereas adpositional or case forms of verbal nouns are a frequent grammaticalization source for converbs (Haspelmath 1995), infinitives in many languages originate as purposive action nouns, i.e. action nouns with the dative or allative marker (Haspelmath 1989). Typical infinitives may in turn be further “nominalized”, e.g., by attaching a definite article, and start being used as action nouns, often acquiring more and more nominal properties. Close-by examples of such a development are provided by the German, Italian and Romanian nominalized infinitives, i.e. infinitives combined with definite articles (see example (5)). This completes “a full cycle of noun-to-infinitive-to-noun” (Disterheft 1980; Gaeta 1998).

3. Internal syntax of action-noun constructions In their semantics and discourse behavior, action nouns occupy an intermediate position between typical nouns and typical verbs. For instance, both John’s singing and John sings can describe one and the same event involving the same participant, and the same goes for the enemy’s destruction of the city and the enemy destroyed the city. However, ANCs, as opposed to normal finite clauses, merely name situations and occur in functions typical of NPs, for instance as arguments to predicates. On the other hand, they differ from typical nouns that refer to things, persons, places, and other more or less concrete objects. This marked combination of the general lexical meaning of the root (actions) and the word’s pragmatic function (reference) tends to be morphosyntactically marked, as compared to the more unmarked combinations of the lexical class meaning and pragmatic functions, underlying parts-of-speech differentiation − object denoting words used for reference (nouns) and action denoting words used for predication (verbs,

67. Action nouns

1199

cf. Croft 2001: 86−92). Derivation of action nouns, and nominalization in general, as well as other word-class changing operations, involve two different operations. The baseverb gets “decategorized” (Hopper and Thompson 2004), i.e. it ceases to be a fullfledged verb and loses some of its verbal properties, but it also gets “recategorized”, i.e. it acquires at least some nominal properties, in accordance with its new functions (Bhat 1994; see Malchukov 2006). Morphosyntactically action nouns often show a mixture of verbal and nominal properties and provide therefore good examples of mixed categories (Lefebvre and Muysken 1987; Spencer 2005). Action nouns have received a lot of attention within various linguistic theories, both formal and functionalist, mainly with respect to their argument structure. The issue at stake is whether and how they can combine with expressions referring to the participants in the situations designated by them, primarily with S (i.e. with the single argument of an intransitive verb and of the corresponding action noun, e.g., with John in John’s running) and with A and P (i.e. with the most agent-like and the most patient-like arguments of a transitive verb and of the corresponding action noun, e.g., with the enemy and the city in the enemy’s destruction of the city). Noteworthy, in a number of languages of the world (in 25 languages among the 168 languages in Koptjevskaja-Tamm 2005), action nouns cannot combine with both A and P within one and the same construction. Sometimes there are strategies for expressing A and P at the same time without, however, making both of them syntactically dependent on the action noun, such as the enemy’s city-destruction (where the P and the action noun build a compound) or the destruction of the city done by the enemy. Although such “valency-lowering” strategies are also occasionally employed in the languages of Europe (cf. a város-nak az ellenség által való pusztít-ás-a the city-DAT the enemy by being destruction-AN-its ‘the destruction of the city being by the enemy’ in Hungarian, see article 70 on nominalization in Hungarian), they are on the whole quite marginal here. The European languages normally allow their action nouns to combine directly with the expressions for S, A and P. There are several cross-linguistically recurrent major patterns of how languages construct correspondences to ‘John’s running’ and ‘the enemy’s destruction of the city’ briefly summarized below (for details and generalisations see Koptjevskaja-Tamm 1993, 2003 and 2005), where the main classification criterion is dependent-marking on the arguments themselves. These major patterns can be ordered as a scale from sentencelike constructions to those that have a NP-syntax, with intermediate patterns combining elements of both sentences and NPs:

Sentential

>

more sentence-like

Poss-Accusative

>

Ergative-Possessive

>

Nominal more NP-like

Fig. 67.1: Scale of sentence-like and NP-like dependent-marking in the major ANC patterns

All these types are represented in the languages of Europe. Example (3) from the Daghestanian language Godoberi illustrates the sentential type, whereby the arguments of the action noun in (3b) show the same case marking as in the corresponding finite clause in (3a) − ergative for A, absolutive for P and dative for the recipient.

1200 (3)

VII. Semantics and pragmatics in word-formation II: Special cases a. aHmadi-di maHamadi-łi rec’i ik̄i. Ahmad-ERG Muhammad-DAT bread:ABS give:AOR ‘Ahmad gave bread to Muhammad.’ b. aHmadi-di maHamadi-łi rec’i ik̄i -ir Ahmad-ERG Muhammad-DAT bread:ABS give-AN ‘Ahmad’s giving bread to Muhammad’ (Kazenin 1994: 51)

The sentential type is mainly found in the ergative North-Eastern Caucasian languages and Basque, but occurs also as a marginal strategy in the Uralic language Mari, in the Altaic language Kalmyk and, even more marginally, in Romance (in combinations with the substantivized infinitives in Catalan, Italian, Portuguese and Spanish, see article 68 on action nouns in Romance). The possessive-accusative type, in which S/A are treated as possessors, while P is marked in the same way as in finite clauses, is illustrated by example (4) from the Altaic (Mongolic) language Kalmyk. Here the A receives the genitive case, typical for adnominal possessors, while the P has the accusative case, typical for Ps in finite clauses. (4)

Zanda [Petʲa-n en degtr-ig umšə-lʁ-n-a] tuskar med-nä. Zanda Petja-GEN this book-ACC read-AN-linking n-GEN about know-PRS ‘Zanda knows that Petja has read this book.’ (lit. ‘about Petja’s reading the book’) (Perkova 2009: 472)

While being the cross-linguistically most frequent ANC type, possessive-accusative ANCs show a relatively limited distribution in Europe. With the exception of English, where they alternate with other patterns, and Romance, where they are restricted to substantivized infinitives, they occur primarily at the European periphery, in a number of Uralic languages (in Mari, Erzya, Komi-Zyrian and Udmurt within the Finno-Ugric and in the Samoyedic language Nenets), in Turkish (and most of the Turkic languages, with Gagauz as the notable exception), in Kalmyk, in Armenian and in the Daghestanian language Agul, as an alternative to the sentential type. The ergative-possessive type, whereby S/P are treated as possessors, while A has a different marking, is exemplified by Romanian in (5) with the nominalized infinitive: (5)

Cumpăra-re-a acestei case de către Ion a fost inutilă. buy-INF-the this.GEN house by Ion was useless ‘The buying (of this house) by Ion was useless.’ (Cornilescu 2001: 469)

This is the most widely spread type of ANCs among the languages of Europe. A large portion of these languages use the same marker for the A in ANCs and for agents in passive clauses − the Slavic languages, English and Dutch, Modern Greek, the Romance languages Catalan, French, Romansh, Romanian and Spanish, the Celtic languages Old Irish, Irish and Scottish Gaelic (marginally, “bookish”), the Iranian language Kirmanji, Albanian, the Finno-Ugric languages Hungarian and Estonian, and the South Caucasian languages Georgian and Megrelian. This suggests a possible connection between this

67. Action nouns

1201

nominalization type and passivization, which, however, on the whole seems to be unwarranted. First of all, it has been argued for at least some languages (for example, for Russian, Georgian, and Estonian in Koptjevskaja-Tamm 1993: 147−155, for Russian and Dutch in Schoorlemmer 1995: 303−306) that there are significant differences in the output of the two processes which are inconsistent with a possible derivation of ergativepossessive ANCs from passives. Second, a few languages use different markers for the As in ergative-possessive ANCs and agents in passive clauses − cf. da parte di in Italian, durch in German and o in Welsh, as contrasted to the agent markers da, von, and gan respectively (Comrie and Thompson 1985: 385−387). Finally, Abkhaz lacks a personal passive and marks the A in ergative-possessive ANCs with the instrumental postposition. In action-nominal constructions of the nominal type all the dependents of the action noun are treated as typical adnominal dependents, which make such constructions look more or less like NPs with non-derived nouns as their head. The S and A look like possessors, whereas P is similar either to possessors (in double-possessive ANCs) or to another adnominal dependent (in possessive-adnominal ANC). Peter’s reading of the book in English is an example par excellence of a double-possessive ANC employing the two distinct possessor slots, the Saxon and the Norman genitive, that English happens to have. Similar examples exist in Irish and Scottish Gaelic, with their synthetic genitives vs. possessive prepositions ag/aig ‘at’. Other languages, such as Estonian, Finnish, Latvian, Lithuanian, Maltese and Georgian (and, marginally, a few other languages), utilize the same genitive marking for both S/A and P. The example in (6) shows a possessiveadnominal ANC in Swedish, in which the prenominal A is marked with an -s-genitive and the postnominal P appears with the preposition av. Note that this preposition, although cognate with the English of, does not mark possessors in Swedish. (6)

1800-talet började med Napoleon-s erövr-ing av nästan hela 1800-years began with Napoleon-GEN conquer-AN of almost whole Europa. Europe ‘The 19th century started with Napoleon’s conquest of almost the whole of Europe.’

Both double-possessive and possessive-adnominal ANCs show a fairly restricted distribution, the former often avoided as clumsy and marginal, and the latter mainly restricted to Scandinavian (Danish, Icelandic, Norwegian, Swedish) and Goidelic Celtic (Scottish Gaelic and Irish). Summarizing, the large-scale typological work has shown that cross-linguistic variation in ANC structure is severely restricted in at least two respects (for details cf. Koptjevskaja-Tamm 2003). First, arguments differ systematically in their proneness to retain sentential marking. In addition, dependent-marking, head-marking and word order form a hierarchy that predicts which combinations of nominal and sentential properties are possible. Specifically for Europe, we can say that its by far largest part is inhabited by people who speak languages with ergative-possessive and nominal ANCs. These are the two types that signal the relation between head and P differently from the corresponding finite clauses. Thus, in these European languages, the ability to inflect according to the nominal pattern is not compatible with verbal government. Words tend to be classified unambiguously as either verbs or nouns according to the combination of these two features, and this sharp distinction is retained even in less prototypical cases.

1202

VII. Semantics and pragmatics in word-formation II: Special cases

4. Verbal and nominal categories in action nouns Action nouns often show a mixture of verbal and nominal properties, but the details of this mixture differ considerably from language to language. As an example, let us take the grammatical category of aspect in the Slavic languages. All Slavic verbs are characterized as imperfective or perfective depending on their lexical properties and the prefixes and suffixes they carry. An interesting question is whether the aspect category is also characteristic of action nouns and whether it is inherited from the corresponding verbs. In Polish, action nouns with the suffix -nie/-cie in nominals may be derived both from perfective and imperfective stems and inherit their aspectual properties. The perfective version of ‘evaluation’ in (7a) is appropriate in a telic context, while its imperfective correspondent would be appropriate in an atelic context (Rozwadowska 1997: 64−68). (7)

a. Ocenie-nie studentów przez nauczycieli nastąpiło teachers.ACC took_place evaluate (PFV)-AN.NOM.SG students.GEN by bardzo szybko. very quickly ‘The evaluation of the students by the teachers took place immediately.’ b. Ocenia-nie studentów przez nauczycieli ciągneło evaluate (IMPFV)-AN.NOM.SG students.GEN by teachers.ACC lasted się przez cały tydzień. REFL by whole week ‘The evaluation of the students by the teachers lasted the whole week.’ (Rozwadowska 1997: 64)

The possibility of inheriting verbal aspectual properties differs across Slavic languages, even when they use a cognate of the Polish -nie/-cie suffix. Serbian seems to be quite similar to Polish, cf. Serbian uručivanje nagrada ‘awarding (IMPFV) prizes’ (e.g., ‘for the whole day’) vs. uručenje nagrada ‘awarding (PFV) prizes’ (e.g., ‘in five minutes’) (Bašić 2010: 43). Russian, on the other hand, is very different. Although it allows -niederivation from both imperfective and perfective stems, the resulting action nouns do not inherit the original aspectual opposition. Examples (8a) and (8b) show the contrast between the imperfective verb pisat’ ‘to write’ and its perfective correspondence napisat’. In example (8c), on the other hand, the action nouns pisanie and napisanie ‘writing’ are used in the same durative context (Pazelskaya 2012). (8)

a. Vasja pisa-l pis’ma dva časa /*za dva Vasja.NOM write-PST.M.SG letters.NOM/ACC two hour.GEN /in two časa. hour.GEN ‘Vasja has been writing letters for two hours.’ b. Vasja na-pisa-l pis’ma za dva časa. /*dva Vasja.NOM PRFV-write-PST.M.SG letters.NOM/ACC in two hour.GEN /two časa. hour.GEN ‘Vasja has written (the) letters in two hours.’

67. Action nouns

1203

c. Pisa-nie /Na-pisa-nie pis-em zanja-l-o u write-AN.NOM.SG /PRFV-write-AN.NOM.SG letter-GEN.PL take-PST-N at Vas-i dva časa. Vasja-GEN two hour.GEN Context: Vasja is involved in a lawsuit and yesterday he had to write ten letters to the other participants of the suit. ‘Writing the letters took Vasja two hours.’ (Pazelskaya and Tatevosov 2006) The examples above lead into the interesting issue of whether there is any general ordering of the features that are acquired and lost during derivation of action nouns, where a partial answer is found among the constraints on the internal syntax of action noun constructions, presented in section 3. The data in Koptjevskaja-Tamm (1993) show that there is some correlation between the different ANC types, on the one hand, and the presence vs. absence of other typical verbal vs. nominal features on their head, such as the ability to express tense/aspect distinctions, voice, negation, and to combine with adverbs/adjectives. For instance, the possessive-accusative ANC in English John’s repairing the house so quickly [came as a big surprise], contains an adverb. Moreover, the AN in the same construction can be negated and/or combine with auxiliaries, again in the way typical of verbs, e.g., John’s having not repaired the house. The nominal (double-possessive) ANC with the same action noun may contain adjectives (and not adverbs), cf. John’s quick repairing of the house, but neither verbal negation nor auxiliaries. Malchukov (2006) suggests that the order in which verbal categories are lost and nominal categories are acquired in nominalization is determined by the interaction between the functionally based hierarchies of categories with such structural factors as morpheme ordering and category cumulation. Partly inspired by Croft’s (2001) theory, he reformulates the two factors behind lexical categorization (the general lexical meaning of roots and the pragmatic function of the words containing them) within the framework of optimality theory as two families of constraints on morphosyntactic marking of lexical categories: FuncFaith: Assign (morphological) categories to a lexical item in accordance with its discourse function; LexFaith: Assign (morphological) categories to a lexical item in accordance with the semantic class of a lexical root. Malchukov further relies on the hierarchies of verbal (or clausal) and nominal categories, that have been suggested in the functionalistic literature (e.g., in functional grammar, role and reference grammar and in a number of cross-linguistic studies, the references too numerous to be listed here), but also in generative approaches. These different layers in the hierarchies host both morphological categories (operators) and their syntactic correspondences (satellites). The general idea is that the internal categories in the hierarchies are more relevant for the semantics of the root, while the external ones are more relevant for syntax and pragmatics, corresponding to the FuncFaith and LexFaith constraints introduced above. The categories on both scales are ordered depending on whether they primarily contribute to lexical semantics (satisfy LexFaith) or to discourse function of a lexical item (satisfy FuncFaith). For instance, valency, voice and aspect in the verbal hierarchy and noun class and number in the nominal one are (more) relevant to the semantics of the

1204

VII. Semantics and pragmatics in word-formation II: Special cases

root (action vs. object), whereas subject agreement and illocutionary force or case and determiners are (more) relevant to syntax and/or pragmatics (predication vs. reference). The verbal (9) and the nominal (10) FuncFaith subhierarchies are represented below: (9)

*Illocutionary Force >> *Subject Agreement >> *Mood >> *Tense >> *Aspect >> *Voice >> *Valency

(10) *-Case >> *-Determiners >> *-Possessor >> *-Number >> *-Class In the nominalization process whenever a verbal category at a certain layer is lost, all the more external verbal categories (i.e. those to the left in the hierarchy in (9) will be lost as well, and whenever a nominal category is acquired, all the more internal nominal categories (i.e. those to the right in the hierarchy in (10)) will be acquired as well. The resultant set of verbal and nominal categories on a nominalized verb can be seen as resulting from interaction between conflicting FuncFaith and LexFaith constraints, depending on the point at which these are interpolated. For instance, if a FuncFaith constraint (e.g., *Aspect) outranks a corresponding LexFaith constraint (*-Aspect), this category will be lost, otherwise it will be retained. It should be emphasized that large-scale generalizations, like those proposed in Koptjevskaja-Tamm (1993, 2003 and 2005) and in Malchukov (2006), should be taken as a bird’s eye view of the domain, of necessity glossing over many important facts that may be unveiled in detailed studies of action nouns in particular languages. Since we are dealing here with lexical categories, different groups of action nouns may show significant differences in their morphosyntactic behaviour. Pazelskaya’s (2006) study of “situation nominals” (another term for action nouns), based on the behaviour of about 3,000 lexemes in the Russian National Corpus (www.ruscorpora.ru), shows, among other things, that action nouns in Russian may contain a negation marker and pluralize, even in their complex-event readings (contra Grimshaw 1990). However, both kinds of phenomena are severely restricted, primarily by the lexical meaning of the verbal stem. Negation (surprisingly not included by Malchukov 2006) is a particularly sensitive phenomenon. Although in action nouns it is expressed by the non-verbal prefix ne-, rather than by the verbal particle ne, Pazelskaya argues that this is a verbal category also in action nouns, as evidenced by its meaning and scope properties (Pazelskaya 2006: 117−148).

5. Inflection vs. derivation vs. transposition There is a long tradition in distinguishing between derivational nominalizations (like conquest, refusal and arrival ) and inflectional nominalizations, like English gerunds (Chomsky 1970). According to this tradition, derivational nominalizations are said to be lexically restricted, may have unpredictable semantic and syntactic properties as compared to the verbs they are derived from, as well as idiosyncrasies in formation; they also combine with their dependents in exactly the same way as prototypical non-derived nouns. Inflectional nominalizations are described as being formed from any verb in a regular way, their meaning is completely predictable and they retain a good portion of verbal syntactic properties. However, as should be evident from the discussion above (and as argued by Comrie 1976, Koptjevskaja-Tamm 1993: 263−266 and Haspelmath

67. Action nouns

1205

1996), action nouns often pose serious problems for a clear-cut distinction here. Koptjevskaja-Tamm (1993: 263−266) argues that the extreme points on the scale of nominalization patterns (Fig. 67.1 in section 3) demonstrate a high degree of consistency among Chomsky’s criteria in that action nouns with a sentence-like syntax normally involve generally applicable and regular markers and show semantic transparency, while action nouns with a nominal syntax often manifest various idiosyncrasies typical of derivation. For the intermediate patterns, however, Chomsky’s criteria often clash. Also Haspelmath (1996: 57−58) suggests a similar universal correlation: − In words derived by inflectional word-class-changing morphology, the internal syntax of the base tends to be preserved. − In words derived by derivational word-class-changing morphology, the internal syntax of the base tends to be altered and assimilated to the internal syntax of primitive members of the derived word-class. The discussion of the inflectional vs. derivational status of action nouns has primarily focused on their formal properties (regularity, productivity and generality in their formation, and their internal syntax) and on their semantics (predictability of meaning). There is, however, another important facet to this issue − their functions. In Bickel and Nichols’ (2007: 169) words, inflection covers “those categories of morphology that are regularly responsive to the grammatical environment in which they are expressed. Inflection differs from derivation in that derivation is a lexical matter in which choices are independent of the grammatical environment” (unfortunately, this definition seems to miss the important distinction between contextual and inherent inflection, cf. Booij 2006). The question is thus to what extent action nouns in a particular language are sensitive to the grammatical environment, in other words, whether there are any functions in the language where action nouns are used obligatorily, regularly or at least frequently. There is considerable cross-linguistic variation in how much and where action nouns are involved in their grammatical systems. In many languages, also in Europe, action nouns are among the predominant forms used in complementation, e.g., in the NorthEastern Caucasian languages (Kazenin 1994; Kalinina and Sumbatova 2007), in Turkish, in the Finno-Ugric languages Mari, Erzya, Udmurt, Komi (Serdobol’skaja et al. 2012), which would make them clearly inflectional according to the definition in Bickel and Nichols (2007). This distinguishes them from action nominals in the more familiar Germanic, Romance or Slavic languages, which also show considerable cross-linguistic variation as to whether action noun constructions are used as a less frequent alternative to another major complementation type, as a regular or even the only complementation type. Thus, in English, stop is one of the verbs that obligatorily take action nouns (gerunds) as complements, e.g., to stop reading / *to stop to read (in the intended meaning), while its close synonym cease takes only infinitives (to cease to read ). In Russian, the verb perestat’ ‘to stop’ cannot take action nouns as complements, the only option for the latter being infinitives, e.g., *perestat’ čtenie / perestat’ čitat’, while its close synonym prekratit’, as well as most of the other phasal verbs, allow combinations both with action nouns and with infinitives, e.g., prekratit’ čtenie / prekratit’ čitat’. However, in contrast to English, no predicates in Russian obligatorily take action nouns as complements, which would count as another argument for treating them as derivation. There is a huge theoretical literature on how to represent the mixture of verbal and nominal properties of action nouns and on how this mixture arises in the process of their formation. Most of it assumes the radically syntactic treatments of word-formation (in

1206

VII. Semantics and pragmatics in word-formation II: Special cases

the terms of Spencer 2006), which places word-formation in the syntax. An action noun starts its life as verb heading a verb phrase, then gradually raises and combines with a nominal head and starts behaving like a noun. Depending on the point at which this merging with a nominal head occurs, the resulting nominalizations may have different combinations of verbal and nominal properties. A few recent examples of such approaches include Borsley and Kornfilt (2000), Kornfilt and Whitman (2001) and the two collections edited by Alexiadou and Rathert (2010a, b). In lexical models of grammar, the categorial mixing takes place at the level of morphological representation. For instance, in the influential type hierarchy model by Malouf (2000), syntactic categories are defined in terms of an inheritance hierarchy and mixed categories are those which inherit from two other categories. Spencer (2013, ch. 8) argues that none of these models are capable of capturing the complex interplay of morphosyntactic factors involved in the derivation of action nouns. Action nouns represent a particularly fascinating example of transposition, which for Spencer is limited to operations that alter the word’s word class without altering its semantic representation and falls precisely between the traditional inflection and derivation. Spencer’s model builds on the notion of “lexical relatedness”, which compares lexemes according to the three cornerstones in their lexical representation − morphophonological form, syntactic category, and a representation of meaning − and is therefore capable of providing a coherent framework for various lexical phenomena (e.g., synonymy, polysemy, inflection, derivation, transposition, argument-structure operations, etc.) that have often been compartmentalized among different linguistic sub-disciplines. Many linguists would, however, contest Spencer’s view by arguing that word classes are not simply labels, but are also associated with a particular semantics (this is, for instance, the standard view within cognitive grammar, cf. Langacker 1987) and that conceptualizing a situation as a noun will necessarily affect its “meaning”.

6. References Alexiadou, Artemis and Monika Rathert 2010a The Syntax of Nominalizations across Languages and Frameworks. Berlin: Mouton de Gruyter. Alexiadou, Artemis and Monika Rathert 2010b The Semantics of Nominalizations across Languages and Frameworks. Berlin: Mouton de Gruyter. Bašić, Monika 2010 On the morphological make-up of nominalizations in Serbian. In: Artemis Alexiadou and Monika Rathert (eds.), The Syntax of Nominalizations across Languages and Frameworks, 39−66. Berlin: Mouton de Gruyter. Bhat, D. N. Shankara 1994 The Adjectival Category. Amsterdam/Philadelphia: Benjamins. Bickel, Balthasar and Johanna Nichols 2007 Inflectional morphology. In: Timothy Shopen (ed.), Language Typology and Syntactic Description. Vol. 3: Grammatical Categories and the Lexicon, 169−240. Cambridge: Cambridge University Press. Booij, Geert 2006 Inflection and Derivation. In: Keith Brown (ed.), Encyclopedia of Language and Linguistics. 2nd ed. Vol. 5, 654−661. Oxford: Elsevier.

67. Action nouns

1207

Borsley, Robert D. and Jacklin Kornfilt 2000 Mixed extended projections. In: Robert D. Borsley and Jacklin Kornfilt (eds.), The Nature and Functions of Syntactic Categories. Syntax and semantics. Vol. 32, 101−131. San Diego: Academic Press. Chomsky, Noam 1970 Remarks on nominalization. In: Roderick A. Jacobs and Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar, 184−221. Boston: Ginn. Comrie, Bernard 1976 The syntax of action nominals: A cross-linguistic study. Lingua 40: 177−201. Comrie, Bernard and Sandra A. Thompson 1985 Lexical nominalization. In: Timothy Shopen (ed.), Language Typology and Syntactic Description. Vol. 3: Grammatical Categories and the Lexicon, 349−398. Cambridge: Cambridge University Press. Cornilescu, Alexandra 2001 Romanian nominalizations: Case and aspectual structure. Journal of Linguistics 37: 467−501. Croft, William 2001 Radical Construction Grammar. Syntactic theory in typological perspective. Oxford: Oxford University Press. Disterheft, Dorothy 1980 The Syntactic Development of the Infinitive in Indo-European. Columbus, OH: Slavica Publishers. Fábregas, Antonio and Rafael Marín 2012 The role of Aktionsart in deverbal nouns: State nominalizations across languages. Journal of Linguistics 48: 35−70. Fortescue, Michael 1984 West Greenlandic. London: Croom Helm. Gaeta, Livio 1998 The inflection vs. derivation dichotomy: The case of German infinitives. In: Bernard Caron (ed.), Proceedings of the XVI International Congress of Linguists (Paris 20−25 July 1997), 1−19. Oxford: Pergamon. Grimshaw, Jane 1990 Argument Structure. Cambridge, MA: MIT Press. Haspelmath, Martin 1989 From purposive to infinitive − a universal path of grammaticization. Folia Linguistica Historica 10: 287−310. Haspelmath, Martin 1995 The converb as a cross-linguistically valid category. In: Martin Haspelmath and Ekkehard König (eds.), Converbs in Cross-linguistic Perspective. Structure and meaning of adverbial verb forms − adverbial participles, gerunds, 1−56. Berlin: Mouton de Gruyter. Haspelmath, Martin 1996 Word-class-changing inflection and morphological theory. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1995, 43−66. Dordrecht: Kluwer. Haspelmath, Martin and Andrea Sims 2010 Understanding Morphology. London: Hodder Education. Hopper, Paul and Sandra Thompson 1984 The discourse basis for lexical categories in universal grammar. Language 60: 703−752. Kalinina, Elena and Nina Sumbatova 2007 Clause structure and verbal forms in Nakh-Daghestanian languages. In: Irina Nikolaeva (ed.), Finiteness, 183−249. Oxford: Oxford University Press.

1208

VII. Semantics and pragmatics in word-formation II: Special cases

Kazenin, Konstantin 1994 Action nominal constructions in Godoberi. In: Aleksandr E. Kibrik (ed.), Godoberi’s Noun Phrase, 50−58. Strasbourg: European Science Foundation. Koptjevskaja-Tamm, Maria 1993 Nominalizations. London/New York: Routledge. Koptjevskaja-Tamm, Maria 2003 Action nominal constructions in the languages of Europe. In: Frans Plank (ed.), Noun Phrase Structure in the Languages of Europe, 723−759. Berlin/New York: Mouton de Gruyter. Koptjevskaja-Tamm, Maria 2005 Action nominal constructions. In: Martin Haspelmath, Matthew Dryer, David Gil and Bernard Comrie (eds.), World Atlas of Language Structures, 254−257. Oxford: Oxford University Press. Kornfilt, Jaklin and John Whitman (eds.) 2011 Nominalizations in Linguistic Theory [= Lingua 121(7): 1159−1314]. Langacker, Ronald W. 1987 Nouns and verbs. Language 63: 53−94. Lefebvre, Claire and Pieter Muysken 1987 Mixed Categories. Nominalizations in Quechua. Dordrecht: Kluwer. Malchukov, Andrej 2006 Constraining nominalization: Function/form competition. Linguistics 44/45: 973−1009. Malouf, Robert 2000 Mixed Categories in the Hierarchical Lexicon. Stanford University: CSLI. Melloni, Chiara 2010 Action nominals inside: Lexical-semantic issues. In: Artemis Alexiadou and Monica Rathert (eds.), The Semantics of Nominalizations across Languages and Frameworks, 141−168. Berlin: Mouton de Gruyter. Pazelskaya [Pazel’skaja], Anna 2006 Nasledovanie glagol’nyx kategorij imenami situacij (na materiale russkogo jazyka). Ph.D. dissertation, Lomonosov Moscow State University. Pazelskaya [Pazel’skaja], Anna 2012 Verbal prefixes and suffixes in nominalization: Grammatical restrictions and corpus data. In: Atle Grønn and Anna Pazelskaya (eds.), The Russian Verb [Oslo Studies in Language 4(1): 245−261 http://www.journals.uio.no/osla]. Pazelskaya [Pazel’skaja], Anna and Sergei Tatevosov 2006 Uninflected VPs, deverbal nouns and the aspectual architecture of Russian. In: James Lavine, Steven Franks, Mila Tasseva-Kurktchieva and Hana Filip (eds.), Formal Approaches to Slavic Linguistics 14. The Princeton Meeting 2005, 258−276. Ann Arbor: Michigan Slavic Publications. Perkova, Natalia 2009 Nominalizacii v kalmyckom jazyke. In: Sergej S. Saj, Vlada V. Baranova and Natalia N. Serdobol’skaja (eds.), Issledovanija po grammatike kalmyckogo jazyka, 464−496. Saint Petersburg: Nauka. Rozwadowska, Bożena 1997 Towards a Unified Theory of Nominalizations. External and Internal Eventualities. Wrocław: Wydawnictwo Uniwersytetu Wrocławskiego. Schoorlemmer, Maaike 1995 Participial passive and aspect in Russian. Ph.D. dissertation, Utrecht University. Serdobol’skaja, Natalia, Anfisa Il’evskaja, Sergej Minor, Polina Miteva, Aleksandra Fajnvejc and Natalia Matveeva 2012 Konstrukcii s sentencial’nymi aktantami v finno-ugorskix jazykax. In: Anna Kuznecova (ed.), Finno-ugorskie jazyki. Fragmenty grammatičeskogo opisanija. Formal’nyj i funkcional’nyj podxody, 382−476. Мoscow: Studia Philologica.

68. Action nouns in Romance

1209

Sleeman, Petra and Ana Maria Brito 2010 Aspect and argument structure of deverbal nominalizations: A split vP analysis. In: Artemis Alexiadou and Monika Rathert (eds.), The Syntax of Nominalizations across Languages and Frameworks, 199−229. Berlin: Mouton de Gruyter. Spencer, Andrew 2005 Word-formation and syntax. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-formation, 73−97. Dordrecht: Springer. Spencer, Andrew 2013 Lexical Relatedness. A Paradigm-based Model. Oxford: Oxford University Press. Spencer, Andrew and Marina Zaretskaya 1998 Stative predicates in Russian and their nominalizations. Essex research reports in linguistics 22: 1−44. Štekauer, Pavol, Salvador Valera and Lívia Körtvélyessy 2012 Word-Formation in the World’s Languages. A Typological Survey. Cambridge: Cambridge University Press. Vendler, Zeno 1967 Facts and events. In: Zeno Vendler (ed.), Linguistics in Philosophy, 12−146. Ithaca, NY: Cornell University Press. Vendler, Zeno 1970 Say what you think. In: Joseph Lloyd Cowan (ed.), Studies in Thought and Language, 79−97. Tucson, AR: University of Arizona Press. Ylikoski, Jussi 2003 Defining non-finites: Action nominals, converbs and infinitives. SKY Journal of linguistics 16: 185−237.

Maria Koptjevskaja-Tamm, Stockholm (Sweden)

68. Action nouns in Romance 1. 2. 3. 4. 5. 6.

Introduction The verb-noun continuum and the morphosyntactic environment Meaning and meaning extensions in Romance action nouns Semantic types and word-formation meaning in Romance action nouns Word-formation types in Romance action nouns References

Abstract Action nouns in Romance will be investigated with particular regard to (i) the format of the deverbal noun and of its morphosyntactic environment with respect to the corresponding verb-centered sentence, (ii) the range of meaning − and more in general of semantic properties − passing from the verbal input to the final noun, and finally (iii) the morphological processes of word-formation involved, namely suffixation and conversion.

1210

VII. Semantics and pragmatics in word-formation II: Special cases

1. Introduction Action nouns (henceforth ANs) are traditionally defined as nouns derived from verbs referring to the event described by the base predicate, i.e. Namen für Satzinhalte lit. ‘nouns for sentence contents’ according to Porzig’s (1930−31) classical definition. Porzig emphasizes the role played by the sentence serving as a reference. This is so because of the essential properties of the ANs which involve not only the predicate but also its arguments, as they surface in the corresponding sentences. In this light, a number of interesting questions arise relating (i) to the format of the deverbal noun and of its morphosyntactic environment with respect to the corresponding verb-centered sentence, (ii) to the range of meaning − and more in general of semantic properties − passing from the verbal input to the final noun, and finally (iii) to the morphological processes of word-formation in contrast with other processes of nominalization typically regarding the verbal infinitive. In what follows these three aspects will be discussed to the extent that they concern the Romance languages as a unitary domain of investigation. In this regard, it should not be forgotten that our perspective takes word-formation as its main object of analysis. Accordingly, the focus will be on those cases which are likely to be assigned to this domain.

2. The verb-noun continuum and the morphosyntactic environment As demonstrated by Koptjevskaja-Tamm (1993), action nominalizations, intended in a broad sense as the results of the transposition of a sentence into the syntactic role of a noun phrase, usually display a syntactic behavior mirroring their corresponding morphosyntactic prototypes (see article 67 on action nouns). Accordingly, they mark their syntactic dependencies adopting respectively a sentential or a nominal type, or a mixture of the two types often with its own special properties. The Romance languages generally select the nominal option for the deverbal nouns resulting from word-formation, while the sentential type commonly occurs with nominalized infinitives. In what follows, Italian will be used for exemplification (cf. Gaeta 2002, 2004), but similar examples can be found in the other Romance languages as well, unless differently specified (cf. Riegel, Pellat and Rioul 2006: 187−188 for French, Picallo 1999 for Spanish, Martí i Girbau 2002 for Catalan, Sleeman and Brito 2010 for Portuguese and Camacho and Santana 2004 for Brazilian Portuguese; for Romanian see Cornilescu 2001 and below). The nominal type directly mirrors the noun phrases encoding possession. This is true both for the way in which the corresponding arguments are encoded using genitive-like phrases (1a) and for their anaphorical reference which is normally accomplished with the help of possessive pronouns (1b): (1)

a. Il rifiuto di Giannii della veritàj the refusal of Gianni of:DET truth ‘Gianni’s refusal of the truth’

68. Action nouns in Romance

1211

b. Il suoi/j rifiuto the his/its refusal ‘his/its refusal’ c. Il suoi rifiuto della verità ‘his refusal of the truth’ d. *Il suoj rifiuto di Gianni the its refusal of Gianni Notice the asymmetry between subject and object reference, which favors the former over the latter when the respective counterparts are explicitly present (1c−d). A different coding for the subject is also common which normally marks the agentive complement in passive clauses (corresponding to Koptjevskaja-Tamm’s 1993 ergative-possessive type): (2)

a. Il rifiuto della verità da parte di Gianni the refusal of:DET truth from part of Gianni ‘Gianni’s refusal of the truth’ b. Il suo rifiuto da parte di Gianni the its refusal from part of Gianni ‘Gianni’s refusal of it’

In contrast to the nominal type, in the ergative-possessive type a possessive pronoun may be used for the object reference (cf. 1d vs. 2b). Furthermore, although it has often been claimed that the ergative-possessive type generally requires the subject complement to occur after the object complement (3a), both orders are actually possible depending on several factors, not the least of which is the length of the constituents involved (3b). A similar variation also regards the strictly nominal type (3c−d) (the examples are from a Google search of the Internet): (3)

a. ??Il rifiuto da parte di Gianni della verità ??the refusal from part of Gianni of:DET truth ‘Gianni’s refusal of the truth’ b. Il rifiuto da parte Gianni della verità riguardo a suo padre the refusal from part of Gianni of:DET truth regard to his father ‘Gianni’s refusal of the truth concerning his father’ c. ho iniziato a sbloccarmi dall’atavico rifiuto dei bambini have begun to become free from:DEF atavistic refusal of:DEF children dei vegetali of:DEF vegetables ‘I’ve begun to become free from children’s atavistic refusal of vegetables’ d. È stato valutato il tasso di rifiuto dei genitori dei bambini is been estimated the rate of refusal of:DEF parents of:DEF children campionati sampled ‘The rate of the sampled children’s refusal of the parents has been estimated’

1212

VII. Semantics and pragmatics in word-formation II: Special cases

On the other hand, the entirely nominal type is often avoided, being really available only with predicates displaying an object which remains unaffected by the consequences of the event (or, in other terms, a comparatively weaker agentive subject). In the other cases, the ergative-possessive type has to be used: (4)

a

La distruzione della città da parte dei Romani the destruction of:DEF town from part of:DET Romans ‘the Romans’ destruction of the town’

b. *La distruzione dei Romani della città the destruction of:DEF Romans of:DEF town It must be added that both complements can be normally dropped while in the presence of only one complement the genitive-like construction is generally preferred (but see Veland 2011: 71 for a contrasting view based on Italian data), even though it gives rise to ambiguous readings: (5)

a. ?La distruzione da parte dei Romani ‘the destruction by the Romans’ b. La distruzione dei Romani ‘the Romans’ destruction/the destruction of the Romans’

In contrast to this, Koptjevskaja-Tamm’s (1993) sentential type characterizes the nominalized infinitives, in which the sentential marking is directly mirrored: (6)

a. L’aver Gianni rifiutato la verità the have:INF Gianni refuse:PSTPTCP the truth ‘Gianni’s having refused the truth’ b. L’aver rifiutato la verità da parte di Gianni the have:INF refuse:PSTPTCP the truth from part of Gianni ‘Gianni’s having refused the truth’ c. *L’aver Gianni rifiutato della verità the have:INF Gianni refuse:PSTPTCP of:DEF truth

Furthermore, the subject can be referred to by means of the agentive complement corresponding to Koptjevskaja-Tamm’s (1993) ergative-accusative type (6b), while the possessive marking of the object in the presence of the sentential marking of the subject is excluded (6c). The difference between the AN and the nominalized infinitive is also expressed by the latter’s more pronounced verbal character given by the occurrence of verbal inflectional properties (see the past inflection in 6a−b) or the absence of nominal inflectional properties like pluralization (*i rifiutari vs. i rifiuti ‘the refusals’). Although the details may slightly vary, the picture is uniform throughout the Romance languages insofar as an AN corresponding to the nominal (or to the ergative-possessive) type is opposed to a nominalized infinitive corresponding to the sentential (or to the ergative-accusative) type and displaying verbal properties. This used to be true also for

68. Action nouns in Romance

1213

Old French, but in modern times the usage of the nominalized infinitive is far more restricted (see Kerleroux 1996: 66−85). Finally, in Romanian the so-called long infinitive has undergone a process of detachment from the verbal paradigm and has lost the typical syntactic properties of the verbal infinitives. Correspondingly, it has acquired nominal properties becoming a true AN and − similarly to its Romance counterparts − conforms to the purely nominal type (cf. Mallinson 1986: 214; see also article 151 on Romanian): (7)

învăţare-a limbilor străine de către englezi learning-DEF languages:DEF:GEN foreign:PL of towards English:PL ‘the acquisition of the foreign languages by the English’

A similar process of detachment from the verbal paradigm has also affected several types of earlier Latin present and past participles, which in their new status as ANs may have also become productive (see section 5). Furthermore, given that the Latin supine, i.e. the verbal noun, has expired in the passage to Romance, its position has radically changed, having only survived in the form of (masculine) ANs. Remarkably, the Latin gerund(ive), which also was originally part of the inflectional paradigm of the verbal noun, has not left any trace in the domain of ANs, except for sparse nominals like Italian/Spanish reprimenda/French réprimande ‘reprimand’, Spanish contienda ‘conflict’, French offrande/Spanish ofrenda ‘offering’, etc.

3. Meaning and meaning extensions in Romance action nouns The range of meanings displayed by the Romance ANs varies considerably. In the face of a basic eventive meaning, more or less mirroring the meaning of the verbal base in a sense to be further specified in section 4, a number of other, more “concrete” meanings are usually found. They are traditionally taken to result from a process of semantic extension, whereby the AN turns out to be adapted to the “concrete” meaning displayed by prototypical nouns. This process usually gives rise to a synchronic polysemy, in which the same noun may or may not display a concrete meaning and accordingly be used in a certain context. Related to this, there is also the question of the occurrence of further prototypical nominal properties like pluralization. Given the importance of the question, it will be treated before entering into the more specific issues concerning the word-formation categories and meanings occurring in the Romance ANs as word-formation patterns. The first aspect of the question concerns the origin of the polysemy. Clearly, the range of meanings to be associated with a word directly increases with the increase of its frequency, and in this sense ANs are not exceptional (cf. Thornton 1991: 86−90 and Gaeta 2002: 216−221 for first attempts to assess the question relying on Italian ANs). The traditional idea explains the polysemy by making reference to the abstract meaning carried by ANs, which renders them unfit with regard to the nominal prototype. Accordingly, a number of concrete meaning extensions come about which have the effect of improving the categorial status of the ANs from the semantic point of view. The extensions follow a metonymic path, insofar as the AN turns out to express the (abstract, i.e. FACT/STATE, or concrete) RESULT of an activity, or the PLACE where it’s supposed to generally take place, or the TIME or MANNER of its unfolding, or the MEANS used to

1214

VII. Semantics and pragmatics in word-formation II: Special cases

carry it out, or the PERSON who normally carries it out, etc. (cf. Rainer 1996 for a historical reconstruction of the question). An open question is whether the most extensively found RESULT meaning is due to an inner ambiguity of the derivative, i.e. as an ambiguity available by virtue of the semantics inherent in the noun itself, which may accordingly exhibit either the EVENT or the RESULT meaning (Spanish examples taken from Picallo 1999: 369): (8)

a. La evaluación de los datos de la encuesta tuvo lugar ayer. ‘The evaluation of the data of the investigation took place yesterday.’ b. La evaluación de los datos de la encuesta se consideró incorrecta. ‘The evaluation of the data of the investigation was considered incorrect.’

In the last years, several attempts have been made to characterize the polysemy in these terms, namely as a semantic shift inherent in the ANs qua semantically complex types of nouns, in the sense of Pustejovsky’s (1995) dot nouns (cf. Jacquey 2006 for French; Jezek and Melloni 2011 for Italian). The latter are introduced to account for the fact (also common to other nouns, see below) that ANs may display more than one meaning at the same time, and in particular the eventive and the resultative: (9)

La evaluación de los datos de la encuesta que tuvo lugar ayer se consideró incorrecta. ‘The evaluation of the data of the investigation which took place yesterday was considered incorrect.’

In this regard, it has to be observed that the word-formation patterns to be surveyed below are not inherently bound to a certain meaning. In other words, the polysemy observed in (8a−b) is not suffix-specific but affects ANs as a whole, although it may be that single suffixes display some of the meaning extensions more frequently than others. Furthermore, the polysemy pattern observed in (9) is not essentially different from those observed in other morphologically complex words (see article 74 on agent and instrument nouns) and even in simplexes (e.g., a word like Italian articolo ‘paper’ in L’articolo che non capisco è sul tavolo ‘The paper that I don’t understand is on the table’). If the EVENT/RESULT polysemy is considered inherent, basically relating to the availability of a direct (affected or effected) object in the argument structure of the verbal base (cf. Picallo 1999: 382−383), then the question arises as to whether this has consequences for the format of the single derivational processes forming ANs or rather is a general feature characterizing natural languages. A similar question concerns the abstract result of an event, namely the FACT/STATE shift, which is also identified by means of a passive(-like) meaning: Italian agitazione ‘act of shaking / fact of being agitated, unrest’ (cf. Fradin 2011 on French; Fábregas and Marín 2012 on Spanish). Probably, the issue will have to be settled empirically. On the one hand, cases can be mentioned of ANs going back to transitive verbs including a potentially effected object, which don’t display any RESULT meaning: Spanish acaparamiento ‘act of hoarding’, but *‘what is hoarded’, and similarly Spanish desmantelamiento ‘dismantling’ (cf. Rainer 1993: 215). On the other, it has been suggested that the inherent shift might be valid only for specific subclasses of verbs, and in particular verbs of creation, redescription,

68. Action nouns in Romance

1215

mental action, emission, appearance and change of state or place (cf. Melloni 2011 for Italian; the NGLE 2009: 357 for Spanish). However, counterexamples like Italian edificazione/French édification ‘act of building’, but *‘what is built’, etc., should make us leery of strong generalizations (cf. Gaeta 2002: 206 for Italian and Defrancq and Willems 1996 for French). The same applies to the FACT/STATE shift, for which conceivable extensions are not attested like Italian maltrattamento ‘act of ill-treating’, but *‘fact of being ill-treated’. A further question relates to how the polysemy should be conceived. Namely, whether the process of meaning extension takes place at the lexical level or at the more general level of the conceptual representation of the word. In this sense, one could think that the lexical status of a complement might be of relevance, whether of argumental or of circumstantial nature. Unfortunately, this question must be answered in negative terms, because on the one hand there are cases like Italian semina ‘sowing season’, affitto ‘rent (cost)’, etc., in which the meaning extension refers to circumstantial information, and on the other there are derivatives like Italian segatura ‘sawdust’, which does not refer to any argumental or circumstantial complement of the verbal base segare ‘to saw’. The specific meaning can only become available if the complete narrative frame of the event of sawing is accessed including the waste produced by the process. This evidence speaks in favor of a (holistic) model of meaning which also includes our world knowledge about the single event and its narrative frame (cf. Rainer 1993: 215 for a discussion on the basis of Spanish examples). Whether this can be done by models inspired by the detailed level of encyclopaedic information included into Pustejovsky’s (1995) qualia structure has to remain an open question (cf. Jezek 2008 for a discussion based on Italian). When the eventive meaning is lost, the reference to the argument structure of the basic verb becomes opaque, as shown by the Italian example in (10a): (10) a. ??La dichiarazione di innocenza da parte di Berlusconi che trovi sulla mia scrivania è ridicola. ‘The declaration of innocence by Berlusconi that you find on my desk is ridiculous.’ b. La dichiarazione di innocenza di Berlusconi che trovi sulla mia scrivania è ridicola. ‘The declaration of innocence of/by Berlusconi that you find on my desk is ridiculous.’ However, this is not necessarily the case when the nominal type is used as shown by (10b), in which Berlusconi can be the author of the declaration as well as simply its object. This clearly derives from the property of showing simultaneously the eventive and the resultative meaning observed above. In this light, the adoption of the ergativepossessive type in (10a) renders the construction containing the AN incompatible with a predicate requiring the resultative meaning. It has been repeatedly claimed that in Spanish only the AN with an eventive meaning may be directly modified by temporal adverbs, while its resultative counterpart needs genitive-like modifiers (cf. Rainer 1993: 214; Picallo 1999: 370): (11) a. La demonstración del teorema de Pitágoras (de) ayer por la tarde nos sorprendió. ‘The demonstration of Pitagoras’ theorem of yesterday evening surprised us.’

1216

VII. Semantics and pragmatics in word-formation II: Special cases b. La demonstración del teorema de Pitágoras *(de) ayer por la tarde es incorrecta. ‘The demonstration of Pitagoras’ theorem of yesterday evening is incorrect.’

Similar examples are also found in Italian and French, which clearly refer to the eventive meaning of the AN (from a Google search on the Internet): (12) a. il Ministro ... ha commentato la pubblicazione oggi sulla Gazzetta ufficiale dell’Unione europea della domanda per il riconoscimento della Vastedda ... un formaggio di pecora prodotto in alcune zone della Sicilia lit. ‘the minister ... has commented the publication today in the official journal of the EU of the application for the acknowledgement of the Vastedda ... a sheep cheese produced in Sicily’ b. Celle-ci a abouti à la publication aujourd’hui de critères pour la conduite du diagnostic et de la protection des enfants ‘This has led to the publication today of criteria for the attitude towards the diagnostics and the protection of children’ c. en attestent les publications aujourd’hui à intervalles réguliers de nombreuses “pièces de jeunesse” ‘this is attested by the publications today at regular intervals of several “youthful plays”’ Notice that the last example shows that an eventive meaning is also available when the AN is pluralized. The pluralization of an eventive AN is generally possible throughout the Romance languages in contrast to what happens with a nominalized infinitive (the Italian examples are from a Google search): (13) a. Numerose furono quindi le uccisioni di partigiani, ma anche di persone indifese come donne, vecchi e bambini. ‘Numerous were then the killings of partisans, but also of helpless people, like women, old persons and children.’ b. Le uccisioni di Bob Kennedy e Martin Luther King da parte di fanatici destrorsi aprirono la strada all’elezione di Nixon. ‘The killings of Bob Kennedy and Martin Luther King by rightist fanatics opened the way to the election of Nixon.’ c. Per fattori “interni”, si intendono gli indebolimenti delle difese dell’organismo, come l’indebolimento della razza per inseguire selezioni genetiche spinte a perseguire record di produzione. ‘By “internal” factors is meant the weakenings of the body’s defences, as for instance the weakening of the race in order to go after genetic selections aiming at pursuing a production record.’ This operation has different properties depending on the semantic profile of the verbal bases involved as well as of the syntactic environment surrounding the ANs (cf. Lombard 1930: 96−105; Lüdtke 1978: 75−77; Roodenburg 2010; Knittel 2011). In general, its

68. Action nouns in Romance

1217

semantic effect normally consists in multiplying the number of the events described by the base predicate with an effect of iterativity or habituality (13a), i.e., it behaves like a pluractional operator (cf. in general Laca 2006 and Iordăchioaia and Soare 2008 on Romanian). A second possible interpretation of a pluralized AN is distributive in the sense that it refers to an action taking place independently with respect to several distinct individuals explicitly mentioned in the text (13b), and finally a pluralized AN can refer to different instantiations of the same action which is decomposed on a scale of nuances (13c).

4. Semantic types and word-formation meaning in Romance action nouns It has been mentioned above that the ANs normally display a basic eventive meaning, more or less mirroring the meaning of the verbal base (cf. Kiefer 1998). The latter varies in a multi-faceted way but can be well tailored adopting Vendler’s (1967) four actional classes, namely states, activities, accomplishments and achievements. They are generally characterized by making reference to three semantic features: [±dynamic], [±durative] and [±telic]: these features define the actionality or aktionsart of the verbal bases, not to be confounded with the aspect, which is its discourse-framed representation. Accordingly, a telic predicate like an accomplishment may be represented as imperfective (and in this sense: detelicized, unbounded) in a given context: Mary was painting her room when her grandfather died. Given its lexical nature, it is not surprising that actionality will be of particular relevance for word-formation. At any rate, it has been pointed out that actionality cannot be measured out exclusively at the lexical level, but requires calling into play the syntactic level, namely the level of argument realization. This is so because the presence of arguments, and typically of objects, may provide a telic value to an activity, especially when they are explicitly “quantized” (cf. Krifka 1992), as in to smoke a cigarette in contrast to to smoke (cigarettes). This opens the still hotly debated question of the so-called optional arguments. For our purposes, in light of what has been observed above concerning the encyclopaedic knowledge necessary to interpret ANs like Italian segatura ‘sawdust’, it will be sufficient to consider whether an argument has to be present or not in the common interpretation of a predicate, and in the latter case, adopting Pustejovsky’s (1995: 62−67) terminology, it has to be considered a default argument which participates in the logical expression but is not necessarily expressed syntactically. The three actional features can help us to better specify what eventive meaning really means when applied to the Romance ANs. In fact, the three features enter both into the selection of the possible verbal bases with regard to the individual suffixes and into the determination of the final meaning of the AN. With their help, the selectional properties of Romance suffixes can be nicely isolated, insofar as the features may either carve out a set of verbal bases sharing the same property which is subsequently inherited by the AN, or form ANs showing sensible shifts of the values displayed by their verbal bases. In the first case, I will speak of the inner actionality of the word-formation pattern which selects a specific set of verbs, and in the second case of the outer actionality which is different from the verbal base (cf. Brinton 1995; Gaeta 2000, 2002, 2005a; Alexiadou 2010).

1218

VII. Semantics and pragmatics in word-formation II: Special cases

Correspondingly, there are in principle four possibilities inasmuch as a word-formation pattern can be (i) quite selective as for the inner actionality but neutral with regard to its effect on the outer actionality; or (ii) neutral with regard to its effect on the inner actionality (i.e. non-selective of a specific verbal set) but feature-shifting with regard to the outer actionality; or (iii) both selective and feature-shifting; or (iv) finally completely neutral in both regards. The Romance languages reflect this complex typology inasmuch as they display a rich set of word-formation patterns largely exploiting the actional features of the verbal bases. Especially when more word-formation patterns give rise to ANs on the basis of the same verb, their semantic profile can be neatly observed. Furthermore, the latter is particularly evident when syntactic tests are employed relating to their possible modifiers or to the predicates they can be combined with (container-predicates, cf. Bartsch 1986; Gaeta 2002: 114−126 for Italian and Huyghe and Marín 2007 for French and Spanish). In spite of a recent increase of attention, this domain remains largely unexplored. Therefore, the quick survey presented below can only constitute the premise of a future systematic investigation. Starting with the first case in which a word-formation pattern displays a specific inner actionality, the stative verbs, characterized by the feature [−dynamic], stand in contrast to the dynamic bases insofar as their corresponding ANs are mostly selected in Italian, French and Spanish by the cognate suffixes -anza/-enza, -ance/-ence and -ancia/-encia (cf. respectively Gaeta 2002: chap. 4, Dal and Namer 2010, and the NGLE 2009: 405): Italian abbondare/French abonder/Spanish abundar ‘to abound’ → abbondanza/abondance/abundancia, French/Italian preferire/préférer/Spanish preferir ‘to prefer’ → preferenza/préférence/preferencia. Although deviations from this generalization are attested (cf. for instance Italian partenza ‘departure’, French délivrance ‘relief, issue’, Spanish influencia ‘influence’), in the case that two homonymous verbs displaying different actional values occur, only the stative ones form the ANs by means of this suffix, while the others select another word-formation pattern: (14) a. competenza ‘competence’ vs. competizione ‘competition’, conseguenza ‘consequence’ vs. conseguimento ‘attainment’, discendenza ‘descent, offspring’ vs. discesa ‘drop’, etc. b. A Mario compete il posto di direttore. ‘The post as director belongs to Mario.’ c. Mario e Andrea competono per il posto di direttore. ‘Mario and Andrea are competing for the post as director.’ d. La competenza/*competizione a Mario del posto di direttore è fuori discussione. ‘Mario’s entitlement to the post as director is beyond discussion.’ e. La *competenza/competizione di Mario e Andrea per il posto di direttore è molto forte. ‘The competition between Mario and Andrea for the post as director is very stiff.’ In these pairs of Italian ANs, the former is based on the stative meaning of the verb (14b), while the latter on the dynamic one (14c). The corresponding AN reflects the

68. Action nouns in Romance

1219

opposition between the stative meaning selecting -enza (14d) and the dynamic one selecting -zione (14e). Moreover, the ANs inherit the stative actionality of the base verbs insofar as they are only compatible with those (imperfective) container predicates which don’t focus on a telic state of affairs: (15) a. La prevalenza del male sul bene è stata interrotta. ‘The prevalence of evil over good has been interrupted.’ b. *La permanenza di Gianni a Roma si compì in due anni. ‘Gianni’s stay in Rome took place in two years.’ In the latter example, the perfective container predicate has to focus on a final state which cannot be brought about by the AN. Clearly, when in this case one speaks of an “eventive” meaning of the AN, one rougly means that the ANs inherit the inner nondynamic actionality of their verbal bases. This has to be kept distinct from what has been said above about the possible semantic shift EVENT > FACT/STATE characterizing ANs that are not necessarily formed from stative verbs. The weakness of this generalization is partly due to the low degree of productivity of this suffix with verbal bases. In this regard, it has to be added that many adjectives that go back to a Latin present participle normally form their abstract noun with the help of the same suffix: Italian arrogante/arroganza, Spanish arrogante/arrogancia, French arrogant/arrogance, etc. In this light, a double motivation lurks here because of the adjectival nature of the present participle which is at the heart of the derivational process (see section 5). Thus, the emergence of this selective preference may be seen (also in diachronic terms) as resulting from the equivalence of the full verb and the periphrasis formed by the present participle and the copula: Mario dipende/è dipendente da sua madre ‘Mario depends on his mother’, which explains the deadjectival/deverbal bivalence of many such abstracts (cf. Rainer 1989: 211). A similar impact of the inner actionality has been suggested for the so-called Romanian long infinitive which selects only telic verbal bases to form an AN while the atelic bases have to resort to the so-called supine (cf. Cornilescu 2001): (16) a a a a

călători ‘to travel’ locui ‘to live’ munci ‘to work’ rîde ‘to laugh’

→ → → →

*călătorirea/călătoritul *locuirea/locuitul *muncirea/muncitul *rîderea/rîsul

The second case may be illustrated by means of the second set of ANs contrasting with the stative ones in (14a) above, namely those formed with the suffixes -zione and -mento. While they don’t seem to be sensitive to the actionality of the verbal bases − i.e. in our terms: their inner actionality is un(der)specified − they generally bring about a shift in the outer actionality insofar as they are positively specified for telicity. Clearly, this effect is only visible with ANs derived from atelic verbal bases, because in the other cases telicity may be said to be inherited from the base. The contrast is particularly well enhanced when ANs based on atelic verbs and combined with a perfective container predicate are compared to their corresponding nominalized infinitives (17a−b), which

1220

VII. Semantics and pragmatics in word-formation II: Special cases

generally show the opposite effect of enhancing the inner dynamics of an event while the final change of state is backgrounded. In fact, the contrast in (17c−d) shows that the nominalized infinitive of the telic verb affondare ‘to sink’ is not compatible with a perfective container predicate: (17) a. L’insegnamento del latino è stato completato. ‘The teaching of Latin has been completed.’ b. *L’insegnare il latino è stato completato the teaching:INF the Latin has been completed c. L’affondamento della nave si compì in mezz’ora. ‘The sinking of the ship took place in half an hour.’ d. *L’affondare la nave si compì in mezz’ora. the sinking:INF the ship took place in half an hour A further case is given again by Romanian, in which the supine can be shown to provide telic verbal bases with a habitual reading which makes them compatible with a prepositional phrase normally focusing on the unbounded character of the event like timp de ‘for (X time)’ (cf. Alexiadou, Iordăchioaia and Schäfer 2011): (18) a. sositul lui Ion cu întîrziere arrive:SUP:DEF his John with delay ‘John’s (habit of ) arriving late’ b. sositul lui Ion cu întîrziere timp de 3 ani for 3 years arrive:SUP:DEF his John with delay ‘John’s (habit of ) arriving late for 3 years’ c. ??sosirea lui Ion cu întîrziere timp de 3 ani for 3 years arrive:INF:DEF his John with delay This stands in contrast to the ANs formed by means of the so-called long infinitive which do not have a habitual reading and accordingly do not admit the prepositional phrase (18b−c). The French suffix -age has been recently analyzed as sensitive not only to the actional property of the verbal base but also to its argument structure, insofar as the prominence of the agent is relevant (cf. Kelling 2003; Martin 2010): (19) a. Le décollement des tuiles par le vent/l’ouvrier ‘The loosening (lit. unsticking) of the tiles by the wind/the worker’ b. Le décollage des tuiles par ??le vent/l’ouvrier ‘The taking off (lit. unsticking) of the tiles by the wind/the worker’ In contrast with the ANs formed with -ment, the ones with -age “are more agentive [...] because they systematically signal the existence of an (intentional) action, either at the beginning of the denoted eventive chain, or upstream” (Martin 2010: 124), and in fact they are not compatible with an impersonal agent in (19b).

68. Action nouns in Romance

1221

The most complex case of interaction of verbal actionality with the word-formation patterns forming ANs is represented by the derivatives based on the feminine form of the past participle as they occur in Italian (cf. Mayo et al. 1995; Gaeta 2000, 2002: chap. 5; Heusinger 2005), Spanish (especially in non-European varieties, cf. Rainer 1993: 437− 440; the NGLE 2009: 382−390), Portuguese (Brito 2005; Vieira 2009), Catalan (Lüdtke 1978: 233) and, more restrictedly, French (cf. Lüdtke 1978: 137: “regressively productive”, but see Ferret, Soare and Villoing 2010; it also stands in competition with the suffix -ade borrowed from Provençal: baignade, grillade, etc. cf. Dubois and DuboisCharlier 1999: 40−41). This case is interesting from several points of view. In diachronic terms, this word-formation pattern results from the reanalysis of an original inflectional form, the feminine form of the (Latin) past participle while the process of regrammaticalization is not yet well understood. In synchronic terms, this type forms ANs displaying a semelfactive meaning (similar to the so-called nomen vicis of traditional Arabic grammar, cf. Gaeta 2009), which has peculiar properties from the point of view of the inner and the outer actionality. In fact, it combines the selection of a specific inner actionality, inasmuch as accomplishments are generally avoided as shown by the Italian examples in (20a), with the introduction of a precise outer actionality whereby the event is represented as bounded as shown by the French examples in (20b−c) (cf. Ferret, Soare and Villoing 2010): (20) a. dormire ‘to sleep’ lavare ‘to wash’ costruire ‘to build’ formare ‘to form’

→ → → →

dormita lavata *costruita *formata

b. Le perçage/??la percée du tunnel a progressé. the drilling/the drill:FEM:PSTPTCP of:DEF tunnel has made progress c. après ??le pesage/la pesée after the weighing/the weigh

FEM:PSTPTCP

du bébé of:DEF child

The derivatives formed by means of the French suffix -age are compatible with imperfective container predicates, which makes them similar to the Italian examples seen in (17b) and (17d) above. They stand in neat contrast with the derivatives formed by means of the feminine past participle which are normally incompatible with imperfective contexts (20b) while they combine with prepositions requiring a clearly bounded interpretation of the event like après ‘after’ (20c). Abstracting away from the details characterizing the single languages, one can tentatively identify for all the Romance languages in which this pattern occurs (with the remarkable exception of Romanian) a common function of the packaging operator (cf. Jackendoff 1991), which has the effect of isolating a single portion of the activity referred to by the verbal base and turning it into a bounded, countable noun. In doing so, the packaging operator gives rise to an AN which describes the isolated portion of the activity as occasional and carried out roughly or in an imprecise and hurried manner. This semantic nuance is particularly evident when these ANs are used in combination with light verbs, typically French faire/Italian fare/Portuguese fazer/Spanish hacer ‘to do’, etc. for intransitive bases and French donner/Italian dare/Portuguese dar/Spanish

1222

VII. Semantics and pragmatics in word-formation II: Special cases

dar ~ pegar, etc. ‘to give’ for transitive bases; cf. the Italian pair in (21) (example from Google), in which the periphrastic structure characterizes the event as imprecise: (21) a. Io gli dò una piegata veloce e le butto nel cassetto. I them:DAT give a fold:AN quick and them throw in:DEF drawer ‘I give them a quick fold and I throw them into the drawer.’ b. ??Io gli dò una piegata accurata e le butto nel I them:DAT give a fold:AN accurate and them throw in:DEF cassetto. drawer One important side-effect of the use of these ANs in verbal periphrases containing light verbs like DO and GIVE is that the ANs can be easily modified by means of evaluative suffixes, very productive at least in Italian, Spanish, Catalan and Portuguese, in this way contributing to the expressive character of the utterance (cf. article 62 on the pragmatics of word-formation): Gianni stava facendo una bella dormit-ina, ma sua moglie lo costrinse a una levat-accia ‘Gianni was having a nice sleep-DIM, but his wife forced him to getting-PEJ up’. In combination with achievements the semelfactive meaning might be seen as directly coming from the base given its punctual nature, but observe that the AN formed in these cases is generally not flanked by other ANs formed with other suffixes, while this is normally possible with the other bases as in the following Spanish examples (cf. Gaeta 2002: 174−176 for Italian examples): (22) a. caer ‘to fall’ llegar ‘to arrive’ salir ‘to leave’ venir ‘to come’

→ → → →

caída/*caimiento/*caición llegada/*llegamiento/*llegación salida/*salimiento/*salición venida/*venimiento/*venición

b. calentar ‘to heat’ investigar ‘to investigate’ llamar ‘to call’ mirar ‘to look’

→ → → →

calentamiento/calentada investigación/investigada llamamiento/llamada miramiento/mirada

It has to be added that the achievements often are also associated with an unaccusative argument structure, which is mirrored in these derivatives when the verbal bases allow an unaccusative alternation, as in the following examples from French (cf. Ferret, Soare and Villoing 2010): (23) a. Marie a percé son abcès. → le perçage de l’abcès ‘Marie burst her abscess.’ b. Son abcès a percé. → la percée/*le perçage de l’abcès ‘Her abscess burst.’ c. L’arrivage/??l’arrivée des ouvriers a été interrompu(e) par un convoi de police. ‘The arrival of the workers was interrupted by a police crew.’

68. Action nouns in Romance

1223

Moreover, the unbounded value of the French suffix -age, already observed above in (20b−c), makes it compatible with punctual verbs but only on the condition that the event is seen as in progress and pluractional: in fact, in (23c) the imperfective container predicate interrompre presupposes an unbounded AN while the plural-dependent genitive requires a pluractional interpretation of the AN, in which several arrivals are understood. The salience of the semelfactive meaning is supported in Spanish by other ANs formed by a number of suffixes, which in general are extremely productive with nominal bases, and in particular -azo: frenar ‘to brake’/freno ‘brake’ → frenada/frenazo ‘sudden braking’, pinchar ‘to prick’/pincho ‘prickle’ → pinchada/pinchazo ‘prick, shot (colloq.)’ (cf. the NGLE 2009: 398), the suffix -ón: acelerada/acelerón ‘burst of acceleration’, atracada/atracón ‘blowout (meal)’, etc. (cf. the NGLE 2009: 397). Furthermore, extensions to nominal bases with the form of the participial ending of the 1st verbal class (Catalan/Spanish/Portugese -ada, Italian -ata, French -ée, etc.) are common and productive (cf. Acquaviva 2005 on Italian; Cabré 2002: 745−746 on Catalan; Dubois and Dubois-Charlier 1999: 207−209 on French; the NGLE 2009: 390−394 on Spanish). They generally share a basic meaning ‘single instantiation typical of N’: thus, ‘a blow of N’ with typical blowing objects (Catalan martell/Spanish martillo ‘hammer’ → martellada/martillada), or body parts (Catalan colze/Italian gomito ‘elbow’ → colzada/gomitata), etc. From a comparative point of view, it is interesting to observe that, even though the two word-formation patterns are cognate, the so-called Romanian supine, which goes back to the masculine form of the Latin past participle, also manipulates inner and outer actionality as seen above in (18), but in the opposite way, namely with the supine giving rise to an unbounded AN while the long infinitive forms a bounded AN (cf. Gaeta 2009). A final word has to be added with regard to the fourth type described above, in which the AN is completely neutral with regard to both the inner and the outer actionality. In this respect, the nominalized infinitive was already mentioned above which is generally possible with any verbal base, while it does not seem to force any shift in its semantic properties, except for a certain effect of backgrounding the telic component if the container predicate focuses on the inner dynamics of the event (Gaeta 2002: 123−124): (24) a. L’ottenere il rimborso ci portò via due ore. the-obtain:INF the reimbursement us brought away two hours ‘Obtaining the reimbursement took us two hours.’ b. ??L’ottenimento del rimborso ci portò via due ore. the-obtain:AN of:DEF reimbursement us brought away two hours This can be observed fairly well with achievements, which combine with the nominalized infinitive emphasizing the duration of the events leading to the resultant state, while the AN, in contrast, is odd. Besides the actional content of the verbal bases, there are other factors tailoring the selective properties of the ANs. One such factor is the denotative or ontological domain normally associated with a specific word-formation pattern. Thus, for instance in Italian the suffix -aggio seems to cover the technical domain while -tura rather expresses the manual/craftsmanship one. Similarly, it has been assumed that part of the diffusion of French -age (for instance in contrast with -ment) is due to a similar specialization. This has not precluded -age from developing a special actional profile, as was briefly dis-

1224

VII. Semantics and pragmatics in word-formation II: Special cases

cussed above (cf. Uth 2011: chap. 5 for a diachronic perspective). To this it should be added that the suffix -age is opposed to the other suffixes also because of the selected ontological domain insofar as it generally combines only with verbs referring to physical events as we have seen above, thus excluding psychological verbs: penser ‘to think’ → ??pensage, préoccuper ‘to worry’ → ??préoccupage, effrayer ‘to frighten’ → ??effrayage, imaginer ‘to imagine’ → ??imaginage, etc. (cf. Martin 2010). Clearly, this factor has effects also on the frequency of the ANs in certain text types and registers, especially under the influence of foreign (English) models. Thus, the French suffix -ance/-ence, as well as its Spanish -ancia/-encia and Italian -anza/-enza cognates, apparently show a significant expansion among verbal bases as a consequence of its usage in certain special contexts (physics: French résistance ‘electric resistence’, perditance ‘leakance’, music: Italian cadenza, Spanish cadencia ‘rhythm’, etc.), giving also rise to interesting cases of blending, as in French varistance ‘resistence to electric (varying) tension’, thermistance ‘resistence to temperature’, etc. (cf. Dubois and DuboisCharlier 1999: 268). Other word-formation meanings are more idiosyncratic. In this regard, the Catalan suffix -era as well as its Occitan cognate are worth mentioning because they normally select unergative verbs of the 1st inflectional class and form a desiderative AN: Catalan plorar ‘to cry’ → plorera ‘desire to cry’, fumar ‘to smoke’ → fumera ‘desire to smoke’, Occitan badalhar ‘to dance’ → badalhèra ‘desire to dance’, parlar ‘to speak’ → parlèra ‘desire to speak’, etc. (cf. Gràcia and Riera 2003).

5. Word-formation types in Romance action nouns Let us now turn to the formal part of the word-formation patterns employed to form ANs in the Romance languages. Generally, suffixation and conversion are used, with the addition of a few (and partly controversial) instances of compounds (see article 39 on verb-noun compounds in Romance), and especially reduplicative compounds like French cache-cache ‘hide and seek; lit. hide-hide’, Italian fuggifuggi ‘stampede; lit. run awayrun away’, etc. (cf. Thornton 2008). Suffixations can be grouped into three further subtypes: (i) suffixes clearly selecting the verbal stem to form the AN; (ii) suffixes selecting another morphome, i.e. a purely morphologically conditioned alternant devoid of any but morphological function (cf. Aronoff 1994: 25); and (iii) derivatives based on participial morphomes diachronically evolved into some sort of suffix-like formative. The last subtype has also been interpreted as a conversion. Let us discuss the three subtypes in a more detailed way. As for the first subtype, the verbal stem, i.e. the verbal root plus the so-called thematic vowel, is selected by the Italian/Portuguese -mento, Spanish -miento suffixes to productively form ANs: Italian reclutare/Portuguese recrutar/Spanish reclutar ‘to recruit’ → reclutamento/recrutamento/reclutamiento, Italian muovere/Portuguese/Spanish mover ‘to move’ → Italian/Portuguese movimento/Spanish movimiento, Italian accudire ‘to attend’/ Portuguese/Spanish seguir ‘to follow’ → Italian accudimento/Portuguese seguimento/ Spanish seguimiento, etc. The thematic vowel varies depending on the verb’s inflectional class: verbs belonging to the 1st class select -a- while the others select -i-. The second subtype gives rise to different allomorphic patterns depending on the morphome selected by the suffix. In general, two main morphomes are at stake, namely

68. Action nouns in Romance

1225

those underlying respectively the present and the past participle of the verbs (see article 44 on paradigmatically determined allomorphy). The third subtype generally underlies the cases of ANs going back to the feminine form of the past participle briefly discussed in section 4. These types might be considered cases of conversion from the past participle, and in fact there is no binding argument speaking against this interpretation, because many ANs can be shown to rely on the morphome of the past participle instead of the Latinate morphome (e.g., Italian comparire ‘to appear’/part. comparso → comparsa, Spanish poner ‘to pose’/part. puesto → puesta, but Spanish ofender ‘to offend’/part. ofendido, AN ofensa). At any rate, the overwhelming productivity of these types with verbs of the 1st class in Catalan, Italian, Portuguese and Spanish makes it possible to isolate a suffix -ada/-ata, which also underlies its denominal cognate briefly seen in section 4. As for the conversions, a masculine and a feminine pattern generally occur, which also display different degrees of productivity (cf. respectively French rêver ‘to dream’ → rêve, Italian saltare ‘to jump’ → salto, chorar ‘to cry’ → choro, Spanish incendiar ‘to burn’ → incendio, and French annoncer ‘to announce’ → annonce, Italian deliberare ‘to deliberate’ → delibera, Portuguese perder ‘to lose’ → perda, Spanish comprare ‘to buy’ → compra). Different interpretations have been provided for these types. Supporters of the suffixal treatment simply assume two different suffixes -a and -o for Italian and Spanish (and a zero suffix for French), in which however a systematic collapse of the inflectional (the gender and inflectional class assignment) and derivational (wordclass change) properties is observed. To cope with this collapse two different processes of conversion have been assumed which are overtly manifested by the different vowels expressing gender and inflectional values: saltare ‘to jump’ → [[salt]V → NMasc -o], deliberare ‘to deliberate’ → [[deliber]V → NFem -a]. An intermediate approach suggested for Italian (cf. Thornton 2004: 516−520, 525, and the discussion in Kerleroux 1996: 235− 236 for French) distinguishes a process of root-based conversion (or suffixation), cf. saltare ‘to jump’ → [salt]Root-o, from a process of stem-based conversion which results from truncation: deliberare ‘to deliberate’ → [[deliber-a]Stem+zione] → [deliber-a]Stem. A final word must be added with regard to ANs as bases for further derivation. Clearly, ANs can normally be further derived into (relational) adjectives (although single patterns may be recalcitrant like the derivatives formed by French -age, Italian -aggio, Portuguese -agem, Spanish -aje: French aborder/Italian abbordare/Portuguese/Spanish abordar ‘to board’ → abordage/abbordaggio/abordagem/abordaje, French recycler/Italian riciclare/Portuguese/Spanish reciclar ‘to recycle’ → recyclage/riciclaggio/reciclagem/reciclaje, etc. which qualify as “closing” suffixes, cf. Gaeta 2005b). More interesting is the issue of the ANs as bases for further verbalization processes, in particular conversions (cf. Iacobini 2005 on Italian). In fact, ANs are possible bases for verbalizations only on the condition that the semantic relation with the extant base has become opaque or bleached: Italian sancire ‘to sanction, establish’ → sanzione ‘sanction’ → sanzionare ‘to sanction, punish’, Spanish sugerir ‘to suggest’ → sugestión ‘suggestion’ → sugestionar ‘to influence’, etc. The limited conversion possibilities of ANs have clearly to do with lexical blocking: their expected meaning would be too similar to the meaning of the verbal base.

1226

VII. Semantics and pragmatics in word-formation II: Special cases

6. References Acquaviva, Paolo 2005 I significati delle nominalizzazioni in -ATA e i loro correlati morfologici. In: Maria Grossmann and Anna M. Thornton (eds.), La formazione delle parole. Atti del XXXVII Congresso Internazionale di Studi della Società di Linguistica Italiana, 7−29. Roma: Bulzoni. Alexiadou, Artemis 2010 Nominalizations: A probe into the architecture of grammar. Part I: The nominalization puzzle. Language and Linguistics Compass 4(7): 496−511. Alexiadou, Artemis, Gianina Iordăchioaia and Florian Schäfer 2011 Scaling the variation in Romance and Germanic nominalizations. In: Petra Sleeman and Harry Perridon (eds.), The Noun Phrase in Romance and Germanic. Structure, Variation, and Change, 25−40. Amsterdam/Philadelphia: Benjamins. Aronoff, Mark 1994 Morphology by Itself. Cambridge, MA: MIT Press. Bartsch, Renate 1986 On aspectual properties of Dutch and German nominalizations. In: Vincenzo Lo Cascio and Co Vet (eds.), Temporal Structure in Sentence and Discourse, 7−39. Dordrecht: Foris. Brinton, Laurel 1995 The aktionsart of English deverbal nominalizations. In: Pier Marco Bertinetto, Valentina Bianchi, James Higginbotham and Mario Squartini (eds.), Temporal Reference, Aspect and Actionality. Vol. 1, 27−42. Torino: Rosenberg & Sellier. Brito, Ana Maria 2005 Nomes derivados de verbos inacusativos: Estrutura argumental e valor aspectual. Revista da Faculdade de Letras − Línguas e Literaturas Modernas [2nd series] 22: 47−64. Cabré, M. Teresa 2002 La derivació. In: Joan Solà, Maria-Rosa Lloret, Joan Mascarò and Manuel Pérez Saldanya (eds.), Gramàtica del català contemporani. Vol. 1: Introducció. Fonètica i fonologia. Morfologia, 731−775. Barcelona: Empúries. Camacho, Roberto G. and Liliane Santana 2007 Argument structure of deverbal nouns in Brazilian Portuguese. Journal of Language and Linguistics 3(2): 229−242. Cornilescu, Alexandra 2001 Romanian nominalizations: Case and aspectual structure. Journal of Linguistics 37: 467−501. Dal, Georgette and Fiammetta Namer 2010 Les noms en -ance/-ence du français: Quel(s) patron(s) constructionnel(s)? In: Franck Neveu, Valelia Muni Toke, Thomas Klingler, Jacques Durand, Lorenz Mondada and Sophie Prévost (eds.), 2 ème Congrès Mondial de Linguistique Française − CMLF 2010, 893−907. Paris: Institut de Linguistique Française. Defrancq, Bart and Dominique Willems 1996 De l’abstrait au concret: Une réflexion sur la polysémie des noms abstraits. In: Nelly Flaux, Michel Glatigny and Didier Samain (eds.), Les noms abstraits. Histoire et théories, 221−239. Villeneuve d’Ascq: Presses Universitaires du Septentrion. Dubois, Jean and Françoise Dubois-Charlier 1999 La dérivation suffixale en français. Paris: Nathan. Fábregas, Antonio and Rafael Marín 2012 The role of Aktionsart in deverbal nouns: State nominalizations across languages. Journal of Linguistics 48: 35−70.

68. Action nouns in Romance

1227

Ferret, Karen, Elena Soare and Florence Villoing 2010 Rivalry between French -age and -ée: The role of grammatical aspect in nominalization. In: Maria Aloni, Harald Bastiaanse, Tikitu de Jager and Katrin Schulz (eds.), Logic, Language and Meaning. Proceedings of the 17 th Amsterdam Colloquium, 284−294. Berlin/Heidelberg: Springer. Fradin, Bernard 2011 Remarks on state denoting nominalizations. Recherches Linguistiques de Vincennes 40: 73−99. Gaeta, Livio 2000 On the interaction between morphology and semantics: The Italian suffix -ATA. Acta Linguistica Hungarica 47(1−4): 205−229. Gaeta, Livio 2002 Quando i verbi compaiono come nomi. Milano: Angeli. Gaeta, Livio 2004 Nomi d’azione. In: Maria Grossmann and Franz Rainer (eds.), La formazione delle parole in italiano, 314−351. Tübingen: Niemeyer. Gaeta, Livio 2005a Thoughts on cognitive morphology. In: Gertraud Fenk-Oczlon and Christian Winkler (eds.), Sprache und Natürlichkeit. Gedenkband für Willi Mayerthaler, 107−128. Tübingen: Narr. Gaeta, Livio 2005b Combinazioni di suffissi in italiano. In: Maria Grossmann and Anna M. Thornton (eds.), La formazione delle parole. Atti del XXXVII Congresso Internazionale di Studi della Società di Linguistica Italiana, 229−247. Roma: Bulzoni. Gaeta, Livio 2009 A metà tra nomi e verbi: I nomi d’azione tra morfologia, sintassi e semantica. In: Edoardo Lombardi Vallauri and Lunella Mereu (eds.), Spazi linguistici. Studi in onore di Raffaele Simone, 111−123. Roma: Bulzoni. Gràcia, Lluïsa and Laura Riera 2003 A propos des noms déverbaux avec le suffixe -era du catalan. Cahiers de Grammaire 28: 153−161. Heusinger, Klaus von 2005 Morphology and the interface between conceptual structure and lexical semantics: The case of Italian nominalization of -ata. Lingue e linguaggio 4(2): 229−242. Huyghe, Richard and Rafael Marín 2007 L’héritage aspectuel des noms déverbaux en français et en espagnol. Faits de Langues 30: 265−274. Iacobini, Claudio 2005 Restrizioni sulla formazione di verbi denominali: Il caso dei nomi di azione. In: Nicola Grandi (ed.), Morfologia e dintorni, 69−85. Milano: Angeli. Iordăchioaia, Gianina and Elena Soare 2008 Two kinds of event plurals: Evidence from Romanian nominalizations. In: Olivier Bonami and Patricia Cabredo Hofherr (eds.), Empirical Issues in Syntax and Semantics 7: 193−216. http://www.cssp.cnrs.fr/eiss7 [last access 20 Sept 2012]. Jackendoff, Ray 1991 Parts and boundaries. Cognition 41: 9−45. Jacquey, Evelyne 2006 Un cas de “polysémie logique”: Modélisation de noms d’action en français ambigus entre processus et artefact. Traitement Automatique des Langues 47(1): 137−166. Jezek, Elisabetta 2008 Polysemy of Italian event nominals. Faits de Langues 30: 251−264.

1228

VII. Semantics and pragmatics in word-formation II: Special cases

Jezek, Elisabetta and Chiara Melloni 2011 Nominals, polysemy, and co-predication. Journal of Cognitive Science 12: 1−31. Kelling, Carmen 2003 The role of agentivity for suffix selection. In: Geert Booij, Janet DeCesaris, Angela Ralli and Sergio Scalise (eds.), Topics in Morphology. Selected Papers from the Third Mediterranean Morphology Meeting, 197−210. Barcelona: Institut universitari de lingüística aplicada, Universitat Pompeu Fabra. Kerleroux, Françoise 1996 La coupure invisible. Villeneuve d’Ascq: Presses Universitaires du Septentrion. Kiefer, Ferenc 1998 Les substantifs déverbaux événementiels. Langages 32 [numéro 131]: 56−63. Knittel, Marie Laurence 2011 French event nominals and number-inflection. Recherches linguistiques de Vincennes 40: 127−148. Koptjevskaja-Tamm, Maria 1993 Nominalizations. London: Routledge. Krifka, Manfred 1992 Thematic relations as links between nominal reference and temporal constitution. In: Ivan Sag and Anna Szabolcsi (eds.), Lexical Matters, 29−53. Stanford: CSLI. Laca, Brenda 2006 Pluriactionnalité. In: Danièle Godard, Laurent Roussarie and Francis Corblin (eds.), Sémanticlopédie. Dictionnaire de sémantique. GDR Sémantique and Modélisation, CNRS, http://www.semantiquegdr.net/dico/ [last access 21 Nov 2014]. Lombard, Alf 1930 Les constructions nominales dans le français moderne. Étude syntactique et stylistique. Uppsala/Stockholm: Almqvist and Wiksell. Lüdtke, Jens 1978 Prädikative Nominalisierungen mit Suffixen im Französischen, Katalanischen und Spanischen. Tübingen: Niemeyer. Mallinson, George 1986 Rumanian. London: Croom Helm. Martí i Girbau, Núria 2002 El SN: Els noms. In: Joan Solà, Maria-Rosa Lloret, Joan Mascarò and Manuel Pérez Saldanya (eds.), Gramàtica del català contemporani. Vol. 2: Sintaxi, 1283−1335. Barcelona: Empúries. Martin, Fabienne 2010 The semantics of eventive suffixes in French. In: Artemis Alexiadou and Monika Rathert (eds.), The semantics of Nominalizations across Languages and Frameworks, 109−139. Berlin/New York: Mouton de Gruyter. Mayo, Bruce, Marie-Theres Schepping, Christoph Schwarze and Angela Zaffanella 1995 Semantics in the derivational morphology of Italian: Implications for the structure of the lexicon. Linguistics 33: 883−938. Melloni, Chiara 2011 Event and Result Nominals. A Morpho-semantic Approach. Bern: Lang. NGLE 2009 = Nueva Gramática de la Lengua Española. Madrid: Asociación de Academias de la Lengua Española, 2009. Picallo, M. Carme 1999 La estructura del sintagma nominal: Las nominalizaciones y otros sustantivos con complementos argumentales. In: Ignacio Bosque and Violeta Demonte (eds.), Gramática descriptiva de la lengua española. Vol. 1: Sintaxis básica de las clases de palabras, 363−393. Madrid: Espasa.

68. Action nouns in Romance

1229

Porzig, Walter 1930−31 Die Leistung der Abstrakta in der Sprache. Blätter für deutsche Philosophie 4: 66− 77. [Repr. in: Hugo Moser (ed.), Das Ringen um eine neue deutsche Grammatik. Aufsätze aus drei Jahrzehnten (1929−1959), 255−268. Darmstadt: Wissenschaftliche Buchgesellschaft, 1969]. Pustejovsky, James 1995 The Generative Lexicon. Cambridge, MA: MIT Press. Rainer, Franz 1989 I nomi di qualità nell’italiano contemporaneo. Wien: Braumüller. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rainer, Franz 1996 La polysémie des noms abstraits: Historique et état de la question. In: Nelly Flaux, Michel Glatigny and Didier Samain (eds.), Les noms abstraits. Histoire et théories, 117− 126. Villeneuve d’Ascq: Presses Universitaires du Septentrion. Riegel, Martin, Jean-Christoph Pellat and René Rioul 2006 Grammaire méthodique du français. 3rd ed. Paris: Presses Universitaires de France. Roodenburg, Jasper 2010 Plurality from a cross-linguistic perspective: The existence of plural arguments supporting nominalizations in French. Lingue e linguaggio 9(1): 41−64. Sleeman, Petra and Ana Maria Brito 2011 Aspect and argument structure of deverbal nominalizations: A split vP analysis. In: Artemis Alexiadou and Monika Rathert (eds.), The Syntax of Nominalizations across Languages and Frameworks, 199−217. Berlin/New York: Mouton de Gruyter. Thornton, Anna Maria 1991 Sui deverbali italiani in -mento e -zione (II). Archivio Glottologico Italiano 76: 79−102. Thornton, Anna Maria 2004 Conversione. In: Maria Grossmann and Franz Rainer (eds.), La formazione delle parole in italiano, 499−533. Tübingen: Niemeyer. Thornton, Anna Maria 2008 Italian verb-verb reduplicative action nouns. Lingue e linguaggio 7(2): 209−232. Uth, Melanie 2011 Französische Ereignisnominalisierungen. Abstrakte Bedeutung und regelhafte Wortbildung. Berlin/New York: de Gruyter. Veland, Reidar 2011 La locuzione preposizionale da parte di tra attivo e passivo. Lingua Nostra 72(3−4): 65−80. Vendler, Zeno 1967 Linguistics in Philosophy. Ithaca: Cornell University Press. Vieira, Ilda M. T. 2009 As nominalizações deverbais em -da no português europeu. MA Thesis, University of Porto.

Livio Gaeta, Turin (Italy)

1230

VII. Semantics and pragmatics in word-formation II: Special cases

69. Verbal nouns in Celtic 1. 2. 3. 4. 5. 6.

Introduction Formal characteristics of verbal nouns Syntactic characteristics of verbal nouns Verbal nouns, verbs, and nouns Conclusion References

Abstract This article considers the category of verbal nouns in Celtic languages. It has been the subject of much debate among linguists as to whether they are to be considered more verbal or more nominal. The structures and syntactic patterns are considered here and the argument assessed. It is argued that they need to be seen in the broader perspective of non-finite verbal formations.

1. Introduction It is conventional for grammars to make a distinction between finite and non-finite parts of the verb, the former being marked for number, person, tense, mood, voice, aspect, etc., and the latter not, but instead typically taking nominal morphology and syntax which would involve, depending on the language, number, gender, declension, etc. However, where we would expect to find the non-finite parts of the verb listed in grammars of the Celtic languages, there is usually very little to be said: no Celtic language has anything like an infinitive in the sense of a nominal form of the verb with tense, mood, aspect, etc. (Disterheft 1980; Jeffers 1975, 1978), though it does have a nominal form, known as the “verbal noun”, e.g., Modern Welsh eistedd ‘sitting’, Modern Irish briseadh ‘breaking’. Celtic languages also lack a full array of verbal adjectives: no Celtic language has a present participle and the only adjectival form is usually tensed as past and voiced as passive (Russell 1995: 258−259), e.g., Old Irish mórthae ‘praised’ (← móraid ‘he praises’), Modern Irish briste ‘broken’ (← bris- ‘to break’), Middle Welsh ysgrifennedic ‘written’ (← ysgrifennu ‘to write’, based on the Latin gerund scribendum), but replaced in Modern Welsh by a periphrasis of wedi ‘after’ and a verbal noun, thus pren wedi cwympo ‘fallen tree; lit. … after falling’. The focus of this article is on the verbal nouns. They have attracted considerable attention in the scholarly literature as they tend to display a wider range of nominal features than might be expected and that makes them less amenable to being absorbed into the standard analytical models. This is particularly true of verbal nouns in the Brittonic languages of Welsh, Cornish and Breton (cf., for example, Willis 1998, Borsley 1993, and more recently Borsley, Tallerman and Willis 2007: 68−103; for further discussion, see section 4 below, and more generally Russell 1995: 258−277; for Middle Breton verbal nouns, Hemon 1975: 264−268). As befits a volume on word-formation, the formal features of verbal nouns are considered first and

69. Verbal nouns in Celtic

1231

then their syntactic characteristics, before turning to the debate over the status of verbal nouns in modern theory.

2. Formal characteristics of verbal nouns In historical terms, specific verbal noun markers have gradually arisen within the history of the Celtic languages, but even in the later stages there are a number of markers, some of which have other nominal functions. In Modern Welsh (on Welsh VN suffixes, see Thomas 1996: 668−764) there are three main endings, -u /ı/ or /ɨ/ (depending on dialect), -i /ı/, and -o /ɔ/ distributed according to the stem vowel: -u after stems containing a, ae, e, and y, e.g., caru ‘loving’, helpu ‘helping’, synnu ‘being surprised’. -i after stems containing o and oe, e.g., torri ‘breaking’, poeni ‘worrying’, etc., and after stems with final -/u/, e.g., llenwi ‘filling’, etc. -o after stems containing a final front vowel, such as -/i/, e.g., cofio ‘remembering’. In Modern Irish the most common verbal-noun marker is -adh, -ú -/ə(γ)/, e.g., pósadh ‘marrying’, briseadh ‘breaking’, which was in its origin a suffix applied to weak verbs in Old Irish (Block-Trojner 2008). In the earlier stages of Irish, verbal nouns of strong verbs were marked by a range of suffixes and often a morphologically different form was used for the simple verb and a prefixed form, e.g., Old Irish bongid ‘to strike’ : búain beside con·boing ‘to smash’ : conbach, gairid ‘to call’ beside gairm ‘call’ : ergaire ‘forbidding’ : ar·gair ‘to forbid’ (Thurneysen 1946: 444−455; it is conventional in Old Irish to use the 3rd sg. present form of the verb as the citation form). In some cases in Old Irish, the verbal noun is suppletive: caraid ‘to love’ : serc ‘loving’, tongaid ‘to swear’ : luige ‘swearing’. One of the most productive suffixes in Modern Irish, especially for verbs based on English loanwords is -áil, e.g., péinteáil ‘painting’, which is derived from the segmented ending of gabáil ‘taking’. All Celtic languages have developed verbal noun markers in the course of their history with the most productive markers tending to be the weak verb markers since new verbs, usually denominal or deadjectival, expand that class. Thus, the older verbal noun patterns are often marginalised and replaced by newer more regular formations, e.g., Old Irish sluindid ‘to call’ : slond but later slondud, etc. Formally, verbal nouns in the Celtic languages are nouns. This is most clearly visible in their syntactic features (see section 3), but it is also evident in their nominal morphology, such as declension in Goidelic languages (Thurneysen 1946: 447−455) and traces of it in Brittonic, and also in derivational patterns where they are treated as nouns and can be used with the same derivational suffixes, e.g., Old Irish marbad ‘killing’ : marbthach ‘killing’, just like macc ‘son’ : maccach ‘having sons’, Welsh gallu ‘being able, being capable’ : galluog ‘capable, clever’ (Russell 1990: 86). Furthermore, verbal nouns can form the basis for new denominative verbs; this is particularly common in the history of Irish where, in the process of regularising what had become a very complex and unpredictable verbal system, new denominative verbs based on verbal nouns were developed to replace older more irregular forms, e.g., Old Irish con·utaing ‘to build’ : ní·cum-

1232

VII. Semantics and pragmatics in word-formation II: Special cases

taig ‘he does not build’ : cumtach ‘building’ > Middle Irish cumtaigid, Modern Irish cumhdaíonn; in Modern Irish ibhid ‘to drink’ : ól ‘drinking’ has been gradually replaced by the verb ólann (McCone 1997: 193−194).

3. Syntactic characteristics of verbal nouns It may be useful to explore the syntactic characteristics of the Celtic verbal noun by dividing them into nominal and verbal features (for reasons of space only the most significant features are discussed; for a more detailed discussion, see Russell 1995: 260− 71).

3.1. Nominal features 3.1.1. Modification by article and adjective Verbal nouns can be used with the article, e.g., Modern Welsh canu ‘singing’: ydych chi’n clywed y canu ‘do you hear the singing?’, Middle Breton hadaff ‘sowing’: amser an hadaff ‘the time for the sowing’, Modern Irish locc in crochda ‘the place of the crucifixion’, etc. They are also modified by adjectives, e.g., Middle Welsh gossot ‘attacking’: y gossot kyntaf ‘the first attack’, Modern Welsh rhedeg ‘running’: rhedeg gyflym ‘swift running’, etc. It has been argued that such usage tells us nothing about their function as verbal nouns but rather that the same forms can be used as action-nouns (Borsley 1993); see section 4 for further discussion.

3.1.2. Modification by noun and pronoun Verbal nouns can be modifiers of other nouns, e.g., Middle Welsh hela/hely ‘hunting’: gwisc hela ‘hunting cloths’, early Modern Irish déanamh ‘making’: fear déanta na mbróg ‘shoemaker; lit. a man of making of the shoes’. In the Goidelic languages the nominal features are clearer since they have retained declension and case up to the modern languages (though in a gradually reduced form) and so the verbal noun in such instances is in the genitive case. The Brittonic languages lost declension and case in the Common Brittonic period and the genitival relationship is marked simply by juxtaposition with the modifier coming in second position (though both Cornish and Breton have secondarily developed a prepositional usage parallel to the use of English of and French de). However, when the modifier is pronominal, possessive pronouns are used, e.g., Modern Welsh gweld ‘seeing’: y mae ef yn gweld Mair ‘he sees M.’ beside y mae ef yn ei gweld hi ‘he sees her’, where ei ... hi ‘her’ is identical to the possessive forms used with a common noun, e.g., llyfr Mair ‘M.’s book’ : ei llyfr hi ‘her book’. The same patterns can be found in early Irish, e.g., Middle Irish tichtu Phádraicc ‘the coming of Patrick’ : a thichtu ‘his coming’, but in the later language with the gradual erosion of case distinctions such patterns can be found with personal names and definite nouns but

69. Verbal nouns in Celtic

1233

are otherwise disappearing; thus, for example, pósadh ‘marrying’: tá sé ag pósadh fear óg ‘she is marrying a young man’ (where fear óg is nom. sg.). There is a similar drift away from using possessive pronouns in front of verbal nouns (Ó Siadhail 1989: 277). There is also a process of “subject-raising” when the subject slot of a copula sentence is empty − a common feature of Irish in sentences like is maith leis ‘he likes; lit. it is good with him’, e.g., péinteáil ‘painting’: ba mhaith lium an doras a phéinteáil ‘I would like to paint the door; lit. the door (nom. sg.), its painting, is good with me’; here an doras which is in a sense the object argument of péinteáil (and in a phrase like péinteáil in doruis ‘painting the door’ would be genitive) has been raised to subject position (Disterheft 1980: 152−155). A further feature of Celtic verbal nouns should be considered, namely transitivity. While Welsh wedi marw pawb o’i deulu ‘after all of this family had died; lit. the dying of all …’, or Middle Irish tichtu Phátraicc ‘the coming of Patrick’ can only be understood as all of this family doing the dying and Patrick doing the coming, and Welsh ei fynd ef ‘his going’ can only be understood as him doing the going. When we are dealing with transitive verbs, matters are more complicated as in theory the same slot would appear to be available for the subject and object argument. However, in practice the object argument takes priority with the subject argument marked by a prepositional phrase. Thus, to take a well-known Old Irish example, serc Dé ‘the love of God’, though apparently ambiguous (either ‘X loves God’ or ‘God loves X’), can only mean in Old Irish the love someone has for God; a subject argument is then framed as a prepositional phrase, e.g., serc Dé do Fergus ‘the loving of God by Fergus’. Conversely, far serc-si do Día means ‘love of you (pl.) by God’.

3.1.3. Government by prepositions The use of verbal nouns in prepositional phrases is another important example of nominal syntax. They frequently occur in Welsh and Early Irish corresponding syntactically to subordinate clauses with finite verbs in other languages, e.g., Modern Welsh mynd ‘going’: prynodd e gar i fynd i waith ‘he bought a car to go to work’ (Evans 1964: 198; Williams 1980: 138). Similarly in Old Irish do + verbal noun can be used to express purpose, e.g., dénam ‘doing’: do dénam uilc ‘in order to do evil’, fortacht ‘helping’: dum fortacht húait-siu ‘that you may help me; lit. for my helping from you’. Verbal nouns can be negated by use of the preposition ‘without’, Welsh heb, Breton hep, Irish cen, but only in circumstances where a negative state is being referred to, e.g., Old Irish is ingir lem cen chretim duib ‘it grieves me that you do not believe; lit. it is grief with me without believing (i.e. a state of non-belief ) by you’, Modern Welsh yr wyf i heb ddarllen y llyfr ‘I have not read the book (i.e. a state of non-reading) by you’.

3.1.4. Verbal nouns as subjects or objects of verbs Verbal nouns can behave as subjects and objects of verbs, e.g., as subject Middle Welsh gwelet ‘seeing’: da yw gennyf dy welet ti ‘I am glad to see you; lit. Seeing you is good with me’; Middle Breton aznavout ‘knowing’ aznavout Doue ... so dleet ‘knowing God ... is a duty’; as object Modern Irish broinn ‘farting’: rinn sé broinn ‘he farted; lit. he

1234

VII. Semantics and pragmatics in word-formation II: Special cases

made a farting’, etc. In the Brittonic languages the ‘doing’ verbs, Middle Welsh gwneuthur, Modern Welsh gwneud, Breton ober, Cornish gruthyl, gul, etc., have developed into a form of auxiliary or light verb with a verbal noun as object, e.g., Middle Cornish dybry ‘eat’: mi a ra y dybry ‘I eat it; lit. I do eating of it’ (Lewis 1946: 49, quoted in Zimmer 1990: 46), Modern Breton selaou ‘listening’: selaou a wreas ar vamm ‘the mother listened; lit. listening did the mother’ (Timm 1990: 192). In Middle Welsh prose the pattern was particularly widespread to express consecutive actions, e.g., bwyta a chyuedach ac ymdidan a wnaethont ‘they ate, caroused and conversed; lit. eating, etc. ... they did’. It continued into the modern language and in the north a slightly different pattern developed using daru (earlier darfu) ‘happened’ with the verbal noun as the grammatical subject.

3.2. Verbal features In the previous section we were concerned with the ways in which verbal nouns displayed nominal features. We turn now to their verbal features.

3.2.1. Modification by adverb A verbal noun can be modified by an adverb, e.g., Modern Welsh cerdded ‘walking’: rhaid yw cerdded yn brysur ‘it is necessary to walk quickly’, where yn brysur is the clearly marked adverb based on prysur ‘quick, hasty’. Adverb usage is compulsory in periphrastic constructions containing verbal nouns (see section 3.2.2) but one difficulty is that adverbs in Celtic languages can be marked formally to distinguish them from adjectives, or not at all; compare Modern Welsh yn + adjective in the example above or the use of go + adjective in Modern Irish, e.g., go cúramach ‘well’. But in other contexts there is no formal marking and so it is not always clear whether we are dealing with an adverb or an adjective, or whether that is always a useful distinction to make.

3.2.2. Periphrastic auxiliary verbs and aspect Celtic languages are rich in periphrastic verbal constructions, most of which involve the use of verbal nouns, and there has been considerable debate over the status of these constructions. They are conventionally regarded as taking the following form: [auxiliary verb, typically ‘to be’, marked for person, number and tense] + [aspect marker] + [verbal noun]; we may take the forms from Modern Welsh and Modern Irish (for the latter, see Block-Trojnar 2003): Modern Irish Progressive: tá sé ag + verbal noun; lit. ‘he is at −’ Perfect: tá sé thar éisi/i ndéid + verbal noun; lit. ‘he is after −’ Prospective: tá sé le + verbal noun; lit. ‘he is with −’ Stative: tá sé i (n)+ possessive pronoun + verbal noun; lit. ‘he is in his −’

69. Verbal nouns in Celtic

1235

Modern Welsh Present (originally progressive): y mae ef yn + verbal noun; lit. ‘he is “in” −’ Perfect: y mae e wedi + verbal noun; lit. ‘he is after −’ At one level, and especially in Irish, these constructions can be analysed as containing a nominal pattern of preposition + verbal noun after the verb ‘to be’, but two factors may give us pause. First, in some instances, especially in the Brittonic languages, the so-called ‘preposition’ does not behave like a preposition; in particular, the marker yn, though formally identical to the preposition meaning ‘in’ and the predicate marker (used inter alia to form adverbs), does not share the mutation patterns of the other homonyms; yn in prepositional function nasalises the following noun, and as the predicate marker lenites, but there is no mutation in, for example, y mae ef yn cerdded ‘he is walking’. It has been sometimes suggested that non-leniting yn is simply an etymologically distinct preposition (Fife 1990: 432−433) but, whatever its origin, it seems to have been grammaticalised as an aspect marker. The equivalent forms in Breton and Cornish (insofar as we can tell for the latter), o and ow, respectively, are strikingly not formally identical to a preposition and do suggest that the Welsh form might be more usefully analysed as an aspect marker than a preposition. These periphrastic constructions have been subject to a variety of analyses. In some cases they have been analysed in parallel to English I am going such that going is the aspectually marked verb with the auxiliary carrying the markers of tense, person and number (see Jones and Thomas 1977: 168−176 on Welsh, McCloskey 1980, 1983 on Irish, Hewitt 1990: 184 on Breton), but it has also been argued that such analyses give rise to their own problems (Stenson 1981: 138−140 on Irish, Fife 1990: 326−398 on Welsh). The difficulty here is that even in synchronic terms it is possible to see that many of these constructions seem to contain grammaticalised prepositional phrases which can be explained historically, and Comrie (1976: 96−105) has pointed to many other languages where aspectual distinctions are expressed through locative expressions, often involving prepositions, and also the use of locative verbs as a progressive auxiliary, such as Spanish estoy cantando, etc. In this respect it is striking that in Irish the form of the verb ‘to be’ used in such expressions in Irish and Scottish Gaelic is the substantive form of the verb ‘to be’ (rather than the copula form), tá and thá respectively, which etymologically is related to Latin stare, etc.

3.2.3. Verbal nouns in subordinate clauses Another feature of verbal noun usage which is arguably more verbal than nominal is their frequent use in subordinate clauses. In Old Irish in certain types of subordinate clauses, typically those expressing purpose and indirect statement, verbal noun constructions can alternate with full finite-verb constructions, e.g., purpose: do·luid do marbad na muicce ‘he came to kill the pig; lit. to killing of the pig’, beside an biuc corot·aicciller ‘wait a moment so that I might speak to you’ (where the clause is marked by co + a subjunctive verb); indirect statement: as·indet-som buith doib i ndoiri ‘he declares that they are in captivity; lit. being by them in captivity’ beside an-as·berid … coni·cloitis geinti tairchetal Críst ‘… what you say … that the gentiles did not hear the prophesying

1236

VII. Semantics and pragmatics in word-formation II: Special cases

of Christ?’. However, in the Modern Irish go + finite verb has become the usual purpose clause pattern, and there is a general tendency in Irish for verbal noun constructions to be gradually replaced by finite verb clauses (see section 3.2.4). Similarly in Welsh prepositional phrases involving verbal nouns, while still productive, are often replaced by finite verb patterns if, for example, they are found in negative, or in some way modal, contexts, and so the verbal sense requires more nuance than is possible with a verbal noun; for example, compare mae hi yn dweud ei fod e’n alluog ‘she says he is clever; lit. ... his being (bod ) clever’, with mae hi yn dweud nad yw e’n alluog ‘she says he is not clever’ (where ... nad yw ... is the neg. + 3rd sg. present ‘is’).

3.2.4. Replacement of finite verbs in co-ordinated strings It was noted in section 3.1.4 that in Middle Welsh a string of verbal nouns could be used with gwnaeth/goruc ‘did’ to express consecutive actions where the verbal noun has to be understood as the object of verb. One of the most verbal characteristics of Middle Welsh verbal nouns was their use in a string following a fully marked finite verb; e.g., Middle Welsh y syrthawd y blew oll … a thorri y croen y amdanei, a syrthaw y holl berued y’r llawr ‘all its hair fell out (syrthawd, preterite 3rd sg. of syrthaw) … and the skin was rent (torri (verbal noun)) from around it, and all its entrails fell (syrthaw, verbal noun) on the floor’; also Middle Breton ma-z huesenn certen ha crenaf, ha coezaf da ndouar ‘so that I sweated (huesenn, imperfect 1st sg.) and trembled (crenaf, verbal noun) and fell (coezaf, verbal noun) to the ground’. In both cases, person and tense are marked on the first verb but the subsequent verbs in the string are verbal nouns, and person and tense are assumed to remain the same.

4. Verbal nouns, verbs, and nouns The previous sections have attempted to set out the nominal and verbal aspects of Celtic verbal nouns. The significance of these characteristics has been debated. Fife (1990: 399−402), who generally prefers a nominal analysis, is inclined to think that such a binary (nominal/verbal) distinction is unhelpful since, apart from modification by adverbs, other verbal features are not characteristic of verbs, but it has been shown above that one of the more verbal aspects of verbal nouns is that there are certain syntactical patterns where they are replaceable with full finite-verb clauses. Borsley’s response (1993; cf. also Borsley, Tallerman and Willis 2007: 68−103) was to argue that verbal nouns are not to be compared with finite verbal forms at all, but rather with the non-finite forms, infinitives and verbal adjectives. He makes a distinction between, for example, the two different uses of Modern Welsh canu ‘singing’, both as a verbal noun in mae’r bachgen yn canu ‘the boy is singing’ and as a noun, e.g., clywais i’r canu ‘I heard the singing’ (where it is used with a definite article as the object of the verb), and argues that the former is verbal (specifically “a non-finite verb”) and the latter nominal. Similarly, Borsley argues that disgrifio ‘describe’ and disgrifiad ‘description’ have different syntactical properties such that the former, the verbal noun or “non-finite verb”, is distinct from nouns (Borsley, Tallerman and Willis 2007: 72−73); for example, in dylai

69. Verbal nouns in Celtic

1237

Gwyn ddisgrifio’r llun ‘Gwyn should describe the picture’, the verbal noun is followed by a noun phrase complement, but in disgrifiad o’r rhes o dai ‘the description of the row of houses’ the noun is followed by a prepositional phrase. However, there is a sense in which the differences are being maximised; for example, the use of a prepositional phrase after disgrifiad is in fact optional and one could say disgrifiad y rhes o dai. It could as well be argued that the distinction is between two types of nouns rather than between nouns and verbs (non-finite or otherwise). Another issue which needs to be raised at this point is that such discussions about verbal nouns in Celtic always tend to focus on the modern stages of the Brittonic languages and on Modern Irish (McCloskey 1980) in which the more nominal aspects of the relationship between the verbal noun and its complements have become less obvious. For example, as was discussed above (cf. section 3.1.2 especially), in the early stages of all the Celtic languages the complement of a verbal noun always took genitival syntax; in Old and Middle Irish the complement is actually in the genitive, e.g., Old Irish marbad na muicce ‘the killing of the pig’ (muicce, gen. sg. of mucc), but in Brittonic languages, where case distinctions have been lost, the same relationship is marked by juxtaposing two nouns, e.g., Middle Breton ar gouzout lenn ‘the ability to read; lit. the knowing of reading’; Middle Welsh ymrwymaw Ynys y Kedeirn ac Iwerdon y gyt ‘the binding of the Island of the Strong (sc. Britain) and Ireland together’ (Evans 1967: 159). In Brittonic languages, therefore, it is easier to argue that the second noun is a noun phrase complement since it is not marked as genitive. This same pattern has continued into the modern Brittonic languages, while in Irish the genitive usage has been gradually replaced (see section 3.1.2). When we place that development alongside the gradual shift in subordinate clause usage from expressions using verbal nouns to those with full subordinate clauses with finite verbs, it is arguable from a historical perspective that there is a shift in progress from a more nominal mode of expression to more verbal one, and it is perhaps the intermediate status between nominal and verbal status which lies at the root of analytic difficulties. But it is worth observing that the issue here is not unique to Celtic. Two brief examples of similar phenomena elsewhere may be useful. Haspelmath (1995: 52−59) has made a useful distinction between “internal” and “external” syntax in his discussion of cases such as German der im Wald laut singende Wanderer ‘the hiker (who is) singing loudly in the wood; lit. the in the wood loudly singing hiker’; here the adverb laut modifies the participle singende in a way to suggest that the internal syntax is verbal, but the external syntax with singende agreeing with Wanderer is nominal. Similarly in cases like Welsh rhaid yw cerdded yn brysur ‘it is necessary to walk quickly’ (see section 3.2.1) it might be argued that the internal syntax of cerdded yn brysur is verbal (hence the use of the adverb yn brysur) but the external syntax is nominal. Haspelmath also discusses the distinctions between the following three sentences: (a) Indonesia annexing East Timor; (b) Indonesia’s annexing East Timor; (c) Indonesia’s annexation of East Timor, which provides a useful corrective that there is a binary distinction between an external verbal and nominal syntax; while (a) is clearly verbal and implying process and (c) is nominal and implies completion of the process. (b) is less easy to analyse and resembles more closely what happens with Celtic verbal nouns (cf. also Grimshaw 1990, esp. 46−48). The second example concerns maṣdar non-finite nominal formations in Arabic (Kremer 2007) which seem to vacillate between verbal and nominal syntax: whereas the subject always takes genitive case, the object can be assigned accusative

1238

VII. Semantics and pragmatics in word-formation II: Special cases

case, but an indirect object can be expressed by the preposition li ‘to, for’; the second of these allows modification by an adverb, but a maṣdar that uses li can be modified by an adjective. Again, we are dealing with nominal formations which fluctuate between verbal and nominal syntax. There is a tendency to regard the workings of verbal nouns in Celtic as something different and in need of special explanation (cf. Russell 1995: 258−277 on the perceived oddities of the Celtic languages), but the above two examples would suggest that analyses of Celtic verbal nouns, and especially those in the Brittonic languages which have received most attention, might benefit from being set in a broader context. It may perhaps be useful at this point to take a step backwards and compare the development of infinitives which are generally much more verbal in their morphology and syntax; for example, they typically mark tense, e.g., English to go, to have gone, to be about to go, Latin ire, ivisse, iturus esse, etc. and voice, e.g., English to take, to be taken, Latin capere, capi, etc., and their complements are usually in the accusative case. Formally though, the markers of infinitives, e.g., Latin -re, Classical Greek -ein, -nai, etc. medio-passive -sthai, Vedic -dhyai, -tave, Hittite -anna, etc. (Disterheft 1980: 12− 17), can only be analysed historically as fossilised case endings of nouns, even though at a later stage they are added to verbal stems marked variously for tense. In other words, as Disterheft has argued (but cf. Jeffers 1975 and 1978), there has been a drift towards the verbalisation of nouns based on verbal roots and it is Celtic (along with a few traces in other languages) that seems to have retained the older more nominal pattern. But even in Celtic, in the later phases of the languages, there seems to be a tendency towards greater verbalisation. One interesting, but fleeting case, can be found in Middle Welsh is a tendency to use the perfective verbal particle ry (usually found with preterite-tense verbs as a prefectivising particle, e.g., Middle Welsh cauas ‘he got’: ry gauas ‘he has got’ (Evans 1967: 166−7)) with verbal nouns in an attempt to introduce tense differences into verbal noun syntax usually in subordinate clauses, e.g., mi a gyfessaf ry bechu ohonaf ‘I confess that I have sinned’ (pechu: verbal noun with preceding particle ry; lit. ‘… a having sinned by me’). Many of the examples occur in texts which have probably been translated from Latin, such as the Welsh translations of Geoffrey of Monmouth’s History of the Kings of Britain, and it is likely, though as yet not demonstrable, that the particle was used to reflect the tense variation in Latin infinitives and in particular to distinguish between present and perfect infinitives. What is clear, however, is that for that to happen there must have been a sense that in certain syntactic contexts a verbal noun had some verbal characteristics.

5. Conclusion Verbal nouns in Celtic present a range of interesting problems for the linguistic analyst. Depending on how one comes to view them, they can seem more or less nominal or verbal. Earlier stages of the languages − and we are fortunate to have evidence for Old Irish from as early as the seventh century with a plentiful supply of evidence for verbal noun usage from the mid-eighth century onwards − seem generally to display a more nominal syntax and that continues in Irish well towards the modern language partly because of the survival of the case-system. In the Brittonic languages, where there is no

69. Verbal nouns in Celtic

1239

indication of a functioning case-system from our earliest evidence onwards, the verbal characteristics of verbal nouns have gradually become more obvious. There is an interesting split in the scholarship where linguists who come at this problem from a historical perspective tend to emphasise the nominal features while those with a synchronic perspective tend to highlight their verbal characteristics. As noted in section 4, however, their surface realisation very often remains nominal, though often sitting on that cusp between the noun and the verb; and that is where comparison with parallel forms in other languages could prove useful.

6. References Anderson, Stephen R. 1981 Topicalization in Breton. Proceedings of the Berkeley Linguistics Society 7: 27−39. Block-Trojner, Maria 2003 The intersection of syntax and morphology in the Modern Irish progressive construction. In: Piotr Bański and Adam Przepiórkowski (eds.), Generative Linguistics in Poland. Morphosyntactic Investigations, 1−14. Warsaw: Polish Academy of Sciences. Block-Trojner, Maria 2008 The morphology of verbal nouns in Modern Irish. Éigse 36: 63−81. Borsley, Robert D. 1993 On so-called ‘verb-nouns’ in Welsh. Journal of Celtic Linguistics 2: 35−64. Borsley, Robert D. and Ian Roberts (eds.) 1996 The Syntax of the Celtic Languages. Cambridge: Cambridge University Press. Borsley, Robert D., Maggie Tallerman and David Willis (eds.) 2007 The Syntax of Welsh. Cambridge: Cambridge University Press. Comrie, Bernard 1976 Aspect. Cambridge: Cambridge University Press. Disterheft, Dorothy 1980 The Syntactic Development of the Infinitive in Indo-European. Columbus, OH: Slavica. Evans, D. Simon 1964 A Grammar of Middle Welsh. Dublin: Dublin Institute for Advanced Studies. Fife, James 1990 The Semantics of the Welsh Verb. Cardiff: University of Wales Press. Fleuriot, Léon 1964 Le vieux breton. Éléments d’une grammaire. Paris: Klincksieck. Gagnepain, Jean 1963 La syntaxe du nom verbal dans les langues celtiques. Vol. 1: Irlandais. Paris: Klincksieck. Genee, Inge 1994 Pragmatic aspects of verbal noun complements in Early Irish: do + verbal noun in the Würzburg Glosses. Journal of Celtic Linguistics 3: 41−73. Grimshaw, Jane 1990 Argument Structure. Cambridge, MA: MIT Press. Haspelmath, Martin 1995 Word-class-changing inflection and morphological theory. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1995, 43−66. Dordrecht: Kluwer. Hemon, Roparz 1941 Grammaire bretonne. Brest: Gwalarn.

1240

VII. Semantics and pragmatics in word-formation II: Special cases

Hemon, Roparz 1975 A Historical Morphology and Syntax of Breton. Dublin: Dublin Institute for Advanced Studies. Jeffers, Robert J. 1975 Remarks on Indo-European infinitives. Language 51: 133−148. Jeffers, Robert J. 1978 Old Irish verbal nouns. Ériu 29: 1−12. Kervella, Frañsez 1947 Yezhadhur bras ar Brezhoneg. La Baule: Skridou Breizh. King, Gareth 1993 Modern Welsh. A Comprehensive Grammar. London: Routledge. Kremers, Joost 2007 Maṣdar formation. In: Everhard Ditters and Harald Motzki (eds.), Approaches to Arabic Linguistics. Presented to Kees Versteegh on the Occasion of his Sixtieth Birthday, 475− 99. Amsterdam. Brill. Lewis, Henry 1946 Llawlyfr Cernyweg Canol. 2nd ed. Cardiff: University of Wales Press. Lewis, Henry and J[ean] R. F. Piette 1966 Llawlyfr Llydaweg Canol. 3rd ed. Cardiff: University of Wales Press. McCloskey, James 1980 Modern Irish nouns and the VP-complement analysis. Linguistic Inquiry 6: 345−357. McCone, Kim 1997 The Early Irish Verb. 2nd ed. Maynooth: An Sagart. Meid, Wolfgang 1990 Handbuch des Mittelbretonischen. (Translation and revision of Lewis and Piette 1966.) Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Ó Siadhail, Micháil 1989 Modern Irish. Grammatical Structure and Dialectal Variation. Cambridge: Cambridge University Press. Rouveret, Alain 1994 Syntaxe du gallois. Principes généraux et typologies. Paris: CNRS Éditions. Russell, Paul 1990 Celtic Word-Formation. Dublin: Dublin Institute for Advanced Studies. Russell, Paul 1995 Introduction to the Celtic Languages. London: Longman. Schumacher, Stefan 2011 Mittel- und Frühneukymrisch. In: Elmar Ternes (ed.), Brythonic Celtic − Britannisches Keltisch, 85−236. Bremen: Hempen. Stenson, Nancy 1981 Studies in Irish Syntax. Tübingen: Narr. Ternes, Elmar (ed.) 2011a Brythonic Celtic − Britannisches Keltisch. Bremen: Hempen. Ternes, Elmar 2011b Neubretonisch. In: Elmar Ternes (ed.), Brythonic Celtic − Britannisches Keltisch, 431− 524. Bremen: Hempen. Thomas, Peter Wynn 1996 Gramadeg y Gymraeg. Cardiff: University of Wales Press. Thurneysen, Rudolf 1946 A Grammar of Old Irish. 2nd ed. Revised and translated by Daniel Binchy and Osborn Bergin. Dublin: Dublin Institute for Advanced Studies. Thorne, David A. 1993 A Comprehensive Welsh Grammar. Oxford: Blackwell.

70. Nominalization in Hungarian

1241

Timm, Leonora A. 1990 Some observations on the syntax of the Breton verbal noun. In: Martin J. Ball, James Fife, Erich Poppe and Jenny Rowland (eds.), Celtic Linguistics. Ieithyddiaeth Geltaidd. Readings in the Brythonic Languages. Festschrift for T. Arwyn Watkins, 189−208. Amsterdam/Philadelphia: Benjamins. Williams, Stephen J. 1980 A Welsh Grammar. Cardiff: University of Wales Press. Willis, Penny 1988 Is the Welsh verbal noun a verb or a noun? Word 39: 201−224. Zimmer, Stefan 1990 Handbuch des Mittelkornischen. (Translation and revision of Lewis 1946.) Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck.

Paul Russell, Cambridge (UK)

70. Nominalization in Hungarian 1. 2. 3. 4. 5. 6. 7.

Introduction -Ás (V to N) -Ó (V to N) -sÁg (A to N) -itás (A to N) Conclusion References

Abstract This article offers an overview of productive Hungarian nominalization phenomena. In addition to describing the most important morphological and morphosyntactic properties of individual suffixes, it concentrates primarily on their semantic and pragmatic traits.

1. Introduction In this article, present-day productive Hungarian nominalization processes are discussed fundamentally from semantic and pragmatic points of view. These processes are all suffixal, which is a general feature of Hungarian inflectional and derivational morphology. The term nominalization is used in the following restricted sense: we only concentrate on cases in which a noun is derived from a word belonging to a different (that is, non-nominal) category. There are four types in this domain: two verb-to-noun and two adjective-to-noun processes. They are relevant morphosyntactically in that the input

1242

VII. Semantics and pragmatics in word-formation II: Special cases

word’s morphosyntactic traits are modified. This typically and generally holds for nominalization processes resulting in a category change, because the category change usually goes hand in hand with a significant semantic change, very often affecting the argument structure and the realization of arguments. Two salient general aspects of this article are as follows. (A) Given the nature and goals of the present volume, we cannot offer a detailed analysis of the phenomena under investigation in any theoretical model. However, we do need to be able to refer to various (rather widely assumed) components and representational levels of grammar. For concreteness, we employ the concepts of lexical-functional grammar. (B) We make systematic comparisons between Hungarian and English for two main reasons. (i) The Hungarian examples are (necessarily) translated into English. This is already a considerable degree of automatic comparison, which can logically be made more detailed and meaningful. (ii) These comparisons can contribute to a fuller understanding of the Hungarian phenomena by placing them in a partial typological context. In this introduction, we first briefly characterize the general descriptive framework for our discussion of the relevant Hungarian nominalization processes, secondly we present our representational convention of the suffixal allomorphy induced by the principles of vowel harmony at work in Hungarian, and thirdly we outline the structure of the article. Following some basic architectural assumptions of lexical-functional grammar (see Bresnan 2001), we adopt in this article the following crucial levels of representation and the following interplay among them. In the lexical entry of a word, the standard types of information are encoded: the phonological form of the word, its syntactic category and the specification of its meaning, which, in the case of predicates, normally requires a semantic structure of varying complexity, and possibly additional (idiosyncratic) information. The semantic structure is called lexical-semantic structure. Certain elements in lexical-semantic structure are mapped onto arguments in the argument structure, also represented in the lexical form of the predicate. This component is designed to mediate between semantics and syntax inasmuch as it is the arguments in the argument structure that are mapped onto syntactic constituents bearing designated grammatical functions. In lexical-functional grammar, grammatical relations are represented in a component called functional structure, which is one of the two parallel levels of syntactic representation, the other being constituent structure. Consider Fig. 70.1 demonstrating this lexicalfunctional grammar style organization of the grammar with a simplified analysis of an English verbal predicate: destroy. Its lexical-semantic structure and argument structure are part of its lexical form, and the arguments’ grammatical functions are represented in the syntax, in particular in functional structure. LSS

LEXICAL-SEMANTIC STRUCTURE

‘x causes y not to exist’

↓

↓ lexico-semantic projection ↓

AS

ARGUMENT STRUCTURE

↓

↓ lexico-syntactic projection ↓

↓

FS

FUNCTIONAL STRUCTURE

(SUBJ)

Fig. 70.1: Mapping in lexical-functional grammar

↓

↓

↓ (OBJ)

70. Nominalization in Hungarian

1243

As Fig. 70.1 also shows, argument structure has a mediating function between the semantics and the syntax of predicates. On the one hand, it represents the fundamental participants of an event, thus it encodes basic information about event structure (a semantic aspect) and, on the other hand, it contains the necessary information about the arguments in order for them to be mapped onto syntactic structure, which is called “functional structure” in Fig. 70.1 (a syntactic aspect). The majority of Hungarian derivational (and inflectional) morphemes exhibit allomorphy due to the rules of vowel harmony (along two dimensions: front vs. back vowels, the main dimension, and rounded vs. unrounded vowels, a subdimension in the front vowel domain). In our representation of the morphemes under discussion we employ the following widely accepted strategy: when there is a front vs. back contrast, the back vowel version is given in a capitalized form, e.g., the allomorphy of the -ság/-ség deadjectival nominalizing suffix is represented as -sÁg. In the glosses of the Hungarian examples we give the actual allomorph of the morpheme in SMALLCAPs. The structure of this article is as follows. In section 2, verb-to-event-noun nominalization involving the -Ás suffix is discussed. Section 3 deals with the other extremely productive and multifunctional deverbal nominalizer, the -Ó suffix, which derives agentive, instrumental and locative nouns. Section 4 characterizes the -sÁg suffix, which turns adjectives into abstract nouns. Section 5 briefly describes the use of the -itás suffix, whose function is the same as that of -sÁg, except that -itás, a suffix of foreign origin itself, is constrained to attach to adjectives of foreign origin. Section 6 offers a summary.

2. -Ás (V to N) Deverbal event nominalization is probably one of the most typical nominalization processes across languages. At the same time, it has several intriguing aspects in various languages. As is well-known, English manifests a special array of phenomena in this domain, and we will set the scene for the discussion of the relevant Hungarian linguistic facts in this broader English context. Consider the following examples, which, for our present purposes, we classify according to a quite widely accepted approach (see Grimshaw 1990 for English, and Szabolcsi 1994 and Laczkó 1995 for Hungarian). (1)

a. The enemy destroyed the city. c. the enemy’s destroying of the city e. all the three destructions of the city

b. the enemy’s destroying the city d. the enemy’s destruction of the city f. The destruction was visible from above.

(1a) is the clausal/verbal point of orientation. (1b) exemplifies a particular “mixed category” typically called verbal gerund (its external syntax is nominal, and its internal syntax is basically verbal). (1c) is an example of nominal gerunds. (1d) contains what Grimshaw calls a complex event nominal, comparable in all relevant respects to the nominal gerund in (1c). In the spirit of Grimshaw’s analysis, it can be assumed that the nominal in (1e) expresses a simple event (hence the possibility of pluralization), and the deverbal noun in (1f) denotes the result of the event described by the input verb. Grimshaw points out that, as is also exemplified in (1d−f ), many derived nominals are ambiguous between the complex event, simple event and result readings and she employs

1244

VII. Semantics and pragmatics in word-formation II: Special cases

a battery of diagnostic tests to tell them apart. The two most important ones, from our Hungarian perspective, are as follows. (i) Complex event nominals inherit the argument structures of the input verbs. However, the argument that in her framework is termed “external argument” (roughly speaking: an agent or an experiencer argument) gets “suppressed”, that is, it becomes a special optional adjunct of the nominal. (ii) Complex event nouns do not pluralize. Grimshaw argues that the similarities among these three types of derived nominals (complex events, simple events and results) can be captured by postulating that they share the same lexical-semantic structural representation, and the differences are due to the fact that only complex events have an argument structure as opposed to simple events and results. She claims that complex event nouns have arguments, and simple event and result nouns have complements directly mapped from the nominal predicate’s lexical-semantic structure, and these arguments and complements very often coincide formally. For concreteness, in Grimshaw’s analysis (and in our terminology), in (1d) the possessor constituent the enemy’s (having a special argument-adjunct status) expresses a suppressed argument of the predicate and the prepositional phrase of the city realizes an oblique argument of the nominal complex event predicate. By contrast, in (1e), in which the noun head has the simple event interpretation because of its plural form among other factors, the prepositional phrase of the city only expresses an adjunct modifier of the noun head. The difference between the statuses of the same prepositional phrase, of the city, can be represented in the following way in the architecture outlined in Fig. 70.1 on page 1242 above (ADJ stands for “adjunct”).

LSS ↓ AS ↓ FS

(1d): complex event

(1e): simple event

‘x causes y not to exist’

‘x causes y not to exist’

↓

↓

↓

↓

(POSS-ADJ) (OBL)

↓ ↓ ↓ (ADJ)

Fig. 70.2: Complex event vs. simple event nominals

The most important point is that of the city is an obligatory oblique argument of the complex event nominal in (1d), while the homophonous simple event nominal in (1e) has no argument structure to begin with, and of the city, as an adjunct, is directly linked to a participant in the lexical-semantic structure of this simple event nominal. In Hungarian there is only one suffix for deriving complex event nouns from verbs productively, the -Ás suffix. Clearly, nouns derived by -Ás are not verbal gerunds, as they cannot take direct object arguments or adverbial modifiers. As regards their characteristic features, -Ás nouns are comparable to both -ing nominal gerunds and -(t)ion nouns in English. They are also similar to nominal gerunds in that, on the one hand, they often allow only the complex event reading in Grimshaw’s (1990) sense and, on the other hand, a great number of them can be used in two or even in all the three senses Grimshaw distinguishes: complex events, simple events and results. Of Grimshaw’s diagnostics for ascertaining the complex event nominal status of a deverbal noun, Szabolcsi (1994) and Laczkó (1995) find the following two efficiently

70. Nominalization in Hungarian

1245

applicable to Hungarian. (i) The arguments of a complex event nominal are obligatory (except for the special status of agentive arguments). (ii) Complex event nominals do not have plural forms (the apparently exceptional cases are due to “institutionalized or ad hoc shifts” to the simple event reading). In addition, there is a very reliable test, proposed by Szabolcsi (1994), which is peculiar to Hungarian nominals. Its essence is as follows. Typically, deverbal nouns in Hungarian take their oblique arguments and adjuncts in “attributivized” forms. The two most productive attributivizing elements are the adjectival derivational suffix -i (glossed as AFF) and the present participial form of the verb van ‘to be’, való ‘being’, glossed as VALÓ. Szabolcsi (1994) shows that when either element can be used, namely, when the argument or adjunct is expressed with certain types of postpositional phrases, the application of való ‘being’ will trigger the complex event reading of an otherwise ambiguous nominal. Consider the following examples (for simplicity, in the glosses we do not indicate the morphologically unmarked nominative case in this article). (2)

a. Hatástalan volt az ebéd után-i beszélget-és. ineffective was the lunch after-AFF converse-ÉS ‘Conversing after lunch was ineffective.’ (complex event) or: ‘The conversation after lunch was ineffective.’ (simple event) b. Hatástalan volt az ebéd után való beszélget-és. ineffective was the lunch after VALÓ converse-ÉS ‘Conversing/*The conversation after lunch was ineffective.’ (complex event/*simple event)

The reason why this is a most useful diagnostic is that practically any derived nominal can be modified with a temporal or locative postpositional phrasal adjunct combined with való ‘being’, and its presence identifies complex event nominals, independently of, and more efficiently than, the other tests. Fig. 70.3 compares some English and Hungarian deverbal nouns. English nominals

Hungarian complex event nominals

Hungarian simple event and/or result nominals

destroying destruction

(3a) el-pusztít-ás away-destroy-ÁS (3b) pusztít-ás destroy-ÁS

(3c) pusztít-ás destroy-ÁS

signing signature

(4a) alá-ír-ás under-write-ÁS

(4b) alá-ír-ás under-write-ÁS

referring reference

(5a) hivatkoz-ás refer-ÁS

(5b) hivatkoz-ás refer-ÁS

loving love

(6a) *szeret-és love-ÉS

(6b) szeret-et love-ÉS

Fig. 70.3: Comparison of English and Hungarian deverbal nouns

1246

VII. Semantics and pragmatics in word-formation II: Special cases

Consider the following examples in (7) illustrating the use of (3a−c). (7)

a. a város-nak az ellenség által-i el-pusztít-ás-a the city-DAT the enemy by-AFF away-destroy-ÁS-its ‘the city’s destruction by the enemy’

(3a)

b. a város-nak az ellenség által való állandó pusztít-ás-a the city-DAT the enemy by VALÓ constant destroy-ÁS-its ‘the city’s constant destruction by the enemy’

(3b)

c. A pusztít-ás látható volt felül-ről. the destroy-ÁS visible was above-from ‘The destruction was visible from above.’

(3c)

The Hungarian example in (7a) is comparable to the English examples in (1c) and (1d), but formally it is even more similar to its English translation in (7a). The reason for this is that only one possessor constituent may occur in a Hungarian noun phrase; therefore, when the input verb is transitive, its object argument is realized as a possessor in the corresponding nominal construction, and its subject argument is encoded as a by-phrase type oblique expression. The nominal in this example strictly has the complex event interpretation due to the presence of the perfectivizing preverb el ‘away’. In (7b) the interpretation of the nominal lacking a preverb is also strictly a complex event owing to the presence of való ‘being’, an alternative device enabling oblique constituents to appear prenominally in Hungarian noun phrases, see the discussion and the examples in (2). The nominal in (7c) is most naturally interpreted as a result just like its English counterpart in (1f ). The following generalizations hold for the nominal types represented by the Hungarian examples in Fig. 70.3. − If the verb base contains a purely perfectivizing preverb, see (3a) in Fig. 70.3, the -Ás noun derived from it is obligatorily a complex event, clearly manifesting, as a rule, all the features generally attributed to such nouns. − An -Ás nominal derived from the corresponding verb base without the preverb is usually ambiguous between the atelic complex event reading and one or both of the other readings, see (3b) and (3c). − If the verb base contains a preverb which also contributes its lexical meaning to the meaning of the complex verb, that preverb must be present in the -Ás nominal in all its possible meanings, see (4). − If the verb base does not contain a preverb and it has no counterpart with a preverb, the -Ás nominal derived from it can be ambiguous between two or more possible readings, including the complex event reading, see (5). − Of the classical Vendlerian verb types (see Vendler 1967), -Ás can productively derive complex event nominals from accomplishments, achievements and processes, but it does not productively create complex-event-like nominals from states. Consequently, the basic empirical generalization is that it can productively attach only to [+dynamic] event verbs. It appears that the reason for this is that when a state verb is nominalized, very often speakers do not separate the “situation” (that is, the [−dynamic] event) and the result readings, and typically it is the result interpretation that dominates this

70. Nominalization in Hungarian

1247

blend. This is clearly indicated by the fact that the nominal counterparts of the majority of state verbs are forms derived not by -Ás but some other (non-productive) suffixes whose primary function is to express a result, e.g., -At, see (6b). As we have pointed out, following Grimshaw (1990) we assume that complex event nominals do inherit the argument structure of the input verb in its entirety, with the exception of the external argument of a transitive base predicate, which is suppressed (or it can be argued that in Hungarian it is realized by a phonologically null pronoun (see Szabolcsi 1994 and Laczkó 1995). However, contrary to Grimshaw’s account, we claim in Laczkó (2000) on the basis of the Hungarian data that even simple event nominals can retain part of the verb’s argument structure. In particular, evidence from Hungarian shows that the oblique arguments can actually be inherited by the simple event nominal, and it is only the core arguments that are suppressed from the argument structure. The rationale for suppressing the core arguments is that in this meaning the function of the nouns is not to denote an eventuality but to name an (institutionalized) event type, a simple event in this sense. Naturally, it will be understood that even such a simple event has participants (represented in the noun’s lexical-semantic structure) but this is background information that need not be encoded in the argument structure. On the other hand, if the meaning of an event type comprises a meaning element expressed by an oblique argument of the input verb, this argument has to be retained, otherwise the nominal will have a different meaning. Consider the following examples. (8)

a. Péter át-repül-t a híd alatt. Peter across-fly-PAST.3SG the bridge under ‘Peter flew across under the bridge.’ b. Péter-nek a híd alatt való át-repül-és-e mindenki-t Peter-DAT the bridge under VALÓ across-fly-ÉS-his everybody-ACC megijeszt-ett. scare-PAST.3SG ‘Peter’s flying across under the bridge scared everybody.’ c. *Péter-nek az átrepül-és-e mindenki-t megijeszt-ett. Peter-DAT the across-fly-ÉS-his everybody-ACC scare-PAST.3SG (without appropriate context) ca. ‘Peter’s flying across scared everybody.’ d. *Péter-nek a híd alatt-i át-repül-és-e-i mindenki-t Peter-DAT the bridge under-AFF across-fly-ÉS-his-PL everybody-ACC megijeszt-ettek. scare-PAST.3pl ‘*Peter’s flyings across under the bridge scared everybody.’

The obligatory nature of the arguments, the impossibility of pluralization and the presence of the való ‘being’ attributivizer show that átrepülés in (8b) is a complex event nominal. But let us now consider the following scenario. There is a special contest for pilots. They have to fly under designated bridges. In this situation flying under a bridge can be conceived of as an institutionalized event and, consequently, realized by a simple

1248

VII. Semantics and pragmatics in word-formation II: Special cases

event nominal converted from the -Ás complex event nominal. And the oblique argument has to be expressed (or it can only be elided if it is recoverable from the context). (9)

a. A bizottság eddig 34 híd alatt-i át-repül-és-t the committee so_far 34 bridge under-AFF across-fly-ÉS-ACC értékel-t. evaluate-PAST.3SG ‘So far, the committee has evaluated 34 flights across under the bridge.’ b. Ez a nap legjobb híd alatt-i át-repül-és-e volt. this the day best bridge under-AFF across-fly-ÉS-its was ‘This was the day’s best flight across under the bridge.’

In these constructions, átrepülés is used as a simple event nominal. This is indicated by its countability in (9a) and the possibility of a non-thematic possessor (9b), that is, the complete suppressibility of the otherwise obligatory intransitive agent argument. On the other hand, the oblique argument is still obligatory. Even this very special example can be considered an ordinary instance of converting a complex event noun into a simple event noun when in a particular community the need for this arises. However, in addition to such a case, there are special cases which are always ad hoc and occasional uses of a typically complex event nominal in a simple event nominal sense. The peculiarities of this usage are as follows. None of the arguments has to be suppressed (other than the agent of a transitive verb base). The noun can be pluralized. The meaning of the whole construction will almost always have a marked, improvizational flavor to it. For instance, if someone says (8d), it will be unacceptable in the ordinary complex event nominal reading (because of the plurality of the nominal). However, marginally it can be reinterpreted in such a way that presumably for the speaker for some reason ‘Péter’s flying under a bridge’ is not a complex event but a simple event, and he is referring to several instances of this simple event. But, as we have pointed out, this is only a peripheral use which can be compared to the following usage: *?I need informations (intended meaning: ‘I need various kinds of information’). As far as the usage of -Ás nominals is concerned, given the central and basically indispensable function of the -Ás suffix in the targeted meanings, they are extremely frequent in their nominalization domain. In the fully productive complex event sense, speakers in spontaneous speech tend to use relatively short, simple noun phrases typically not containing more than a possessor argument and possibly at most one oblique argument attributivized by the való ‘being’ element or by some other attributivizing device admitted by the syntactic category of the oblique argument in question, and some possible additional adjectival modification. It is also to be noted that a significant number of speakers try to avoid using való ‘being’ constituents by (linguistically speaking) erroneously assuming that they are instances of “foreignism”. This assumption is erroneous, because this construction type is not at all alien to the nature of Finno-Ugric languages; moreover, there are cases in which the való ‘being’ construction has no alternative (when the oblique constituent is expressed by a case-marked noun phrase). Syntactically, considerably more complex noun phrases headed by complex event nominals are frequently used in formal or written technical Hungarian.

70. Nominalization in Hungarian

1249

3. -Ó (V to N) As the following examples show, an -Ó word can be used as a (“present”) participle (10), as an adjective (11), and as a noun denoting the doer (12)−(13), the instrument (14) or the typical location of an action (15), and, occasionally, the action itself (16). (10) a mindenki-t meglep-ő hír the everybody-ACC surprise-Ő news ‘the news surprising everybody’ (11)

a még meglep-ő-bb hír the even surprise-Ő-COMP news ‘the even more surprising news’

(12) a. ír-ó write-Ó ‘writer’

b. a névtelen levél ír-ó-ja the anonymous letter write-Ó-its ‘the writer of the anonymous letter’

(13) a. *a kiagyal-ó the conceive-Ó ‘*the conceiver’

b. a terv kiagyal-ó-ja the plan conceive-Ó-its ‘the conceiver of the plan’

(14) a. számlál-ó count-Ó ‘counter’

b. az üveg-ek számlál-ó-ja the bottle-PL count-Ó-their ‘the counter of the bottles’

(15) a. kávéz-ó have.coffee-Ó ‘café’

b. internetez-ő use.the.internet-Ő ca. ‘the designated place for using the internet’

(16) a. találkoz-ó b. esküv-ő vow-Ő meet-Ó ‘(official) meeting’ ‘wedding’ The -Ó participle is comparable to English -ing (as used in noun phrases) and the -Ó adjective can be assumed to be derived from it by conversion, cf. (10) and (11). For the postulation of participle-adjective conversion in English in the framework of lexicalfunctional grammar, see Bresnan (1982). Komlósy (1992) proposes a battery of tests for telling participles and adjectives apart in Hungarian. For instance, participles cannot be used predicatively, and they cannot have comparative and superlative forms. On the other hand, adjectives cannot have object arguments, as opposed to (active) participles. Thus, the -Ó form in (10) is definitely a participle, and the -Ó form in (11) is unquestionably an adjective. Note that (12a) and (14a) express a profession and an instrument, respectively, while (12b), (13b) and (14b) exemplify the productive use of -Ó nominalization that inherits the argument structure of the input verb. In present day Hungarian, -Ó, in addition to its agentive and instrumental use, can be productively employed to derive nouns that denote the typical/institutionalized location of an event. This productivity is shown by the fact that new predicates can also take the -Ó suffix to express a location, see (15b). The “simple event” use of -Ó, illustrated in (16), is no longer productive.

1250

VII. Semantics and pragmatics in word-formation II: Special cases

As regards the derivation of -Ó nouns, at first sight it would appear simpler and more straightforward to assume participle-to-noun conversion because it is a view generally held that -Ó adjectives are converted from -Ó participles, see Komlósy (1992), Kiefer (1998) and Laczkó (2006). However, in Laczkó (2000) we assume that in present-day Hungarian there is an -Ó deverbal nominalizing suffix homophonous with the participial suffix. In actual fact, we claim that the three productive uses of -Ó nouns are so distinct that it is even more appropriate to posit homophonous -Ó nominalizing suffixes. Our major motivation for postulating a (multifunctional) nominalizing suffix is as follows. In the case of the -Ó participle, the justification for assuming a participle-to-adjective conversion process is definitely beyond question. In this context, the postulation of yet another conversion, especially one with a different output category (participle-to-noun) would be rather implausible. Moreover, as regards the three productive uses of -Ó nouns, at least two radically distinct functions would have to be associated with this participleto-noun conversion. Let us take a closer look at these three principal cases. The productive agentive and instrumental uses of the -Ó forms in question can be lumped together, and it can be assumed that the -Ó1 nominalizing suffix targets verbs whose subject is an “external argument”, which is typically an agent, but it can also be an instrument if there is no agent in the argument structure. It is noteworthy that the agentive/instrumental use of -Ó in Hungarian is directly and straightforwardly comparable to the relevant use of the -er suffix in English, cf. the English translations in (12) and (14). For a detailed generative linguistic analysis of -er, see Rappaport Hovav and Levin (1992), and for a detailed generative linguistic analysis of -Ó, see Laczkó (2000, 2001). In the locative use of -Ó nouns, it can be postulated that the -Ó2 nominalizing suffix attaches to a verbal stem, eliminates its argument structure and creates a noun that refers to the typical location of the action designated by the input verb. In Hungarian, there is practically no verb-noun conversion in either direction, as opposed to English (cf. look (V) → look (N), email (N) → email (V)). However, it is a generally accepted assumption in Hungarian historical linguistics that in early Hungarian the majority of derivational suffixes were multi-functional, or, rather, category-neutral (for discussion and references, see Károly 1956). For instance, the -Ó suffix in early Hungarian can be considered to have had both participle forming and nominalizing functions simultaneously. In present day Hungarian, the nominalizing function of this suffix is no longer productive, but it used to be. In Laczkó (2000), we discuss the emergence of the locative use of -Ó and we also propose a more principled explanation for the alleged “passive” use of -Ó participles in compounds. In both cases, our conclusion is that the source of these developments is the -Ó “event” noun and not the -Ó participle.

4. -sÁg (A to N) This suffix productively attaches to adjectives, and turns the input adjective’s attribute meaning into the corresponding abstract notion meaning. It is straightforwardly comparable to English -dom, -ness and -ity. Consider the following examples. (17) a. szabad-ság free-SÁG ‘freedom’

b. kék-ség blue-SÉG ‘blueness’

c. buta-ság stupid-SÁG ‘stupidity’

70. Nominalization in Hungarian

1251

It is generally assumed that it derives nouns from adjectives with an argument structure, which minimally have a subject argument when they are used predicatively (and assumptions differ as to whether they have a subject argument when they are used attributively). When -sÁg is attached to the adjective, the derived noun can normally occur on its own (that is, the expression of the subject of the input adjective becomes optional); however, when the derived noun occurs in a possessive construction, the possessor, as a rule, has to be interpreted in the same manner as the subject of the input adjective. Thus, it is plausible to assume that the derived noun inherits the argument structure of the input adjective in such a way that the expression of the “subject” argument as a possessor (just like in the case of deverbal nouns, see section 2) becomes optional. Moreover, several adjectives have an additional oblique argument, and this argument is also inherited by the derived noun in an optional fashion, and when it is present in the construction, it basically has the same primary form of realization as the oblique arguments of nouns derived from verbal predicates: the való ‘being’ construction, cf.: (18) a. A döntés független volt a külső tényező-k-től. the decision independent was the external factor-PL-from ‘The decision was independent of the external factors.’ b. a döntés-nek a külső tényező-k-től való független-ség-e the decision-DAT the external factor-PL-from VALÓ independent-SÉG-3SG ‘the decision’s independence of the external factors’ Note, however, that there is a viable alternative analysis available in this case, given the optionality, in combination with the -sÁg suffix, of all the arguments of the input adjectives. It can also be assumed that the derivation removes the argument structure of the adjective, and, thus, the optional constituents in question are not arguments, instead, they are only complements in Grimshaw’s (1990) terminology. Basically, the former account is along the lines of Grimshaw’s treatment of complex event nominals, whereas the latter is in the same vein as her analysis of simple event nominals.

5. -itás (A to N) The input adjective expresses some abstract attribute and the resulting noun denotes the corresponding abstract notion. The suffix is of foreign origin, and it is directly relatable to -ity in English. It can only attach to adjectives of foreign origin, and its function is the same as that of its English counterpart. Given the foreign word orientation of the suffix, the derived noun typically occurs in either ordinary or formal speech. (19) a. mobil-itás mobile-ITÁS ‘mobility’

b. modern-itás MODERN-ITÁS ‘modernity’

c. kommunikativ-itás COMMUNICATIVE-ITÁS ‘communicativity’

6. Conclusion This article has discussed productive present-day Hungarian nominalization phenomena. After pointing out the most important morphological and morphosyntactic properties of

1252

VII. Semantics and pragmatics in word-formation II: Special cases

individual suffixes, it concentrated on their crucial semantic and pragmatic traits. In Fig. 70.4 we offer a summary of the most significant properties of these suffixes. The representation has the following general format: suffix (i) X[input category] (most salient meaning(s) of X) (ii) Y[output category] (most salient meaning(s) of Y) (iii) pragmatic information (iv) additional information (if any). -Ás

(i) V (dynamic predicate) (ii) N (complex event nominal) (iii) neutral (iv) argument structure of input verb retained; non-productive types: simple event and result nominal (without argument structure); some speakers avoid using való ‘being’ as a means of attributivizing prenominal adjuncts and oblique arguments

-Ó

(i) V (predicate) (ii) N (doer, instrument, location of activity) (iii) neutral (iv) verbto-noun derivation and not adjective-to-noun conversion; simple event nominal use no longer productive

-sÁg

(i) A (attribute) (ii) N (abstract notion) (iii) neutral (iv) all arguments of input adjective become optional, two possible analyses: with or without argument structure

-itás

(i) A (attribute) (ii) N (abstract notion) (iii) neutral or formal (iv) foreign origin of both suffix and input adjective

Fig. 70.4: Verb-to-noun and adjective-to-noun nominalization

Acknowledgements The author gratefully acknowledges that the research reported here was supported by the OTKA (Hungarian Scientific Research Fund) project entitled Comprehensive Grammar Resources: Hungarian (grant number: NK 100804).

7. References Bresnan, Joan 1982 The passive in lexical theory. In: Joan Bresnan (ed.), The Mental Representation of Grammatical Relations, 3−86. Cambridge, MA: MIT Press. Bresnan, Joan 2001 Lexical-Functional Syntax. Oxford: Blackwell. Grimshaw, Jane 1990 Argument Structure. Cambridge, MA: MIT Press. Károly, Sándor 1956 Igenévrendszerünk a kódexirodalom első szakaszában. Nyelvtudományi Értekezések 10. Budapest: Akadémiai Kiadó. Kiefer, Ferenc 1998 Alaktan. In: Ferenc Kiefer, Katalin É. Kiss and Péter Siptár (eds.), Új magyar nyelvtan, 185−289. Budapest: Osiris Kiadó. Komlósy, András 1992 Régensek és vonzatok. In: Ferenc Kiefer (ed.), Strukturális magyar nyelvtan 1. Mondattan, 299−527. Budapest: Akadémiai Kiadó.

71. Result nouns

1253

Laczkó, Tibor 1995 The Syntax of Hungarian Noun Phrases. A Lexical-Functional Approach. Frankfurt/M.: Lang. Laczkó, Tibor 2000 Az ige argumentumszerkezetét megőrző főnévképzés. In: Ferenc Kiefer (ed.), Strukturális magyar nyelvtan 3. Morfológia, 293−407. Budapest: Akadémiai Kiadó. Laczkó, Tibor 2001 A comprehensive analysis of -ó/-ő, a multifunctional deverbal nominalizer in Hungarian. In: István Kenesei (ed.), Argument Structure in Hungarian, 13−49. Budapest: Akadémiai Kiadó. Laczkó, Tibor 2006 Melléknév, ige vagy melléknévi igenév? In: László Kálmán (ed.), KB 120. A titkos kötet. Nyelvészeti tanulmányok Bánréti Zoltán és Komlósy András tiszteletére, 193−212. Budapest: Tinta Könyvkiadó. Rappaport Hovav, Malka and Beth Levin 1992 -er nominals: Implications for the theory of argument structure. In: Tim Stowell and Eric Wehrli (eds.), Syntax and the Lexicon, 127−154. New York: Academic Press. Szabolcsi, Anna 1994 The noun phrase. In: Ferenc Kiefer and Katalin É. Kiss (eds.), The Syntactic Structure of Hungarian, 179−274. New York: Academic Press. Vendler, Zeno 1967 Linguistics in Philosophy. Ithaka: Cornell University Press.

Tibor Laczkó, Debrecen (Hungary)

71. Result nouns 1. 2. 3. 4. 5. 6.

Introduction Result nouns in the generative tradition Morphological properties in a cross-linguistic perspective Stricto sensu results: a heterogeneous class Semantic properties of the base verb References

Abstract Result nouns are deverbal nouns denoting the object or the state produced by the event expressed by the base verb. Interestingly, many Indo-European languages possess a limited amount or lack altogether dedicated and productive morphological means to form result nouns. The polysemy of deverbal action nouns emerges as an important tool for expressing result objects or states, also in a cross-linguistic perspective. The present article provides the reader with an overview of some influential studies dedicated to the event/result polysemy of action nouns and a brief survey of the morphological, syntactic and interpretive properties of result nouns on the basis of data drawn from European languages (English, German, Swedish, Italian, Russian).

1254

VII. Semantics and pragmatics in word-formation II: Special cases

1. Introduction Looking across several studies on nominalization, two main ways of approaching the class of result nouns neatly emerge. On the one hand, we find a syntactic notion of result noun, which defines a semantically heterogeneous class of deverbal nouns derived by means of so-called transpositional affixes (see Beard 1995). These nouns, though morphologically indistinguishable from event-denoting nominalizations, which are argument-taking and possess other verbal features (see article 67 on action nouns and article 68 on action nouns in Romance), have the morphosyntactic behavior of absolute nouns and denote several types of referential entities encompassing, beyond results or effects, means/instrument (e.g., heating), location (e.g., encampment), agent (e.g., governance) and further meanings. On the other hand, there is a stricto sensu definition of result noun, which groups together deverbal nouns naming the outcome/effect/byproduct of the event expressed by the base verb, hence overlapping with the category of nomina acti (see, among others, the onomasiological classification of deverbal nouns in German proposed by Motsch 1999 and relevant discussion in Osswald 2005: 256−257). Specifically, result nouns can be entities that come into existence during the event denoted by the base verb (e.g., creation, formation) or may refer to the resultant state of a telic event (e.g., contamination, segregation). Under both definitions, result nouns are complex word forms derived from verbs and formed by means of the same range (or a subset) of affixes involved in the formation of action nominals. They exhibit yet an exceptional and, arguably, derivative morphosyntactic and semantic characterization with respect to event-denoting nouns. Further, apart from lexicalizations (e.g., E. information, G. Wohnung ‘apartment’, It. calzatura ‘shoe’), a result noun tends to have an event denotation too. As we will see in section 3, exceptions to this claim are deverbal nouns obtained through zero suffixation, which can unambiguously point to results, contrary to nominals derived by means of overt suffixation (see the doublet squarcio/squarcia-mento ‘laceration’ ← squarciare ‘to lacerate’ in Italian, where the zero-affixed form unambiguously denotes the result object). Quite the reverse, many action nouns lack a corresponding result interpretation, e.g., E. competition, It. inseguimento ‘pursuit’, Russian uničtoženie ‘annihilation’. Besides the event/result polysemy of action nominals, however, further morphological means are cross-linguistically attested for expressing the objects (but not the states) resulting from the events named by the base verb. On the one hand, in Italian we find nominal conversions of (some) past participles, e.g., frullato ‘(milk) shake’ ← frullare ‘to blend’, filmato ‘short film’ ← filmare ‘to film’, etc. As with event/result nouns, we are not dealing with a phenomenon dedicated to the formation of result nouns only, since It. -to nouns generally express the object of the verb, be it a theme (venduto ‘goods sold’ ← vendere ‘to sell’) or a patient (assistito ‘patient/client’ ← assistere ‘to assist’). On the other hand, Russian (not alone among Slavic and non-Slavic languages) has some suffixes which, combined with verbal bases, unambiguously express the byproducts or resulting objects of the event, e.g., carapina ‘scratch, abrasion’ ← carapat’saˆ ‘to scratch’, obrubok ‘stump, stub’ ← obrubit’ ‘to chop off ’, vyzˇimki ‘squeezings’ ← vyzˇimat’ ‘to squeeze’ (see article 162 on Russian). These affixes have other functions too, being employed for the formation of object, place, instrument nouns, etc. These and comparable data will not be addressed in this article, which is dedicated to result nouns as semantic extensions of action nouns.

71. Result nouns

1255

Against this background, it should be clear that result nouns represent a special case among nominalization phenomena. In particular, the general paucity of dedicated and productive derivational means, especially lacking in the Romance and Germanic languages, correlates with the fact that result nouns mainly represent a robust and (semi-)regular case of the polysemy displayed by action nominals. This ambiguity has not only attracted semanticists, but syntacticians too have studied this aspect of nominalization with the aim of capturing the non-trivial morphosyntactic correlates of the event/ result polysemy. Before examining in more detail data drawn from some Indo-European languages, we will briefly see how generative linguists have dealt with event/result deverbal nouns, paying special attention to their understanding of the result class.

2. Result nouns in the generative tradition In the generative framework, the issue of idiosyncratic semantics in action nominals dates back, at least, to Chomsky’s (1970) seminal investigation Remarks on Nominalization, which called attention to several peculiar properties of this derivational mechanism not easily explainable in terms of transformations and, so doing, paved the way towards lexicalist approaches to word-formation (see article 7 on word-formation in generative grammar and article 45 on rules, patterns and schemata in word-formation). Among the first generative studies pointing to the dichotomy action vs. result nominals (see Anderson 1983; Lebeaux 1986; Malicka-Kleparska 1988; Roeper 1987 and Zubizarreta 1987), Grimshaw (1990) is arguably the most influential. In chapter 3 of her book, she establishes a correspondence between a specific array of morphosyntactic properties of nominals and their semantic characterization. Grimshaw, in particular, identifies a set of diagnostics able to reveal the underlying nature of a deverbal noun and to isolate the class of result nouns with respect to the others. Since Grimshaw (1990), however, the mainstream morphosyntactic literature has adopted the label “result” as a cover term for nominals lacking argument structure and other verbal properties; hence this label has gradually lost its semantic motivation, as explained in section 2.1. Other studies, shortly overviewed in section 2.2, have specifically targeted the semantic ambiguity of nominals and surveyed the meaning values of result nouns more attentively. In these works, attempts have been made to define a finer-grained taxonomy of the meanings typically grouped together in this class, and to understand the source of the event/result polysemy in deverbal nominals.

2.1. Result nouns in a morphosyntactic perspective Conceived in a lexicalist framework, Grimshaw’s analysis identifies three classes of nominals: on the one hand, she identifies “complex event” nominals, manifesting an array of verb-related properties; on the other, she groups together “simple event” and “result” nouns, since they both lack the peculiar properties of the former class despite a seemingly identical morphological structure. In order to account for the event/result dichotomy in particular, she postulates the existence of ambiguous affixal entries en-

1256

VII. Semantics and pragmatics in word-formation II: Special cases

dowed with different binding properties with respect to the argument structure of the base verb. Specifically, an English suffix like -(at)ion (or -ment/-ure, etc.), comes with an ambiguous specification of its external argument, i.e. Ev (event) or R (referential) argument, and this has serious drawbacks for the morphosyntactic behavior of the corresponding nominal (see Grimshaw 1990: 63−68). On this basis, Grimshaw’s rationale captures the fact that result nouns refer to entities, either concrete or abstract, since they are obtained through the binding of an argument of the base verb (one of Grimshaw’s examples is observation, where “-ion binds something like a theme”, Grimshaw 1990: 66). On the other hand, the binding of the Ev argument of the verb generates a complex event nominal, which preserves the internal argument and other verbal properties. It remains unclear, in this account, how to derive the meaning of result nouns that do not correspond to argument-structure participants of the base verb (e.g., translation, correction, etc.). More recent attempts to account for the event/result puzzle in nominalization are Borer (2003) and Alexiadou (2001), both conceived in constructionist frameworks “which revive in various forms the sentential Generative Semantics analyses of event nominals” (Newmeyer 2009: 91), and share the assumption that “the structure […] determines not only grammatical properties, but also the ultimate fine-grained meaning of the lexical items […]” (Borer 2003: 33). Borer’s (2003) approach to nominalization theorizes that an English noun such as formation can be formed in one of two ways: either the suffix -ation attaches on top of a complex chunk of hierarchical structure crucially including verb-related functional projections, or it merges directly with a categorially neutral lexical item, which gets its category through the nominalizing affix. In the former case, the derivation generates an argument-supporting nominal (roughly corresponding to a complex event nominal in Grimshaw’s account); in the latter it delivers a referential noun. Hence, in this analysis the lack of verbal properties of result/referential nouns is parasitic on the lack of a verb in their syntactic derivation. Result nouns are therefore structurally different from event nouns. As with most syntactic approaches to this phenomenon, lexical-semantic issues in result nominals are completely overlooked; it is not accidental that, whereas most studies address result nouns, Borer (2003) more appropriately targets the class of referential nouns, a label grouping together all non-argument taking nominals, independently of their semantics. Alexiadou’s (2001) analysis shares with Borer’s (2003) the common basis of imputing the verbal properties of action nominals to the presence of an extended verb phrase (with non-trivial differences in its internal hierarchical constitution) within the structural makeup of the derived noun; analogously, the behavior of result nominals as absolute nouns would depend in her analysis on the lack of verb-related functional projections within the derivation of nominals. All in all, being focused on the morphosyntactic derivation of result nouns, these analyses neglect the existence of essential semantic distinctions in the class of result nouns, which are mainly defined on the grounds of their morphosyntactic properties rather than on lexical-semantic ones. Moreover, both Grimshaw and Borer (the former explicitly, the latter implicitly) assume that the very existence of result nouns follows from the ambiguity of the nominalization process itself. Specifically, it is a consequence of affixal ambiguity in Grimshaw’s analysis, and it depends on the locus of merger of the nominalizing affix (i.e. at different levels in the derivation, according to Borer 2003 and Alexiadou 2001). However, although they target the source of this ambiguity, the issue of its type is not further specified in these accounts: that is, it is not specified

71. Result nouns

1257

whether result nouns are homonymous forms or an interpretive extension of (polysemous) action nouns. A step forward towards the integration of a finer-grained semantic analysis in the morphosyntactic treatment of result nouns can be found in Alexiadou (2009) and Roßdeutscher and Kamp (2010). Both approaches attribute an important role to the semantics of the root, intended as a category neutral item in distributed morphology, which might be held responsible for the seemingly unpredictable and idiosyncratic meaning of the referential/result noun. On the basis of a detailed morphological analysis of deverbal nouns in Greek, Alexiadou (2009) suggests that the meaning type of the root should play a role in determining the semantic type of a result nominal. In particular, she identifies nominals having as their bases stative roots or roots denoting manner, instrument or entities. On this, Alexiadou specifically argues (2009: 271): “[…] the availability of the result interpretation will always be dependent on a particular combination of v and the different types of roots. This might explain why certain derived nominals are ambiguous between event and result interpretations, while others are ambiguous between event and object interpretations. The latter contain roots that are not stative, but rather instruments or entities”. A finer-grained semantic analysis, combining assumptions and tools of distributed morphology and discourse representation theory, can be found in Roßdeutscher and Kamp (2010), who attribute the readings of -ung nominals in German to the type of root involved in their formation. Specifically, if the root identifies a sortal noun, like in Bestuhlung ‘seating/installation of seats’, the expected readings of the nominal are event, state and object; if it denotes a property, e.g., Schwa¨chung ‘weakening’, the nominal should convey event and result-state readings. However, as they admit, the picture is complicated by the presence of property and sortal root-based nominals lacking the predicted result-state reading, e.g., Sa¨uberung ‘cleaning’ or Mischung ‘mixture’, and, in particular, by property-root based nominals conveying an unpredicted result reading, e.g., A¨nderung ‘change’. Therefore, despite being focused on structural properties, syntactically oriented analyses too have started to consider the semantic role of the core derivational source, i.e. the root, as a crucial factor for determining the type of result noun.

2.2. Result nouns in a lexical-semantic perspective Lexical-semantic studies on result nouns are not lacking but, in this case too, result nouns are targeted in the wider perspective of the semantic variation exhibited by action nominals. Among the semantic studies developed in the generative and lexicalist tradition is Bierwisch’s (1989) proposal, which represents one of the first attempts to account in a principled way for the ambiguous interpretation of deverbal (and deadjectival) nominalizations (cf. also article 61 on word-formation and metonymy). First of all, mainly on the grounds of German data, Bierwisch identifies two types of result noun: result states and result objects, either physical or abstract (e.g., Ordnung ‘arrangement’, Bebauung ‘construction’, Übersetzung ‘translation’). Further, Bierwisch insists on the role of conceptual and encyclopedic knowledge associated with the event codified by the base verb as the main source for conceptual shifts of semantic interpretation in deverbal nouns. For result interpretations, in particular, he claims that “the character of the result is determined by the conceptual knowledge related to the type of event” (1989: 38).

1258

VII. Semantics and pragmatics in word-formation II: Special cases

Accordingly, nominalizations derived from verbs like jump, ride, recite cannot denote resulting entities or states, because the actions expressed by these predicates do not lead to the formation of results. In order to account for the semantic variation of nominalizations, Bierwisch suggests the existence in the lexical system of semantic templates, which “provide the systematic patterns channeling the flexibility of conceptual interpretation” (1989: 49). In this approach, result nouns are thus not specified as such in the lexicon but are the output of a semantic shift applying to event nouns and taking place in specific contextual settings. Further, conceptual shifts instantiate a polysemy that pervades the simplex lexicon too, and, as such, they cannot be treated as a specific property of nominalization or of the type of derivational morphology involved in these phenomena. Along the lines of Bierwisch (1989), Brandtner (2011), though recognizing the role of lexical restrictions on the interpretation of deverbal nouns, focuses on the interpretations of nominals in context, specifically by paying attention to co-predication structures, wherein the conflicting event and result readings are simultaneously selected by modifiers and predicates (e.g., Die fünfminütige Messung ist auf zwei Stellen genau ‘the fiveminute measurement is accurate to two decimal places’, see Brandtner 2011: 15). In her account, result readings and other meaning shifts of -ung nominals in German are analyzed as the outcome of meaning transfers à la Nunberg (1995). On the basis of English nominalizations, Pustejovsky (1995) insists on the lexicalsemantic features of the base verb as the source for result readings. In his work, the polysemous character of process/result nouns is understood as a case of logical polysemy, accordingly defined as “dot objects” in the generative lexicon framework. Restricted to accomplishment-based nominals, Pustejovsky’s analysis rests on the hypothesis that resultstate and result-object readings are associated to the state subevent in the complex event structure of the base verb. Hence, both types of result meanings should not be available to nominals derived from process verbs (like walk), while result-object reading should be inaccessible to nominals like destruction, since they describe events directed towards the destruction of preexisting objects. Therefore, Pustejovsky calls attention to the semantics of the base verb, considering the event-structural properties and root/idiosyncratic properties of the base verb. Lastly, on the grounds of Italian data and within the lexicalist framework of word-formation developed by Lieber (2004), Melloni (2007, 2010, 2011) develops a theoretical account of the event/result polysemy exploring the semantic features of the morphological building blocks in deverbal nominals. Specifically, the formation of result nouns is attributed to the inherent polysemous characterization of the transpositional affixes involved in deverbal nominalizations. Further, special attention is dedicated to the full range of polysemy patterns of deverbal nominals, encompassing agentive, means, manner, and temporal denotations. Concerning stricto sensu results, as referring to entities that come into existence during the event denoted by the base verb, it is explained how it is possible to predict this interpretive value on the grounds of a semantic analysis of the verbal base (we will come back to this issue in section 5). To sum up, several important issues have been discussed in these accounts, which might be summarized as follows: the affinity in nominal polysemy between morphologically simplex and complex lexical items, the crucial import of verb semantics for triggering a result noun, the study of the inherent character of this polysemy and its actualization in sentential contexts.

71. Result nouns

1259

3. Morphological properties in a cross-linguistic perspective In the preceding sections, we have seen that result nouns have been understood as an elusive category from a semantic point of view. In accordance with the aims of the present contribution, however, we will consider as result nouns only those nominals identifying the effect or by-product of an event; other meaning extensions of action nouns are addressed in further articles of this volume (see articles 67, 68 on action nouns, and 73−76 for other nominalization types). Furthermore, though an adequate cross-linguistic analysis exceeds the limits of this contribution, we will try to highlight the properties of result nouns through data from different languages. On the grounds of several investigations dedicated to action nouns in various European languages, it seems feasible to argue that their polysemy patterns are very similar and, as we will see in section 5, semantic considerations with respect to their verbal sources, in particular, seem to be cross-linguistically valid. Important differences arise when considering the morphological properties of result nouns, both with respect to the bases and to the head suffixes. It should be borne in mind that result nominals are formed by means of the same suffixes heading action nouns; hence, it is interesting to see whether there are morphological restrictions on or affixal specializations/preferences for the formation of result nouns. Starting from the latter, the suffixes involved manifest different properties as to the (often idiosyncratic) selection of their verbal bases and to their chance of heading ambiguous event/result nouns. First, it should be noticed that in general there is not just one, but several transpositional suffixes yielding nominals with event/result interpretations. These affixes in fact cover approximately the same semantic space or, to say it with Booij and Lieber (2004), they constitute a single paradigmatic cell of semantic derivation and are semantically interchangeable. Let’s start from English: here we find several affixes of Latinate origin, -ment, -ion, -ure, -ance, -al, which can be the head of event/result deverbal nouns. Further, Grimshaw (1990: 67) argues that -ing forms event nominals, whereas the zero suffix should give unambiguous result nouns only (loosely defined as such, and not necessarily as effects or products). However, exceptions to these generalizations can be found in both directions, showing that strong generalizations about the polysemy of deverbal action nouns are hardly tenable: (1)

John endeavored to publicize his findings in the local media. → Result

(2)

The frequent release of the prisoners by the governor. → Event (Newmeyer 2009: 101)

In German, suffixation with -ung covers approximately 80 % of the deverbal nouns (Osswald 2005). The meaning options of -ung nouns are various, and encompass locative, manner, and temporal meanings, beyond the stricto sensu event/result polysemy. As to its selection properties, -ung preferably selects transitive, dynamic and prefixed base ¨ bersetz-ung ‘translation’, Bepflanz-ung verbs (e.g., in Absperr-ung ‘obstruction’, U ‘planting’). Nevertheless, there are many exceptions to these general selection tendencies, disclosing -ung’s sensitivity to more subtle semantic features in the base verb (the reader is referred to Roßdeutscher and Kamp 2010 for a thorough discussion). In Swedish (see Tenev 2008), three suffixes compete for the formation of event/result nouns: -ning/ing, -ande/ende, zero suffix. The suffix -ning is widely used to form deverb-

1260

VII. Semantics and pragmatics in word-formation II: Special cases

ala nouns with a (secondary) result meaning: o¨ppning ‘opening’ ← o¨ppna ‘to open’, byggning ‘large building’ ← bygga ‘to build’. Interestingly, the suffix -ande is semantically specialized, since it primarily denotes the result of a mental activity in speech or in writing: meddelande ‘a message’ ← meddela ‘to send a message’, påstående ‘a statement’ ← påstå ‘to state’, uttalande ‘an utterance, a statement’. Zero suffixation (as in English, the nominal use of the verbal stem), whose productivity is increasing, produces result nouns too: fo¨rho¨r ‘interrogation’ ← fo¨rho¨ra ‘to interrogate’, bidrag ‘contribution’ ← bidraga ‘to contribute’. Among Romance languages, Italian exhibits a pattern comparable to English, with -mento, -zione and -tura having similar effects on their bases in that nouns derived by means of these affixes exhibit the polysemy at issue (see Melloni 2006 on these affixes). Zero derived nominals, in the masculine and feminine forms (e.g., tagli-o ‘cut’ ← tagliare ‘to cut’, rettific-a ‘amendment’ ← rettificare ‘to amend’) are also attested and prone to produce result nouns. Other affixes are either attested to a lesser degree or manifest different properties: -aggio is hardly found in result nouns; -ata possesses inherent quantitative/aspectual properties, specifically − as a “packaging operator” (see Gaeta 2004) − it temporally delimits the situation expressed by the base verb (nuotata ‘(a) swim’); -ìo mainly heads unambiguous event nominals or nominals with event/sound polysemy (sgocciolio ‘drip’). Moreover, the same base verb can be nominalized by two or more affixes, which either select different meanings of the base verb or form nominals with different polysemy options. As an instance of the former case, consider the pair trattazione/tratta-mento ‘investigation/treatment’, where each member has been lexicalized with distinct senses dependent on the selection of a specific meaning of trattare, the base verb. As an instance of the latter, consider the pair divarica-mento/divarica-zione ‘divarication/divergence’, where only the latter conveys a result-state reading (see Gaeta 2004: 318). Among the Indo-European languages, Slavic represents a very interesting case because, contrary to Romance and Germanic, grammatical aspect is morphologically encoded on the verb, especially by means of affixal devices. Let us consider Russian nominalizations: as in Romance and Germanic, there exist several productive nominalizing suffixes attaching to simplex and complex verbal bases: -ka, zero suffix, -niе (and its many allomorphs), and -stvo are among the most productive ones, and all compete in principle for the same base. The situation is complicated here by the fact that the base verb usually has two forms, i.e. one for the imperfective (often derived through so-called secondary imperfectivization) and the other for the perfective. In many cases, derived action nouns preserve this opposition, but only morphologically, since the aspectual opposition gets semantically neutralized under nominalization, see Schoorlemmer (1995). As an example, consider the pair perepiska/perepisyvanie ‘copying’ from perepisat’PF and perepisyvat’IPF ‘to copy’ respectively. In other cases, the opposition is also morphologically neutralized, since nominalizations are obtained from either the imperfective or the perfective stem only. For example, from the paired verb kroit’IPF/skroit’PF ‘to cut out’, we get three nominals but all derived from the imperfective, i.e. kroj, krojka, kroenie ‘cutting’. In general, there is a tendency for perfective bases to yield nominals with result readings, while imperfectives tend to yield event-denoting nominals. However, several exceptions can be found in both cases (see Valdivia, Castellví and Taulé 2013 for a thorough discussion and for the data reported here). A quite robust generalization concerns nominalization of secondary imperfectives by means of -nie (e.g., raspisyvanie

71. Result nouns

1261

‘writing out’ ← raspisyvat’ ‘to write out’). These nominals tend to block the result meaning and other semantic extensions, suggesting that, contrary to other imperfectivebased nominals, the unbounded aspectuality of the base verb is retained under nominalization (see Sadler, Spencer and Zaretskaya 1997). Furthermore, although there is no strictly predictable or regular correspondence between morphological aspect and the interpretation of the corresponding nominal, the general availability of more than one base allows for a greater specialization of the nominal meaning, with the consequence that some of these nouns are unambiguously specified as result-denoting in context, e.g., izdanie ‘publication/edition’ or zakaz ‘order’. All in all, it seems that the distribution of the suffixes within the domain of result nouns cannot be described in terms of strictly morphological factors; the availability of result nouns does not depend on the use of specific affixes, but on a complex interplay of lexical and semantic factors, largely dependent on the semantics of the base verb. Interestingly, however, zero-derived nominals (e.g., E. cut, Ru. zakaz ‘order’, It. squarcio ‘laceration’) are those that more consistently tend to give nominals with result (or object) meaning only. As a general consideration, this tendency, which is manifested crosslinguistically, might be imputed to the lack of an overt transpositional marker, arguably allowing for the concomitant lack of event denotation despite the verbal nature of the root.

4. Stricto sensu results: a heterogeneous class Though restricted to the subset of stricto sensu result nouns, the present use of result groups together ontologically and linguistically elusive categories: results or effects of an event can be objects, either abstract or concrete, but they can be states too. Hence, we find temporal, concrete and abstract entities clustered in the same class. Consider the following examples: (3)

Economic sanctions and diplomatic isolation must be maintained and intensified.

(4)

What has intrigued scholars about Da Vinci’s inventions is that when assembled, many do not work.

(5)

13,000 becquerels per cubic meter is a measurement close to the authorized maximum level.

In (3), isolation refers to the result state of the corresponding dynamic event; in (4), inventions unambiguously identifies material objects, while in (5) measurement refers to an abstract (numerical) value. Some of these nouns can actually have both senses: obstruction, for example, beyond the expected event denotation, can stand for the result state and the resulting material object, while translation primarily refers to an abstract object and, by metonymic extension, to the concrete object containing it (e.g., a faulty translation / a twelve page translation). This semantic variation has non-trivial linguistic correlates. When they denote states (isolation, pollution), result nouns typically are mass; when they denote objects, either physical or abstract, they typically are count nouns. However, in the latter case also,

1262

VII. Semantics and pragmatics in word-formation II: Special cases

they can be mass (stuffing) and collectives too (collection), depending on the nature of the result associated with a certain event type. Therefore, no unifying feature can be found at this level. As noted in section 2.1, result nouns do not correspond to a fixed position in the argument structure of the base verb. As observed by Grimshaw (1990), result nominals can refer to the verb internal argument, as in the following Italian nouns: costruzione ‘construction’ ← costruire ‘to build’, ritrovamento ‘find’ ← ritrovare ‘to find’, expressing the built and found objects respectively. But this is not the rule, as one might easily notice considering cases like traduzione ‘translation’ or cambiamento ‘change’, which refer to a ‘new’ version of the object and to a modified element in the preexisting object, respectively. Although these are not argument structure participants, they could nonetheless be interpreted as referring to non-projected semantic or implicit participants, arguably playing a role at a presyntactic level of representation (e.g., lexical conceptual structure, see Jackendoff 1990). This explanation however cannot account for all the cases, as with the emblematic case of It. segatura ‘sawdust’ ← segare ‘to saw’, whose result meaning is the effect of a conceptual shift potentially applicable to many event types, and not clearly predictable from verb semantics. Other challenging cases are those of nominals referring to the means by which the event is accomplished: It. imbottitura ‘stuffing/padding’ ← imbottire ‘to pad’, G. Abdeckung ‘coverage’ ← abdecken ‘to cover up’. These nouns refer to the means employed to accomplish the event and, at the same time, to the resulting entity. To conclude on this, result states, by definition, cannot refer to argument-structure or thematic participants; they are temporal objects which express a subevent in the complex event structure of the base verb (i.e. pollution ← to pollute, a causative verb). Related to this interpretive variation, result nouns manifest quite different argumenttaking properties. When they denote result states, they project syntactic satellites retaining their original thematic interpretation. Specifically, they tend to be “intransitive” nouns as they optionally express the experiencer by means of genitival expressions variously codified across languages. However, with psychological states (see frustration in (6) from the causative psychological verb frustrate), the target of emotion, as well, can be projected as an optional syntactic satellite. (6)

Rachel’s frustration with people who don’t appreciate her

Entity-denoting nouns manifest a less consistent behavior: they either block the projection of the internal argument, when they refer to the verbal theme (7); or they optionally retain full argument structure with thematic interpretation (8). More often, however, a result noun can take modifiers amenable to various interpretations and, especially in the case of material objects, the original verb-internal argument might be expressed as the “place” onto/into which the new object is located (9): (7)

The construction (*of the bridge) collapsed unexpectedly.

(8)

John’s discussion (of the data) was published in the journal.

(9)

When he smiled, I could see the fillings in his teeth.

All in all, this inconsistent morphosyntactic and interpretive characterization points to the impossibility of analyzing result nouns along the same lines as agent or patient

71. Result nouns

1263

nouns, which have often been understood as operations on the argument structure of the base verb (see article 74 on agent and instrument nouns). More specifically, in the case of result nouns, the chance of binding or, in more neutral terms, denoting different argument-structure or lexical conceptual structure participants or even adding a participant to the thematic grid of the base verb (see It. segatura ‘sawdust’) is responsible for the varied behavior manifested at the morphological, syntactic and semantic levels, as illustrated in this section.

5. Semantic properties of the base verb As recently observed by Harley (2009: 322) for English nominalizations, “one type of meaning shift − from event to result readings − seems to be quite productive and predictable, and hence hardly idiomatic”. This standpoint is actually not new in the literature on result nouns and has led the spirit of much research on the topic since the last century. Specifically, the observation that the semantics of the base verb plays a crucial role in the formation of a result nominal dates back to Gamillscheg (1921) and was later developed by Ermecke (1929) and Baldinger (1950) (the reader is referred to Rainer 1996 for a detailed historical perspective; for recent approaches along this line of analysis, see among others Ehrich and Rapp 2000 on German, and Melloni 2007, 2011 on Italian). However, the debate on this issue has not always made it clear which semantic features of the base verb are decisive for deriving a noun with result reading. Intuitively, to produce a result, a verb should be dynamic and telic, that is to say, able to produce a result state and/or an associated result object. Hence, it would be reasonable to ask whether generalizations on the availability of result meanings can be expressed in terms of traditional aktionsart classes.

5.1. Aktionsart types On the basis of the intuition that a result is the causal byproduct of a dynamic and telic event, we predict that states and processes should be unavailable as bases of stricto sensu result nouns. This expectation is at least partially confirmed: states, i.e. non-dynamic events, cannot derive result nouns (interestingly, nominals with stative base verbs can have other concrete readings: It. giacenza ‘unsold goods’ ← giacere ‘to lay’). However, a more careful look at dynamic verbs in general shows that generalizations expressed on the other aktionsart classes are neither sufficient nor necessary for determining a result interpretation of the deverbal noun. On the one hand, while it seems cross-linguistically confirmed that accomplishment and achievement verbs are the best candidates for forming result nouns, activity/process nominals can convey a result reading too: It. lavatura ‘dirty water’ (especially, dishwater) ← lavare ‘to wash’, frustata ‘trace/sign of a whip’ ← frustare ‘to hit with a whip’. In these cases, the existence of a result/effect is rooted in the conceptual knowledge associated with the event type and is thus independent of the aspectual characterization of the verb, which, under nominalization, comes to be conceived of as a marginal type of creation verb. On the other hand, many accomplishments and achievements lack a result object reading altogether, see destruction ← destroy or It. annichilimento ‘annihilation’ ← anni-

1264

VII. Semantics and pragmatics in word-formation II: Special cases

chilire ‘to annihilate’, whose base verbs describe actions directed towards the annulment of a pre-existing object. Other examples are G. Entleerung ‘drainage’ ← entleeren ‘to empty’ or Erreichung ‘attainment’ ← erreichen ‘to attain’. It thus seems that conceptual, rather than aspectual, features can be held responsible for the lack of a result associated with these events (see discussion in section 5.2). Interestingly, most accomplishment and achievement nominals lack a result-state reading too. This is the case for creation nominals, for which a corresponding resultstate interpretation is systematically lacking (e.g., construction or creation cannot refer to the state of being constructed or created respectively, see Ježek and Melloni 2011). Further, many nominals derived from verbs describing a change of state cannot refer to the result state, see It. asciugatura ‘drying’ ← asciugare ‘to dry’. As observed by Osswald (2005), these restrictions might have to do with the distinction between resultant states and target states proposed by Parsons (1990, ch. 12), in that only target states − i.e. defined as independently identifiable states − are candidates for the result-state interpretation of a deverbal nominal. In particular, verbs for which a target state passive is available can (but need not!) create nominals with result-state reading. On the contrary, nominals obtained from verbs for which the target state is missing are unable to form result-state nominals. To uncover the type of result state, Kratzer (2000) employs the Zustandspassiv test with modification by immer noch ‘still’ in German. The verb obstruct positively answers to the test: Die Ausfahrt ist immer noch versperrt ‘The driveway is still obstructed’. The verb empty, on the contrary, does not: Der Briefkasten ist (*immer noch) geleert ‘The letterbox is (*still) emptied’, indicating that leeren implies a resultant state of its object, rather than a target state (see Kratzer 2000: 385−386). We refer the reader to the discussion in Osswald (2005) for further details.

5.2. Types of creation verbs The data exemplified so far suggest that we should pay attention to the conceptual semantic features encoded in the verb root, i.e. the more idiosyncratic building blocks of semantic meaning. Specifically, the best candidates for result-object nouns are verbs that describe the coming into existence of an entity. In this domain, however, a first distinction should be drawn between verbs of explicit creation and implicit creation. Melloni (2007, 2011) is based on this dichotomy and argues that − beyond creation verbs projecting the effected entity onto the object position, i.e. “result-object verbs” − there are verbs implying the creation of an element which remains syntactically unexpressed. Verbs such as build, coin, compose, create, form, invent, produce and the like belong to the former group (see also image creation verbs, Levin 1993: 169−172). Nominals derived from result-object verbs denote either the event or the abstract/concrete entity resulting from it. In the latter case, the nominal identifies the internal argument of the base verb. Examples of this class are available in all the languages at issue: E. creation, G. Produktion ‘production’, Sw. byggning ‘(large) building’, It. incisione ‘engraving/ recording’, Ru. stroenie ‘building’, etc. Many other verbs, however, are likely to yield corresponding nominals with event and result meanings, see in particular verbs of assembling and combining, speech act and mental action verbs (and in general verbs with predicative complements), verbs of sound and light emission, verbs of appearance, verbs of cooking (see Levin 1993), etc.

71. Result nouns

1265

Turning to verbs of implicit creation, these are of two types: “creation through representation” and “creation through modification” verbs. Representation verbs lexicalize situations expressing the coming into being of an entity which is a representation of the source argument, typically mapped onto the direct object position. Verbs in this class are copy, falsify, imitate, represent, translate, transcribe, and so forth (see also performance verbs, Levin 1993: 178). While these verbs project internal arguments corresponding to what Dowty (1991) defines as “representation source themes”, their derived nominals lexicalize the representation of the source theme. Let us consider the emblematic case of the resultobject noun translation: it is worth emphasizing that this noun lexicalizes a non-syntactic participant, the translated text, which could suitably be defined as a semantic, implicit or “root” argument. As noticed in section 4, this state of affairs is directly reflected in the valency properties of these nominals, which optionally retain the verb-internal argument, as in a flawless translation (of the source text). Other prototypical instances of nominals pertaining to this class are the following: E. representation, G. Übertragung ‘transcription’, Sw. sammanfattning ‘summary’, It. imitazione ‘imitation’, Ru. perevod ‘translation’. Finally, modification verbs describe a situation bringing about a modification to the referent of the patient (syntactically, an internal argument); this modification is conceived of or conceptualized as an effected entity or result. Verbs in this class lexicalize events that bring about a modification in/onto an existing object, by addition or subtraction of material, breaking or fracturing of the referent of the theme argument, etc. The class is thus highly heterogeneous from a semantic viewpoint and mainly describes concrete actions: alter, amend, break, cover, cut, enlarge, extend, lacerate, modify, etc. The corresponding nominals lexicalize the modification itself (as the event and the result) and accordingly tend to denote concrete referents: E. extension, G. Beschädigung ‘damage’, Sw. beriktigande ‘correction’, It. lacerazione ‘laceration, wound’, Ru. izmenenie ‘change’. It is important to emphasize that this and other attempts to classify verbs on similar semantic bases can only be used as directing factors, rather than strong determinants of the result meaning for the derived nominals. In fact, a complex interplay of interpretive, morphological and lexical factors (such as numerous instances of semantic blocking, not discussed here) should be held responsible for the (un)availability of a result interpretation in deverbal nominals.

6. References Alexiadou, Artemis 2001 Functional Structure in Nominals. Amsterdam/Philadelphia: Benjamins. Alexiadou, Artemis 2009 On the role of syntactic locality in morphological processes: The case of (Greek) derived nominals. In: Anastasia Giannakidou and Monika Rathert (eds.), Quantification, Definiteness and Nominalization, 253−280. Oxford: Oxford University Press. Anderson, Mona 1983 Prenominal genitive NPs. The Linguistic Review 3: 1−24. Baldinger, Kurt 1950 Kollektivsuffixe und Kollektivbegriff. Ein Beitrag zur Bedeutungslehre in Französischen mit Berücksichtigung der Mundarten. Berlin: Akademie Verlag. Beard, Robert 1995 Lexeme-Morpheme Base Morphology. A general theory of inflection and word formation. New York: State University New York Press.

1266

VII. Semantics and pragmatics in word-formation II: Special cases

Bierwisch, Manfred 1989 Event nominalizations: Proposals and problems. Linguistische Studien. Reihe A: Arbeitsberichte 194: 1−73. Booij, Geert and Rochelle Lieber 2004 On the paradigmatic nature of affixal semantics in English and Dutch. Linguistics 42(2): 327−357. Borer, Hagit 2003 Exo-skeletal vs. endo-skeletal explanations: Syntactic projections and the lexicon. In: John Moore and Maria Polinsky (eds.), The Nature of Explanations in Linguistic Theory, 31−67. Chicago: CSLI and University of Chicago Press. Brandtner, Regine 2011 Deverbal nominals in context. Meaning variation and copredication. Ph.D. dissertation, Universität Stuttgart. Chomsky, Noam 1970 Remarks on nominalization. In: Roderick Jacobs and Peter Rosenbaum (eds.), Readings in English Transformational Grammar, 184−221. Waltham: Ginn. Dowty, David 1991 Thematic proto-roles and argument selection. Language 67: 574−619. Ehrich, Veronika and Irene Rapp 2000 Sortale Bedeutung und Argumentstruktur: -ung Nominalisierungen im Deutschen. Zeitschrift für Sprachwissenschaft 19(2): 245−303. Ermecke, Gustav 1929 Das Wesen der sprachlichen Abstrakta und ihre Bildung durch Suffixe im Romanischen, nebst einem Hinweis auf den Einfluß dieser Art Suffixbildung auf das Englische und Deutsche. Langendreer: Pöppinghaus. Gaeta, Livio 2004 Nomi di azione. In: Maria Grossmann and Franz Rainer (eds.), La formazione delle parole in italiano, 314−351. Tübingen: Niemeyer. Gamillscheg, Ernst 1921 Grundzüge der galloromanischen Wortbildung. In: Ernst Gamillscheg and Leo Spitzer (eds.), Beiträge zur romanischen Wortbildungslehre (= Biblioteca dell’Archivum Romanicum 2), 1−80. Genève: Olschki. Grimshaw, Jane 1990 Argument Structure. Cambridge, MA: MIT Press. Harley, Heidi 2009 The morphology of nominalizations and the syntax of vP. In: Monika Rathert and Anastasia Giannankidou (eds.), Quantification, Definiteness and Nominalization, 320−342. Oxford: Oxford University Press. Jackendoff, Ray 1990 Semantic Structures. Cambridge, MA: MIT Press. Ježek, Elisabetta and Chiara Melloni 2011 Nominals, polysemy and co-predication. Journal of Cognitive Science 12: 1−31. Kratzer, Angelika 2000 Building statives. In: Lisa J. Conathan, Jeff Good, Darya Kavitskaya, Alyssa B. Wulf and Alan C. Yu (eds.), Proceedings of the 26 th Annual Meeting of the Berkeley Linguistic Society, 385−399. Berkeley: Berkeley Linguistic Society. Lebeaux, David 1986 The interpretation of derived nominals. In: Anne M. Farley, Peter T. Farley and KarlEric McCullough (eds.), Parasession on Pragmatics and Grammatical Theory at the 22nd Regional Meeting of the Chicago Linguistic Society, 231-247. Chicago: Chicago Linguistic Society.

71. Result nouns

1267

Levin, Beth 1993 English Verb Classes and Alternations. A preliminary investigation. Cambridge: Cambridge University Press. Lieber, Rochelle 2004 Morphology and Lexical Semantics. Cambridge: Cambridge University Press. Malicka-Kleparska, Anna 1988 Rules and Lexicalisations. Selected English nominals. Lublin: Redakcja Wydawnictw Katolickiego Uniwersytetu Lubelskiego. Melloni, Chiara 2006 Logical polysemy in word formation: E and R suffixes. Lingue e Linguaggio 5(2): 281− 308. Melloni, Chiara 2007 Logical polysemy in word formation. The case of deverbal nominals. Ph.D. dissertation, University of Verona. Melloni, Chiara 2010 Action nominals inside: Lexical semantic issues. In: Artemis Alexiadou and Monika Rathert (eds.), The Semantics of Nominalizations across Languages and Frameworks, 139−166. Berlin/New York: Mouton de Gruyter. Melloni, Chiara 2011 Event and Result Nominals. A Morpho-semantic Approach. Bern: Lang. Motsch, Wolfgang 1999 Deutsche Wortbildung in Grundzügen. Berlin/New York: Mouton de Gruyter. Newmeyer, Frederick J. 2009 Current challenges to the lexicalist hypothesis: An overview and a critique. In: William D. Lewis, Simin Karimi, Heidi Harley and Scott O. Farrar (eds.), Time and Again. Theoretical perspectives on formal linguistics. In honor of D. Terence Langendoen, 91− 117. Amsterdam/Philadelphia: Benjamins. Nunberg, Geoffrey 1995 Transfers of meaning. Journal of Semantics 12(2): 109−132. Osswald, Rainer 2005 On result nominalization in German. In: Emar Maier, Corien Bary and Janneke Huitink (eds.), Proceedings of Sinn und Bedeutung 9, 256−270. Nijmegen: Nijmegen Centre for Semantics. Parsons, Terence 1990 Events in the Semantics of English. Cambridge, MA: MIT Press. Pustejovsky, James 1995 The Generative Lexicon. Cambridge, MA: MIT Press. Rainer, Franz 1996 La polysémie des noms abstraits: Historique et état de la question. In: Nelly Flaux, Michel Glatigny and Didier Samain (eds.), Les noms abstraits. Histoire et théories, 117− 126. Villeneuve d’Ascq: Presses Universitaires du Septentrion. Roeper, Thomas 1987 Implicit arguments and the head-complement relation. Linguistic Inquiry 18(2): 267− 310. Roßdeutscher, Antje and Hans Kamp 2010 Syntactic and semantic constraints in the formation and interpretation of ung-nouns. In: Artemis Alexiadou and Monika Rathert (eds.), Nominalizations across Languages and Frameworks, 169−214. Berlin/New York: Mouton de Gruyter. Sadler, Luisa, Andrew J. Spencer and Marina Zaretskaya 1997 A morphomic account of a syncretism in Russian deverbal nominalization. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1997, 181−215. Dordrecht: Kluwer.

1268

VII. Semantics and pragmatics in word-formation II: Special cases

Schoorlemmer, Maaike 1995 Participial Passive and Aspect in Russian. Ph.D. dissertation, Utrecht University. Tenev, Ivan 2008 Deverbal nominalization in Swedish and Norwegian: Nomina actionis and nomina acti. Constrastive Linguistics 33(1): 5−24. Valdivia, Glòria de, Joan Castellví and Mariona Taulé 2013 Morphological and lexical aspect in Russian deverbal nominalizations. In: Irina Kor Chahine (ed.), Current Studies in Slavic Linguistics, 267−280. Amsterdam/Philadelphia: Benjamins. Zubizarreta, Maria Luisa 1987 Levels of Representations in the Lexicon and in the Syntax. Dordrecht: Foris.

Chiara Melloni, Verona (Italy)

72. Quality nouns 1. 2. 3. 4. 5. 6. 7. 8. 9.

Terminological preliminaries Problems of delimitation Quality nouns cross-linguistically The semantics of quality nouns Semantic extensions The syntactic functions of quality nouns Affix rivalry and restrictions The origin of quality-noun suffixes References

Abstract Many languages have dedicated derivational means for turning predicative adjectives into quality nouns, among them almost all European languages. After discussing terminological issues and delimiting quality nouns from the neighboring categories of status and action/result nouns, the present article provides an overview of the most important issues concerning the semantics and functions of this derivational category. In the last two sections, the intense affix rivalry among quality-noun suffixes and their origin are addressed.

1. Terminological preliminaries The derivational category referred to here as quality nouns has received a variety of names in the course of time, most of which are still in current use. The term quality noun itself is a loan translation of Latin nomen qualitatis, which reaches back to the Middle Ages. Its original meaning was simply ‘name of a quality’, the term therefore

72. Quality nouns

1269

could also refer to an adjective. In modern linguistic terminology, however, it is reserved for nouns denoting a quality, generally both derived (e.g., ugliness ← ugly) and underived (e.g., beauty). (The use of nomen qualitatis as a name for personal nouns such as softie or drunkard, as in chapter 15 of Baeskow 2002, is quite idiosyncratic.) Instead of quality noun, we also find property noun, and similarly nom de propriété besides nom de qualité in French. A synonym of nomen qualitatis is nomen essendi, which nowadays is rarely used outside Slavic, especially Polish and Slovak, philology. A second terminological tradition is based on the classification of quality nouns as a subclass of abstract nouns. Ultimately, this tradition can also be traced back to medieval philosophy, where quality nouns were considered to be abstracta, i.e. a kind of universals (in the philosophical sense of the word), in opposition to the corresponding adjectives, which were considered to be concreta (cf. Gruppe 1954 [1831]: 154). Abstract and abstraction are vague concepts that have been applied to many other types of nouns beyond those mentioned here (cf. Heinimann 1963, Mikkola 1964, Flaux, Glatigny and Samain 1996). Trost (1976: 238) therefore proposed to reserve the term Abstrakta for derivatives only. In order to distinguish, in this latter category, quality nouns from action nouns and status nouns, deadjectival abstract nouns are opposed to deverbal and denominal abstract nouns. This terminology seems to go back to the German triptych Adjektivabstraktum, Verbalabstraktum, Substantivabstraktum. Adjektivabstraktum, for example, can already be found in Becker (1824: 61). In line with the Latin tradition of subsuming nouns and adjectives under a single category of nomina, Adjektivabstrakta and Substantivabstrakta are sometimes conflated in one category, Nominalabstrakta. Instead of Adjektivabstraktum, we also find Eigenschaftsabstraktum ‘quality abstract’ in German (at least since Brugmann 1906: 641 ff.). The third terminological tradition is of a more recent vintage, going back to the early days of generative grammar. In this tradition, the process of forming a noun (or nominal) is called nominalization. When this term is used as a result noun, deadjectival nominalization (or property nominalization) is a synonym of quality noun. For recent research on nominalizations in the generative tradition, cf. Alexiadou and Rathert (2010), Roy and Soare (2011), Spencer (2013, ch. 8), article 60 on word-formation and argument structure, section 8, and article 61 on word-formation and metonymy, section 5.

2. Problems of delimitation 2.1. Quality nouns vs. status nouns vs. action/result nouns As we have seen in the preceding section, the category of quality nouns is relatively well established in morphological terminology, and there is a consensus about the fact that its nucleus is constituted by formations such as ugliness, stupidity, accuracy, etc. The boundaries of the category, however, are not sharply delimited, especially towards the related abstract categories of action/result nouns and status nouns. The close relationship between q u a l i t y n o u n s and s t a t u s n o u n s is reflected by the fact that many suffixes can be found in both categories (cf. also article 73 on status nouns). Here are first some illustrative examples taken from the monographic articles of the handbook: Hungarian lusta-ság ‘laziness’ (← lusta ‘lazy’) vs. menedzserség ‘manager’s profession’, Budugh çetin-uval ‘difficulty’ (← çetin ‘difficult’) vs. de-

1270

VII. Semantics and pragmatics in word-formation II: Special cases

mirçi-tuval ‘occupation of a blacksmith’ (← demirçi ‘blacksmith’), Aghul reʜet-ʕ˳el ‘easiness’ (← reʜet ‘easy’) vs. pːačːah-ʕ˳el ‘kinghood’ (← pːačːah ‘king’), etc. In Latin, denominal Deitas (← Deus ‘God’), a calque of Greek θεότης, has been attested alongside deadjectival divinitas (← divinus ‘divine’) since St Augustine, sparking off a range of similar formations in philosophical jargon (e.g., hominitas ← homo, -inis ‘man’). In English, the status-noun suffixes -dom and -hood also occasionally occur with adjectival bases: artistdom vs. freedom, priesthood vs. falsehood, etc. However, as pointed out in Bauer, Lieber and Plag (2013: 261), overall the two categories can be distinguished quite neatly in semantic terms. When -ness, for example, is applied to nominal bases as in babeness, “it highlights or picks out the significant characteristics that make that entity what it is”, while in babedom or babehood the suffix “denotes a status rather than a set of qualities”. A different kind of overlap between deadjectival and denominal abstract nouns can be found in several Romance languages (cf. Dal and Namer 2010 for French). Quality nouns referring to local identity can be derived both from ethnic adjectives (e.g., Spanish africanidad ‘Africanness’ ← africano ‘African’) and from the geographical names themselves (e.g., portugalidad ‘Portugueseness’ ← Portugal, instead of portugués ‘Portuguese’). The reason for this is purely formal. The country name is chosen just in case this choice provides a final sequence more in line with the phonological restrictions imposed by the suffix: the final sequence -alidad is well attested with -idad, while this is not the case for -esidad. The distinction is equally blurred between q u a l i t y n o u n s and a c t i o n n o u n s . Pinault (1996: 202) has pointed out that throughout the history of Indo-European, suffixes have migrated in both directions between these two categories. Two interfaces between action and quality nouns deserve to be distinguished here. On the one hand, we have to keep in mind that some languages of the world have no or only very few adjectives. In such languages, concepts corresponding to adjectives in the more common European languages are mostly expressed by stative verbs, which in turn can be nominalized like other verbs, yielding “action” nouns very similar semantically to quality nouns. In such languages, a concept such as ‘to be lazy’ is expressed by a verb comparable to English to laze, and ‘laziness’ therefore corresponds to a deverbal abstract, again vaguely comparable to the English noun laze (cf. a quiet laze). An English suffix apt to illustrate the continuum between quality and action nouns with stative verbs is -ance, which allows both a deverbal and a deadjectival analysis when joined to stative verbs: abundance, for example, can refer either to the state of ‘abounding’ or of ‘being abundant’, dominance to the fact of ‘dominating’ or of ‘being dominant’. The reason for this ambiguity is that the Latin suffix that is at the base of -ance and its Romance cognates (cf. article 68 on action nouns in Romance, section 4) was the result of attaching the abstract suffix -ia to the present participle in -ns/-ntis (cf. abundare ‘to abound’ → abundans ‘abounding, abundant’ → abundantia ‘abundance’; Daude 2002: 237−241). The second area where quality and action nouns, or more precisely result nouns (cf. article 71), overlap is that of resultant states. In Kalmyk, a Mongolian language spoken in Russia (cf. article 191, section 4.1.3), the word for ‘blueness’, kôkra˝n, is derived from the verb kôkr(h) ‘to become blue’ by means of a suffix deriving result nouns. The equivalent of ‘blueness’ in that language therefore literally means something like ‘state resulting from having become blue’. In many European languages result nouns formally derived from the present stem, not from the past participle, serve as quality-noun substi-

72. Quality nouns

1271

tutes for past participles expressing resultant states. Spanish aislado ‘isolated’, for example, is nominalized by means of the result noun aislamiento (← aislar ‘to isolate’), while *aisadez, formed with the quality-noun suffix -ez, is decidedly odd. Malkiel (1946: 32− 35) has already pointed out that the exact nature of the trade-off between quality nouns and result nouns with past participles differs from language to language. In German or English, for example, quality-noun suffixes are more easily combined with past participles expressing resultant states than in the Romance languages: the quality nouns Isoliertheit and isolatedness, for example, are possible options besides Isolation/Isolierung and isolation. For a recent intent to pin down the semantic conditions for obtaining a state reading in French, cf. Fradin (2011).

2.2. Types of quality nouns Even clearly deadjectival abstract nouns do not form a completely homogeneous derivational category, a fact that has received little attention in the literature (but cf. Daude 2002: 288−304 on Latin). First there is the question of whether a distinction should be drawn between quality nouns in the literal sense of the word quality and state nouns, as suggested, among others, by Trost (1976: 225): “Bei den Adjektivabstrakta kann man im Russischen und Deutschen Eigenschafts- (vgl. dobrota ‘Güte’) und Zustandsbezeichnungen (vgl. zdorov’je ‘Gesundheit’) unterscheiden.” [Among Russian and German deadjectival abstracts we can distinguish designations of qualities (cf. dobrota ‘goodness’) and designations of states (cf. zdorov’je ‘health’).] I am not aware, however, of any suffix that would be rigorously limited to either permanent or episodic qualities, which is why it seems preferable to keep quality noun as a cover term for both kinds of situation. English -ness, for example, can take as bases both adjectives expressing permanent qualities such as clever, complicated, etc. and adjectives expressing states such as drunken, isolated, etc. On the other hand, it is true that at least some suffixes forming quality nouns seem to shun bases expressing states. Such is the case for Italian -ezza (cf. Rainer 1989: 183−193), which can only be attached productively to participles denoting permanent qualities (e.g., controllatezza ← controllato ‘self-controlled’), but not to those denoting states (e.g., *insonnolitezza ← insonnolito ‘drowsy’). In the case of ordinary adjectives, however, some lexicalized formations based on adjectives expressing states can also be found (e.g., tristezza ← triste ‘sad’). As far as English is concerned, there has been some discussion about whether the suffixes -ness and -ity are really synonymous. Riddle (1985) has made a case for a systematic difference on the basis of contrasts such as hyperactivity (a diagnosable condition) vs. hyperactiveness (an episodic property), but in their critical appraisal Bauer, Lieber and Plag (2013) prefer to attribute occasional contrasts of this kind “to the greater propensity of forms in -ity to be high frequency established forms and to have lexicalized meanings” (p. 258). In many cases, both suffixes can be used interchangeably. For another attempt to differentiate the English quality-noun suffixes semantically, cf. Hamawind (2008, ch. 3). The necessity of an even more fine-grained typology has been argued for recently by Martin (2013) in her study of French “dispositional nouns”:

1272

VII. Semantics and pragmatics in word-formation II: Special cases To summarize, D[ispositional] N[oun]s differ from each other by the range of readings they can have. Permanent readings: All can denote a disposition, but only a subset can also denote habits. Transient readings: Some denote events (indiscrétion), others denote action dependent states (discrétion, inconstance), those like despotisme can denote temporary dispositions.

Though no French suffix is confined to one of these readings, Martin nevertheless is able to show that statistical preferences do exist, for example, of -erie for the eventive reading or of -isme for the dispositional reading. A second factor that contributes to the internal heterogeneity of the category of quality nouns is constituted by the fact that quality nouns are sometimes enriched with additional shades of meaning that set them apart from the pure prototype. The Italian suffix -aggine, for example, only expresses negative qualities (cf. Rainer 1989: 232−251). The same, according to article 182, section 4.1.2, is true of Basque -keria as opposed to neutral -tasun (cf. handitasun ‘greatness, enormity’ vs. handikeria ‘arrogance, excessive grandeur’ ← handi ‘big’). A somewhat different case is represented by French -itude, which is a pure quality-noun suffix in formations like exactitude ‘exactness’, but has acquired implications of oppression or aspiration towards freedom or group identity in formations like corsitude ‘Corsicanness’, féminitude ‘womynhood’, etc. This second “identity” series arose in analogy to négritude (lit. ‘negroness’), a noun itself forged after servitude ‘id.’ and originally referring to a literary movement focused on black identity (cf. Bourquin 1979; Koehl 2012; the conjecture in Martin 2013: 175 that this specific shade of meaning is the result of the “irradiation” of the meaning of the base of habitude ‘habit’ and attitude ‘id.’ is unable to explain why this semantic extension occurred just after the diffusion of négritude and not before). A third special case is constituted by formations in -ismus such as Latin strabismus ‘id.’ (← strabus ‘squint-eyed’). The noun here refers to a disease which consists in being squint-eyed, to “squint-eyedness”, as it were. Such terms therefore come close to quality nouns, but the suffix -ismus nevertheless in the first place signals membership in the referential category of diseases (cf. Roché 2007: 48). A last kind of word that is sometimes assimilated to quality nouns are color-referring terms: the green of the apple is thought to be equivalent to the greenness of the apple. Moltmann (2013: 222−224), however, has shown that their behavior is different in important respects. She argues that in Green is nice the color as a whole is evaluated, in Greenness is nice, on the contrary, its instances (cf. also article 61 on word-formation and metonymy, section 5). Though this seems to be true for English, it cannot be denied that in other languages at least some conversions of adjectives can be used as substitutes for quality nouns, for example, French (le) sérieux lit. ‘the serious’, which is in all respects equivalent to a quality noun in uses such as le sérieux de l’affaire ‘the seriousness of the affair’ (note that sérieux has no suffixal quality noun at its side). But even in French, this use is lexically restricted (cf. Malkiel 1938: 67): one cannot say, for example, *le beau de la fille ‘the beauty of the girl’, instead of la beauté de la fille. The semantics of this kind of conversion and its relationship to quality nouns therefore has to be described at the level of the individual language.

3. Quality nouns cross-linguistically In his sample of 42 genetically unrelated languages, Bauer (2002: 40) has found that 14 have a category of “abstract nouns” based on adjectives, occupying the fourth rank

72. Quality nouns

1273

among nominal derivational categories behind deverbal abstract nouns (32), personal or agentive nouns (24) and diminutives (15), but before denominal abstract nouns (12) and instrument nouns (12). In the European languages represented in this handbook all seem to possess this derivational category, except Nenets, a Samoyedic language (cf. article 175). This shows that quality nouns are relatively common cross-linguistically, but at the same time it is also clear that languages can do perfectly without them. In fact, even in those European languages that have quality nouns, these have syntactic equivalents, as the following examples show: (1)

John has wisdom ≈ John is wise

(2)

No one doubts John’s sincerity ≈ No one doubts that John is sincere

(3)

Wisdom is better than strength ≈ Being wise is better than being strong

(4)

They love him for his wisdom ≈ They love him for his wise behavior

(5)

John’s helpfulness makes him popular ≈ John’s helpful nature makes him popular

(6)

Honesty is a virtue ≈ The property of being honest is a virtue

These syntactic constructions are fully, or at least roughly, equivalent semantically to the corresponding abstract nouns. The question of how to account for these equivalences is still debated; cf., for English, Reichl (1982), Spencer (2013), Moltmann (2013), among others. It goes without saying that the structure and use of alternative means for expressing the content of quality nouns are highly language-specific. I would just like to point out here that in German the equivalent of English being + adjective, which undoubtedly has the status of a noun phrase in uses such as the one in (3), is a construct with the infinitive sein ‘to be, being’ as second member. At the moment I am writing these lines, de Gruyter has just announced a book about German identity by Maja Figge with the somewhat clumsy title Deutschsein (wieder-)herstellen. Weißsein und Männlichkeit im bundesdeutschen Kino der fünfziger Jahre. ‘Germanness’, in this title, is rendered as Deutschsein, lit. ‘German-to be’, and ‘whiteness’ as Weißsein, lit. ‘white-to be’. Such German formations are sometimes dubbed “compounds”, but probably an analysis as cases of conversion of a predicate consisting of an adjective followed by the infinitive sein would be more adequate. One way of forming quality nouns deserves special mention here because it looks so outlandish from a European perspective. In some languages of East Asia, quality nouns are formed by compounding, viz. by joining the antonymous adjectives (or nouns) of a particular scale: Tibetan srab-mthug ‘density; lit. thin-thick’, Khalkha xaluun-xüjten ‘temperature; lit. heat-cold’, Tokharian tsopats mkältö ‘size; lit. big-small’ (cf. Wälchli 2005: 152−154). The relatively high cross-linguistic frequency of quality nouns can be interpreted as an argument against the view that the use of abstract nouns is necessarily bound to higher levels of civilization. Herder, in his Abhandlung über den Ursprung der Sprachen (1772), has argued that the language of every people has recourse to contain abstractions − in a more general sense than that of quality nouns −, since these are essential for reasoning, and that every people is able to create abstract nouns, but at the same

1274

VII. Semantics and pragmatics in word-formation II: Special cases

time he thought that the number of abstract terms depended on the needs of the people, establishing a correlation with the level of cultural development (cf. Herder 1953, vol. I: 785−788). This idea has still been endorsed during the 20th century by linguists such as Ermecke (1929), Szadrowsky (1933: 97−98), or Lazar (1975). The latter, for example, says about the Uralic language family: It can be stated as a general rule valid for all languages treated that the occurrence of abstract nouns or the presence of abstractive formation is to a great degree dependent on the level of development attained by the society in which the language in question is spoken. The closer to nature a people lives the more concrete its conceptual world and the linguistic reflection of this world. (Lazar 1975: 308)

In the case of Uralic languages, Lazar has observed that the Samoyedic languages had the smallest number of abstract nouns, while “the further to the west (towards Western Europe) a people lives, the more common the abstracts in its language” (Lazar 1975: 308). A further caveat against a certain interpretation of Herder’s hypothesis is constituted by the fact that already the Indo-European protolanguage had half a dozen of qualitynoun suffixes according to Pinault (1996: 201−202).

4. The semantics of quality nouns The semantics of quality nouns became early on the focus of philosophical controversies, most famously during the dispute over universals in the Middle Ages. In linguistics, one popular semantic characterization of quality nouns claims that they present qualities as things, substances, entities, or the like. The exact wording differs from author to author. Paul (1920: 363 [1880: 215]), for example, has claimed that the origin of quality nouns as well as action nouns is due to a metaphor that presents the quality or action “unter der Kategorie des Dinges” [under the category of thing]. In a similar vein, Langacker (1987: 208) still proposes to view (deverbal) abstract nouns such as love, envy, etc. as “abstract ‘substances’ analogous in many ways to physical substances”. In Goddard and Wierzbicka (2014: 205−237) the explications of the meanings of abstract nouns, couched in terms of their natural semantic metalanguage, are introduced by the “reifying word ‘something’”. From a model-theoretic perspective, Jenks, KoontzGarboden and Makasso (submitted) also reach the conclusion that the nominal encoding of property concepts “entails a semantics of substances” (abstract). Another strand of the linguistic literature claims that, from a semantic point of view, there is no difference at all between an adjective and the corresponding quality noun in its pure reading. The adjective is said to be simply “transposed” into another part of speech. This way of presenting things goes back at least to Bally (1965 [1932]: 116 f.) and has gained great popularity especially in Slavic linguistics thanks to Miloš Dokulil (cf. Dokulil 1968: 209−210). A recent defense of this position can be found in Spencer (2013), who holds that “the property nominalization is a form of the adjective lexeme, that is, a genuine transposition rather than a genuine derivational process” (p. 338). Additional semantic readings of the nominalization, such as the fact-reading, in Spencer’s view are derived from the construction in which the quality nouns occur and are not due to the process of suffixation itself.

72. Quality nouns

1275

In the philosophical literature, the identification of parts of speech and ontological categories intrinsic in the first of these positions has repeatedly been criticized since Strawson (1959) at least (for an application of Strawson’s ideas to French, cf. Riegel 1985: 75−108). A recent account of the semantics of quality nouns in this tradition is Moltmann (2013). According to this philosopher, only in a “reifying term” like the property of wisdom is a property as an abstract object, as a universal, referred to, while in other contexts such as Socrates’ wisdom, John has wisdom, Wisdom exists, etc. the quality noun is said to refer to so-called “tropes”, i.e. particularized properties. “Wisdom”, as the author puts it (p. 2), “is in fact not a term referring to a property, but rather a term plurally referring to the various instances of wisdom, namely tropes.” The reifying use is by and large restricted to a technical register, especially philosophical discussions, while in everyday usage quality nouns have a particularist semantics, i.e. refer to a particular (in the philosophical sense of the word) or to a plurality of particulars. In Socrates’ wisdom, for example, wisdom refers to the particular way in which wisdom is manifested in Socrates’ behavior. This behavior is perceivable, and hence concrete, though abstraction in a more traditional sense is also involved, as Moltmann observes (p. 3), one that “involves (psychologically speaking) attending to only one aspect of a particular and abstracting from all others”. If Socrates’ wisdom refers to his wisdom in general and not to his behavior on one specific occasion, reference to a plurality of instances of wisdom is involved. In Wisdom is rare the plurality of instances referred to is even wider, including actual and possible instances. In that respect, quality nouns resemble mass nouns like water or gold, and in fact they are treated similarly by the syntax. Moltmann has identified a number of criteria that show that property-referring and trope-referring quality nouns behave indeed differently (the acceptability judgments in the examples below are hers). Generosity exists and The property of generosity exists, for example, have different truth conditions: the first sentence is true if there is at least one instance of generosity, while the second one is true “just in case the abstract object as such exists, regardless of its instantiations” (p. 15). In other cases, the different reference of both types of usage manifests itself in differences of acceptability, as in Honesty is rare vs. ??The property of honesty is rare (p. 16), The property of honesty is complex vs. ??Honesty is complex (p. 42), or The concept of poverty is vague vs. ??Poverty is vague. Moltmann further shows that tropes are maximally specific, grounded in the real world, and in that respect differ from states and facts (p. 56 f.). This difference is said to explain the different behavior with respect to the combinability with a verb like describe: John described Mary’s beauty vs. ??John described Mary’s being beautiful/?John described the fact that Mary is beautiful. Tropes, states and facts also differ in many other respects, e.g., gradability: the extent of John’s happiness vs. ???the extent of John’s being happy/ ???the extent of the fact that John is happy. A last observation of Moltmann’s (cf. p. 76 f.) that I would like to point out concerns the existence of two kinds of quality nouns based on gradable adjectives, those based on the comparative content of the adjective (e.g., height), and those based on the positive form (e.g., tallness). This distinction accounts for contrasts such as the following: John’s height is 2 meters vs. ???John’s tallness is 2 meters; John’s height has changed vs. ??John’s tallness has changed. Both types of nominalization, however, have in common

1276

VII. Semantics and pragmatics in word-formation II: Special cases

that they do not refer to simple tropes, but “to complex entities that incorporate the way the trope is ordered with respect to tropes of the same sort” (p. 78).

5. Semantic extensions The semantic characterization of quality nouns in the preceding section only covers the prototypical uses of quality nouns in natural languages. Besides these prototypical uses, however, many quality nouns have developed secondary uses, generally on a metonymic basis. As observed by Lüdtke (1978: 69), the result of the extension is not predictable in the strict sense of the word, but nevertheless related to the semantics of the adjectival base. Here are the most common extensions of this kind that can be found in Italian (cf. Rainer 1989: 356−368): (7)

a. b. c. d. e. f. g. h. i. j. k. l.

mortalità ‘mortality, death rate’ (← mortale ‘mortal’) immensità ‘immensity, huge amount’ (← immenso ‘immense’) Antichità ‘Antiquity, Ancient times’ (← antico ‘antique’) lontananza ‘distance, area that is far away’ (← lontano ‘distant’) antichità ‘antique (object)’ (← antico ‘antique’) bellezza ‘beauty, beautiful woman’ (← bello ‘beautiful’) sporcizia ‘filth, dirt; lit. dirtiness’ (← sporco ‘dirty’) cristianità ‘Christianity, Christians as a group’ (← cristiano ‘Christian’) assurdità ‘absurdity, something absurd’ (← assurdo ‘absurd’) avversità ‘adversity, calamitous event’ (← avverso ‘adverse’) atrocità ‘atrocity, atrocious act’ (← atroce ‘atrocious’) (Vostra) Altezza ‘(Your) Highness’ (← alto ‘high’)

As the English glosses show, the Italian situation is quite representative of other European languages as well (although, according to Lüdtke 1978: 68−69, metonymic extensions are much more common in Romance languages than in English or German). The commonalities are the result of both the naturalness of some of these extensions and the massive borrowing among European languages over the centuries. We would expect the more straightforward metonymic extensions to potentially be found in all natural languages, while others are clearly culture-bound (e.g., the type Your Highness, which has Latin antecedents). But even the more natural extensions may vary in productivity from language to language due to a variety of factors. In some instances, blocking seems to be involved: if English dirtiness, contrary to Italian sporcizia, does not occur in the concrete sense ‘dirt, filth’, this could simply be a consequence of the existence of the nouns dirt and filth, which already express the same concept (dirty is derived from dirt, while sporco is a primary adjective). In other cases, the greater restrictedness of the rules of pluralization seems to hamper metonymic extensions: while in Italian, nouns like infedeltà ‘infidelity’ and disubbidienza ‘disobedience’ can easily be used in the plural in the sense of ‘acts of infidelity/disobedience’ (le sue infedeltà, disubbidienze), such a use is excluded for the German equivalents Untreue and Ungehorsam (seine *Untreuen, *Ungehorsame). That this restriction cannot be formulated on a more general, semantic level is shown by the acceptability of similar extensions with quality nouns in -heit and

72. Quality nouns

1277

-keit: seine Bosheiten ‘his malicious acts, or remarks’ (← böse ‘malicious’), Treulosigkeiten ‘acts of disloyality’ (← treulos ‘disloyal’), Unbotmäßigkeiten ‘acts of insubordination’ (← unbotmäßig ‘insubordinate’), etc. But undoubtedly many differences must be considered as conventional facts about single languages or even varieties of a language: in Swiss dialects (cf. Szadrowsky 1933: 25), for example, the use of quality nouns to refer to persons (type 7f above) is much more common than in Italian (or in the German standard language, for that matter): ist dër e n Tümmi vs. *ist der eine Dummheit ‘is this guy stupid; lit. a stupidity’. Overall, the productivity and the restrictions on the productivity of metonymic extensions, not only in the domain of quality nouns, are still neglected topics of linguistic research. One related topic that has attracted quite some attention concerns differences in pluralization between German and Romance languages (cf., among others, Zindel 1958; Wald 1990). The Serbo-Croatian suffix -ost is worth mentioning since it seems to cause different stress patterns in the prototypical, trope-referring use and in more concrete extensions (cf. the somewhat unclear treatment in Arsenijević 2011). Once semantic extensions of this kind have become established in the lexicon of a language, they can act as direct models for similar formations and take on a life of their own, constituting independent patterns of word-formation. The collective extensions of the type cristianità (7h), for example, have given rise to an autonomous collective pattern already in Latin (cf. nobilitas ‘noblity’, christianitas ‘Christendom’, humanitas ‘mankind’, etc.; article 66 on collectives, section 2.3); furthermore, they seem to have been reanalyzed as denominal. In modern Italian we find neologisms such as intellettualità ‘intelligentsia’, cattolicità ‘the Catholics (as a group)’, musulmanità ‘Muslimdom’, etc. A second example of this kind is the type mortalità (7a). Just like English mortality, it not only refers to the condition of being mortal, which is the prototypical quality-noun meaning, but also to a statistical quantity, viz. the death rate. This statistical type certainly came into being on the basis of the extent reading of quality nouns (e.g., John’s velocity surprised everyone), but constitutes by now an independent pattern that mostly operates on the basis of relational, not predicative, adjectives in Italian: natalità ‘birth rate’ (← natale ‘birth-’), nuzialità ‘marriage rate’ (← nuziale ‘wedding-’), alcolicità ‘alcoholic strength’ (← alcolico ‘alcoholic’), salinità ‘salt content’ (← salino ‘saline’), piovosità ‘rain rate’ (← piovoso ‘rainy’), etc. Similar patterns are attested for other Romance languages (cf. Koehl 2009 on French), but also for Slavic languages, e.g., Cz. nehodovost ‘accident rate’ ← nehodový, relational adjective of nehoda ‘accident’. In Russian, the “statistical” reading can be found with quality nouns derived from passive present participles of imperfective verbs: poseščaemost’ ‘number of visitors’ ← poseščaemyj ‘being visited’.

6. The syntactic functions of quality nouns As we have seen in section 3 many languages do well without a category of quality nouns. These are therefore not a necessary part of language. So why do they exist at all? A fully satisfactory answer to this question would probably have to take the form of a genetic explanation, showing how pathways of grammaticalization or semantic change

1278

VII. Semantics and pragmatics in word-formation II: Special cases

further the coming into being of quality-noun affixes. From a functional point of view, one might want to speculate about why quality-noun affixes tend to stay in a language once they have come into being. Is it mere inertia, or is their permanence due to the benefits that speakers can derive from their existence? Linguists who practice this functional kind of reasoning normally insist on the usefulness of quality nouns, and abstract nouns in general, for compressing whole predicates in cases of anaphorical reference: Quality nouns are very useful for taking up complex predicates in discourse. Their usefulness in anaphoric reference … This idea is strongly associated in German-speaking countries with the pioneering work of Walter Porzig who, in his 1942 book, characterized abstract nouns as “Namen für Satzinhalte”, i.e. names for the content of sentences. In Porzig (1930−31: 72), the function of abstract nouns had already been defined as follows: “Das echte Abstraktum stellt sich sprachlich-deskriptiv immer dar als Vergegenständlichung eines Satzinhaltes vom Prädikat aus.” [From the point of view of descriptive linguistics, the genuine abstract noun always presents itself as a reification of the content of a sentence’s predicate.] The realization of the arguments of the adjectival base is optional with quality nouns. If, however, the realization is deemed convenient in a certain discourse context, both the external and the prepositional or clausal complement of the base can be realized. The external argument appears as a genitive or as a possessive adjective, while the prepositional and clausal complements normally remain unchanged: (8)

a. b. c. d.

John John John John

is is is is

honest ≈ John’s honesty, the honesty of John, his honesty happy about his new car ≈ John’s happiness about his new car ready to leave ≈ John’s readiness to leave glad that you are here ≈ John’s gladness that you are here

However, the identity of the preposition introducing the internal argument is not a necessity: (9)

a. John is fond of chocolate vs. John’s fondness for chocolate b. John is keen on sport vs. John’s keenness for sport

Measure phrases are also realized differently: (10) John is two meters tall vs. John’s height of two meters

7. Affix rivalry and restrictions In principle, one universally applicable pattern for forming quality nouns per language should be sufficient, as is the case in Kabardian (cf. article 194, section 4.1). Supporters of Humboldt’s universal would even claim that to be the ideal situation for which languages should strive. In most European languages, however, a range of different patterns, all of them suffixal, compete for the realization of our derivational category. In English, the suffix -th only survives in a few lexicalized formations like length (← long), while both -ness and -ity are synchronically productive. Due to its origin, the latter by and

72. Quality nouns

1279

large remains confined to Latinate bases, while -ness, of Germanic origin, has extended its reach even to some Latinate base types (cf. aggressiveness besides aggressivity, ubiquitousness besides ubiquity, etc.). Adjectives in -ate, -ant and -ent prefer the suffix -cy. Just like in English, also in German (cf. Oberle 1990; Doerfert 1994) the synchronic rivalry between suffixes reflects different diachronic layers of formations. First the old suffix -ida was replaced by -î, which then in turn was replaced by -heit. -ida (cf. Middle High German dünnede ← dünn ‘thin’) is by now extinct as a quality-noun suffix, while -î survives in a large number of formations such as Treue (← treu ‘loyal’), Röte (← rot ‘red’), etc. and has even remained productive in Alamannic dialects (cf. Szadrowsky 1933; Hausser 1959). Such formations normally block the attachment of the highly productive suffix -heit, but not necessarily (cf. Schlauheit besides Schläue ← schlau ‘smart’). The suffix -keit originally was a phonologically conditioned variant of -heit, and -igkeit originated from -keit through the reanalysis of formations with bases ending in -ig. Lithuanian (cf. article 169, section 4.1.2) has ca. 15 suffixes that compete for the formation of quality nouns (for a diachronic account, cf. Bammesberger 1973). The situation is even more complex in Romance languages (see the monographic treatment of Italian in Rainer 1989; of Portuguese in Correia 2004; as well as the detailed descriptions of French, Catalan and Spanish in Lüdtke 1978). Depending on one’s definition of suffixal allomorphy, up to 34 suffixes can be counted in Italian. What fundamentally distinguishes the situation in Italian from that in English and German is the fact that, despite the huge number of suffixes, there is no default suffix, which is why a number of adjectives that do not fall into the domain of one of the suffixes are condemned to remain without a corresponding quality noun. Most Italian suffixes are unproductive, i.e. lexically governed, the others obey phonological, morphological, semantic and stylistic restrictions, often more than one of these at the same time. From such affix-specific conditions we have to distinguish two general semantic conditions that seem to hold for quality nouns more generally. The first one is well-known. It requires bases to be predicative adjectives: relational and uniquely attributive adjectives are excluded, as the contrast between the relational and the predicative use of German gestrig shows. In its relational use in (11a), which is confined to the attributive position, the adjective, derived from gestern ‘yesterday’, means ‘yesterday’s, of yesterday’, while in the figurative, predicative use of (11b) it means ‘old-fashioned’. Hence the difference in acceptability of the quality noun Gestrigkeit in (11a; ‘the remark of yesterday’) and (11b; ‘the old-fashioned remark’): (11) a. die gestrige1 Bemerkung vs. *die B. ist gestrig/*die Gestrigkeit der B. b. die gestrige2 Bemerkung ≈ die B. ist gestrig/die Gestrigkeit der B. A second general semantic constraint was first proposed in passing by Malkiel (1945: 173). After having observed that the formation of quality nouns is in principle unconstrained from a referential point of view, he adds the following proviso: “Nur zu modifizierten, d. h. gedanklich nicht vollwertigen Formen wie it. bellino, sp. bonito fehlen Entsprechungen.” [Only modified, conceptually incomplete forms such as It. bellino, Sp. bonito lack corresponding quality nouns.] Malkiel’s observation was tested against a wider range of Italian data in Rainer (1989: 55−59) and found to be essentially correct. The same restriction seems to hold for Russian, although occasional exceptions occur

1280

VII. Semantics and pragmatics in word-formation II: Special cases

on the Internet (e.g., mil-yj ‘dear’ → mil-en’kij ‘very dear’ → milen’k-ost’). There are, however, also potentially problematic examples from other languages. In English and German, the approximative suffixes -ish and -lich seem to tolerate quality-noun suffixes: bluishness, for example, is listed in several dictionaries, and Bläulichkeit is well attested on the Internet. The same situation seems to obtain in Polish (e.g., niebieski ‘blue’ → niebieskawy ‘bluish’ → niebieskawość ‘bluishness’) and Russian (e.g., sin-ij ‘blue’ → sin-evat-yj ‘bluish’ → sinevat-ost’ ‘bluishness’). At the present state of our knowledge it is unclear whether this should be interpreted as evidence that the constraint in question is language-specific, or whether the different behavior can be attributed to semanticpragmatic factors in a principled way. In the latter case, the degree of stability of the concept could well be the decisive factor. Cross-linguistic variation is also observable with respect to the compatibility of quality-noun suffixes with comparative and superlative suffixes. While German *Größerkeit or *Größerheit (← größer ‘bigger’) sound decidedly odd to my ears, English biggerness enjoys some popularity on the Internet, though it may not be to everybody’s taste. Derivatives from superlatives such as dearestness are considered “exceptional” by Marchand (1969: 335). The same seems to be true for Russian (e.g., milyj ‘dear’ → nai-mil-ejš-ij ‘dearest’ → naimilejš-est’). In Finnish, on the contrary, according to Lazar (1975: 96), “[a]bstract nouns can theoretically be formed from any comparative form”: vanha ‘old’ → vanhempi ‘older, senior’ → vanhemmuus ‘seniority, greater age’, etc.

8. The origin of quality-noun suffixes Although a study with a broad coverage remains a desideratum, the eclectic evidence that I have gathered is sufficient to show that the diachronic pathways leading to qualitynoun suffixes are highly diverse. First, such suffixes are sometimes the result of g r a m m a t i c a l i z a t i o n , i.e. of the bleaching and subsequent analogical extension of the second constituent of compounds. The case of the status-noun suffixes -hood and -dom, which also marginally function as quality-noun suffixes, is well-known and need not be rehearsed here (cf. Trips 2009; on the analogous German suffixes, Piltz 1951). In the Uralic languages, some quality-noun suffixes can be shown to derive from nouns with a temporal meaning such as ‘day’, ‘time’, ‘age’ (cf. Lazar 1975: 311 ff.; e.g., Zyrian pemyd lun ‘dark day’ > pemydlun ‘darkness’). Lazar also endorses a similar origin for the Balto-Finnic quality-noun suffix -us, tracing it back to a word meaning ‘year’. While this temporal pathway is quite intuitive, one would decidedly like to have more details about the exact semantic changes involved in Hungarian, where “-ság/-ség can be traced back to the noun ság ~ szág ‘mound, hill’” (Tamás Forgacs, article 118, section 3.1). In Budugh (cf. article 196, section 4.1.2) the quality-noun suffix -xhın is said to be related to the verb yıxhar ‘to be’, which again makes much sense. In a second scenario, the noun in an adjective-noun sequence is not bleached but suppressed through e l l i p s i s . The meaning of the suppressed noun is then absorbed by the ending of the adjective. According to Porzig (1951: 153), Ancient Greek γηϑοσύνη ‘joy’ was shorthand for a phrase such as γηϑοσύνη φρήν ‘joyful heart’, giving rise to a long series of similar quality nouns in -οσύνη. The same kind of origin is also attributed to the Greek quality-noun suffix -ία.

72. Quality nouns

1281

A third mechanism leading to quality-noun suffixes also involves s e m a n t i c c h a n g e of an existing affix, but this time no ellipsis is involved. The pejorative Italian quality-noun suffix -aggine (cf. Malkiel 1976), for example, derives from the Latin suffix -ago, -aginis which occurs in some nouns denoting diseases (e.g., lumbago ← lumbus ‘loin’). This origin explains why the suffix is still restricted to negative qualities in Italian. A fourth kind of origin is a f f i x c o a l e s c e n c e , of which there are several varieties. In Akhwakh (cf. article 205, section 4.1.2), the quality-noun suffix goes back to a combination of a suffix deriving intransitive verbs plus the masdar suffix. New quality-noun suffixes can also arise as a result of pleonastically adding younger, more productive suffixes to older ones. Such has been the origin of the Irish suffix -aige (< -ach + -e) as well as -adas (< -atu + -us) and -achas (< -ach + -us) (cf. article 114 From Old Irish to Modern Irish, section 3.3.1). Reanalysis, in the form of boundary shift, can also lead to new forms of suffixes, the staple case being German -igkeit, whose -ig- originally was a suffix of the base. Last but not least, quality-noun suffixes seem to be prone to b o r r o w i n g . This certainly has to do with the fact that quality nouns are particularly frequent in philosophical and other elaborated kinds of discourse and are often transferred to other languages in the process of transferring such discourses to language communities that lack them. This is the way in which Ancient Greek -ία entered the Latin language, and how Latin -itas was transferred to many European languages, including Hungarian (-itás) and Breton (-ded ). All Western European languages have been deeply influenced by Latin in the domain of quality nouns, even where these use native suffixes: it is generally thought, for example, that German Freiheit (Old High German frîheit) was coined by Notker of Saint Gall in order to render Latin libertas. Seen in this light, Lazar (1975) is undoubtedly correct in attributing the much higher number of quality nouns in Balto-Finnic languages and Hungarian with respect to Uralic languages situated farther to the East or North to the deep influence of Western European languages: even if the suffixes themselves are of native origin, many formations must be considered as calques.

9. References Alexiadou, Artemis and Monika Rathert (eds.) 2010 The Syntax of Nominalizations across Languages and Frameworks. Berlin: De Gruyter Mouton. Arsenijević, Boban 2011 The semantic ontology of deadjectival nominalizations in Serbo-Croatian. Recherches linguistiques de Vincennes 40: 53−72. Baeskow, Heike 2002 Abgeleitete Personenbezeichnungen im Deutschen und Englischen. Kontrastive Wortbildungsanalysen im Rahmen des Minimalistischen Programms und unter Berücksichtigung sprachhistorischer Aspekte. Berlin/New York: de Gruyter. Bally, Charles 1965 [1932] Linguistique générale et linguistique française. 4 th ed. Berne: Francke. Bammesberger, Alfred 1973 Abstraktbildungen in den baltischen Sprachen. Göttingen: Vandenhoeck & Ruprecht. Bauer, Laurie 2002 What you can do with derivational morphology. In: Sabrina Bendjaballah, Wolfgang U. Dressler, Oskar E. Pfeiffer and Maria D. Voeikova (eds.), Morphology 2000. Selected

1282

VII. Semantics and pragmatics in word-formation II: Special cases

papers from the 9 th Morphology Meeting, Vienna, 24−28 February 2000, 37−48. Amsterdam/Philadelphia: Benjamins. Bauer, Laurie, Rochelle Lieber and Ingo Plag 2013 The Oxford Reference Guide to English Morphology. Oxford: Oxford University Press. Becker, Karl Ferdinand 1824 Die Deutsche Wortbildung oder die organische Entwicklung der deutschen Sprache in der Ableitung. Frankfurt/M.: Hermannsche Buchhandlung. Bourquin, Jacques 1979 Remarques sur la formation néologique récente de substantifs en -itude dérivés ou non. In: Roseline Adola (ed.), Néologie et lexicologie. Hommage à Louis Guilbert, 47−66. Paris: Larousse. Brugmann, Karl 1906 Grundriss der vergleichenden Grammatik der indogermanischen Sprachen. 2 nd ed. Strassburg: Trübner. Correia, Margarita 2004 Denominação e construção de palavras. O caso dos nomes de qualidade em português. Lisboa: Colibri. Dal, Georgette and Fiammetta Namer 2010 French property nouns based on toponyms or ethnic adjectives. In: Franz Rainer, Wolfgang U. Dressler, Dieter Kastovsky and Hans Christian Luschützky (eds.), Variation and Change in Morphology. Selected papers from the 13 th International Morphology Meeting, Vienna, February 2008, 53−74. Amsterdam/Philadelphia: Benjamins. Daude, Jean 2002 Les substantifs abstraits de qualité. In: Chantal Kircher-Durand (ed.), Grammaire fondamentale du latin. Tome IX. Création lexicale: La formation des noms par dérivation suffixale, 225−305. Louvain: Peeters. Doerfert, Regina 1994 Die Substantivableitung mit -heit/-keit, -ida, -î im Frühneuhochdeutschen. Berlin/New York: de Gruyter. Dokulil, M[iloš] 1968 Zur Theorie der Wortbildung. Wissenschaftliche Zeitschrift der Karl-Marx-Universität Leipzig. Gesellschafts- und Sprachwissenschaftliche Reihe 17: 203−211. Ermecke, Gustav 1929 Das Wesen der sprachlichen Abstrakta und ihre Bildung durch Suffixe im Romanischen nebst einem Hinweis auf den Einfluss dieser Art Suffix-Bildung auf das Englische und Deutsche. Langendreer: Pöppinghaus. Flaux, Nelly, Michel Glatigny and Didier Samain (eds.) 1996 Les noms abstraits: histoire et théories. Actes du colloque de Dunkerque (15−18 septembre 1992). Lille: Presses Universitaires du Septentrion. Fradin, Bernard 2011 Remarks on state denoting nominalizations. Recherches linguistiques de Vincennes 40: 73−99. Goddard, Cliff and Anna Wierzbicka 2014 Words and Meanings. Lexical semantics across domains, languages, and cultures. Oxford: Oxford University Press. Gruppe, Otto Friedrich 1954 [1831] Wort und Begriff Abstraktum. In: Heinrich Junker (ed.), Sprachphilosophisches Lesebuch, 148−165. Heidelberg: Winter. Hamawand, Zeki 2008 Morpho-Lexical Alternation in Noun Formation. Basingstoke: Palgrave Macmillan. Hausser, Eberhard 1959 Abstrakta im Schwäbischen. Ph.D. dissertation, University of Tübingen.

72. Quality nouns

1283

Heinimann, Siegfried 1963 Das Abstraktum in der französischen Literatursprache des Mittelalters. Bern: Francke. Herder, Johann Gottfried 1953 Werke in zwei Bänden. Ed. Karl-Gustav Gerold. München: Hanser. Jenks, Peter, Andrew Koontz-Garboden and Emmanuel-Moselly Makasso submitted Basaá and the lexical semantics of property concept nouns. Available at: http:// linguistics.berkeley.edu/~jenks/Research_files/basaa-pc-words-submitted-for-circ.pdf [last access 8 Jan 2014]. Koehl, Aurore 2009 Are French -ité suffixed nouns property nouns? In: Fabio Montermini, Gilles Boyé and Jesse Tseng (eds.), Selected Proceedings of the 6 th Décembrettes. Morphology in Bordeaux, 95−110. Somerville, MA: Cascadilla. Koehl, Aurore 2012 Altitude, négritude, bravitude ou la résurgence d’une suffixation. In: Congrès Mondial de Linguistique Française. SHS Web of Conferences 1 (2012), 1307−1323. Available at: http://www.shs-conferences.org Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Vol. 1: Theoretical prerequisites. Stanford: Stanford University Press. Lazar, Oscar 1975 The Formation of Abstract Nouns in the Uralic Languages. Stockholm: Almqvist & Wiksell. Lüdtke, Jens 1978 Prädikative Nominalisierungen mit Suffixen im Französischen, Katalanischen und Spanischen. Tübingen: Niemeyer. Malkiel, Jacques [Yakov] 1938 Das substantivierte Adjektiv im Französischen. Berlin: Speer & Schmidt. Malkiel, J. [Yakov] 1945 Probleme des spanischen Adjektivabstraktums. Neuphilologische Mitteilungen 46: 171− 191 and 47: 13−45. Malkiel, Yakov 1976 One characteristic derivational suffix of literary Italian: -(t)aggine. Archivio Glottologico Italiano 61: 130−145. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. Martin, Fabienne 2013 Stage level and individual level readings of dispositional nouns. In: Nabil Hathout, Fabio Montermini and Jesse Tseng (eds.), Morphology in Toulouse. Selected proceedings of Décembrettes 7 (Toulouse, 2−3 December 2010), 155−183. München: LINCOM Europa. Mikkola, Eino 1964 Die Abstraktion: Begriff und Struktur. Eine logisch-semantische Untersuchung auf nominalistischer Grundlage unter besonderer Berücksichtigung des Lateinischen. Helsinki: Suomalainen Kirjakauppa. Moltmann, Friederike 2013 Abstract Objects and the Semantics of Natural Language. Oxford: Oxford University Press. Oberle, Birgitta 1990 Das System der Ableitungen auf -heit, -keit und -igkeit in der deutschen Gegenwartssprache. Heidelberg: Winter. Paul, Hermann 1920 [1880] Prinzipien der Sprachgeschichte. 5 th ed. Tübingen: Niemeyer.

1284

VII. Semantics and pragmatics in word-formation II: Special cases

Piltz, Gunter 1951 Die Bedeutungsentwicklung der Substantiva auf -heit, -schaft und -tum. Ph.D. dissertation, University of Hamburg. Pinault, Georges-Jean 1996 Aspects de la reconstruction de l’abstrait en indo-européen. In: Nelly Flaux, Michel Glatigny and Didier Samain (eds.), Les noms abstraits: Histoire et théories. Actes du colloque de Dunkerque (15−18 septembre 1992), 199−211. Lille: Presses Universitaires du Septentrion. Porzig, Walter 1930−31 Die Leistung der Abstrakta in der Sprache. Blätter für deutsche Philosophie 4: 66− 77. Porzig, Walter 1942 Die Namen für Satzinhalte im Griechischen und im Indogermanischen. Berlin: de Gruyter. Porzig, Walter 1951 Die Entstehung der abstrakten Namen im Indogermanischen. Studium Generale 4: 145− 153. Rainer, Franz 1989 I nomi di qualità nell’italiano contemporaneo. Wien: Braumüller. Reichl, Karl 1982 Categorial Grammar and Word-Formation. The de-adjectival abstract noun in English. Tübingen: Niemeyer. Riddle, Elizabeth M. 1985 A historical perspective on the productivity of the suffixes -ness and -ity. In: Jacek Fisiak (ed.), Historical Semantics. Historical Word-Formation, 435−461. Berlin: Mouton. Riegel, Martin 1985 L’adjectif attribut. Paris: Presses Universitaires de France. Roché, Michel 2007 Logique lexicale et morphologie: La dérivation en -isme. In: Fabio Montermini, Gilles Boyé and Nabil Hathout (eds.), Morphology in Toulouse. Selected proceedings of the 5 th Décembrettes, 45−58. Somerville, MA: Cascadilla. Roy, Isabelle and Elena Soare (eds.) 2011 Nominalizations. Saint Denis: Presses Universitaires de Vincennes. Spencer, Andrew 2013 Lexical Relatedness. A paradigm-based model. Oxford: Oxford University Press. Strawson, Peter F. 1959 Individuals. An essay in descriptive metaphysics. London: Methuen. Szadrowsky, Manfred 1933 Abstrakta des Schweizerdeutschen in ihrer Sinnentfaltung. Frauenfeld: Huber. Trips, Carola 2009 Lexical Semantics and Diachronic Morphology. The development of -hood, -dom and -ship in the history of English. Tübingen: Niemeyer. Trost, Klaus 1976 Zur Definition der Abstrakta (am Beispiel von Russisch und Deutsch). Indogermanische Forschungen 81: 221−239. Wälchli, Bernhard 2005 Co-Compounds and Natural Coordination. Oxford: Oxford University Press. Wald, Lucia 1990 Some observations concerning the plural of abstract nouns in Romanian and other Romance languages. Revue Roumaine de Linguistique 35: 389−393. Zindel, René 1958 Des abstraits en français et de leur pluralisation. Bern: Francke.

Franz Rainer, Vienna (Austria)

73. Status nouns

1285

73. Status nouns 1. 2. 3. 4. 5. 6. 7. 8.

Introduction Terminology Categorial indeterminacy Morphological patterning Semantic iridescence Meaning, senses and diachrony Final remarks References

Abstract The category of status nouns (nomina status), stipulated mainly on the basis of suffixal formations in Romance languages, is not firmly established in the study of word-formation. The discussion in this article is therefore problem-oriented, aiming at the clarification of terminology, the etching of conceptual criteria for definition and the assessment of cross-linguistic evidence for the acknowledgement of status nouns as an independent derivational category. From a theoretical point of view, status nouns, as a subtype of abstract nouns denoting typically a property (e.g., motherhood), display interesting interactions between the lexical meaning of the base form, typically a personal noun, and the semantics of the respective marker (typically a suffix).

1. Introduction Unlike inflectional categories, for which a well-established system of classification and nomenclature has been available since the times of the ancient Greek and Roman grammarians, derivational categories do not rest on a solid conceptual framework where phenomena can be sorted out and filed away in a straightforward manner. Thus, while linguists might be in disagreement on the function and use of, say, the dative in languages with oblique subject marking, as in the German phrase Mir träumt von etwas ‘I am dreaming of something’, there is no disagreement on the fact that mir is the dative singular form of the personal pronoun in Standard German. Furthermore, there is tacit unanimity among linguists that, in spite of the fact that the term dative, inherited from Latin (casus) dativus (translated from Greek dōtikḗ [ptõsis] ‘giving [case]’), was illcoined and misleading right from its invention two-and-a-half millennia ago, there is no point in making up a new technical term that would capture the function of this category more appropriately. In derivational morphology the situation is very different. Only the most basic and widespread categories are labeled consistently across different theoretical frameworks and language-related fields, so that, e.g., it always seems to be clear what is at issue when a derivative is called an agent noun (nomen agentis) or an action noun (nomen actionis). But beyond the cultivated landscape of the most common derivational categories, many of which are expressible in terms of case functions or argument rela-

1286

VII. Semantics and pragmatics in word-formation II: Special cases

tions, lurks the wilderness of conceptual and terminological proliferation, where one and the same term is used for different phenomena, or conversely, similar phenomena are termed differently in different philological fields and/or by different authors. A rather cross-grained species of the conceptual and terminological rank growth in the morphological jungle is the category of status nouns (nomina status), the taming and domestication of which has been chosen as the topic of the present article. Status nouns are a subtype of abstract nouns. Unlike other classes of abstract nouns, like action nouns (cf. article 67), which are derived from verbs, and quality nouns (cf. article 72), which are derived from adjectives, the formation of status nouns involves no change with respect to word class. This makes status nouns similar to other denominal nouns, such as collectives, diminutives, or denominal nouns with temporal or locative meaning (nomina loci and temporis in the traditional terminology). However, while the latter types of nouns have concrete meaning, status nouns refer to abstract concepts, which makes them similar to deadjectival nouns denoting properties (nomina qualitatis in the traditional terminology, also called nomina essendi). Since neither the Latin term nomen status nor its English equivalent status noun is firmly established in the nomenclature of word-formation, it is necessary to first discuss the meaning and use of this element of metalanguage before addressing the entity it is supposed to denominate.

2. Terminology Although listed among “common derivational meanings of nouns” in introductory textbooks (e.g., Haspelmath and Sims 2010: 87), the item status in the term status noun is bound to provoke confusion, since status and state are partially synonymous English words, stemming ultimately from the same Latin source, i.e. status ‘manner, position, condition, attitude’ (originally an action noun in -tus derived from the verb stem sta- ‘to stand’, homonymous with the past participle of sistere ‘to put’). A first distinction to be paid attention to is therefore the one between status noun as a technical term for a noun denoting a status, and terms for nouns denoting a state or condition, the latter corresponding to quality nouns. Another distinction that should be borne in mind refers to the term stative as used for the classification of verbs lacking a semantic value of action (as opposed to dynamic verbs). The meaning of stative (or static) in this case is closely related to the meaning of state (in the sense of a condition), but both of them are less closely related to what is meant by status when speaking of “status nouns”. Besides this use of status, in Semitic linguistics, the term status (of a noun) is used for a morphosyntactic distinction between nouns governing a noun phrase in a possessive or relational construction (status constructus), predicative nouns (status absolutus or indeterminatus), and all other nouns (status rectus or determinatus). It goes without saying that this use of the term status bears no relationship to ‘status’ as a semantic category in word-formation, nor do other uses of status as a term in theoretical and descriptive linguistics, e.g., Bech’s degrees of status for different types of infinitival constructions in German (Bech 1983), the so-called status suffix in Mayan languages (Verhoeven 2007), or the use of status for reference to the position of discourse participants on a scale in terms of sociolinguistics and personal deixis (cf. Lyons 1977: 108; 574–576; Hunston 2011).

73. Status nouns

1287

Inasmuch as ‘status’ is neither a functional or thematic role like ‘agent’, ‘patient’ or ‘instrument’, nor a circumstantial role, nor is a change of word-class involved in the formation of status nouns, the specification of a lexical item as a status noun relies on criteria that are not based on grammatical information, but on lexical meaning alone. As a consequence, the delimitation of the semantic range of status nouns is handled rather loosely in the literature, and there will hardly be two authors who agree on the subject, apart from the fact that many works on word-formation lack any reference to such a derivational category at all. It is beyond the scope of the present article to discuss in detail the different uses of the term status, also in relation to state and stative, across the history of research, but it should be remembered that for historical linguists of the nineteenth century the term nomen status was a label for deadjectival nouns denoting a state of being. For example, the Sanskrit nouns jayá ‘victory’ (from an adjective meaning ‘winning, defeating’, ultimately based on the verbal root ji- ‘to win’) and javá ‘hurry’ (from an adjective meaning ‘hurried, fast’, ultimately based on the verbal root jū- ‘to hurry’) were classified as nomina status by Benfey (1866: 292), who also used the notion Nominalabstractum (on which see below), but primarily when writing about Semitic languages (e.g., Benfey 1844). Contemporary authors still speak of “stative nouns” when referring to deadjectival formations like wickedness, illness or sickness (e.g., Bauer 2001: 164), or to substantivized participles like the beating (Brinton and Akimoto 1999: 103), or use stative noun as an antonym of event noun or eventive noun (e.g., Dal and Namer 2012). Furthermore, “stative nouns” (originally termed “inactive”) are set off from “active nouns” in the typological conception of “active languages” (as opposed to accusative and ergative languages; see Donohue and Wichmann 2008 for an overview). Although the adjective stative is clearly derived from state and not from status, a quantum of ambiguity remains, the more so in view of the fact that a broader understanding of the meaning ‘status’ is implied in the nomenclature still current, e.g., in Uralic linguistics, where the term nomen status is used for nominalized participles and adverbs (e.g., Udmurt užam ‘worked, something already worked up’, from uža- ‘to work’, cf. the action noun užan ‘work [in general]’; Napol’skich 2003: 299; cf. also Lazar 1975). As for the Latin term nomen status, terminological variegation seems to be due to the lack of a distinction, in Latin, between the meanings conveyed by the English words status and state, respectively. Although Latin, besides status ‘manner, position, state, condition, attitude’, provides other derivatives from stare ‘to stand’ with abstract meaning, e.g., statio ‘standing (still), station, post’ but also ‘office, position’, the only nominalization of stare ‘to stand’ that made its way into linguistic terminology is status, and so the concept of nomen status inherits its entire polysemy, ranging from general abstraction, i.e. ‘noun denoting a condition or quality’, to the more specific sense ‘noun denoting a position in a hierarchy or classification’. In the Romance languages, though, regularly developed outcomes of Latin status contrast with Latin status borrowed as a mot savant, so that French now has état (Old French estat, whence English estate and state), with the general abstract meaning ‘condition’ (plus various extensions like ‘public household’), besides statut, with the more specific meaning, as in statut social ‘social status’, statut de la femme mariée ‘legal status of a married woman’, etc. Similarly, Italian has stato, with the general abstract meaning ‘condition’ (plus various extensions like ‘sovereign political entity’), besides status, with the more specific meaning, as in status diplomatico ‘diplomatic status’, status giuridico ‘legal status’, etc. Likewise, Spanish has estado, with the general abstract meaning ‘condition’ (plus various extensions like estado

1288

VII. Semantics and pragmatics in word-formation II: Special cases

mayor ‘general staff’), besides status (pronounced, and sometimes also written, with a prothetic e-), with the more specific meaning, as in estatus civil ‘marital status’, and so forth. In German, the distinction between nouns denoting a state or condition and nouns denoting a status can be expressed by using the label Zustandsbezeichnung for the former and Statusbezeichnung for the latter (cf. also the term Standesbezeichnung used, e.g., by Meyer-Lübke 1921: 58 for nouns of the type Latin consulatus ‘dignity or office of a consul’). Although ‘status’, like ‘state’, is undisputedly an abstract concept (cf. the German terms Nominalabstraktum and Substantivabstraktum, covering, among other denominal types of derivatives, the category of status nouns in the sense intended here), the notion of ‘status’ appears also in the characterization of the meaning of denominal derivatives that are concrete nouns. For example, the English suffix -ess (as in actress) is described by Bauer (1983: 221) as “showing either professional status or the status of the woman’s husband”. Bauer does not use the term status noun in his account of English wordformation, but does refer to the concept of status for capturing the semantic value of the suffix in question. A similar role of ‘status’ meaning as a connotation of personal nouns is assigned to the suffix -ee by Mühleisen (2010: 176). Likewise, a treatise on “status nomination” in Russian by Migirina (1980), in spite of its title, deals not with status nouns as such but with concrete nouns, derived as well as underived ones, denoting persons pertaining to groups that can be defined in terms of status, thus constituting potential bases for the derivation of status nouns, but not being status nouns themselves. Among these nouns are underived kinship terms like snoxá ‘daughter-in-law’, but also various types of derived nouns denoting persons according to physical and other properties (e.g., gorbún ‘hunchback’, from gorb ‘hump’), names of inhabitants, and so forth. Still another use of the notion ‘status’ occurs in article 137 on Frisian (section 4.1.1), where the prefix haad- (as in haadstêd ‘capital city’) is characterized as denoting status. From these examples it can be seen that the notion ‘status’ in the metalanguage of derivational morphology is far from being reserved for the characterization of a specific type of denominal abstract nouns. The German notions Nominalabstraktum and Substantivabstraktum mentioned above are both current since the early nineteenth century. Nominalabstraktum, which appears, e.g., in Franz Bopp’s Sanskrit grammar (Bopp 1833: 1216), remains ambiguous with respect to the basis of derivation, since in Indo-European linguistics adjectives traditionally count as a nominal category because of their inflectional properties. For example, Friedrich Kluge speaks of “denominativabstracta zu adjectiven” (Kluge 1899: 55) and refers to deadjectival nouns like Gothic junda ‘youth, young age’ (Kluge 1899: 62), Old Norse bernska ‘childhood’ (from bernskr ‘childish’), fyrnska ‘(old) age’ (also forneskja, from forn ‘old’), mennska ‘human nature, humanity’ (from mennskr ‘human’) as “nominalabstracta”. The somewhat less ambiguous designation “denominative abstracta”, applied alternatively by Kluge, could have been made use of when referring to nouns like Old Norse bróðerne ‘brotherhood’, faðerne ‘fatherhood’ (Kluge 1899: 80), in order to make sure that these are derived from real nouns and not from adjectives, but no such terminological differentiation is detectable in the literature. The term Substantivabstraktum, which appears, e.g., in Jacob Grimm’s German grammar (there applied, however, to substantivized personal pronouns, Grimm 1837: 293), is more precise, specifying the category of the derivational base unequivocally. It has gained currency in Romance linguistics through influential authors like Meyer-Lübke

73. Status nouns

1289

(e.g., Meyer-Lübke 1890: 118). Still, many authors neglect the necessity of keeping the nomenclature as descriptive as possible, or ignore the righthand head rule holding for compounds, and refer to abstract nouns of any kind as “Substantivabstrakta”, just because they are nouns. For example, Hennings (2012: 212) labels Middle High German suht ‘infirmity’ as a “Substantivabstraktum”, at the same time explaining it correctly as deadjectival (from siech ‘ailing’). There are even authors who indulge in formulations like “Substantivierung von Substantiven” when writing about denominal nouns (e.g., Bendel 2006: 177–181). Some authors have invented alternative designations, e.g., “desubstantivische Prädikatsnominalisierung” (Lüdtke 1998: 376–378; cf. already Lüdtke 1978), but the propensity for brevity in terminology seems to prevent this creation from gaining general acceptance. Still others refrain from naming derivatives of the type fatherhood explicitly, characterizing their meaning with reference to “social relations” and the like (e.g., Friedländer 1992: 205). In view of this state of affairs regarding the terminology, it comes as no surprise that the status of status nouns as a category distinct from other abstract nouns is in need of clarification. In the following section, the prerequisites for such a clarification will be examined.

3. Categorial indeterminacy The category of status nouns is not constituted by a sharply outlined set of derivatives with a homogeneous semantic value but rather by an agglomeration of derivational types centering around a prototypical core. As pointed out by Štekauer, Valera and Körtvélyessi (2012: 6), “[f ]uzziness appears as a natural consequence of the scalar nature of linguistic facts, frequently reflected in vague boundaries between word-formation processes and categories on the one hand, and between word-formation per se and other levels of language on the other (e.g., inflection, syntax, phonology)”. The most basic type of status nouns is embodied by denominal nouns expressing the quality of being what the meaning of the respective base noun is and nothing else. Examples of such pure status meaning are not easy to find. English citizenship seems to be a good candidate, since all semantic aspects of this derivative are purely abstract, i.e. no reference to concrete dimensions like space or time is implied. The best proof for pure status meaning of a noun is the unacceptability of combining it with prepositions establishing concrete relations. Consider the sentence The officer had a look at my citizenship. This can only be a meaningful utterance if citizenship is used metonymically for the documents that verify the status of being a citizen. This kind of metonymic use cannot be excluded in principle, but in the case at hand it seems rather unlikely to occur in everyday conversation. Similarly, an utterance like After my citizenship I could not register for social health care any longer could only make sense if enounced by a person who has lost citizenship, but then it would rather be understood as elliptical for After the loss of my citizenship … or After I had lost my citizenship … or a similar construction. Pure status nouns like citizenship are incompatible with concreteness also in that they are not likely to take the subject position of transitive verbs with concrete meaning. An utterance like My citizenship asks for prolongation only makes sense with the abstract reading of to ask for, as in This problem asks for a solution. Typically, the most common

1290

VII. Semantics and pragmatics in word-formation II: Special cases

collocations involving nouns like citizenship are those with verbs of attainment or attestation and their opposites, i.e. citizenship can be granted, acquired, received, revoked, recognized, filed, given up, renounced, lost and so on, but it cannot be heard or seen (in the basic, non-metaphorical sense of the respective verbs). Status nouns are a showcase example for the issue of how much of the meaning of a derivative depends on the lexical meaning of the base. Since the meaning of citizen is a highly abstract one, referring to a convention or construct and not to something natural, the status noun citizenship has a relatively abstract meaning, too. This becomes clear by comparing it to status nouns like childhood. Since the meaning of child is more concrete than that of citizen, childhood turns out to be less prototypical as a status noun, allowing for more semantic content than just ‘the quality of being a child’. In fact, the pure status meaning of childhood seems not to be the focus of the semantic shades of this derivative, but the temporal meaning as an age bracket is prevalent. We will take up this example again at the end of this section. The most straightforward criterion for establishing a derivational category is the regularity of the correlation between the meaning and the form of the derivatives in question. If every word derived with a certain affix would mean something substantially different, one would hardly think of establishing a derivational category uniting the different meanings under a common semantic denominator, just because of the formal resemblance of the words generated by the derivational process. For example, if the only English words derived with the suffix -dom were wisdom, christendom, popedom and gaydom, it would be audacious to establish a common derivational category for these derivatives (leaving aside the issue of lexicalization), since wisdom denotes a mental quality (or ability, to be more precise) of human individuals, while christendom denotes a religion (plus its cultural implications and geographical extension), popedom denotes the jurisdiction of an ecclesiastical institution, and gaydom denotes a community of people with a particular sexual orientation (including their world-view). Attempting to bring all four semantic values, i.e. quality, creed, institution, and community, under one conceptual denominator would result in a very abstract idea of something like ‘having to do with humans’, a notion that can be dismissed as unhelpful at best and certainly nothing that might count as a semantic basis for a derivational category. If our hypothetical sample of formations were real (as a sample), there would be two solutions to the problem of semantic disparity: either a set of different suffixes happens to sound alike (homonymy), or one original suffix has developed different meanings over time (polysemy). The former case would constitute an endpoint for any further analysis, except for the study of the sound changes that led to the homonymy of the formerly distinct suffixes. In the latter case, the shift of meaning implied by the synchronic disparity should be recapitulated as a series of plausible transitions that can be captured in terms of metaphor, metonymy or any of the common pathways of semantic change, and under favorable circumstances the original meaning of the suffix could be reconstructed. Still, there would be no reason to establish a common derivational category for the synchronic set of forms, since it is theoretically odd to posit an anachronistic entity, i.e. a reconstructed anterior meaning, as a synchronic motive for the patterning of forms. So while “nothing doing” is the keynote resounding from our artificial example, as far as the synchronic account is concerned, there is news from the real world that conveys a more varied picture of the interplay of morphological patterning and derivational meaning (apart from the fact that nothing needs to be reconstructed here, the

73. Status nouns

1291

history of the suffix -dom and its cognates being perfectly well attested in all Germanic languages except Gothic, cf. Dietz 2007). The good news is that derivatives are hardly ever distributed in the way depicted above for the sake of argumentation. For a given derivational process, the normal state of affairs is a much larger number of derivatives, in the first place, with an at least partly cohesive range of meanings. Thus for the English suffix -dom we find parallel formations in all four semantic fields represented by the examples given in the preceding paragraph (although the patterns differ considerably with respect to productivity, cf. Bauer 2001: 167–172). Alongside wisdom there is deadjectival freedom, with a roughly congruent meaning of ‘condition’ (at least in the sense of ‘ability’ or ‘capacity’ wisdom and freedom share a semantic value). Alongside christendom there is heathendom; popedom is paralleled by nouns like kingdom, both agreeing in the semantic values of ‘jurisdiction’ as well as ‘domain’, and the categorial semantic value of gaydom is roughly the same as that of, say, artistdom, meaning a way of life (plus the community leading it). Special senses like the use of kingdom as a taxonomic level in biology need not be considered for the present discussion. The bad news is that the existence of a wealth of derivatives with a more or less cohesive range of meanings generated by a given derivational process by no means warrants the establishment of a derivational category. For example, the meanings of the nouns in -dom cited and classified by Marchand (1969: 262–264) across the attested history of English are pinned down by this author with the following categorial notions: ‘jurisdiction’, ‘state’, ‘condition’, ‘statute’, ‘dignity’, ‘status’, ‘authority of’, ‘position of’, ‘realm’, ‘territory’, ‘domain’, ‘region’, ‘land of’, ‘world of’, ‘collectivity of’, ‘inhabitants of’, ‘community’, ‘fraternity’, ‘class’, ‘group united by a common interest’, ‘collective body of people representative of’, to which is to be added the evaluative sense of depreciation (as in whoredom). Marchand’s approach to derivational semantics can be characterized as a bottom-up typology (or a semasiological approach, in terms of Štekauer 2005: 207). Instead of establishing a top-down system of derivational categories, where derivatives of various kinds are pigeonholed, he provides an alphabetical list of “principal sense groups” capturing the conceptual range covered by the derivatives of a certain type (Marchand 1969: 516–522), where the suffix -dom appears under the headings of ‘collectivity’, ‘condition of’, ‘depreciation’, ‘state’, and ‘territory’. Furthermore, according to Marchand, the following nouns in -dom are “type-words” considered to be representative for a derivational type reflecting a particular conceptual sphere: freedom, martyrdom, kingdom, duncedom, stardom, boydom, theaterdom, and deverbal listendom (not in the OED). Of these, only stardom has the meaning of ‘status’ as its primary semantic value (cf. OED s. v.). In Bauer’s account of Marchand’s examples this type corresponds to group (c), i.e. “nominal bases with the noun in -dom denoting a state or condition” (Bauer 2001: 166). According to Marchand (1969: 263), for the suffix -dom “the neutral shade of ‘status’ is no longer the leading one”, and even for more far-fetched formations like saintdom other senses than ‘the condition or quality of being a saint’ are attested (collectivity in this case, cf. OED s. v.). The semasiological approach, taking into account the full spectrum of polysemy and semantic extensions of a given derivational pattern, seems not to lead to something like the formation of a clear-cut system of derivational categories, but rather produces a fuzzy set of fuzzy concepts, structured mainly in quantitative terms by the tendencies of analo-

1292

VII. Semantics and pragmatics in word-formation II: Special cases

gy and productivity. This does not mean, however, that the bottom-up way of assessing derivational meaning must be abandoned for the sake of bringing the gamut of senses into the idealized form of a category. If a derivational category of nomina status cannot be established on the basis of a bottom-up account of a single suffix like English -dom (or any of its cognates in Germanic languages, cf. Bauer 2001: 163–172), this approach can still turn out to be fruitful if applied to a whole group of functionally related suffixes with overlapping meanings. If the overlapping semantic field would turn out to be ‘status’, this could be taken as evidence in favor of nomina status as a derivational category, from which the other senses are secondarily derived by semantic extension. As a matter of fact, a brief perusal of the categorial notions employed by Marchand to capture the semantic range of nouns in -dom, -hood and -ship reveals that all three groups of derivatives overlap only in one semantic sphere, namely that of ‘state, condition (of being)’. This is in accordance with Lehrer’s (2003: 225–228) analysis of the polysemy patterns of the three suffixes in question, where for all of them ‘state/condition’ is posited as the “general sense” from which shifts towards more concrete or other abstract meanings like ‘territory’ (e.g., dukedom), ‘collectivity’ (e.g., clerkdom) or ‘skill’ (e.g., penmanship) set off. If one would agree to accept the notion ‘state or condition’ as the essential semantic value of status nouns, nothing would prevent us from including deadjectival formations under the definition of this category, e.g., a good deal of the nouns in -ness. It would be difficult then, however, to draw a dividing line between nomina status and nomina qualitatis, and since the latter, unlike the former, are a well-established derivational category, there is a tacit consensus that the two should be kept apart by reserving the caption of status nouns for denominal derivatives only. It is at this point where the bottom-up approach meets its opposite in the guise of a pragmatic deliberation. A semasiological account of derivatives in -ness would never lead to a categorial distinction between, e.g., wetness meaning ‘the quality of being wet’ and wetness meaning ‘the state or condition of being wet’. Imposing a categorial difference onto the notions of ‘quality’ and ‘state or condition’ is as arbitrary as imposing such a difference onto the notions of ‘quality or state’ and ‘condition’, or on any other arrangement of these labels. It is the top-down approach to derivational categories that elevates such labelings from their ancillary and purely descriptive role to an ontological level of higher order. For any top-down categorial system to be enlightening, however, the choice of semantic labels has to be made with circumspection regarding the meanings of the labels themselves. Using ‘state’ and ‘status’ interchangeably, as if they were synonymous, has left the concept of nomina status poorly understood thus far, and, besides that, the rich polysemy of markers expressing status meaning has not contributed to clarification. This state of affairs is reflected also in introductory literature. For instance, in Haspelmath and Sims (2010: 87), the showcase example (actually the only example) for the category of status nouns is childhood. Under the assumption that a watertight specimen has been chosen by the authors of an introductory textbook, it is tempting to explore how this item is handled elsewhere in the literature. For Marchand (1969: 293), who declares that “the current meaning of derivatives [in -hood] is ‘status of −’”, childhood is to be classed among the nouns “with the nuance ‘time, period’ of the respective state”, together with babyhood, boyhood, girlhood. This means that childhood, displaying what Marchand calls a “nuance”, is not a representative of the pure ‘status’ meaning of a

73. Status nouns

1293

status noun. Since the semantic extension to a temporal meaning ‘period of being N’ is quite common for nouns like childhood, the equivalents of this noun in other languages are consequently not classified as status nouns in the few handbooks of word-formation where this category is considered explicitly. For Spanish, Rainer (1993) does not list infancia ‘childhood’ among the status nouns derived with the suffix -ia (such as comandancia [a military rank], gerencia ‘management, direction’, intendencia ‘directorship’), but as a temporal extension, together with various other extensions (Rainer 1993: 510– 511): “Ein Lebensalter bezeichnen adolescencia und infancia” [An age bracket is denoted by adolescencia and infancia]. Another Spanish word for ‘childhood’, niñez, is formed with a suffix that is not even listed by Rainer among those used for the derivation of status nouns in Spanish, its main function being the formation of deadjectival quality nouns (but note that niñez is denominal, from niño ‘child’), and here niñez appears in a particular group: “Eine Spezialität von -ez sind die Bezeichnungen von Altersstufen: madurez, niñez, vejez u. a.” [A peculiarity of -ez are the designations for age brackets: madurez, niñez, vejez and others] (Rainer 1993: 506). Thus, given the synonymy of infancia and niñez, if the latter is not a status noun, the former can hardly be subsumed under this heading. It is maybe for this reason that the Italian equivalent infanzia is not mentioned in Rainer’s account of “nomi di status” in Italian (Rainer 2004: 241–242), if not because of it being a lexicalized formation lacking a proper base (in the contemporary language the noun infante appears with the meaning ‘small child’ only in the formulaic expression il Divino Infante ‘Child Jesus’). Still, infanzia is akin to derivatives in -anza that constitute a pattern for status nouns derived from kinship terms, e.g., cuginanza ‘cousinhood’, and in Old Italian infanza was the current form. Notice also that the meaning of the German cognate of English childhood, viz., Kindheit, is restricted to the temporal aspect only, i.e. ‘the period of being a child’, the sense of ‘status’ being reflected here indirectly at best, while Kindschaft is monosemous for the status meaning as such, paralleling Vaterschaft ‘fatherhood’ and Mutterschaft ‘motherhood’ in this respect (with no collective extension as in the case of the English translations given in the glosses, but see below, section 5). This is interesting also with respect to an observation made by Aronoff and Cho (2001) regarding the sensitivity of suffixes to the kind of predicate level inherent to a derivational base: child can either be a stage-level predicate (as an age bracket) or an individual-level predicate (as a kinship term), and this difference can manifest itself in the selection of suffixes. On the other hand, as pointed out in article 135 on English word-formation (section 4.1.3), the semantic similarity of status suffixes can lead to “triplets or doublets with the same base, for example studenthood, studentdom, studentship, […] with no apparent difference in meaning” (cf. also Lieber 2010: 70–71). As can be seen, because of its entanglement with neighboring and partially overlapping meanings, the concept of ‘status’ does not lend itself easily to the establishment of a derivational category without some semantic refinement. It turns out that this refinement also facilitates the assessment of the category in terms of morphological patterning, to which we will turn in the next section.

4. Morphological patterning In view of the fact that a purely conceptual distinction between derived nouns with the meaning ‘condition or state of being A’ (where A is an adjective) and nouns with the

1294

VII. Semantics and pragmatics in word-formation II: Special cases

meaning ‘condition or state of being N’ (where N is a noun) is somewhat artificial, an additional semantic feature is advocated in order to establish status nouns as a derivational category sui generis. What makes a denominal abstract noun a status noun is the implication of, or reference to, a hierarchy or a classification of some kind. A status noun can thus be defined as an abstract noun denoting a position or range in a hierarchical system, in a classification, or in any ordered set of items. With respect to this requirement, Haspelmath and Sims’ (2010) type-word childhood would be a valid example for this category, since childhood is a range on the scale of age brackets, which can be seen as a hierarchy made up, in ascending order, of the sequence babyhood, toddlerhood, childhood, adulthood and old age (as a minimum; enlargements and refinements like prenatality, puberty, teenhood, boyhood or girlhood are of course possible). Notice that basically there is no need for the items on such a scale to be morphologically analogous. The notion of hierarchy underlying the definition of status noun is purely semantic, so the positions or ranges building up the hierarchy should in principle be expressible by means of any linguistic form, be it a simplex noun, a derived noun, a compound word or a phrase. For example, dictionaries provide no English noun in -hood that would convey the meaning of French vieillesse, Italian vecchiaia, Russian stárost’ or German Alter, simply because there is no simplex noun that could serve as a basis of derivation. For whatever reason, the existing substantivized old, as in young and old will learn together, seems not to qualify as a basis for a status noun *oldhood. But since -hood is applicable to adjectival bases as well (cf. falsehood, likelihood, etc.), *oldhood would not be ruled out in principle (in fact there are plenty of records on the Internet). It looks as if a paradigmatic effect caused by the semantic proximity of the items building up the hierarchy were strong enough to block a deadjectival derivative from joining the paradigm of denominal formations, at the same time allowing for a morphologically heterogeneous item, i.e. a noun phrase acting like a compound (old age), to fill one of the slots. However, considering the fact that the English equivalent of French vieillard, Italian vegliardo, Russian starík, German Greis, etc. is a bipartite lexeme, viz. old man, another resolution of the inconsistency appears possible. If in a case like English old age, where a morphologically analogous set of denominal derivatives in -hood contrasts with a constructionally deviant type, i.e. a phrase, a certain pressure of leveling were exerted on the deviant slot to be filled with an analogous form, old age could tend to be replaced by a formation analogous to the others in the hierarchy − old-manhood in this case. There seems to be no need to put an asterisk in front of this form in order to mark its non-existence, since it is already firmly established in language use (probably enhanced by the existence of manhood), yielding over ten million records in a quick-search on the Internet (carried out in September 2012). The same analogical pressure seems to keep the existing noun adultness from competing with adulthood in the lexical field of age brackets. The idiosyncrasy of semantic values assigned to items in this lexical field manifests itself, by the way, in the fact that although coming of age is opposed to growing old on the time scale of the life cycle of individuals, ageism is understood as the prejudice against old age and not against adulthood. On the one hand, the example adduced in the preceding paragraph suggests that status nouns do pattern, with all the entailments of patterning, like analogical pressure and alignment of form and function, at least at a language-specific level, in this case English. On the other hand, if the meaning of age brackets were a typical meaning of status nouns, the restriction of this category to denominal formations would have to be abandoned in

73. Status nouns

1295

a cross-linguistic perspective, since French jeunesse ‘youth’ and vieillesse ‘old age’ are deadjectival, so are Italian gioventù ‘youth’, adultezza ‘adulthood’ (the latter not in dictionaries, but recorded) and vecchiaia ‘old age’ (denominal derivation from substantivized adjectives jeune, giovane ‘young’, vecchio, vieux ‘old’ is possible, but improbable), and the same holds for Russian mólodost’ ‘youth’ and stárost’ ‘old age’. Also the morphological analogy mirroring the semantic adjacency of age brackets, as observed in English, is absent in other languages, e.g., in German, where the items on the scale are morphologically completely heterogeneous: Kindheit ‘childhood’ is denominal, Jugend ‘youth’ is derivationally opaque (originally deadjectival), Erwachsenenalter ‘adulthood’ is a compound, and Alter ‘old age’ in its turn is an originally deadjectival but derivationally opaque formation with no constructional parallels. It seems to be the case that the lexical field of age brackets is associated with “real” status nouns for three reasons. Firstly, the meanings of the items in this lexical field reflect a graded scale, thus fulfilling the definitional requirement for status nouns expressed at the beginning of this section. Secondly, some of the derivational bases are generic personal nouns, which is another important ingredient of statushood, the core of which is formed by nouns derived from nominal bases denoting ranks, titles, functions and the like, e.g., monkdom, directorship, or fatherhood. Thirdly, some of the terms for age brackets are derived with suffixes that are used for deriving status nouns. While the third criterion is plainly tautologic, the first two are essential, but not exhaustive, for a derivative to qualify as a status noun. A preliminary conclusion to be drawn from the discussion so far is that Haspelmath and Sims (2010) have chosen a rather problematic example for illustrating an equally problematic derivational type (notice that neither status nor status noun appears as a lemma in their glossary, although it is stated in a footnote [p. 87] that “the glossary gives definitions of the derivational meanings” mentioned in the text).

5. Semantic iridescence Age brackets are not the only borderline case revealing the problematic nature of the concept of status nouns as a derivational category. The tentative definition proposed at the beginning of this section hardly captures the derivative fatherhood just presented as a typical example of a status noun, since the status of being a father, based on biological facts from which juridic and social status follow, does not imply any hierarchy or classification, but is a matter of either being it or not. If the binary relation of class membership vs. non-membership in a class would count as a classification, then the definition of status noun would extend over all nominal bases as possible inputs, since for each conceivable entity a statement can be made about whether it has the status of being what it is called or not. If a sentence like The computerhood of my laptop is questionable because of software problems is acceptable, and there are pieces of evidence that it is (Gerald Duffett’s book The computerhood of God [Welsh Riviera Press, 2009] may count as one of many), then our discussion of status nouns ends up right where it started, i.e. with the distinction between status and state. One of the best-proven ways out of circular reasoning is to expand the dimensional setup of the respective topic. For the case at stake this means not to think of derivational

1296

VII. Semantics and pragmatics in word-formation II: Special cases

categories as if they were plates on a table where derivatives are placed like French fries, with some plates closer together than others and some food being untidily served, hanging over from one plate to an adjacent one. Dwelling a bit more on the culinary metaphor, it could be said that derivational categories of the kind at issue here are not so much the plates containing the food but rather the system of meals and courses behind the menu, in the sense that one can have French fries for breakfast, lunch or supper, depending on the time of the day or the time one got out of bed, just as one can have them as a starter, as a main course or for dessert, or all three at once, depending on the gastronomic (and economic) circumstances. This metaphor, which is ready to give way to a better one, is meant to point out that, in addition to the general fact that the meaning of a suffix may vary depending on the lexical meaning of the base, a given derivative can function as a noun denoting a state in one utterance and a status in another, depending on the context and the pragmatic setting. Under particular discourse-specific conditions, any state of being can be conceived as a status, which means that the essence of status nouns is not so much a matter of lexical classification but of semantic potential based on the latitude of derivational abstraction. This is in accordance with Baeskow’s (2010) account of English nouns in -ship, -hood and -dom, where the view is held, against proponents of so-called neo-constructionist models of grammar, that suffixes do convey conceptual information, but that the microstructure of the senses, captured by Baeskow in terms of Pustejovskian qualia-structures (Pustejovsky 1995), depends on, and develops under, specific contextual conditions. While for some derivatives the categorial information is provided by the base (e.g., status meaning if the base denotes a rank), for others it is claimed to come from the suffix (e.g., collective meaning). The arguments for the latter case might not appear fully convincing, since even the semantic divergence of doublets like kingdom vs. kingship or motherhood vs. motherdom, claimed by Baeskow to “exclusively result from the suffixes” (Baeskow 2010: 19), could as well be due to analogy modeled on derivatives where the respective meanings are provided by a derivational base. The two factors are not mutually exclusive, however, but may act synergetically. Our German example Kindheit vs. Kindschaft adduced at the end of section 3 above leads to a case in point, being one of the rare instances where the explicit coining of a new formation can be observed directly in a text: in a commentary to a psalm, Luther writes “als manschaft heiszt versamlung (gesamtheit) der menner, priesterschaft der priester, also sind kindschaft die ganz gemein, seine (gottes) söne und töchter” (italics in the source) [as manschaft means assembly (totality) of the men, priesterschaft of the priests, so are kindschaft the whole community, his (god’s) sons and daughters]. Rudolf Hildebrand, the author of the volume of the dictionary from which this quotation is taken (DWB, vol. 11, published 1873) comments on this innovation (which after all has not gained acceptance in the language community) that “bei dem mangel eines einfachen collectivs verdiente das pflege” [in view of the lack of a plain collective this would deserve attendance]. Notice that the base of the collective nouns in -schaft is in the singular (Mannschaft, not *Männerschaft), which speaks for assigning plural meaning to the suffix, there being no other source for the plural meaning of the derivative. Moreover, the idea of ‘child’ as such, in terms of stereotypic cognition, was probably closer to plurality, in Luther’s days, than that of ‘man’ or ‘priest’, since the normal state of affairs in a family was a plurality of children. Thence an additional motivation for the coining of collective kindschaft seems possible. Historically it can be shown that among the oldest formations in -schaft there

73. Status nouns

1297

are collectives where the plural meaning is inherent to the base (e.g., Old High German heriscaf ‘troops’, from heri ‘army’; kunniscaft ‘generation, descendants’, from kunni ‘kin, kinsfolk’). The rise of collective meaning, which was not inherent to Old High German scaf(t) as a free noun, was surely enhanced by such items. From this perspective it is not unexpected that there are relatively few examples of lexically established status nouns with a meaning strictly confined to this specification alone, even in languages where status meaning is expressed by a wealth of markers. For Spanish, Rainer (1993: 231) enumerates no less than seventeen suffixes deriving nouns with status meaning (see also article 145 on Spanish word-formation, section 4.1.1), none of which is monosemous except -uría, which can hardly be assigned the status of a suffix (due to lack of frequency it is not treated in Pharies 2002), there being only one derivative containing it (agregaduría ‘office or position of an attaché’). Another suffix with two instantiations only, -idumbre, already shows rich polysemy: one of the derivatives, (in)certidumbre ‘(un)certainty’, is a quality noun, and the semantic spectrum of the other one, servidumbre, embraces the senses ‘servants’ (as a collective), ‘attendance’, ‘serfdom’, ‘statute labor’, ‘bondage’, and ‘subserviency’. Italian has a similar amount of suffixes serving for the derivation of status nouns, but some of them are extremely marginal, with only one or two derivatives, and the majority of them deviating semantically from the notion ‘status’, so that the category of status nouns in Italian boils down to the only one productive type of formation, viz., -ato, as in cardinalato ‘title or office of a cardinal’. An inspection of the semantic range of Italian nouns in -ato reveals that the productive suffix is the one that expresses status meaning most consistently. Rainer (2004: 243–244) cites a wealth of neologisms in -ato, like praticantato ‘unpaid traineeship’ or cantautorato ‘singer-songwritership’, that fit the semantic scheme ‘condition or quality of being N’. However, the additional requirement for a noun to qualify as a status noun proper, i.e. the implicit reference to an ordered system or hierarchy, is hardly fulfilled by most of the formations. In Rainer’s words, the common denominator of status nouns is “che esprimono un determinato ruolo sociale inserito in una gerarchia o comunque classificazione” [that they express a certain social role in terms of a hierarchy or classification] (Rainer 2004: 241). If this requirement were applied rigorously to all examples, only few of the neologisms would remain within the categorial scope envisaged in the formulation just quoted. Therefore Rainer pleas for a more loose definition of what counts as ‘status’, implicitly suspending the criterion of hierarchy previously invoked. A formally more rigid but semantically indeterminate stance is taken by Dubois and Dubois-Charlier (1999: 201–205) in their account of status nouns in French. These authors refrain from making terminological use of the distinction between état ‘state’ and statut ‘status’ provided by the vocabulary of their language, establishing instead a category of “nominalisation de statut ou d’état” in a transformational manner, whereby derivational meaning results from the nominalisation of an underlying phrase. By this mechanism, denominal nouns are treated in the same way as nominalized verbs and all other types of derived nouns. Status meaning is thus assigned as if it were a syntactic category reducable to an adjective or an attributive noun. As with the cognate Italian formations in -ato and Spanish formations in -ado, status meaning is postulated as fundamental (“sens fondamental”) for French nouns in -at (e.g., bénévolat ‘volunteerdom’). This is a highly productive pattern, whereas formations in -age (e.g., esclavage ‘slavery’), -ion (e.g., cognation [same meaning as in English]), -ure (e.g., candidature [same

1298

VII. Semantics and pragmatics in word-formation II: Special cases

meaning as in English]), -ise (e.g., maîtrise ‘mastership’) and -ie (e.g., bigamie ‘bigamy’) are not productive for this meaning (notice that the suffixes -age and -ion are predominantly used in the derivation of derverbal nouns, e.g., bafouillage ‘stammering’, from bafouiller ‘to stammer’; captation ‘legacy hunting’, from capter ‘to obtain by devious means’. The transformational account of suffixal derivation offers a motivation for categorial meaning of derivatives within a limited set of logical formulae based on predicative relations (along schemes like Homer is the author of the Iliad → Homer’s authorship of the Iliad). However, all semantic values beyond the “fundamental sense” remain to be accounted for in terms of semantic extensions. For example, it is impossible to derive the meaning of French personnage ‘personage’ in any other way than along diachronic pathways of semantic change that display the fuzziness of derivational categories and their boundaries: Personnage was originally a typical status noun meaning roughly ‘office or position in the ecclesiastic hierarchy’, relying on the sense ‘dignitary’ of Old French persone that is still reflected in English parson. While English parsonage acquired locative meaning, French personnage developed a sense that can only tentatively be captured in terms of a mixture of genericity and typicality that is quite unique to this particular derivative. In the final section, only the most common semantic extensions irradiating from the core meaning of ‘status’ will be sketched briefly.

6. Meaning, senses and diachrony Few languages seem to have dedicated suffixes expressing status meaning exclusively, e.g., Botlikh -łi (see article 204 on Botlikh, section 4.1.1). The normal situation is that of several suffixes with wider meaning but intersecting in this semantic sphere, as is the case in Romance, Germanic and Slavic languages, or of one general suffix expressing various senses, among which ‘status’, as is the case with Turkish -lik/-lık/-lük/-luk (see article 184 on Turkish, section 4.1.1) and its cognates in other Turkic languages (see section 4.1.1 in articles 188–190 on Gagauz, Karaim and Chuvash), or Hungarian -ság/ -ség (see article 181 on Hungarian, section 4.1). Suffixes containing ‘status’ as one of their semantic values are very often multifunctional in the sense that they can attach to nominal as well as to adjectival bases, as is the case, e.g., with Basque -tasun (see article 182 on Basque, section 4.1.1), or to nominal as well as to verbal bases, e.g., Kabardian -ʁa (see article 194 on Kabardian, section 4.1), or are free from any selectional restrictions regarding the category of the base, as is the case with the Udi suffix -luɣ (see article 197 on Udi, section 4.1). The diachronic source for the semantic widening of derivational markers is mostly metonymic extension. The principal pathways leading to this dispersal will briefly be sketched in the remainder of this article. The extension from status meaning to temporal meaning has already been discussed in section 4 with respect to nouns denoting age brackets. It can be stated schematically as {condition or quality of being N}N' > {duration of being N'}N'' whereby N is a personal noun, N' is a status noun and N'' a nomen temporis.

73. Status nouns

1299

This extension, besides the age brackets issue already mentioned, is motivated in the first place by status nouns referring to political and administrative offices. Already in Latin a noun like consulatus ‘function, office of a consul’ could easily be reinterpreted as a period of time, i.e. ‘term of office of a consul’, since such offices were assigned with limited duration (one year in the case of the consulate). The sarcastic passage in Cicero (Epistulae ad familiares, VII, 30) “suo toto consulatu somnum non viderit” [he would not find sleep during his whole consulate], said of a certain Caninius who held the office for one day only, is a clear instance of extension of the status meaning to a temporal one. Once established as a semantic value of nouns in -atus and their descendants in Romance languages, temporality could become the primary semantic value of formations like French quinquennat ‘period of five years’ or septennat ‘period of seven years’, which besides the time span may also denote the office held for that particular period, so that the status meaning comes in through the back door again. The example of consulate leads to another principal pathway of metonymy involving status meaning. Since the fulfilment of an office is usually connected to certain locations or premises, the transition from status meaning to locative meaning is likewise motivated by nouns denoting political, administrative, military or ecclesiastic functions. Notice that the locative meaning of office is itself an instance of metonymic extension from a noun whose primary meaning was that of an action (Latin officium ‘service, duty, business’, from the verbal compound *opificium, from opus ‘work’ and facere ‘to make’). Schematically, the transition can be formulated as follows: {condition or quality of being N}N' > {place where the fulfilment of N' is being carried out}N'' whereby N' is a status noun and N'' a nomen loci. In the case of consulate, the locative meaning cannot be traced back to Latin, but arose in the late Middle Ages, when the office of consul as the plenipotentiary of a government was established under the label of the ancient Roman title. Since the premises where a consul held his office were extraterritorial in the respective states, the semantic value of ‘place’ could easily come into play, leading to a locative reading of the derivative. By this it is not implied that consulatus was the first status noun to adopt a locative meaning. The metonymic vicinity of ‘office or function’ with the place where the holder of the office or function resides is too obvious as to erect any obstacles for the understanding of how the diachronic extension from the abstract concept to the concrete implementation might have come about. Locative extension is of course not restricted to status nouns denoting offices or functions. The case of English neighbourhood can be cited as an instance where the locative meaning, inherent to the base, has become the primary one and is even spreading to German through semantic borrowing: until recently, German Nachbarschaft, besides its abstract meaning, was used only as a personal collective noun and not in the sense of a district or an area, as it is nowadays under the influence of English. In French, the twofold continuation of Latin -ātus as vernacular -é and “learned” -at has led to lexical doublets like locative évêché ‘bishopric, diocese; bishop’s seat, cathedral city’ besides the status and collective noun episcopat ‘episcopate, episcopacy’, but most words in -at lack a respective counterpart).

1300

VII. Semantics and pragmatics in word-formation II: Special cases

Significantly, the example of consulate is not a case for this third extension pattern, viz., the transition from status to collective meaning, since the number of holders of the office of a consul was limited to two in Ancient Rome, and the plenipotentiaries bearing the title of consul in modern times are singletons exclusively. Collective meaning arises, however, in many instances of status nouns where the holders of the respective rank or dignity form a body, like bishops, hence the collective meaning of episcopate (or episcopacy). An early impulse for the extension towards collectivity must have been the Roman institution of the triumviratus ‘triumvirate’, where number is already inherent in the status noun (which is a compound made up of the numeral trium ‘of three’ and vir ‘man’). The extension can be stated as {condition or quality of being N}N' > {group of persons holding or embodying N'}N'' whereby N' is a status noun and N'' a nomen collectivum. For some Latin nouns in -atus, collective meaning is the primary one, with status meaning sometimes even unattested, e.g., comitatus ‘escort, company’, equitatus ‘cavalry’, peditatus ‘infantry’. This might have been enhanced by the presence of a verb underlying the nominal base with a regularly formed participle in -atus: equitatus could also be read as ‘mounted’ (from equitare ‘to ride a horse’) and as an attribute be associated with a body of troops. But in view of the fact that the derivational bases of status nouns are mostly nouns denoting persons, a shift from abstract status to concrete collective plural meaning need not be supported by formal momenta: the same shift to collective meaning occurred in English (e.g., playerdom, brotherhood, membership) and German (e.g., Christentum ‘christianity’, Menschheit ‘humanity’, Dienerschaft ‘domestics’) without any motivation except semantic contiguity between class meaning and the fact that classes are made up of a plurality of individuals. Once a collective reading of nouns in -atus was firmly established, the pattern could of course be analogically extended to nominal bases without intermediary status meaning. Thus in Middle Latin, from syndicus ‘syndic’ (from the Greek word for ‘public advocate’) syndicatus ‘syndicate’ was derived as a collective noun which at no point in time seems to have expressed the sense ‘condition or quality of being a syndic’. In languages with a variety of suffixes deriving status nouns, metonymic extension can lead to functional differentiation. For example, German Beamtenschaft expresses the collective sense (‘the body or class of civil servants’), while Beamtentum continues the status meaning (‘officialdom’, neutral, not pejorative as in English). The reverse holds in Dutch, where “-dom is used to form nouns that denote a group of people” (cf. article 136 on Dutch, section 4.1). Extension can of course be blocked by existing formations or underived lexemes. In French, for example, the notion of priesthood is expressed by two monosemous status nouns, prêtrise and sacerdoce, denoting only the office or dignity of a priest, while the collective meaning is covered by clergé. A fourth extension pattern frequently attested for status nouns is the shift to a sense of ‘activity’, as in censorship, sponsorship, leadership. Significantly, most of the nouns that have this reading as part of their polysemy are derived from agent nouns, also in German, e.g., Mittäterschaft ‘complicity’, but items like Spanish abogacía ‘legal profession’ show that agentivity of the base form is not a precondition for the derivative to denote an activity (abogado goes back to a Latin noun in -atus, but here this is the

73. Status nouns

1301

passive participle ending of the verb advocare ‘to call upon’, not the suffix deriving denominal status nouns). On the other hand, not every status noun derived from an agent noun necessarily develops a sense of activity, as can be seen in the case of German Dienerschaft cited above, where the collective meaning is so prevalent that the possible reading ‘activity of N’ is not actuated. Notice also that in article 134 on German, German formations in -erei are classed as status nouns (section 4.1.1), although the example provided is analyzed as deverbal (Paktiererei ‘deal making’ ← paktieren ‘to strike a deal’). In the interest of space, only two of the more particular senses hived off from the original status meaning shall be mentioned here as concluding specimens. One is the extension towards a meaning of ‘attitude, style, conduct’, as manifested in formations like duncedom (assigned the status of a “type-word” by Marchand 1969: 263; OED: “The domain of dunces; dunces collectively; a dunce’s condition or character”) or German Dandytum ‘dandyism’. It is possible, however, that this sense developed out of the collective meaning and is thus only secondarily related to the status meaning as such. In French, formations in -tude with a base denoting a class of humans are closely related to this semantic type (cf. Dal and Namer 2010). The other one is due to the fact that semantic extension is strongly determined by the meaning of the derivational base. For example, if the base is a noun implying a social relationship, then the derivative may denote not only ‘the condition or quality of being N’ but the relationship as such, as in the case of Englisch friendship or its German equivalent Freundschaft. Depending on the number of bases with similar semantic implications, sets of formations can easily arise, establishing new semantic values of the respective suffix, as in German Kameradschaft ‘camaraderie’, Partnerschaft ‘partnership’, Gefolgschaft ‘retinue’, Gegnerschaft ‘antagonism’, Feindschaft ‘enmity’ and so on.

7. Final remarks As mentioned in the introduction, the category of status nouns is not firmly established in the systematics of derivational morphology. Therefore the present treatment is only a tentative approach, with the restricted objective of sizing up the outlines of the conceptual bits and pieces that need to be considered in the study of this type of derivatives. Some relevant derivational markers, especially English -hood, -dom and -ship, are wellstudied both synchronically and diachronically (cf., e.g., the monograph by Trips 2009), but much work has still to be done in order to complete the picture of diachronic developments in this semantic field with data from a broader range of languages. Crosslinguistic study of status nouns, which still remains a desideratum, will be greatly enhanced by the material collected in chapter XVI of this handbook.

8. References Aronoff, Mark and Sungeun Cho 2001 The semantics of -ship suffixation. Linguistic Inquiry 32(1): 167–173.

1302

VII. Semantics and pragmatics in word-formation II: Special cases

Baeskow, Heike 2010 His lordship’s -ship and the king of golfdom: Against a purely functional analysis of suffixhood. Word Structure 3(1): 1–30. Bauer, Laurie 1983 English Word-Formation. Cambridge: Cambridge University Press. Bauer, Laurie 2001 Morphological Productivity. Cambridge: Cambridge University Press. Bech, Gunnar 1983 Studien über das deutsche Verbum infinitum. Tübingen: Niemeyer. Bendel, Christiane 2006 Baskische Grammatik. Hamburg: Buske. Benfey, Theodor 1844 Über das Verhältnis der ägyptischen Sprache zum semitischen Sprachstamm. Leipzig: Brockhaus. Benfey, Theodor 1866 Review of: Graziadio Isaia Ascoli 1865 Studj Ario-Semitici. Parte seconda. Memorie del Reale Istituto Lombardo, classe di lettere, e scienze morali e politiche, volume 10, anno 1865/1867, fasc. III: 13–36. Göttingische gelehrte Anzeigen Jg. 1866, 1. Band, 8. Stück: 281–293. Bopp, Franz 1833 Vergleichende Grammatik des Sanskrit, Zend, Griechischen, Lateinischen, Litthauischen, Gothischen und Deutschen. Berlin: Dümmler. Brinton, Laurel J. and Minoji Akimoto 1999 Collocational and Idiomatic Aspects of Composite Predicates in the History of English. Amsterdam/Philadelphia: Benjamins. Dal, Georgette and Fiammetta Namer 2010 French property nouns based on toponyms and ethnic adjectives: A case of base variation. In: Franz Rainer, Wolfgang U. Dressler, Dieter Kastovsky and Hans Christian Luschützky (eds.), Variation and Change in Morphology. Selected papers from the 13 th International Morphology Meeting, Vienna, February 2008, 53–73. Amsterdam/Philadelphia: Benjamins. Dal, Georgette and Fiammetta Namer 2012 When an emerging form drives another word’s meaning: The case of French abstract nouns. Paper read at the 15 th International Morphology Meeting, Vienna, February 9− 12, 2012. Dietz, Klaus 2007 Denominale Abstraktbildungen des Altenglischen: Die Wortbildung der Abstrakta auf -dōm, -hād, -lāc, -rǣden, -sceaft, -stæf und -wist und ihrer Entsprechungen im Althochdeutschen und im Altnordischen. In: Hans Fix (ed.), Beiträge zur Morphologie. Germanisch, Baltisch, Ostseefinnisch, 97–172. Odense: University Press of Southern Denmark. Donohue, Mark and Søren Wichmann (eds.) 2008 The Typology of Semantic Alignment. Oxford: Oxford University Press. Dubois, Jean and Françoise Dubois-Charlier 1999 La dérivation suffixale en français. Paris: Nathan. DWB = Deutsches Wörterbuch von Jacob und Wilhelm Grimm. 16 Vol. 1854−1971. Leipzig: Hirzel. Friedländer, Marianne 1992 Lehrbuch des Malinke. Leipzig: Langenscheidt/Verlag Enzyklopädie. Grimm, Jacob 1837 Deutsche Grammatik. Vierter Theil. Göttingen: Dieterichsche Buchhandlung. Haspelmath, Martin and Andrea D. Sims 2010 Understanding Morphology. 2nd ed. London: Hodder Education.

73. Status nouns

1303

Hennings, Thordis 2012 Einführung in das Mittelhochdeutsche. 3rd ed. Berlin/Boston: de Gruyter. Hunston, Susan 2011 Corpus Approaches to Evaluation. Phraseology and evaluative language. New York/ Abingdon: Routledge. Kluge, Friedrich 1899 Nominale Stammbildungslehre der altgermanischen Dialekte. 2 nd ed. Halle/S.: Niemeyer. Lazar, Oscar 1975 The Formation of Abstract Nouns in the Uralic Languages. Stockholm: Almqvist & Wiksell. Lehrer, Adrienne 2003 Polysemy in derivational affixes. In: Brigitte Nerlich, Zazie Todd, Vimala Herman and David D. Clarke (eds.), Polysemy. Flexible patterns of meaning in mind and language, 217–232. Berlin/New York: Mouton de Gruyter. Lieber, Rochelle 2010 Towards an OT morphosemantics: The case of -hood, -dom, and -ship. In: Susan Olsen (ed.), New Impulses in Word-Formation, 61−79. Hamburg: Buske. Lu¨dtke, Jens 1978 Pra¨dikative Nominalisierungen mit Suffixen im Franzo¨sischen, Katalanischen und Spanischen. Tu¨bingen: Niemeyer. Lu¨dke, Jens 1998 Romanische Abstrakta bzw. Pra¨dikatsnominalisierungen mit Suffixen im Franzo¨sischen, Katalanischen und Spanischen. In: Udo L. Figge, Franz-Josef Klein and Annette Martinez Moreno (eds.), Grammatische Strukturen und grammatischer Wandel im Franzo¨sischen. Festschrift fu¨r Klaus Hunnius zum 65. Geburtstag, 359−381. Bonn: Romanistischer Verlag. Lyons, John 1977 Semantics. 2 Vol. Cambridge: Cambridge University Press. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2 nd ed. München: Beck. Meyer-Lu¨bke, Wilhelm 1890 Italienische Grammatik. Leipzig: Reisland. Meyer-Lu¨bke, Wilhelm 1921 Historische Grammatik der franzo¨sischen Sprache. Zweiter Teil: Wortbildungslehre. Heidelberg: Winter. Migirina, Nina Josifovna 1980 Tipy nominacij dlja oboznačenija statusov lica v sovremennom russkom jazyke. Kišinev: Štiinca. Mühleisen, Susanne 2010 Heterogeneity in Word-Formation Patterns. A Corpus-Based Analysis of Suffixation with -ee and its Productivity in English. Amsterdam/Philadelphia: Benjamins. Napol’skich, Vladimir 2003 Review of: Eberhard Winkler 2001 Udmurt. München: LINCOM Europa. Linguistica Uralica 39(4): 288–304. OED = Oxford English Dictionary. http://www.oed.com [last access 1 Jan 2015]. Pharies, David 2002 Diccionario etimológico de los sufijos españoles y de otros elementos finales. Madrid: Gredos.

1304

VII. Semantics and pragmatics in word-formation II: Special cases

Pustejovsky, James 1995 The Generative Lexicon. Cambridge, MA: MIT Press. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rainer, Franz 2004 Nomi di status. In: Franz Rainer and Maria Grossmann (eds.), La formazione delle parole in italiano, 241–244. Tübingen: Niemeyer. Štekauer, Pavol 2005 Onomasiological approach to word-formation. In: Pavol Štekauer and Rochelle Lieber (eds), Handbook of Word-Formation, 207–232. Dordrecht: Springer. Štekauer, Pavol, Salvador Valera and Lívia Körtvélyessy 2012 Word-formation in the World’s Languages. A typological survey. Cambridge: Cambridge University Press. Trips, Carola 2009 Lexical Semantics and Diachronic Morphology. The development of -hood, -dom and -ship in the history of English. Tübingen: Niemeyer. Verhoeven, Elisabeth 2007 Experiential Constructions in Yucatec Maya. Typologically based analysis of a functional domain in a Mayan language. Amsterdam/Philadelphia: Benjamins.

Hans Christian Luschützky, Vienna (Austria)

74. Agent and instrument nouns 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Introduction Terminological observations “Agent noun” or “subject name” ? The polysemy of agent nouns Deverbal and denominal agent nouns Semantic subtypes of agent and instrument nouns “Agent” in portmanteau morphs Valency and syntax of deverbal agent and instrument nouns The origin of agentive and instrumental patterns References

Abstract Agent nouns are the second most frequent semantic category realized by means of wordformation in the languages of the world. The present article provides an overview of the central problems linked to this category, including terminology, language-internal and cross-linguistic variability of the category, its relation to instrument nouns, its valency and syntax, as well as its origin.

74. Agent and instrument nouns

1305

1. Introduction The cognitive category “agent” is of such fundamental importance in the life of human beings that it must be expressed somehow in any natural language. According to a typological survey conducted on the basis of 42 genetically unrelated languages (Bauer 2002: 40), agent nouns, in fact, turn out to be one of the most frequent categories realized morphologically, second only to action nouns: 24 of the languages of the sample, i.e. a respectable 57 %, were found to possess at least one agentive word-formation pattern (cf. also Baker and Vinokurova 2009: 544). The second conceptual category which will be dealt with in this article, viz. instrument nouns sensu lato, i.e. including means of action in general, is attested in 12 of the languages of the sample, occupying the 6 th rank among all derivational categories. The category “instrument”, by the way, must also be deeply rooted in our conceptual system, since even some animals are able to manipulate tools.

2. Terminological observations The two categories at issue are generally referred to as agent nouns (or agentive nouns) and instrument nouns (or instrumental nouns). The related word-forming processes are called agent(ive) and instrument(al) nominalization, terms which are also occasionally used in a result reading as synonyms of agent/instrument noun. The terms agent noun and instrument noun, attested in English since the 19th century, were loan translations of the corresponding Latin terms nomen agentis (or actoris) and nomen instrumenti, which had been created by Orientalists (cf. “Participium activum, seu nomen Agentis”, Antonius ab Aquila Arabicae linguae novae et methodicae institutiones. Rome 1650, Index, s. v., “Nomina Instrumenti, quo fit actio”, Vassalli, M. A. Mylsen phoenico-punicum. Rome 1791, p. 110). French scholars also resorted to loan translations, coining nom d’agent and nom d’instrument, while Germans adopted the Neo-Latin terminology directly. The term agent noun was first applied to patterns forming prototypical agents. Scholars, however, soon realized that the same patterns were also used, in many languages, for expressions which could not be classified as agentive in any natural manner. Faced with this situation, some opted for extending the range of concepts covered by the term agent noun beyond prototypical agents, while others maintained the term’s literal meaning, introducing instead new terminology for the non-canonical cases. Unfortunately, the rather confused terminological situation as described for the beginning of the 20th century in Kärre (1915: 5−17) has not yet been overcome.

3. “Agent noun” or “subject name”? The term agent noun, if taken literally, obviously implies that the noun in question refers to an agent. But what is an agent? This notion has been the subject of much discussion in the linguistic literature. For our present purposes, it may suffice to follow Fradin

1306

VII. Semantics and pragmatics in word-formation II: Special cases

(2005: 162), who establishes the following scale of agentivity: The referent of the subject of the base verb of the agent noun (1) “causally affects other participants” (e.g., a killer), (2) is “volitionally involved in the event” (e.g., a swimmer), (3) “has a notion or perception of other participants in the event or state” (e.g., a believer), (4) “possesses another participant” (e.g., a house owner) or (5) at least “effectuates or performs the event” (e.g., a squinter). Only the first two cases constitute what Fradin calls “strong agents”, i.e. agents in the strict sense of the word. The vast majority of formations classified as agent nouns in Indo-European languages are strong agents in this sense. Latin (Torrego 1996), for example, remained quite close to the agentive core as far as the suffix -tor is concerned. The same is still true for Romance nouns which descend from this Latin suffix, though not for the corresponding adjectives (Rainer 1993: 447 for Spanish; Bisetto 1995: 51−54 for Italian; Fradin 2005 for French). The examples situated lowest on the agentivity scale which I have been able to find for Spanish -dor were conocedor ‘expert’ (← conocer ‘to know’; level 3) and tenedor (de acciones) ‘(stock) holder’ (← tener ‘to hold’; level 4). The second most important agentive suffix of Spanish, viz. -nte, which goes back to the Latin present participle, also occasionally occurs with strong agents (cf. asaltante ‘attacker’ ← asaltar ‘to attack’), but in general seems to have a greater affinity to low-agentivity verbs (cf. amante ‘lover’ ← amar ‘to love’, oyente ‘hearer’ ← oír ‘to hear’, habitante ‘inhabitant’ ← habitar ‘to inhabit’, descendiente ‘descendant’ ← descender ‘to descend’, etc.). With respect to Dutch, it has been pointed out that even verbs whose subject corresponds to a theme argument can occasionally be nominalized with -er: uitloper ‘offshoot’ (← uitlopen ‘to sprout’), uitvaller ‘dropout’ (← uitvallen ‘to drop out’), etc. In order to be able to give a unitary definition of Dutch nouns in -er, Booij (1986: 507) therefore proposed to speak of “subject names” instead, “because the basic effect of the suffix -er is that it binds whatever θ-role is linked to the subject position of the base verb”. Basically the same idea was also put forward in the mid-eighties by American scholars with respect to English -er, as Levin and Rappaport (1988: 1068) recall (in reality, the idea of linking agent nouns and the notion “subject” has a long tradition). These authors adduce as an argument in favour of the subject nominalization hypothesis further verbs with non-agentive subjects which nevertheless license nouns in -er: The new gadget opens cans / opener; This book sells well / bestseller ; This lottery ticket scratches / scratcher; This chicken broils well / broiler. The discussion concerning the subject nominalization hypothesis has focussed almost exclusively on English and Dutch -er (but cf. also Laca 1986 on Spanish, as well as Szigeti 2002 on German and Hungarian). It could certainly be worthwhile to broaden the perspective (cf., for a first overview, Baker and Vinokurova 2009: 549−552). In many non-European languages properties and states are also expressed by means of verbs. Now, one observes that such stative verbs normally form the corresponding personal noun in the same way as do action verbs. In Standard Malay, for example, the prefix pəN- is not only attached to action verbs (cf. pəmantu ‘assistant’ ← bantu ‘to help’), but also to stative verbs such as takut ‘to be afraid’ or malu ‘to be shy’, forming designations for persons with these characteristics (cf. pənakut ‘coward’, pəmalu ‘modest person’). Another type of verb which normally does not enter into agentive patterns in Indo-European is constituted by change-of-state verbs. Such verbs are, however, licensed in Pipil (cf. kukuyani ‘sick person’ ← kukuya- ‘to get sick’, mikini ‘dead person’ ←

74. Agent and instrument nouns

1307

miki- ‘to die’). The suffix used is the normal agentive suffix of that language (cf. takwi:kani ‘singer’ ← takwi:ka- ‘to sing’). The subject nominalization hypothesis has not been welcomed in all quarters, especially among cognitive linguists. Implicit in this critique is the assumption, which is not obviously true, that the proponents of the subject nominalization hypothesis claimed that it could account for all uses of -er. Ryder (1991: 299), for example, raised the following two objections: “[I]n addition to their failure to account for the complete range of present data, none of them [sc. accounts based on argument structure; F. R.] provides motivations for the expansion in the use of these expressions since the Old English period.” The first of these objections alludes to the fact that some types of formations, for example non-automatic instruments such as poker, terms of clothing such as waders, locative nouns such as diner, or causative event nouns such as laugher ‘game which causes spectators to laugh’ are not covered by the subject nominalization hypothesis, since the referents cannot act as a subject of the base verb (in the relevant sense!): *the poker pokes the fire, *the waders wade in the water, *the diner dines, *the laugher laughs. Ryder’s second remark aptly highlights the fact that the subject nominalization hypothesis also fails to provide insight into the diachronic processes by which the amazing array of uses of Modern English -er developed out of the almost exclusively agentive core of Old English (Kastovsky 1971). Admittedly, the subject nominalization hypothesis was not designed to account for these diachronic processes, but one should nevertheless expect a synchronic theory to be interpretable in a plausible manner as the last stage of diachrony. Cognitive accounts see the polysemy of -er as the result of a series of conceptual shifts and reanalyses, which are still reflected in synchrony as metaphoric or metonymic ties between single uses, though the relationship between synchrony and diachrony, of course, need not be one of isomorphism. I will come back to this issue in section 4. The subject nominalization hypothesis becomes even less attractive if, as Ryder (1999: 277) does, denominal and other formations in -er are also taken into account: birder ‘person who watches birds’, Fullbrighter ‘person receiving a Fullbright scholarship’, New Englander ‘person from New England’, upper ‘event producing positive feeling in participants’, overnighter ‘luggage pieces intended for overnight trips’, etc. Ryder’s solution consists in treating formations in -er in analogy to nominal compounds. Under her analysis, the suffix -er only indicates “that the whole word is a noun” (p. 278). The concrete reading of a derivative is said to be inferable from the interaction of this general meaning of the suffix with the meaning of the base and the wider context. To give just one example of Ryder’s strategy: The absence − in English, but apparently not in Dutch or Pipil, as we have seen! − of derivatives in -er from unaccusative verbs such as disappear or die is attributed by Ryder to the fact that nouns prototypically refer to referents stable across time, which she thinks sufficiently explains their preference for habitual and durative verbs. This is not the place for a detailed evaluation of cognitive accounts of affix polysemy of the type proposed by Ryder (for a cognitive analysis similar in spirit cf. Panther and Thornburg 2002). I will come back to the question in the next section, where it will be suggested that more thorough analyses of the diachronic development of polysemous patterns might be helpful in this connection.

1308

VII. Semantics and pragmatics in word-formation II: Special cases

4. The polysemy of agent nouns It has long been observed that the same pattern is often used in languages for deriving agent and instrument nouns. According to Bauer (2002: 43), 3 out of 10 languages of his sample which have both agent and instrument nouns, i.e. 30 %, use the same pattern. In a broader typological survey based on 62 languages, Luschützky and Rainer (2011) have found that this is even the case for 47 %. Starting with Meyer-Lübke (1890: § 498), this agent-instrument syncretism has traditionally been explained as the result of a conceptual shift − metaphorical for Meyer-Lübke, metonymical for others − from agent to instrument. Dressler’s (1986) and Booij’s (1986) extension schemes, for instance, can be considered as modern versions of Meyer-Lübke’s idea. Romance languages constitute an interesting test case in this respect, since Latin -tor was uniquely agentive, while its Romance descendants all also have an instrumental (and sometimes locative) meaning. The conceptual-shift story has always been taken for granted also with respect to Romance. In Rainer (2011), however, I have been able to show that no conceptual shift has ever taken place on the way to Romance. In Catalan and Occitan, as well as in a series of Italian and French dialects, the instrumental and locative meanings of the descendant of -tor were simply due to a conflation of -tor with the Latin instrumental-locative suffix -torium. The dialects of Florence and Paris did not conflate these two Latin suffixes, which is why Standard Italian and Standard French continue to distinguish them as -tore/-toio and -eur/-oir respectively. The reason why Modern Italian and French nevertheless have lots of instrumental derivatives in -tore and -eur is that these entered the language later on as the result of borrowing from English (cf. Engl. ventilator > It. ventilatore, Fr. ventilateur) or of ellipsis (of the type It. apparecchio misuratore ‘measuring instrument’ > misuratore ‘id.’), especially after 1750. Spanish instrument nouns in -dor are mostly due to these same two sources, but this language had also borrowed instrument and place nouns in -dor from Catalan (and Occitan) beginning with the Middle Ages (note that the regular outcome of Latin -torium in Spanish is -dero). This is why we find locative nouns in -dor in Spanish, but not in Standard Italian or Standard French. It would be worthwhile to investigate to what extent the polysemy of agent nouns in other languages is amenable to similar non-conceptual causes (i.e. homonymization as a consequence of sound change, borrowing, ellipsis). Müller (2011), for example, argues that the agent-instrument polysemy of German -er is due to the conflation of Latin -arius and -arium. Couldn’t this explanation be extendible to other Germanic languages? In the case of English, one would furthermore have to take into account the undeniable influence of French on Middle English, when the French instrumental-locative suffix -oir, which in the Anglo-Norman dialect had been conflated with -o(u)r (= French -eur), was also adapted as -er (cf. Fr. comptoir > Engl. counter); cf. Luschützky and Rainer (2013: 1350−1351). Thorough diachronic investigations could turn out to be a healthy antidote against cognitive analyses as the ones mentioned at the beginning of this section, which often are hardly more than diachronic guesswork based on synchronic data. It remains to be seen whether we need mechanisms of conceptual extension at all, and if such is the case, under what conditions exactly they operate.

74. Agent and instrument nouns

1309

5. Deverbal and denominal agent nouns The term agent noun is normally applied to deverbal nouns only, keeping denominal (or deadjectival) nouns separate under labels such as personal noun. Some linguists, however, also include denominal nouns in the category of agent nouns (cf. Ryder, as well as Panther and Thornburg, in section 3). Both positions can be defended with good arguments, depending on the goals of one’s investigation. At any rate, agent nouns and agentive denominal personal nouns are intimately linked. Both categories in fact designate a participant in a more or less complex scenario. In the case of a prototypical action scenario, we have as central ingredients the action itself, an actor, a patient (the person or thing affected or created), often an instrument with which the action is carried out, and a general setting in space and time. The purpose of an agent noun is to designate the actor of such a scenario. Depending on the saliency of the other ingredients of the scenario, the base chosen for the agent noun will be a verb or a noun. If the action itself is considered most salient, we normally get a verbal base (cf. hunter ← hunt), though some languages prefer to represent the action in form of an action noun (cf. Bashkort zagovorsy ‘conspirator’ ← zagovor ‘conspiracy’; YasinBurushaski daróskuin ‘hunter’ ← darú ‘hunting’; Miya bá kír ‘thief ’ ← ákir ‘theft’; residually in German agent nouns such as Wächter ‘guardian’ ← Wacht ‘guard’ ← wachen ‘to guard’). Alternatively, one can also opt for one of the other participants, choosing as a base, for example, the object affected (cf. Spanish carterista ‘pickpocket’ ← cartera ‘wallet’) or the instrument (cf. Bashkort baltasy ‘joiner’ ← balta ‘axe’). In a language where such an option is available, one could also choose to resort to a compound expressing both the action and one of the non-agentive participants (cf. cattlestealer). Languages often have special patterns for denominal and deverbal agent nouns, but sometimes one and the same pattern can be used for both purposes, as is the case with English -er. The intimate relationship between deverbal and denominal agent nouns is also apparent from the fact that, in diachrony, the one has often been derived from the other through reanalysis. The reanalysis of German -er, which had been borrowed from denominal Latin -arius, as a deverbal suffix must have taken place with nouns flanked by homonymous noun and verb stems (cf. Old High German helfari ‘helper’ ← helfa ‘help’ or helfan ‘to help’). In the case of the Ancient Greek denominal suffix -istes, the ancestor of English -ist, on the contrary, it was the denominal use which was due to the reanalysis of deverbal agent nouns (cf. kithara → kithar-izo ‘to play the kithara’ → kithar-is-tes ‘kithara player’, then: kithara → kithar-istes). Instrument nouns, on the contrary, seem to be overwhelmingly deverbal (but cf. examples such as Spanish paellera ‘paella pan’ ← paella ‘id.’, etc.).

6. Semantic subtypes of agent and instrument nouns When talking about degrees of agentivity in section 3, we have already seen that deverbal agent nouns do not constitute a uniform category from the point of view of that scale, and sometimes are not even agentive at all. But this is not the only conceptual dimension along which agent nouns may differ.

1310

VII. Semantics and pragmatics in word-formation II: Special cases

One further dimension concerns aspect. The referent of an agent noun may carry out the action either habitually, be it as a profession (cf. hunter) or as a habit (cf. [heavy] smoker), or on a single occasion (cf. The gapers standing around burst into laughter). As we can deduce from these examples, English -er is able to cover all three shades of meaning. Many languages, however, use different patterns for the occasional and the habitual reading. In Nganasan (Wagner-Nagy 2001: 142), for example, the occasional reading ‘the one hunting (over there)’ is expressed by the nominalized present participle basutuə (← basud’a ‘to hunt’), while the name for the professional hunter, basu’’śi, is derived with the agentive suffix -’’śi. Tübatulabal (Voegelin 1935: 100) has the same semantic distinction, but uses two specialized suffixes, -(i)bï˙l (habitual) and -(a)pï˙l (occasional). Where a formal distinction is made at all, the divide, like in Nganasan and Tübatulabal, is normally between the habitual and the occasional reading. Benveniste (1948), on the contrary, argued that in the oldest stages of Indo-European the habit reading had to be conflated with the occasional reading, while the professional reading coincided, at the level of the language system, with the instrumental reading (cf. Balles and Lühr 2005 for an assessment of this dubious hypothesis). Nganasan even has a third agentive pattern, which expresses the additional feature that the agent enjoys carrying out the action (cf. basugutə ‘(one) who enjoys hunting’). This pattern, however, is more frequent in attributive than nominal function. Another similar feature which often occurs in combination with agentive patterns is that of excess. Latin, for example, besides the neutral agentive suffix -tor, had a special suffix, viz. -o, -onis, for this excess reading (cf. bibo ‘drunkard’ ← bibere ‘to drink’). A fourth dimension relevant for the formation of agent nouns in some languages is tense. Even English -er, by the way, has a temporal component in examples such as the winner of the last championship, definable as ‘the person who won the last championship’. But this feature of anteriority or perfectivity is entirely dependent on the aktionsart of the verb and on the context. It disappears, for example, in John is a winner. What we are interested in here are languages where tense is a stable feature of agentive patterns. This is the case, for example, in Tarascan (Foster 1969: 82−83), which uses nominalized active participles to refer to agents. Since this language has two kinds of active participles, present and past, this temporal distinction is straightforwardly carried over to agent nouns (cf. ešéri ‘seeing’/‘one who sees (it)’ vs. ešérini ‘having seen’/‘one having seen (it)’ ← ešé ‘to see’). More to the point is the case of Tübatulabal which, according to Voegelin (1935: 161−162), besides the habitual and occasional affix we have already seen above, also has a “past agentive suffix” (-(a)pïgana-) and a “recent past agentive suffix” (-(a)pïna). On the basis of a verb whose meaning is ‘to roast in the ground’, for example, these suffixes derive agent nouns denoting ‘the one who roasted it in the ground’ and ‘the one who just roasted it in the ground’ respectively. Voegelin thinks that these suffixes were the result of the amalgamation of formerly independent morphemes, of which the first one is the past tense suffix -pï˙-, while -ga- is glossed as ‘to own’. Other languages for which a tense distinction in agent nouns is reported are Estonian (Kerge 1996) and Bhumij (Ramaswani 1992: 69−70). Instrument nouns are much simpler conceptually than agent nouns. The event performed with an instrument is necessarily an action, which severely limits the possible verbal bases. Additional features such as “fondness” or “excess” are irrelevant, since they can only be attributed to human beings. Temporal distinctions are theoretically

74. Agent and instrument nouns

1311

conceivable (‘an instrument which has been used/is used/will be used for Ving’), but do not seem to occur. Nor does any language community seem to consider it useful to have a separate category of “occasional instruments”, though such a distinction is also conceptually possible (a stone used for driving in a nail would be an occasional instrument as compared with a hammer). An internal division which does play a role in natural languages is the one between concrete instruments (tools, machines, etc.) and abstract “instruments” (means). Some languages use one and the same instrumental pattern for both. In Classical Quiché (Friedrich 1955: 437−438), for example, the instrumental suffix -bal not only derives names of concrete instruments such as etzabal ‘toy’ (← etz ‘to play’), but also names for “means” such as michbal ‘trick’ (← mich ‘deceive’), tohbal ‘punishment’ (← toh ‘to pay for’) or maihabal ‘marvel’ (← maihah ‘to be astonished’). In other languages, on the contrary, instrumental patterns essentially remain limited to the designation of concrete objects (French -oir, for example). In languages of this kind, designations of means are often formed by metonymically extending action nouns (cf. English punishment). A second differentiation which is relevant in some languages is that between traditional tools and more modern instruments (mostly machines): Spanish, for example, uses − now largely unproductive − -dera for the first category (cf. podadera ‘pruning shears’ ← podar ‘to prune’; cf. Rainer 1993: 440), while -dor dominates the second one.

7. “Agent” in portmanteau morphs Portmanteau morphs are often thought to occur only in inflection. In reality, they also show up in derivation, though much less frequently so. As far as the notion “agent” is concerned, it is sometimes combined in one morph together with the notion “female”. Familiar cases are the Latin couple -tor (masc.) vs. -trix (fem.), Old Spanish -dor (masc.) vs. -dera (fem.), or their Dutch counterpart -er (masc.) vs. -ster (fem.). Wayuu (Olza Zubiri and Jusayu 1979: 165) even has four portmanteau morphs, one for masculine singular agents (-i ), two for feminine singular agents (-lü and -t), and one for masculine plural agents (-lí ). Sierra Popoluca (Foster and Foster 1948: 22) has a special suffix that “is attached to verb themes to indicate the habitual ‘doers’ as a class”. Ebermann (1983: 155), finally, reports a portmanteau morph combining “agent” and “negation” from Mahou (‘one who cannot V’).

8. Valency and syntax of deverbal agent and instrument nouns A question which has attracted considerable, though still insufficient attention in the literature on agent and instrument nouns, especially from a cross-linguistic perspective, is in how far these preserve the valency and syntax of their base verbs. Concerning the valency and syntax of agent nouns, a distinction has to be made between nominalized active participles and genuine agent nouns (Baker and Vinokurova 2009: 544). Nominalized participles often carry over their verbal syntax to their nominal use, as in the following Italian example, where esercenti is the plural form of the active present participle esercente (← esercire ‘to practice’): gli esercenti la professione medica

1312

VII. Semantics and pragmatics in word-formation II: Special cases

‘those (lit. the) practicing the medical profession’. As one can see from the absence of a preposition, the construction is entirely parallel to that of the verb: esercire la professione medica ‘to practice the medical profession’. Even in the case of nominalized participles, however, lexicalized nouns such as insegnante ‘teacher; lit. (a person) teaching’ require the syntax typical of noun phrases: gli insegnanti di latino ‘the teachers of Latin’/ *gli insegnanti il latino vs. insegnare il latino ‘to teach Latin’. Verbal syntax, on the contrary, seems to be extremely rare with genuine agent nouns, which normally follow, as expected, the syntax of the noun phrase. A direct object of the base verb, for example, must be introduced in English by of and cannot be added directly as in a verb phrase: The Vandals destroyed Rome / the destroyers of Rome / *the destroyers Rome. This, however, does not seem to be universally true. Bresnan and Mugane (2006), for example, have shown that in Gikuyu agent nouns have a mixed syntactic behaviour, partly nominal and partly verbal, much like gerunds in English (but cf. Baker and Vinokurova 2009: 547 for discussion). This language allows noun phrases which correspond to an ungrammatical English construction such as *the driver a rusty truck to Arizona reluctantly. Complements introduced by prepositions generally can be “inherited” more freely, as is the case for at (cf. to look at) and across (cf. to swim across) in the following two examples: She is nearly as eager a looker at women as am I; Who was the first swimmer across lake Ontario? However, even in the case of prepositional complements the valency of agent nouns is heavily restricted in comparison with that of their base verbs. Instrument nouns, especially those which do not designate automatic, agent-like devices, seem to be more restricted with respect to “inheritance”. It has been observed (Rappaport Hovav and Levin 1992) that in English they cannot realize externally the arguments of the base verb: While coffee grinder can refer both to an agent and an instrument, grinder of coffee can only refer to an agent. This generalization does not hold, however, for all languages, as is apparent from Spanish contador de visitas ‘visit counter; lit. counter of visits’, which coexists with the synthetic verb-noun compound cuentavisitas ‘id.’ (← contar ‘to count’ [cuenta is the 3rd person singular form], visita ‘visit’, -s ‘plural’).

9. The origin of agentive and instrumental patterns The last question I would like to address is where agentive and instrumental patterns come from. From a grammaticalization perspective, one would surmise that agentive patterns ultimately derive from those noun phrases which allow designating agents: a man/he who kills another man, those/somebody stealing money, etc. The antecedents of today’s affixes should consequently be the nominal or pronominal heads of such participial or relative constructions. (In the case of headless relative clauses, the resulting word-formation pattern is one of conversion; cf. Jacobi 1897 on Indo-European, aptly summarized in Kastovsky 2009: 333−337.) This indeed turns out to be one possible source of agent and instrument nouns. In Kisi (Childs 1995: 199), for example, the agentive prefix wànà“is related to the word for ‘someone’”. The agentive prefix là- of Acholi, on the other hand, is traced back by Crazzolara (1938: 35) to a word meaning ‘person’. In Supyire

74. Agent and instrument nouns

1313

(Carlson 1994: 115), the agentive suffix foo means ‘owner, possessor, person in charge’ as an independent noun. It appears as the second constituent of noun-noun compounds such as kànhà fòò ‘village chief ’, but “[w]hen affixed to a verb, the resulting nominalization means ‘one who verbs’ ”. As this example suggests, another source of affixes could be the bleaching of the head of a compound (cf. also “semi-suffixes” such as -zeug lit. ‘stuff ’ in German: Spielzeug ‘toy’ ← spielen ‘to play’). It would be misleading, however, to think that grammaticalization is the only way for agent and instrument nouns to arise. An even more important mechanism seems to be reanalysis. In many languages, as we have seen (cf. also German formations of the type Vorsitzender ‘chairman’ ← vorsitzen ‘to chair’, which are still inflected as adjectives), nominalized active participles do the job of agent nouns (so-called participial nouns). Now, if the affix of such a noun is directly related to the verb, an agentive affix is born. In many cases, both patterns, participial and agentive, continue to be used side by side for some time, as is the case with French -ant, for example (cf. une femme enseignant le français ‘a woman teaching French’ vs. un enseignant/une enseignante ‘a teacher m./f.’). Reanalysis has also been seen to play a role in the transition from denominal to deverbal agent nouns, for example in Old High German. Matzinger (2005: 256−263) has shown that Hittite agent nouns in -ala- and -t(t)alla- arose as a consequence of the reanalysis of nominalized relational adjectives whose base was an action noun. Another frequent source for reanalysis are metonymical extensions of action nouns, but this source is more important for instrument (and place) nouns than for agent nouns. German, for example, has formed the instrument noun Kupplung ‘clutch’ directly from the verb kuppeln ‘to operate the clutch’ by means of -ung, an action noun suffix reanalysed as an instrumental suffix on the basis of the many instrumental metonymic extensions in -ung (the corresponding operation is referred to with the nominalized infinitive Kuppeln). A further case of reanalysis in the domain of instrument nouns is the Latin suffix -orium, which goes back to relational adjectives in -ius derived from agent nouns in -tor. Such adjectives were first used in syntagmas of the type opus tectorium ‘plaster-work’ (← opus ‘work’, tectorium ‘concerning plaster’ ← tegere ‘to cover’, derivational stem: tect-), then − after ellipsis had taken place − in isolation as nouns. At that point, tectorium could be reanalysed as ‘that which serves for covering’, thereby giving rise to an instrumental suffix -orium. A last kind of source for reanalysis seems to be constituted by diminutives, which have come to designate (small) instruments in some languages (cf. French calculette ‘pocket calculator’ ← calculer ‘to calculate’, Italian frullino ‘hand blender’ ← frullare ‘to blend’). As discussed in section 4, the view that instrument nouns derive from agent nouns through metaphorical or metonymic extension has always been quite popular, though little evidence has been presented up to now for such a diachronic pathway. Beard (1990: 119) has pointed out that in Serbo-Croatian the opposite extension seems to have taken place; this language, in fact, “uses the traditional inanimate instrumental affix, -l(o), to mark animate subjective (agentive) derivations”. A last immediate source is borrowing. The Creole language Tetun Dili, for example, after having lost almost all of its derivational morphology, has borrowed the Portuguese suffix -dor for the formation of agent nouns (Hayek and Williams-van Klinken 2003).

1314

VII. Semantics and pragmatics in word-formation II: Special cases

10. References Baker, Mark C. and Nadya Vinokurova 2009 On agent nominalizations and why they are not like event nominalizations. Language 85: 517−556. Balles, Irene and Rosemarie Lühr (eds.) 2005 Indogermanische nomina agentis. Leipzig: Universität Leipzig, Institut für Linguistik. Bauer, Laurie 2002 What you can do with derivational morphology. In: Sabrina Bendjaballah, Wolfgang U. Dressler, Oskar E. Pfeiffer and Maria D. Voeikova (eds.), Morphology 2000. Selected papers from the 9 th Morphology Meeting, Vienna, 24−28 February 2000, 37−48. Amsterdam/Philadelphia: Benjamins. Beard, Robert 1990 The nature and origins of derivational polysemy. Lingua 81: 101−140. Benveniste, Émile 1948 Noms d’agent et noms d’action en indo-européen. Paris: Adrien-Maisonneuve. Bisetto, Antonietta 1995 Il suffisso -tore. Quaderni Patavini di Linguistica 14: 39−71. Booij, Geert E. 1986 Form and meaning in morphology: The case of Dutch “agent nouns”. Linguistics 24: 503−517. Bresnan, Joan and John Mugane 2006 Agentive nominalizations in Gĩkũyũ and the theory of mixed categories. In: Miriam Butt, Mary Dalrymple and Tracy H. King (eds.), Intelligent Linguistic Architectures. Variations on themes by Ronald M. Kaplan, 201−234. Stanford: CSLI. Carlson, Robert 1994 A Grammar of Supyire. Berlin/New York: Mouton de Gruyter. Childs, G. Tucker 1995 A Grammar of Kisi. Berlin/New York: Mouton de Gruyter. Crazzolara, Joseph P. 1938 A Study of the Acooli Language. London: Oxford University Press. Dressler, Wolfgang U. 1986 Explanation in natural morphology: Illustrated with comparative and agent-noun formation. Linguistics 24: 519−548. Ebermann, Erwin 1983 Die Sprache der Mauka. Ph.D. dissertation, University of Vienna. Foster, Mary L. 1969 The Tarascan Language. Berkeley/Los Angeles: University of California Press. Foster, Mary L. and George M. Foster 1948 Sierra Popoluca Speech. Washington, D.C.: United States Government Printing Office. Fradin, Bernard 2005 On a semantically grounded difference between derivation and compounding. In: Wolfgang U. Dressler, Dieter Kastovsky, Oskar E. Pfeiffer and Franz Rainer (eds.), Morphology and Its Demarcations. Selected papers from the 11th Morphology Meeting, Vienna, February 2004, 161−182. Amsterdam/Philadelphia: Benjamins. Friedrich, Johannes 1955 Kurze Grammatik der alten Quiché-Sprache im Popol Vuh. Wiesbaden: Steiner. Hayek, John and Catharina Williams-van Klinken 2003 Um sufixo românico numa língua austronésia: -dor en tetum. Revue de Linguistique Romane 67: 55−65.

74. Agent and instrument nouns

1315

Jacobi, Hermann G. 1897 Compositum und Nebensatz. Studien über die indogermanische Sprachentwicklung. Bonn: Cohen. Kärre, Karl 1915 Nomina agentis in Old English. Uppsala: University Press. Kastovsky, Dieter 1971 The Old English suffix -er(e). Anglia 89: 285−325. Kastovsky, Dieter 2009 Diachronic perspectives. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 323−340. Oxford: Oxford University Press. Kerge, Krista 1996 The Estonian agent nouns: Grammar versus lexicon. Sprachtypologie und Universalienforschung 49: 286−294. Laca, Brenda 1986 Die Wortbildung als Grammatik des Wortschatzes. Untersuchungen zur spanischen Subjektnominalisierung. Tübingen: Narr. Levin, Beth and Malka Rappaport 1988 Nonevent -er nominals: A probe into argument structure. Linguistics 26: 1067−1083. Luschützky, Hans Christian and Franz Rainer 2011 Agent-noun polysemy in a cross-linguistic perspective. STUF − Language Typology and Universals 64(4): 287−338. Luschützky, Hans Christian and Franz Rainer 2013 Instrument and place nouns: A typological and diachronic perspective. Linguistics 51(6): 1301−1359. Matzinger, Joachim 2005 Nomina agentis im Hethitischen: Skizze eines Wortbildungstyps. In: Irene Balles and Rosemarie Lühr (eds.), Indogermanische nomina agentis, 253−267. Leipzig: Universität Leipzig, Institut für Linguistik. Meyer-Lübke, W[ilhelm] 1890 Italienische Grammatik. Leipzig: Reisland. Müller, Peter O. 2011 The polysemy of the German suffix -er: Aspects of its origin and development. STUF − Language Typology and Universals 64(1): 33−40. Olza Zubiri, Jesús and Miguel Ángel Jusayu 1979 Gramática de la lengua guajira. Caracas: Universidad Católica Andrés Bello. Panther, Klaus-Uwe and Linda L. Thornburg 2002 The roles of metaphor and metonymy in English -er nominals. In: René Dirven and Ralf Pörings (eds.), Metaphor and Metonymy in Comparison and Contrast, 279−319. Berlin/ New York: Mouton de Gruyter. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rainer, Franz 2011 The agent-instrument-place “polysemy” of the suffix -TOR in Romance. STUF − Language Typology and Universals 64(1): 8−32. Ramaswani, N. 1992 Bhumij Grammar. Manasagangotri: Central Institute of Indian Languages. Rappaport Hovav, Malka and Beth Levin 1992 -Er nominals: Implications for the theory of argument structure. In: Tim Stowell and Eric Wehrli (eds.), Syntax and the Lexicon, 127−153. Orlando, FL: Academic Press. Ryder, Mary Ellen 1991 Mixers, mufflers and mousers: The extending of the -er suffix as a case of prototype reanalysis. In: Laurel A. Sutton, Christopher Johnson with Ruth Shields (eds.), Proceed-

1316

VII. Semantics and pragmatics in word-formation II: Special cases

ings of the Seventeenth Annual Meeting of the Berkeley Linguistic Society, February 15−18, 1991. General Session and Parasession on the Grammar of Event Structure, 299−311. Berkeley, CA: Berkeley Linguistic Society. Ryder, Mary Ellen 1999 Bankers and blue-chippers: An account of -er formations in Present-day English. English Language and Linguistics 3: 269−297. Szigeti, Imre 2002 Nominalisierungen und Argumentvererbung im Deutschen und Ungarischen. Tübingen: Niemeyer. Torrego, M. Esperanza 1996 Conditions syntaxiques pour la formation des noms d’agent en latin. In: Hannah Rosén (ed.), Aspects of Latin. Papers from the Seventh International Colloquium on Latin Linguistics, Jerusalem, April 1993, 181−192. Innsbruck: Institut für Sprachwissenschaft. Voegelin, Charles F. 1935 Tübatulabal Grammar. Berkeley: University of Berkeley Press. Wagner-Nagy, Beáta 2001 Die Wortbildung im Nganasanischen. Szeged: Universitas Szegediensis.

Franz Rainer, Vienna (Austria)

75. Patient nouns 1. 2. 3. 4. 5.

Introduction Heterogeneous patterns in patient nominalizations Recent semantic and syntactic perspectives on patient nouns Polysemy and ambiguity in patient nouns: diachronic and cognitive perspectives References

Abstract The patient role seems to be less salient and cognitively identifiable than the agent role in the world’s languages. This contribution looks at the various ways in which patient nouns are formed in selected European languages, describes recent semantic approaches to the characterization of patient nouns (with particular reference to English -ee suffixation) to then focus on polysemy and ambiguity in patient nominalizations. The challenges of an apparently instable relationship between derivational marker and patient role will be discussed in the light of current diachronic and cognitive approaches.

1. Introduction One of the semantic (or thematic) roles of entities which are affected by the action of the verb is the patient role. “Patient” as ‘undergoer of an action’ can be contrasted with “agent” (‘doer of an action’) or “instrument”, which at times overlaps with agent in their

75. Patient nouns

1317

causal force in an event. In generative grammar patient is recognized as one of the theta roles which represents the argument structure syntactically required by a verb (cf. Lieber 2004; Olsen 1986; Rappaport and Levin 1992). As part of cognitive semantic categories in morphology, patient, along with agent and instrument, is one of the major types of “first order entity” (persons, things, and animals) marked morphologically (Mackenzie 2004). As a cognitive category, “patient” seems to be of less importance as part of human experience than “agent” (see article 74 on agent and instrument nouns) and, arguably, “instrument” and is therefore less often marked in natural languages than agent. Ryder (1999: 284) sees a hierarchy of salience, i.e. the degree to which something is noticeable in comparison with its surroundings, between agent, instrument and patient role with agent as most and patient as least salient role. This is supported by typological investigations of morphological markings where agent nouns have been found to be much more likely to be marked than patient nouns. In Bauer’s (2000: 40–41) comparison of 42 unrelated languages more than half, i.e. 24 languages have agent nouns derived from verbs but he finds not a single patient noun in his sample. Lack of salience and “identifiability” (Ryder 1999: 284) may have contributed to the fact that, if there are ways to morphologically mark a patient noun (partially: “object noun”) in a language, it often involves heterogeneity in form and ambiguity in meaning. The differentiation between patient nouns and result nouns (cf. article 71 on result nouns) is not always clear.

2. Heterogeneous patterns in patient nominalizations There are cases where a language has one marker for agent, instrument, and patient. Mackenzie (2004: 978), for example, cites Babungo, a Niger-Kongo language in the Northwest of Cameroon, where a prefix N- can indicate all three thematic roles. More typically, however, those languages which have nominalizing word-formation patterns for indicating the type of role expressed in the noun, use more specific ones. In derivational morphology, ideally one affix would be used consistently for indicating one particular semantic role. In most real language situations, however, the relationship between derivational marker and semantic role specification and meaning is more complex. Languages like German, Dutch, English or French use one or more dominant derivative patterns for forming identifiable agent nouns, e.g., the deverbal noun suffix -er in German (fahren ‘to drive’ → Fahr-er ‘driver’), Dutch (speel ‘to play’ → speel-er ‘player’), English (dance → dancer), -eur/-euse in French (danser → danseur/-euse ‘dancer’), etc. Corresponding easily identifiable derivative patterns for patient nouns are less common. The suffix -ee, which is used for patient nominalizations in English, seems to be an exception rather than the rule. This rather special status is also reflected in the fact that -ee nouns in English have received a lot of attention in current scholarly work, and especially the particular syntactic and semantic properties of this word-formation pattern (Bauer 1983; Barker 1998; Baeskow 2002; Portero Munoz 2003; Booij and Lieber 2004; Mühleisen 2010). Despite the fact that there is a particular article on the semantics of the English -ee suffix in this handbook (see article 52 on semantic restrictions on wordformation: the English suffix -ee) this contribution will also refer to -ee nouns in English as an example of one of the most discussed and complex patient nominalizations in a European language.

1318

VII. Semantics and pragmatics in word-formation II: Special cases

In much of the recent literature on patient nouns in selected European languages, -ee serves as a basis of comparison with which other more heterogeneous ways to mark patient nouns are contrasted. Booij and Lieber (2004) compare English affixation patterns for agent and patient nouns with Dutch. One of the most comprehensive studies on nomen agentis and nomen patientis in English and German is Baeskow (2002) who lists eight different options to form patient nouns in German (2002: 590–591). While there are Germanic and Latinate derivational suffixes (-ling, -at, and -and/-end) which correspond loosely to -ee in English, their occurrence is restricted to a very limited number of lexical items and none of them is productive. (1)

-ee derivation in English and corresponding derivation in German -ee derivation examinee detainee trainee refugee arrivee

-ling derivation Prüfling Häftling Lehrling Flüchtling Ankömmling

addressee grantee

-at derivation Adressat Stipendiat

experimentee *no -ee equivalent

-and/-end derivation Proband Promovend ‘doctoral candidate’ (cf. also Baeskow 2002: 590 f.)

The Germanic suffix -ling is also present in English (e.g., suckling) and Dutch (zuigeling ‘infant’ ← zuig ‘to suck’ or bekeerling ‘convert’ ← bekeer ‘to convert’). As the alleged means of patient noun derivation in German, -ling shares a number of problems with -ee derivation: firstly, the patient meaning in words like Flüchtling (← flüchten ‘to flee’) or Ankömmling (← ankommen ‘to arrive’) is questionable since the referents are the actual doers of the action. The same applies to the English counterparts and a number of examples (e.g., absentee, escapee, standee) which have given rise to debates about the semantic properties and specific characteristics of -ee words (Barker 1998, Baeskow 2002, Portero Munoz 2003, Booij and Lieber 2004; see also section 3 of this article). A second shared problem is the possibility of forming nouns which are non-human and even non-animate, e.g., Fäustling ‘mitten’ (← Faust ‘fist’) in German or benefactee in English as description of grammatical role. The last two examples can also serve to illustrate a third problematic point: the base of derivation is by no means always a verb. In the case of -ling, the base can be a verb, a noun, an adjective, or even a numeral: (2)

-ling suffixation with different bases V

lehren prüfen

‘to teach/train’ ‘to examine’

Lehrling Prüfling

‘trainee’ ‘examinee’

75. Patient nouns

1319

N

Gunst Junge

‘patronage’ ‘boy’, ‘young man’

Günstling Jüngling

‘protegé’ ‘young man’ (iron. or pej.)

A

dumm schön

‘stupid’ ‘beautiful’

Dümmling Schönling

‘fool’ ‘beau’, ‘glamour boy’

Num.

erst zwei

‘first’ ‘two’

Erstling Zwilling

‘firstling, first-born’ ‘twin’

The -ling examples above possess an additional feature: the suffix here also includes the notion of a diminutive which, at least in contemporary usage, conveys a more or less belittling attitude towards the referent. In a more obvious way, this is also expressed in formations like Jüngling (← Junge ‘boy’) or even Dichterling (← Dichter ‘poet’) which indicate that the referent is not to be taken seriously or also not quite finished (cf. also Baeskow 2002: 504 f.) in his role as a young man or as a poet. As for the Germanic suffix -ling in English, the OED notes that “the personal designations in -ling are now always used in a contemptuous or unfavourable sense (though this implication was not fully established before the 17th c.), as courtling, earthling, groundling, †popeling (= papist), vainling, worldling”. While there is no equivalent direct overlap of patient suffix -ee and diminutive -ie or -y in English, the “lack of control” entailed in the patient suffix is sometimes playfully exploited for hypocoristic meaning (cf. Mühleisen 2010: 74, 177– 179). In addition to the -ling suffix for (mostly) person referents, the suffix -sel creates patient nouns for (mostly) inanimate referents in German. Yet again, the patient role goes hand in hand with a diminutive and often negative meaning: In Anghängsel (← anhängen ‘to append’), Einschiebsel (← einschieben ‘to insert’), Gereimsel (← reimen ‘to rhyme’) or Geschreibsel (← schreiben ‘to write’), the suffix -sel adds the connotation of ‘small and unimportant’ to an appendix or insertion and ‘unfinished and bad’ to a rhyme or piece of writing. In German, the notably negative connotations associated with -ling and -sel as a consequence of this overlap of patient and diminutive meaning have no doubt contributed to the lack of productivity of both suffixes in contemporary German and even led to recent substitutions of Auszubildende/r f./m. for Lehrling (both meaning ‘trainee, apprentice’). The former is formed as a gerund (participium necessitatis) of the verb ausbilden ‘to train’ and, in contrast to Lehrling, is not burdened with a diminutive or pejorative meaning. Nominalized participle constructions are, in fact, the more productive and often preferred form of forming nouns with a patient role meaning in German but also in a number of other European languages. Apart from the nominalized gerund forms mentioned before, nominalized past participles or present participles are used. An example of a past participle construction in German would be der Befragt-e ‘interviewee’ from befragen ‘to interview’ via the past participle befragt ‘interviewed’. Genesender ‘recoveree’ presents an example of a noun formation on the basis of a present participle, genesend, formed from the verb genesen − another patient noun construction, both in English and in German, where the referent is actually the experiencer of the process. With regard to Dutch, Booij and Lieber (2004: 332) note that “there is no specific suffix in Dutch which forms personal object-related nouns […]”. The principal strategy of creating patient nouns from verbs is, similar to the German participle constructions, by substantivizing adjectives that are converted past participles by means of -e suffixation.

1320 (3)

VII. Semantics and pragmatics in word-formation II: Special cases verb

adjectival participle

noun

addresser ‘to address’

geadresseerd ‘addressed’

geadresseerd-e ‘addressee’

ontsnap ‘to excape’

ontsnapt ‘escaped’

ontsnapt-e ‘excapee’

steel ‘to steal’

gestolen ‘stolen’

gestolen-e ‘what has been stolen’ (Booij and Lieber 2004: 333)

For the Dutch examples, the nouns based on past participles can be either animate (e.g., geadresseerde, ontsnapte) or inanimate (e.g., gestolene) which makes this pattern only partly comparable to English -ee suffixes where the feature ‘animate’ or ‘sentience’ plays a central role. Participle constructions for the formation of patient nouns are also commonly found in languages other than German or Dutch. In Czech and Russian, patient nouns are not listed as a separate grammatical category. However, patient meaning can be found in noun derivations from deverbal adjectives (diachronically: adjectival participles), e.g., in the Czech examples odsouzenec ‘the convicted’ (← odsouzený ← v. odsoudit), ozbrojenec ‘the armed’ (← ozbrojený ← v. ozbrojit), vzdělanec ‘the educated’ (← vzdělaný ← v. vzdělat) or zaměstnanec ‘the employed’ (← zaměstnaný ← v. zaměstnat). In Russian, derivatives from present passive participles like ljubimec ‘loved one’ (← ljubimyj ← v. ljubit’) can also be used. The English deverbal noun suffix -ee has its origin in French legal loan words formed as French past participles in the 14th and 15th century when the derivational affix was nativized in English and gradually became productive in the formation of English patient nouns in legal and more general domains. French has continued to use mainly nominalized past participles for expressing patient role, making French and English patient nouns at times remarkably similar in form despite divergent processes of formation: (4)

French nom. past participle (specified for gender)

English deverbal noun (verb + -ee)

employé m., employée f. invité m., invitée f. retraité m., retraitée f.

employee invitee retiree

The last examples illustrate that the similarity is not based on borrowings but on -ee as a nativized suffix in English which has its origin in a morphological borrowing. In principle, passive participles in English may serve the same function of denoting ‘undergoer of an action’, e.g., in examples like the beloved, the aforementioned, the cursed or the intended which, however, are rather frozen expressions. While the employed, the invited or the retired as alternative to employee, invitee and retiree may sound somewhat elliptical and clumsy, it is nevertheless a possible construction in English. The relationship including the semantic feature relationship between various types of past participle

75. Patient nouns

1321

constructions in English and -ee nominalizations will be discussed in more detail in section 3. The apparent diversity and heterogeneity of morphological means for the creation of nouns with the idealized meaning of ‘affected object of the action expressed by the verb’ is remarkable in languages like German, Dutch and English. The fact that in these languages, noun derivations are also possible which have a non-verbal base (e.g., in biographee in English, therapee − cf. Mühleisen 2010: 29 − or Günstling ‘protegé’ in German) adds to the rather blurred impression. In addition, there are also examples of patient nouns which are formed with the (typically agentive) suffix -er in English (e.g., broiler or roaster for particular types of chicken, Ryder 1999: 286) − puzzlements which lead Booij and Lieber (2004: 334) to raise the question of whether or not the affixes themselves make any semantic contribution to their bases. This is an especially critical question to ask when we realize that there are so many -er and -ee words formed from nonverbal bases that nevertheless have a sort of dynamic or situational meaning in spite of the lack of a verbal base. Second, we should ask why the meanings of -er and -ee can sometimes overlap: although -er most often forms subject-oriented nouns and -ee object-oriented nouns, there is nevertheless a significant number of -er forms which are object-oriented and -ee forms which are subject-oriented. (Booij and Lieber 2004: 334)

Recent semantic approaches to patient nouns, and especially with reference to nominalizations with -ee in English have attempted to throw light on the specific characteristics that make a patient noun recognizable and identifiable.

3. Recent semantic and syntactic perspectives on patient nouns What are the most central semantic features of a patient noun? The thematic role has traditionally been associated with notions like ‘passivity’, ‘undergoer of an action indicated by the verb’ and ‘human’. As noted in the discussion of the German suffix -ling in section 2, there are counter-examples for each one of these notions in such German patient noun derivations. The same applies to English -ee nominalizations in that some are active rather than passive (e.g., escapee), some have nominal rather than verbal bases (e.g., aggressee − as opposed to aggressor) and others still are non-human (brushee − ‘a pet who is brushed’ − cf. Mühleisen 2010: 3). One of the most discussed descriptions of the semantic characteristics of patient nouns of the -ee nominalization type in English is Barker (1998). In order to overcome the problems outlined above, he attributes basically three semantic properties to -ee words; (i) the referent of an -ee word must be sentient, (ii) the use of an -ee noun entails lack of volitional control on the part of its referent, and (iii) the denotation of an -ee noun must be episodically linked to the denotation of its stem. The notion of ‘sentience’ rather than ‘human’ or ‘animate’ in patient nouns implies that a sense of feeling or experience is attributed to the referent in an -ee formation. Non-human animate referents in brushee or cleanee are animals or pets to which the speaker attributes this capacity. In contrast, animals whom we do not see as feeling or experiencing beings are not a possible patient noun referent. In an example from a recent corpus of new -ee formations (Mühleisen 2010), eatee is neither a slaughtered animal nor the meat on the plate but a human being who is also a participant in the action −

1322

VII. Semantics and pragmatics in word-formation II: Special cases

the example is taken from a report on a trial on cannibalism (see also semantic property iii in Barker’s list of characteristics): (5)

Ex.

Eatee:

Afterwards, the Eatee was in a bathtub full of water, so he could bleed to death. The Eater would check up on him every so often. On the video, the Eatee said, “If I am still alive in the morning, let’s cook and eat my testicles”. (cf. Mühleisen 2010: 45)

Inanimate referents are also a common phenomenon in both established -ee words (e.g., benefactee) and in the corpus of recent -ee formations (e.g., decodee). This might be accounted for in the framework of sentience in that role behaviour in human interaction is semantically extended to instruments and abstract concepts. A further explanation might be seen in section 4 where the role of analogous formations and collocations, for instance on the basis of -er words (decoder − decodee, cf. Mühleisen 2010: 47, 136– 139) are highlighted. “Lack of volitional control”, the second feature which Barker singled out as decisive for English patient nouns, may furthermore serve to resolve some of the cases where the referent is the “doer of the action” − e.g., resignee, escapee − but involuntarily, unconsciously or reluctantly so. We can see the same phenomenon in German patient nouns formed with -ling, e.g., Flüchtling ‘escapee’. For numerous -ee nouns, the “lack of volitional control” feature is persuasive. The most obvious cases of patient nouns are those where “something beyond control” happens to the -ee noun referent in the course of the action, e.g., as in muggee, blackmailee, murderee, amputee. But some of the more jocular formations like kissee, huggee, squeezee also play with this notion of “lack of volition”. On the other hand, the element of lack of volition or lack of control is not given in other subject formations like attendee, dinee, expressee or parkee. Furthermore, for -ee words in the legal sphere, such as debtee, obligee, pledgee or trustee, the semantic feature concerned seems to be more one of commitment and obligation rather than lack of volition. “Episodic linking”, the third of Barker’s requirements for English patient nouns, is defined as “the referent of a noun phrase headed by an -ee noun must have participated in an event of the type corresponding to the stem verb” (Barker 1998: 711), e.g., a muggee is only a muggee if he or she has participated (or is even a regular participant) in a mugging event. Episodically linked patient nouns do not describe an individuallevel property of the referent (e.g., intelligence, height, etc.) but a “stage-level” property of an individual, i.e. something that may be true for an individual at one point in time but false for another point in time, for instance, the stage of being a huggee is linked to the duration of the hugging event. Other stages are altered more permanently through the event, for instance, a retiree remains a retiree after the event of retiring has taken place. Barker suggests that episodic linking in -ee nouns “does not depend on the syntactic argument structure of the stem” (Barker 1998: 713). Instead, the meaning of the verb and its aspectual properties (punctual vs. non-punctual) determine whether or not the event in question causes a permanent or temporary change of state for the participant. Among the -ee words in which a permanent change of state can be seen are, for instance, adoptee, amputee and divorcee − or, in more recent formations, circumcisee or convertee (cf. Mühleisen 2010). Others are anti-punctual (e.g., employee), naturally punctual (pho-

75. Patient nouns

1323

tographee, assassinee, burglaree) or naturally non-punctual (borrowee, trainee), etc. Barker (1998: 717) here seeks to overcome the problem that the majority, but by no means all, -ee patient nouns are verb-derived in that he notes that “all that is required in order to satisfy the definition of episodic linking given […] is that the stem be associated with a set of eventualities that can serve as qualifying events, and the attested uses of nominal stem -ee nouns satisfy this requirement”. Despite some obvious exceptions to the model pointed out above, Barker’s (1998) description of the semantic features of the English patient noun formed with -ee is one of the most elaborate explanations of the heterogeneity of this word-formation pattern. In Barker’s view, -ee nouns are exceptional and he even argues for a classification of -ee as a separate thematic role, not quite “on a par in importance with notions like Agent and Patient, which pervade language (or at least, pervade linguistic description)” (Barker 1998: 723) − an idea which has not been taken up with much enthusiasm in scholarly debates. In an attempt to “move beyond the deadlock created by purely semantic approaches to -ee derivation” (2006: 340), Heyvaert (2006) comes back to the widespread role of participle constructions in patient nominalizations and claims that English -ee nouns, both non-agentives and agentives, are systematically related to the English past participle morpheme -ed. Building on Langacker (1991), she distinguishes between four types of past participles in English: PERF1 and PERF2 are adjectival and derive from stative participles with PERF1 attaching to intransitive verbs (e.g., “a newly enlisted friend”) and PERF2 to transitive (e.g., “the more motivated patient”) ones. PERF3 is the passive variant of the past-participle morpheme (e.g., “my arm was burned”) while PERF4 represents its perfect variant (e.g., “he has been promoted to the main Australian team”). Each type differs in profile with regard to stative and resultative relations as well as the type of involvement participants have in the process of change of state. The network of related meanings of the different variants links up the various uses of -ed and constitutes an essential part of the analysis of the past participle constructions which function as agnates of -ee nominalizations, so Heyvaert (2006: 353). The “most obvious relation of what has always been perceived as the prototypical ‘passive’ core of -ee nominalizations and the passive (PERF3) use of the past participle”, are exemplified here: (6)

-ee nominalizations and employee detainee payee

− − −

PERF3

past participle

(s)he is employed (s)he is detained (s)he is paid

Closer to PERF2 past participles are those -ee nominalizations where an adjectival version of the past is expressed: (7)

-ee nominalizations and adoptee electee

− −

PERF2

past participle

an adopted child, or: someone who has been adopted an elected member, or: someone who has been elected

With regard to the agentive type of -ee nominalizations, Heyvaert (2006) points out that the greater part expresses meanings which are similar to those realized by PERF1 and

1324

VII. Semantics and pragmatics in word-formation II: Special cases

PERF4. Some relate to PERF1 or the adjectival use of the past participle (8). Others are less stative or adjectival in nature but resemble the present perfect use of the past participle realized by PERF4 (9):

(8)

agentive -ee nominalizations and retiree escapee enlistee

(9)

− − −

− − − −

past participles

a retired officer an escaped prisoner an enlisted soldier

agentive -ee nominalizations and resignee returnee forgettee deferee

PERF1

(s)he (s)he (s)he (s)he

has has has has

PERF4

past participles

resigned returned forgotten deferred

Both Barker (1998) and Heyvaert (2006) have attempted to explain the apparent heterogeneity of English patient nouns with -ee suffixation in unified models, the former with a claim to exceptional status as a thematic role, the latter linking the various meaning types to past participle constructions which, as was demonstrated in section 2, are also commonly used to form patient nouns in languages like German, Dutch and French. While both models convincingly account for the majority of cases of -ee nouns, there remain some more marginal specimen which do not seem to fit any unified model.

4. Polysemy and ambiguity in patient nouns: diachronic and cognitive perspectives So far, it has been outlined that the patient role is cognitively less important and less recognizable (or less salient and identifiable, cf. Ryder 1999: 284) than the agent role in person nominalizations. This may have led on the one hand to a greater heterogeneity in the formation of patient nouns in languages like English, German and Dutch and, on the other hand, to a semantically fuzzier notion of the criteria which define a patient noun. Some recent approaches, notably Baeskow (2002) and Mühleisen (2010) have therefore included cognitive models like prototype theory to account for more central and more marginal examples of patient nouns. In her highly comprehensive treatment of the semantics of agent and patient nouns in English and German, Baeskow (2002: 64 ff.) uses and expands Dowty’s (1991) model of thematic proto-roles in order to overcome problems of using fixed theta-grids for suffixes (noun based, non-event character of the verb). In Dowty’s model, only two proto-roles exist, Proto-Agent and Proto-Patient, which are each defined by different verbal entailments. Proto-patient entailments are: a) undergoes change of state, b) incremental theme, c) causally affected by another participant, d) stationary relative to movement of another participant, and e) does not exist independently of the event, or not at all. The semantic characteristics do not all have to apply in order to qualify for a particular proto-role. Rather, this model allows for degrees of representativeness of items within a proto-role like proto-patient role. More marginal items within this proto-role might

75. Patient nouns

1325

then only have one of the characteristics above, more central items would share all or most of them. Baeskow (2002) furthermore includes a diachronic dimension in her contrastive analysis of English and German patient nouns and also highlights the historical development of loan suffixes like the French derived English -ee suffix. Mühleisen (2010) also uses a diachronic perspective to account for the heterogeneity of patterns for English patient nouns and demonstrates that a number of parameters have changed from the early beginnings of -ee suffixation in the 15th century until today (2010: 61–90): a) the word class of the base of the derivation (correlative noun, verb), b) the type of passive object noun (indirect, direct), c) the passive character (object, subject), d) the semantic field of occurrence (legal, general, ironic/jocular), e) human reference (human, non-human, including animals and technical components), and f) the specific use in particular varieties of English (Britain, U. S.A., Australia, Scotland). Because of the simultaneous presence of existing words from all periods, new words may be formed in analogy to a variety of template words − be it on the basis of a correlative passive of agent nouns (predominant in the 16th century in words like debtee, vendee) or a deverbal patient noun (representee − ‘one who is represented’, 1624 OED) which changed its meaning to agent noun (representee − ‘a parliamentary representative’, 1644) in the course of just a few decades. Analogical coinings based on the semantics of another, similar base (“base-substitution”, Schröder and Mühleisen 2010: 49) may thus be taken as one explanatory factor for polysemy and ambiguity in -ee words. A second possibility is analogical formation by “affix-substitution” (van Marle 1985: 256). In the coinage of -ee patient nouns affix-substitution seems to be a common phenomenon. Usually, they are triggered by the co-occurrence of the relational opposite -er agent noun which demonstrates that the creation of patient nouns is perhaps cognitively more dependent on agent nouns than previously assumed. In a corpus study of 1,000 potential new -ee words searched on the world wide web (Mühleisen 2010: 121–164), an astonishing number of 748 items were actually attested. Out of these successful -ee patient nouns, 552 words (73.8 %) occurred at least in one instance in “lexical solidarity” with an -er word, e.g., typically in examples like “the question here is who is the anointer and who is the anointee”, “[…] what is more important was that I was more of a bruiser than a bruisee”, “it’s better to be the fusser than the fussee” or “one former whistler insisted he never did so with the intent of actually embarrassing the whistlee” (Mühleisen 2010: 136–137). Analogical coining from an existing -er word even extends to instances where the -er word and -ee word have (more or less intentionally) slightly incongruous meanings − as in the instrument vs. person pairing of elevator − elevatee below (10). However, affixsubstitution is not the only kind of analogical coining process that is noticeable here: in a process of “base-substitution”, in the example of failee (11), it seems to be the semantics of another, similar noun − trainee − which is responsible for the creation process. The context in which the neologism was formed best reveals the specific analogical coining process: (10) Elevatee:

With the Secret Floor Ballot, each employee in a large office building has a small remote control in his pocket, with which he can signal “3” without the elevator button lighting up. By the time the thing stops at the third floor, he is out the door before he can see the grimaces of the elevatee he left behind.

1326 (11) Failee:

VII. Semantics and pragmatics in word-formation II: Special cases This will ensure the other trainees don’t get bored and the “failee” doesn’t get disheartened; you can return to this exercise later, […] (Mühleisen 2010: 140, 40, italics added, S. M.)

While the examples in (10) and (11) are certainly more marginal items, the study of new -ee words also shows that the most frequent formations are relatively rule-governed and comply with “prototypical” characteristics of -ee words (Mühleisen 2010: 143), i.e. they are a) verb-derived (i.e. verb exists), b) with existing correlative -er noun, c) in direct object relation to the verb, d) sentient and probably human, e) role participant, f) nonvolitional and non-active part in the event, g) can be used in a legal as well as more general contexts. However, both prototypical and more marginal -ee suffixed patient nouns continue to exist and to provide templates for the formation of new words. The relationship between patient noun and word-formation patterns associated with patient role has been shown to be rather heterogeneous and sometimes ambiguous, not least because of the low degree of salience patient role seems to have. It is suggested that context and especially the correlation with the more salient agent noun in context is cognitively more decisive for an identification of the particular patient noun than the type of word-formation pattern or derivative marker.

5. References Baeskow, Heike 2002 Abgeleitete Personenbezeichungen im Deutschen und Englischen. Kontrastive Wortbildungsanalysen im Rahmen des Minimalistischen Programms und unter Berücksichtigung sprachhistorischer Aspekte. Berlin/New York: de Gruyter. Barker, Chris 1998 Episodic -ee in English: A thematic role constraint on new word formation. Language 74(4): 695–727. Bauer, Laurie 1983 English Word-Formation. Cambridge: Cambridge University Press. Bauer, Laurie 2000 What you can do with derivational morphology. In: Sabrina Bendjaballah, Wolfgang U. Dressler, Oskar E. Pfeiffer and Maria D. Voeikova (eds.), Morphology 2000. Selected papers from the 9 th International Morphology Meeting, Vienna 24–28 February 2000, 37–48. Amsterdam/Philadelphia: Benjamins. Booij, Geert and Rochelle Lieber 2004 On the paradigmatic nature of affixal semantics in English and Dutch. Linguistics 42(2): 327–257. Dowty, David R. 1991 Thematic proto-roles and argument selection. Language 67: 547–619. Heyvaert, Lisbeth 2006 A symbolic approach to deverbal -ee derivation. Cognitive Linguistics 17(3): 337–364. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar. Descriptive application. Stanford: Stanford University Press. Lieber, Rochelle 2004 Morphology and Lexical Semantics. Cambridge: Cambridge University Press.

76. Place nouns

1327

Mackenzie, Lachlan 2004 Entity concepts. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphologie. Ein internationales Handbuch zur Flexion und Wortbildung. Vol. 2, 973–983. Berlin/New York: de Gruyter. Mühleisen, Susanne 2010 Heterogeneity in Word-Formation Patterns. A corpus-based analysis of suffixation with -ee and its productivity in English. Amsterdam/Philadelphia: Benjamins. OED = Oxford English Dictionary. 2nd ed. Olsen, Susan 1986 Wortbildung im Deutschen. Stuttgart: Kröner. Portero Munoz, Carmen 2003 Derived nominalizations in -ee: A role and reference grammar based on semantic analysis. English Language and Linguistics 7(1): 129–159. Rappaport, Hovav Malka and Beth Levin 1992 -Er nominals: Implications for the theory of argument structure. In: Tim Stowell and Eric Wehrli (eds.), Syntax and Semantics. Vol. 26: Syntax and the Lexicon, 127–153. San Diego: Academic Press. Ryder, Mary Ellen 1999 Bankers and blue-chippers: An account of -er formations in present-day English. English Language and Linguistics 3(2): 269–297. Schröder, Anne and Susanne Mühleisen 2010 New ways of investigating morphological productivity. Arbeiten aus Anglistik und Amerikanistik 35(1): 35–60. van Marle, Jaap 1985 On the Paradigmatic Dimension of Morphological Creativity. Dordrecht: Foris.

Susanne Mühleisen, Bayreuth (Germany)

76. Place nouns 1. 2. 3. 4. 5. 6. 7. 8.

Introduction The grammatical status and semantics of place nouns Deverbal and denominal place nouns Semantic subtypes of place nouns Overlap between place nouns and other nominalization patterns Place nouns in English Expanding the class of place nouns References

Abstract Derived words which represent the category of place nouns denote places. Prototypically, such nouns are products of deverbal locative nominalization. There are also well attested patterns of place nouns based on other nouns. The article gives a survey of selected

1328

VII. Semantics and pragmatics in word-formation II: Special cases

examples of the category in question taken from a variety of European languages (including Czech, English, Hungarian, Irish, Polish, Serbo-Croatian), in order to determine the scope of the process of place-noun derivation as well as its principal characteristics.

1. Introduction In word-formation, the term place noun denotes nouns referring to places. In English usage, this designation has been employed more commonly than its Neo-Latin equivalent nomina loci, which is still popular in some European traditions. The concept of place nouns has been called upon in descriptions of the word-formation systems in a variety of languages, especially those which show overt morphological marking (e.g., affixes) in the derivation of so-called locative nominals based on verbs. The meaning of such a product of nominalization can then be paraphrased, generally, as “a location where the activity described by the verb tends to occur” (Payne 1997: 229) or, according to a shorter formula, “a place where ‘verb’ happens” (Comrie and Thompson 1985: 355). The idea can be illustrated with the following English example: refine → refinery. Conceived of in this manner, the standard process deriving place nouns presents itself as a type of deverbal nominalization. It ought to be noted that, cross-linguistically, there are several other, usually more prominent, types of participant nominalizations (like agentive, patientive, instrumental, resultative), plus action or event nominalizations (for details, see, e.g., Comrie and Thompson 1985; Payne 1997: 223−231). Accordingly, locative nominals can be regarded as defining the core of a corresponding derivational category, termed “place nouns”. Naturally, this categorial designation is of relevance for those languages which make use of place-denoting nominalization. While the derivational category in question is not universally attested, a large number of languages have been reported to possess more or less productive patterns of locative nominalization, i.e. they can be argued to have the derivational category of place nouns. The list includes many Indo-European languages (Slavic in general, French, German, English, Irish, etc.) as well as numerous languages representing other families and areal divisions: Hungarian (Comrie and Thompson 1985: 355), various Asian languages (Zeitoun 2002), many Bantu languages (Comrie and Thompson 1885: 355), Nahuatl (Stiebels 1999), and so on. See Luschützky and Rainer (2013) for a more representative list. The informal label “place names” (or “names of places”) that is sometimes used in the literature to designate certain common nouns, including derived locative nominals (i.e. place nouns), must be distinguished from the identical term place names often employed in onomastics to denote a type of proper nouns (toponyms like Oxford, Wimbledon, etc.; see, e.g., Coates 2006: 335). However, the distinction may be hard to draw, given the fact that certain onomastic terms, based on toponyms, are themselves products of morphological derivation, including place-noun formation; cf. Russian Volga → Povol’ž’e ‘the Volga region’ or Polish Wisła ‘Vistula’ → Powiśle ‘the Vistula region’. From the formal viewpoint, the Slavic pattern illustrated here represents the method of prefixal-suffixal (alternatively: “prefixal-paradigmatic”) derivation. This can be further exemplified with the following Polish pairs (involving common nouns): góra ‘mountain’ → pogórze ‘foothills’, gleba ‘soil’ → podglebie ‘undersoil’, brzeg ‘shore, bank’ → nabrzeże ‘wharf ’, etc. In a broader, cross-linguistic perspective, suffixation alone emer-

76. Place nouns

1329

ges as the principal method of deriving place nouns (cf., however, prefixation of ma- as a locative marker in Arabic).

2. The grammatical status and semantics of place nouns Derived words which represent the category of place nouns can be viewed as a subclass of all common nouns that denote locations, including a large number of simplex names (plus originally complex names that are hardly analyzable in synchronic terms; cf. English dairy). Thus, for instance, in English we have nouns like field, desert, city, flat, etc. which denote different kinds of open or enclosed areas and other sorts of locations. Therefore, the broad concept of place nouns is also relevant as a major lexico-semantic category. Besides, cross-linguistic comparisons reveal that a particular locative term that in one language has the status of a monomorphemic (underived) item, e.g., flat in English, may have a derived translation equivalent in another language; cf., in Polish, mieszkanie ‘flat’ from the verb mieszkać ‘to live, dwell’. From the morphological viewpoint, only the Polish lexeme will qualify as an instance of a place noun (since it is a derivative; its locative reading is due to semantic concretization of the basic actional semantics − mieszkanie ‘living’). More broadly speaking, such comparisons also reveal why the classes of place nouns to be found in various languages are not co-extensive: a nominal locative concept that in one language happens to be encoded by a derived form is sometimes expressed by simplex nouns in other languages (if it is lexically expressed at all). Moreover, it appears that the derivational category of place nouns (no matter how productive it is in a given language) is well-motivated in a broader grammatical context, if we go beyond morphology and the lexicon: “location” is one of the fundamental categories of human language and, as such, it reveals itself at different levels of grammatical structure. Suffice it to say, that locations or, more generally, spatial relations are encoded at the level of syntactic structure (cf., for instance, adverbs of place or the locative alternation in English). Ultimately, “location” is one of the basic categories of human cognition. Hence, it can be argued that place nouns are rooted in a major cognitive category. When conceived of as a conceptual category that is of relevance in wordformation (noun derivation, in particular), “location” has been defined as follows: “Lexical items that belong to the conceptual category of Location are those which denote a place in space related to a particular activity and/or an enclosed place surrounding individuals in particular situations” (Haselow 2011: 66). It is a secondary issue what sort of a name we wish to choose in order to be able to refer to this conceptual category − in English or in any other language; whether it is to be labeled as “location”, “place”, “space”, “where”, etc. (cf., for instance, the function “Place” in Jackendoff 1990, the semantic primitive “Where” in Wierzbicka 1996, the case function [+Locus] in Beard 1995 or the feature [+Loc] in the framework developed in Lieber 2004).

3. Deverbal and denominal place nouns In many languages that have the category of place nouns encoded by the method of affixation, the process operates on input forms representing two classes: verbs and nouns.

1330

VII. Semantics and pragmatics in word-formation II: Special cases

This is one of the reasons for the semantic diversification in the output (see section 4). While a deverbal process, in regular instances, calls for a paraphrase like ‘the place of a particular action, process or state’, a derivative based on a noun requires a different explanation of its meaning. Consider, respectively, the following English pairs: refine → refinery ‘a place (factory) where sth is refined’ vs. swan → swannery ‘a place where swans are kept’. In order to reflect the observed bifurcation in the potential input, a generalized semantic formula for the whole class of place nouns should specify disjunctively members of both subclasses; for example: in English, “-(e)ry nouns may […] name locations connected in some way with the entities or activities indicated by their bases” (Adams 2001: 65). However, since the exact mode of derivation is sometimes hard to determine, certain locatives are interpreted as instances of parallel (double) motivation, i.e. they are viewed as coined on verbs and nouns. This manner of treatment is usually advocated for English locatives like printery, brewery, bindery, etc. (cf. Marchand 1969: 284). A complex lexeme of this structural type points, first, to the corresponding agent noun as its immediate base (thus printery ‘a place where a printer works’); secondly, the derivative in question is ultimately based on the verb itself (accordingly: printery ‘a place where sb prints’). As a consequence of this semantic ambivalence, the locative suffix emerges as having two allomorphs, viz. -y and -ery, respectively. Apart from the allomorphs -y and -ery, the suffix in question has one more allomorph, -ry, as in rabbitry (for details, see Marchand 1969: 285). Further details on English locatives are presented in section 6. Similarly, within the class of place nouns in Polish, there are instances of double (or even triple) motivation; cf. druk-arnia ‘printery, print shop’ ← druk ‘print’ / druk-ować ‘to print’ / druk-arz ‘printer’. Moreover, as the evidence from Slavic languages demonstrates, locatives are occasionally based on adjectives rather than verbs or nouns; cf. Polish pusty ‘empty’ → pustynia ‘desert’, Czech úzký ‘narrow’ → úžina ‘strait, inlet’, etc. To sum up, the existence of deverbal as well as denominal locatives and, in particular, cases of double motivation complicate the semantic analysis of place nouns. In the next section, we turn our attention to further problems in their semantics.

4. Semantic subtypes of place nouns The notional boundaries of the category “place noun” in word-formation are not particularly well-defined, especially when evidence from a variety of languages is taken into account (see Luschützky and Rainer 2013 for further details and examples). Suffice it to say that this is due, in part, to the abstractness and relative vagueness of the conceptual category “location”. Besides, specific, narrowly defined, categories like “containers” (e.g., English trough) may be conceptualized differently: some languages view them as, basically, places while others treat them as instruments. It turns out that some examples of the broadly defined category “place noun” (cf. section 2) are better than others, and still other instances are assignable to specific semantic subtypes. As has been mentioned (section 1), the class is represented, prototypically, by products of deverbal nominalization, i.e. locative nominals, which denote the place of a particular action, process or state. However, many deverbal locatives instantiate lexicalization or semantic narrowing

76. Place nouns

1331

and so, accordingly, they must be provided with a variety of specialized paraphrases. Thus, for instance, the Polish noun pływalnia is based on the verb pływać ‘to swim’, but it does not denote any place where one can swim but rather, specifically, ‘a swimming pool’. Crucially, the notion of ‘place’ that is supposed to be the key element in the definition of any place noun, turns out to be rather ambiguous and underspecified. This is because, in many languages, the class of locative nominals subsumes several groups of derivatives which range over related concepts like “open area” (e.g., a field), “building” (e.g., a factory or plant), “room”, etc. Consider, respectively, Polish łowisko ‘fishing ground’ ← łowić (ryby) ‘to fish’, montownia ‘assembly plant’ ← montować ‘to assemble’, poczekalnia ‘waiting room’ ← poczekać ‘to wait’. The semantic interpretation of place nouns gets even more complicated if the admissible input forms are nouns as well as verbs. Indeed, numerous locatives in Polish are based on nouns; consider the following pairs: kartofel ‘potato’ → kartoflisko ‘potato field’, cement ‘cement’ → cementownia ‘cement plant’, mleko ‘milk’ → mleczarnia ‘dairy’, trup ‘corpse’ → trupiarnia ‘morgue’. Overall, such denominal formations can be paraphrased very crudely with the formula ‘place of N’, to be elaborated upon so as to reflect the implicit verbal component. Then the particular semantic subclasses can receive more adequate circumscriptions; e.g., ‘a place where N is grown’, ‘a place where N is sold’, ‘a place where N is stored’, etc. For more examples and discussion on place nouns in Polish, see, for instance, Górska (1982), Grzegorczykowa and Puzynina (1999: 413−414, 447−448), Szymanek (2010: 56−61). Slavic locative formations point to yet another problem of semantics: “[u]nlike other IE [Indo-European] languages, which exhibit a single Locative nominal referring to a location ‘in’, for example, English bakery, Slavic languages have two locative nominalizations, an in-Locative, meaning ‘place in which’ and an on-Locative, meaning ‘place on which’” (Beard 1995: 87−88). This semantic division can be illustrated with the following Serbo-Croatian pairs: čekao-nic-a ‘waiting room’ vs. čekal-išt-e ‘waiting area, hunting blind’, both based on the past tense verbal stem čekal- ‘wait’, igrao-nic-a ‘dance hall, casino’ vs. igral-išt-e ‘playground’, both based on the past tense verbal stem igral‘play, dance’, etc. (Beard 1995: 88). Significantly, “[t]he on-derivation is consistently associated with the Neuter suffix -ište” while the in-derivation is associated with a “range of consistently Feminine affixes” (Beard 1995: 88), exemplified with the following pairs: igl-a ‘needle’ → igl-ar-a, igl-ar-ic-a ‘pin cushion’, pepeo ‘ashes’ → pepel-jar-a ‘ashtray’, pil-a ‘saw’ → pil-an-a ‘sawmill’, račun-i- ‘calculate’ → račun-ic-a ‘arithmetic notebook’, rafin-ir-aj- ‘refine’ → rafin-er-ij-a ‘refinery’ (Beard 1995: 88). Additionally, the class of in-locatives subsumes the type represented by items like mes-ar-a ‘meatshop’ (with feminine declension), related to mes-o ‘meat’ and mes-ar ‘butcher’. There are also masculine markers that spell out in-locative nominalizations in Serbo-Croatian; these are suffixes like -ac, -n-jak, -in-jak, -ar-nik, -nik (see Beard 1995: 89 for a full account, highlighting certain exceptions to, and theoretical consequences of the division outlined here). It is worth stressing that the derivation of place nouns is remarkably productive in Serbo-Croatian, which is additionally evidenced by the existence of parallel forms; cf., for instance, gusk-a ‘goose’ → gušč-ak / gušč-ar-nik / gus-in-jak ‘goose pen’. Likewise, Czech has a range of more or less productive suffixes which partly specialize in rendering specific types of locative functions. Consider, for instance, the suffix -ina, found in some de-adjectival place nouns denoting natural features of the countryside (e.g., rov-n-ý ‘flat, level’ → rov-ina ‘plain’, hlub-ok-ý ‘deep’ → hlub-ina ‘depth’,

1332

VII. Semantics and pragmatics in word-formation II: Special cases

úz-k-ý ‘narrow’ → úž-ina ‘strait, inlet’; all with truncation of the adjectival formative). Several other suffixal patterns exist in the case of place nouns referring to man-made objects (buildings, etc.); for example, the suffixes -na (-árna, -írna, -ovna) in, e.g., herna ‘casino, gaming-house’ ← hrát ‘to play’, kovárna ‘forge’ ← kovat ‘to forge’, válcovna ‘mill’ ← válcovat ‘to roll’, also -iště (hřiště ‘playing field, playground’ ← hrát ‘to play’), -inec (sirotčinec ‘orphanage’ ← sirotka ‘orphan’), -ník (chodník ‘footway, pavement’ ← chodit ‘to go, walk’), -nice (zbrojnice ‘armoury’ ← zbroj ‘armour’), etc. (for details, see Šmilauer 1971: 48−52). It remains to be seen to what extent the particular sense groups, of the sort illustrated above, stand for any wider patterns distributed across the languages of Europe.

5. Overlap between place nouns and other nominalization patterns There are well attested cases of formal overlap between different types of derived nominals, in various languages. The most commonly cited case of this sort involves the categories of agentive and instrumental nominalizations; cf. Payne (1997: 228): “[i]nstrument nominalizations are often identical to agent nominalizations”. This is exemplified with the English suffix -er which appears in both agent nouns like painter and instrument nouns like slicer (see also Beard 1984; Booij 1986, 2010: 77; article 74 on agent and instrument nouns). Moreover, the kind of formal overlap evidenced here can be extended, occasionally, to cover locative nouns as well; cf. painter vs. slicer vs. diner ‘a small restaurant’. According to Kastovsky (1986a: 597), in English, “there is a clear hierarchy of productivity with the morphological pattern V-er of the type Agent − Instrument − Experiencer − Patient − Locative − Action […]”. This triple overlap of nominalization functions (agent / instrument / location), quite marginal in English, appears to be more common in Hungarian: “the same suffix used for agentive and instrumental nominalization can form place nouns (-ó or -ő depending on vowel harmony)” (Comrie and Thompson 1985: 355). The case is illustrated with the following pairs: ír ‘to write’ → író ‘writer’ (agentive); hegyez ‘to sharpen’ → hegyező ‘sharpener (for pencils)’ (instrumental); társalog ‘to converse’ → társalgő ‘place of conversing = parlour’ (locative), mulat ‘to have fun’ → mulató ‘place for having fun = bar’ (locative). Consider, additionally, a few more examples of the locative use of the formatives -ó and -ő: megáll ‘to stop’ → megálló ‘(bus)stop’, vizsgál ‘to examine’ → vizsgáló ‘examination room’, pihen ‘to rest’ → pihenő ‘resting place’, ebédel ‘to have lunch’ → ebédlő ‘dining room, canteen’, műt ‘to operate’ → műtű ‘operating room’. Besides, Hungarian has the suffix -ede/-oda, used by nineteenth-century language reformers in order to coin suitable locatives. For example: étkez(ik) ‘to eat’ → étkezde ‘canteen (pejorative)’, ír ‘to write’ → íroda ‘office’, óv ‘to protect’ → óvoda ‘kindergarten’ (Anikó Szalay, personal communication). Evidently, the formal overlap between place nouns and other classes (categories) of derived nominals, in various languages, is far from accidental. This relationship seems to have a semantic/functional basis and may be due to historical developments. Thus, on the one hand, a diachronic mechanism of semantic extension has been evoked, by virtue of which transpositional action nominals, with originally eventive reading, come to acquire a locative sense. A development of this sort has been reported for Italian

76. Place nouns

1333

(Melloni 2007: 110−114). In Italian, suffixes like -zione, -mento, -ata, -aggio, etc., that are typical of action nouns are also found in nominals with the secondary, and contextdependent, sense of location. For instance, consider the meanings of the nouns entrata ‘entrance’ or rianimazione ‘resuscitation ward’ in, respectively, L’entrata di questo palazzo è maestosa ‘The entrance to this palace is magnificent’ and Il paziente è stato trasferito in rianimazione ‘The patient was transferred to the resuscitation ward’ (Melloni 2007: 112). It is worth noting in this context that “Italian lacks productive systematic morphological means for the formation of locative nouns from verbal bases” (Melloni 2007: 114) and so, in a sense, the process of semantic extension under discussion takes over the function of strictly locative derivation. Melloni (2007: 114) adds that Italian also has a few locative suffixes that combine mainly with nominal bases, including the suffix -eria (see also Melloni 2011). On the other hand, synchronic accounts of the functional affinities between place nouns and other classes of derived nominals often rely on the notion of derivational polysemy where, in fact, certain observed instances of derivational polysemy may be due, precisely, to the workings of semantic extension as a historical process (on derivational polysemy, see, e.g., Beard 1990). Luschützky and Rainer (2013) demonstrate that the perceived polysemy may be due to a variety of factors like reanalysis, ellipsis, homonymization and borrowing. “A well-known example of polysemy in the realm of word-formation is the set of deverbal nouns ending in -er in Dutch, English, and German” (Booij 2010: 77). Nouns in -er have a range of interpretations, including such concepts as agent (both animate and non-animate), instrument, object, event, and causer. The existence of forms like English diner (cf. above) suggests that the polysemy is even more far-reaching: the corresponding “conceptual extension schema” should cover locative uses as well. To take another example: the English noun smelter is dictionary-defined as ‘a factory or machine in which metal is smelted’ (cf. also smeltery (locative) and smelter with agentive meaning). Thus, apart from suffix polysemy, we are dealing here with the polysemy of a particular derivative (whereas “a factory” may be assigned to the class of locatives, “a machine” represents instrumental nouns). Cases of this sort clearly demonstrate that the boundary separating place nouns from instrument nouns is sometimes hard to draw. Let us consider one more example. Dutch has the noun-forming suffix -ij/-er-ij which is unproductive but can be found in some attested locatives, e.g., abd-ij ‘abbey’ ← abt ‘abbot’ (Geert Booij, personal communication). Instead, so to speak, the locative meaning can be conveyed by other, polysemous suffixes; cf. the deverbal noun won-ing ‘lodging’, incorporating the polysemous suffix -ing which normally creates action nouns (cf. Booij and Lieber 2004 on the notion of paradigmatic extension). A number of Dutch words in -erij (or: -er-ij ) can be analyzed in two ways, due to double motivation, i.e. as either deverbal or denominal formations. Thus, for instance, bakkerij ‘bakery’ could be bakk-erij ‘bak-ery’ or bakker-ij ‘baker-y’ (cf. also German Bäckerei ). For more examples of Dutch locatives in -erij (e.g., brouwerij ‘brewery’, drukkerij ‘printing office’) as well as a general account on the suffix -erij, see Hüning (2004). Due to space limitations, we must leave undiscussed other examples of derivational polysemy (and semantic extension) evidenced by locative formations in various languages. The cases that have been reported reveal diverse links between place nouns and agentives, instrumentals (e.g., Rainer 2011 on the agent / instrument / place “polysemy”

1334

VII. Semantics and pragmatics in word-formation II: Special cases

in Romance) as well as collectives (e.g., Adams 2001: 65 and Lieber 2004 on -ery in English; Melloni 2007, 2011 on Italian).

6. Place nouns in English The derivational category of place nouns is weakly productive in present-day English. This becomes apparent, first, when we compare the modest set of English locative nouns with the impressive class of locatives in any Slavic language where, admittedly, the term place noun accounts for a major and full-blown category of derivation. Secondly, within English morphology alone, it is enough to compare words tentatively designated as place nouns to certain other classes of derived nouns like, for instance, agentive or instrumental nominalizations, to realize that patterns of the latter type are far more productive and occupy a more central position in the word-formation system. Diachronically, the situation is fairly complicated. According to Haselow (2011: 119), the category “location” was represented by as many as seven native suffixes in Old English. The suffixes were as follows: -ærn (-ern), -d (-d, -t, -Þ), -dom, -el (-l ), -en, -ung, -0̸ (zero-derivation). However, in Middle English, “suffixation ceased to be used for the extension of the lexicon with respect to nouns denoting locations” (Haselow 2011: 181). In other words, the category “location” was lost altogether, as it “became entirely unexpressed by morphological means during the transition from OE to ME” (Haselow 2011: 211). However, “[a]s a consequence of the Norman Conquest in 1066, English came into close contact with Norman French from which it borrowed a large number of morphologically related words, which over time were segmented morphologically” (Haselow 2011: 265). One of the newly established suffixes was the morpheme -ery (from French -erie; e.g., nun → nunnery). Of course, quite a few locative nouns in -ery are still in use in English and neologisms are occasionally added to the class (cf. Marchand 1969: 284; Adams 2001: 65), but it is open to question, according to Haselow (2011: 269), whether this element can be properly classified as a location-encoding formative, given the fact that it is multifunctional (polysemous) as it encodes a range of meanings like collectivity, acting / behaviour (especially undesirable), characteristic of N, place which is connected with N, V, etc. (cf. also Marchand 1969: 282). Leaving aside the historical controversies surrounding the category “place noun” in English, below we give a brief overview of locative formations in the modern lexicon. Thus, it is remarkable that the following three major process types − suffixation, conversion and compounding − have all given rise to complex nouns that can be assigned to the category in question by virtue of their semantics. First and foremost, English locatives are coined by suffix attachment. As has been mentioned, the principal exponent is the multifunctional suffix -(e)ry (see, e.g., Marchand 1969: 284). Synchronically, the class of relevant output forms may be divided into three patterns: (a) forms that are exclusively deverbal (e.g., boilery), (b) forms that are exclusively denominal (e.g., piggery), and (c) forms that are interpretable as instances of double motivation, being motivated by the basic verb (or noun) as well as the corresponding agentive nominal in -er (cf. bakery, pottery ‘a potter’s workshop or factory’). A few of the synchronically analyzable nouns in -ery ultimately derive from Latin or French, e.g., spicery. Related to -ery is the “unconventional suffix” -(e)(t)eria, with by-forms such as -eria, -teria, -eteria (Baldi and Dawar 2000: 968). This formative − extracted from earlier

76. Place nouns

1335

established forms like cafeteria (cf. Kastovsky 1986b: 417) − has given rise to a locative pattern of creative word-formation. However, it is restricted in meaning as it normally denotes names of shops or retail outlets; cf. basketeria, chocolateria, garmenteria, candyteria, caketeria, cookieteria, drygoodsteria, etc. Since such names often arise as jocular and eye-catching neologisms, the termination serves an expressive function as well (as opposed to the “neutral” suffix -ery). Another suffix of relevance is -age which forms “both collectives and place nouns (the former more productively than the latter)” (Lieber 2004: 42). The following words are instances of place nouns: anchorage, moorage, orphanage, hermitage. By contrast, the multifunctional suffix -dom can hardly be regarded as a locative marker, even though among its several meanings, there is the sense of ‘territory, domain, region’ (Marchand 1969: 263), as in kingdom. Next, there are the English nouns in -(at)ory. This type goes back to Latin nouns in -atorium and, indirectly, to adjectives in -atorius. Some of these forms can be analyzed, synchronically, as place nouns. Consider the following pairs: observe − observatory, purge − purgatory. This sporadic evidence, however, does not stand for a live derivational pattern, even though a few further examples might be added, of place nouns in -ory that are either non-compositional or unanalyzable synchronically (e.g., conservatory, laboratory, lavatory, promontory, territory). Similarly, a handful of loan words in -ary can be assigned to the class of place nouns on semantic grounds, regardless of whether they are synchronically analyzable; cf. granary ‘place for (storing) grain’, library ‘place for books’, apiary ‘place where bees are kept’, etc. There are also attested English place nouns in -arium, e.g., vivarium / vivary ‘game preserve, place for raising live animals’. Finally, one of the major English nominalizers, the suffix -er − productively used in agentive and instrumental derivation − appears, quite exceptionally, in a handful of locative nouns; cf. the relevant senses of the nouns diner ‘a place where you can dine’ and sleeper ‘a train you can sleep in; a bed or sofa in a train, in which you can sleep’ (Heyvaert 2003: 100). Secondly, the locative function in question is a feature of some deverbal nouns derived by conversion (zero-derivation). Even though conversion, as a type of morphological process, is highly productive in English (particularly noun-to-verb conversion), the number of relevant examples of locative nominalization appears to be relatively small. Consequently, Marchand (1969: 375) devotes just a single sentence to the “place” group of deverbal nouns formed by zero-derivation, illustrated with a handful of examples: bend, dump, hangout, hideout, lounge, sink, stand, stop, turn. Further examples and discussion of the pattern in question may be found elsewhere. Thus, for instance, Kastovsky (1986b: 413) juxtaposes the class of ordinary zero-derived locative nouns (e.g., to dump − a dump, to stop − a stop, to turn − a turn) with those that are based on particle verbs (e.g., to hide out − a hideout, to hang out − a hangout). Cetnarowska (1993: 103) considers verb/noun pairs like to carry − a carry ‘a place where a boat is carried across the land between two rivers or lakes’, to hunt − a hunt ‘the area where people regularly hunt (foxes)’, etc. On this basis, a locative division is added to the list of semantic functions attributable to “bare” nominalizations in English. A handful of other relevant examples are listed in Adams (2001: 29): the zero-derived locatives haunt, hide, lounge, pass, retreat, stop. To sum up, it appears that zero-derived nouns account for a relatively small class of place nouns in present-day English. In older periods, this pattern showed changeable productivity: “zero-derivation was the most frequent word-formation type

1336

VII. Semantics and pragmatics in word-formation II: Special cases

used for the formation of nouns denoting locations in OE, but it fell entirely out of use in early ME […]” (Haselow 2011: 180). Thirdly, the concept of location is indispensable in interpreting the semantics of several types of compound nouns in English, notably those which are headed by nouns like shop (cf. bookshop, coffee shop), house (powerhouse, slaughterhouse), room (cloakroom, smoking-room), yard (shipyard, vineyard ), etc. Admittedly, compounding in English ought to be set apart from derivation (by affix or by zero-affix). However, it is sufficient to compare the relevant locative compounds in English with their functional equivalents in languages with a more robust category of place nouns in order to realize that the exact mode of encoding the locative sense is, in fact, a secondary issue. For example, consider the fact that a number of locative compounds of the kind just illustrated have their suffixally derived analogues in Polish, a language where the formation of place nouns is far more productive (while compounding, on the whole, is less productive than in English); to give just a few examples: E. bookshop − P. księgarnia (← książka/księga ‘book’), E. toolshop − P. narzędziownia (← narzędzie ‘tool’), E. smoking-room − P. palarnia (← palić ‘to smoke’). Indeed, the alternative use of either compounding or suffixation is occasionally attested in English as well; cf. vineyard and vinery, printing house (also printhouse) vs. printery.

7. Expanding the class of place nouns So far, we have marginalized one key factor that is normally brought up in descriptive studies of particular derivational categories and their exponents, that is productivity. The limits set for this survey do not allow for any detailed examination of the productivity of individual locative affixes, not to mention the drawing of cross-linguistic comparisons and generalizations. Instead, we take a look, in this section, at a language whose repertoire of locative word-formation has been fairly modest, so that efforts have been made to enrich the category in question. The case in point is Irish. (The account to be given below draws on data and interpretations supplied by Aidan Doyle, personal communication.) In general, place nouns are far less common in Irish than in many other European languages. In order to compensate for the lack of modern terminology in 20th century Irish, the state set up a number of committees for coining new words and expressions corresponding to those current in English. The members of the committees were mostly L2 speakers of Irish. One of the expedients they resorted to was the revival of obsolete affixes, or the extension of the range of existing ones. One such affix is connected with place nouns. In Old Irish the noun lann ‘house’ occurs as the second element of a compound, e.g., bech ‘bee’ + lann → bechlann ‘beehive’. By the beginning of the 20 th century lann survived only in learned formations, most of them coined by lexicographers in the period 1600−1900. It was no longer a separate word, and hence words containing it, like leabharlann ‘library’, from leabhar ‘book’ + lann, were only partially transparent in their semantics, with lann having more the status of a cranberry morph than the second element of a compound. Twentieth-century committees decided to revive -lann as a suffix, and coined in this way a number of locative expressions. For example: bia ‘food’ → bialann ‘restaurant’, pictiúr ‘picture’ → pictiúrlann ‘cinema’.

76. Place nouns

1337

The spoken language of the 20 th century had its own productive device for deriving expressions with the meaning ‘place connected with activity’; namely, the word teach ‘house’, used as the first element in a left-headed phrasal compound, with the second element in the genitive: teach ‘house’ + ól ‘drinking’ → teach óil ‘public-house, tavern’, teach ‘house’ + bainne ‘milk’ → teach bainne ‘dairy’, etc. Thus instead of bialann ‘restaurant’, the traditional compound would have been teach itheacháin ‘house of eating’. It appears that this was a highly productive rule, in that virtually any verbal noun could be compounded with teach as the need arose. To sum up, we are dealing here with two processes, one reflecting the learned register of L2 scholars, the other reflecting the unlearned register of L1 speakers. However, the former process, i.e. the one that is due to the language-planning efforts of L2 speakers, can hardly be said to represent a productive rule. Another point worth noting is that the lexical items produced with the suffix -lann and by compounding rarely have the same meaning. The compounds reflect a pre-industrial way of life, while the 20 th century formations reflect modern, industrialized urban life. Besides, a third pattern, used almost exclusively by L2 speakers, produces locative expressions that are a calque on the corresponding English compounds; for instance: English swimming-pool > Irish linn snámha (linn ‘pool’ + snámha ‘swimming’), English dining-room > Irish seomra bia (seomra ‘room’ + bia ‘food’). Finally, one may note the very limited occurrence of a suffix-like element -ús, which derives from English house in compounds; for instance: bakehouse > bác-ús, cookhouse > coc-ús. The element -ús occurs in a handful of words, and it is highly debatable whether we can really call it an affix in Irish. It is certainly not productive.

8. References Adams, Valerie 2001 Complex Words in English. Harlow: Longman. Baldi, Philip and Chantal Dawar 2000 Creative processes. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 963−972. Berlin/New York: Mouton de Gruyter. Beard, Robert E. 1984 Generative lexicalism. Quaderni di Semantica 5: 50−57. Beard, Robert E. 1990 The nature and origins of derivational polysemy. Lingua 81: 101−140. Beard, Robert E. 1995 Lexeme-Morpheme Base Morphology. A General Theory of Inflection and Word Formation. Albany: State University of New York Press. Booij, Geert 1986 Form and meaning in morphology: The case of Dutch “agent nouns”. Linguistics 24: 503−518. Booij, Geert 2010 Construction Morphology. Oxford: Oxford University Press. Booij, Geert and Rochelle Lieber 2004 On the paradigmatic nature of lexical semantics in English and Dutch. Linguistics 42: 327−357.

1338

VII. Semantics and pragmatics in word-formation II: Special cases

Cetnarowska, Bożena 1993 The Syntax, Semantics and Derivation of Bare Nominalizations in English. Katowice: Uniwersytet Śląski. Coates, Richard 2006 Names. In: Richard Hogg and David Denison (eds.), A History of the English Language, 312−351. Cambridge: Cambridge University Press. Comrie, Bernard and Sandra A. Thompson 1985 Lexical nominalization. In: Timothy Shopen (ed.), Language Typology and Syntactic Description. Vol. 3: Grammatical Categories and the Lexicon, 349−398. Cambridge: Cambridge University Press. Górska, Elżbieta 1982 Formal and functional restrictions on the productivity of word formation rules (WFRs). Nordic Journal of Linguistics 5: 77−89. Grzegorczykowa, Renata and Jadwiga Puzynina 1999 Rzeczownik. In: Renata Grzegorczykowa, Roman Laskowski and Henryk Wróbel (eds.), Gramatyka współczesnego języka polskiego. Morfologia. 3rd ed., 389−468. Warsaw: Wydawnictwo Naukowe PWN. Haselow, Alexander 2011 Typological Changes in the Lexicon. Analytic Tendencies in English Noun Formation. Berlin/New York: Mouton de Gruyter. Heyvaert, Liesbet 2003 A Cognitive-Functional Approach to Nominalization in English. Berlin/New York: Mouton de Gruyter. Hüning, Matthias 2004 Woordensmederij. De geschiedenis van het suffix -erij. Utrecht: LOT. Jackendoff, Ray 1990 Semantic Structures. Cambridge, MA: MIT Press. Kastovsky, Dieter 1986a The problem of productivity in word formation. Linguistics 24: 585−600. Kastovsky, Dieter 1986b Diachronic word-formation in a functional perspective. In: Dieter Kastovsky and Aleksander Szwedek (eds.), Linguistics across Historical and Geographical Boundaries. In Honour of Jacek Fisiak on the Occasion of His Fiftieth Birthday, 409−421. Berlin/New York: Mouton de Gruyter. Lieber, Rochelle 2004 Morphology and Lexical Semantics. Cambridge: Cambridge University Press. Luschützky, Hans Christian and Franz Rainer 2013 Instrument and place nouns: A typological and diachronic perspective. Linguistics 51(6): 1301−1359. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A SynchronicDiachronic Approach. 2nd ed. München: Beck. Melloni, Chiara 2007 Polysemy in word formation. The case of deverbal nominals. Ph.D. dissertation, University of Verona. http://www.univr.it/documenti/AllegatiOA/allegatooa_03454.pdf [last access 25 Nov 2014]. Melloni, Chiara 2011 Event and Result Nominals. A Morpho-semantic Approach. Bern: Lang. Payne, Thomas E. 1997 Describing Morphosyntax. A Guide for Field Linguists. Cambridge: Cambridge University Press.

77. Intensification

1339

Rainer, Franz 2011 The agent-instrument-place “polysemy” of the suffix -TOR in Romance. In: Hans Christian Luschützky and Franz Rainer (eds.), Agent noun polysemy in Indo-European languages (= STUF − Language Typology and Universals 64(1)), 8−32. Berlin/New York: de Gruyter. Šmilauer, Vladimír 1971 Novočeské tvoření slov. Prague: Státní Pedagogické Nakladatelství. Stiebels, Barbara 1999 Noun-verb symmetries in Nahuatl nominalizations. Natural Language and Linguistic Theory 17: 783−836. Szymanek, Bogdan 2010 A Panorama of Polish Word-Formation. Lublin: Wydawnictwo KUL. Wierzbicka, Anna 1996 Semantics. Primes and Universals. Oxford: Oxford University Press. Zeitoun, Elizabeth 2002 Nominalization in Mantauran (Rukai). Language and Linguistics 3(2): 241−288.

Bogdan Szymanek, Lublin (Poland)

77. Intensification 1. 2. 3. 4. 5. 6. 7.

Terminological preliminaries Delimiting intensification Scales Intensifiable and non-intensifiable bases Degrees of intensity The origin of intensifying patterns References

Abstract Intensification, which is a universal feature of language, is realized by morphological means in many languages. The present article first addresses problems of delimitation with respect to neighbouring categories such as quantification, augmentation/diminution, qualification, and emphasis. The main discussion is dedicated to an overview of types of scales and degrees of intensity. The final section contains hints at the origin of patterns of intensification in word-formation.

1340

VII. Semantics and pragmatics in word-formation II: Special cases

1. Terminological preliminaries The word family intensive, intensify, intensifier, intensification occurs in different senses in the linguistic literature. In the present article these terms will only be used to refer to the expression of degree, which is undoubtedly their most common use. Another aspect of our terminology worth highlighting at the outset is that, contrary to everyday usage, intensification will not only refer to the expression of a high degree, but to any kind of degree. Though this terminology is by now relatively well established in the literature concerning the expression of degree, it is by no means universally accepted nor always used with the precision desirable in scholarly works (for some interesting clarifications, cf. Kleiber 2007). There can be no question of reviewing all competing terms found in the literature, an endeavour which alone could easily fill the space allotted to this article, especially if languages other than English were taken into account (on Romance, see Kiesler 2000). In his classical article, Sapir (1944) used the term grading, though in a somewhat more comprehensive sense. Another general term of some currency in the literature is emphasis (cf. section 2.4), but this term is even more polysemous than intensification. Traditional grammar has no term subsuming the whole range of intensification as understood here, but normally focuses only on the expression of a high degree of adjectives, which is called absolute superlative, or elative.

2. Delimiting intensification In the sense of ‘expression of degree’ intensification is a linguistic universal. In her survey of semantic primitives, Wierzbicka (1996: 67) points out that “evidence suggests that all natural languages have a word corresponding to the English word very”. In ontogeny, the ability to distinguish degrees of qualities also manifests itself early on. In the case of one German-speaking girl (Rainer 2010: 166−167), comparatives of the type Denda no gut. Lat besser ‘That no good. Salad better’ occurred already at the age of 2;2, and intensifiers like ganz ‘all’ (ganz nass ‘all wet’) at 2;6. Though intensification as such is undoubtedly a language universal, this of course does not mean that all languages resort to word-formation for its expression. On the contrary, at least in the more familiar Indo-European languages (cf. Rainer 1983 on Italian), analytic expressions of degree by far outnumber synthetic patterns, which nevertheless constitute a substantial part of word-formation. In a typological survey of 42 genetically unrelated languages (Bauer 2002: 41−42), “intensive” verbs were found in 11 languages and “attenuative” adjectives in 9. Typological surveys in this area are hindered, apart from descriptive neglect, by the vagaries of terminology, which are in turn tied to substantial problems of demarcation. Intensification, in fact, shades off imperceptibly into quantification, augmentation and diminution, qualification, and emphasis.

2.1. Intensification and quantification Intensification and quantification are closely related concepts. This is suggested, for example, by the observation that “many degree expressions that are used outside of the nominal

77. Intensification

1341

system start out historically as measure phrases indicating high degree of quantity in the nominal system” (Doetjes 2008: 132). Relevant English adverbs are all, a lot, most, quite or somewhat. In word-formation, the intimate relationship between intensification and quantification is most obvious in the phenomenon of so-called “verbal plurality” (cf. section 4). Several scholars, in the wake of Gary (1979), have therefore called for a unitary account of intensification and quantification. An explicit attempt at defining parallelisms between quantifiers and degrees of intensification can be found in Os (1989: 212).

2.2. Intensification and augmentation/diminution Nominal diminutives and augmentatives, in their referential use, serve the purpose of referring to small or big exemplars of the entity designated by the base noun: Spanish taza ‘cup’ → tacita ‘small cup’, tazón/tazota ‘big cup’, charco ‘puddle’ → charquito ‘small puddle’, charcón/charcote ‘big puddle’, etc. As one can see, nominal diminutives and augmentatives also involve the assignment of a degree to a quality, just like in the case of intensification. The relevant dimension has to be inferred from the base noun (‘volume’ or ‘height’ in the case of a cup, ‘extension’ in the case of a puddle, etc.). Under certain circumstances nominal diminutives and augmentatives can encroach in an even more direct manner on the domain of intensification, viz. when languages allow them to be added also to intensifiable bases, as is the case in Spanish to some extent: susto ‘scare’ → sustito ‘light scare’, sustón/sustote ‘strong scare’, etc. Here, the relevant dimension is ‘scaredness’, a state directly expressed by the base. On the augmentation/ intensification cline in German, cf. Ruf (1996). The relationship between diminution/augmentation on the one hand and intensification on the other is even closer in languages where diminutives and augmentatives are allowed to take adjectival bases. As expected, in such languages adjectival diminutives come close in function to approximative/attenuative patterns, and augmentatives to intensive ones: Spanish tonto ‘silly’ → tontito ‘somewhat silly’, tontorrón ‘very silly’, etc. At least in Spanish, however, it would be inadequate to equate adjectival diminutives with approximative/attenuative patterns, and adjectival augmentatives with intensive ones. Augmentative grandote (← grande ‘big’), for example, is not simply a synonym of intensive grandísimo ‘very big’. Even in its adjectival use, the suffix -ote normally conserves secondary meanings and connotations typical of its augmentative use which explain why its distribution is quite different from that of the pure intensive suffix -ísimo. In the same vein, in Ni azulito ni rosita ‘Neither blue-DIM nor pink-DIM’, the name of a shop for unconventional baby clothing, the diminutive -ito does not have so much a denotative function (‘light A’) as the function of evoking the world of babies. Petersen (1916: 440) has claimed that Latin adjectival diminutives could also express intensification. This, at first sight surprising, function is said to have arisen with bases meaning ‘small’, like Latin parvus, whose diminutive parvulus, according to Petersen, was given the interpretation ‘very small’ rather than ‘somewhat small’. A similar case is made by Rūķe-Draviņa (1953) for Latvian. In her analysis of the corresponding French diminutives, however, which are sometimes also glossed as ‘very, too A’, Delhay (1996: 112−117, 291−307) prefers to view a possible intensive meaning as a side-effect of the primarily subjective meaning of the diminutive. The diminutive suffix -ot of petiot

1342

VII. Semantics and pragmatics in word-formation II: Special cases

‘small-DIM’ (← petit ‘small’), for example, primarily expresses endearment, but since very small children or very tiny things more readily provoke reactions of endearment, an intensive implicature may arise.

2.3. Intensification and qualification Students of degree adverbs have long pointed out the difficulty of separating qualitative from intensive adverbs (Os 1989: 91−92; Peters 1993: 31), since the former are a major source for the latter diachronically. An analogous case from the realm of word-formation are German compounds with the first element bitter ‘bitter(ly)’: bitterkalt ‘bitterly cold’ (← kalt ‘cold’), bitterböse ‘furious’ (← böse ‘angry’), bitterernst ‘extremely serious’ (← ernst ‘serious’). The same kind of smooth transition from one category into the other can also be observed in intensifying noun-adjective compounds of the Germanic languages. Many of the first elements of these adjectives once expressed a live metaphor, but then underwent a process of semantic bleaching turning them into pure intensifiers. While in German stocksteif ‘as stiff as a poker; lit. stick stiff ’, the metaphor is still retrievable even synchronically, stock- has become a kind of intensive prefix in stockbesoffen ‘dead drunk; lit. stick drunk’, stockkatholisch ‘Catholic through and through; lit. stick Catholic’, and similar formations.

2.4. Intensification and emphasis It has already been mentioned that the term emphasis is used in linguistics in different senses. In one of these, it is a rough synonym of intensification, as when patterns expressing a very high degree, such as Italian adjectives in -issimo (cf. bellissimo ‘very very beautiful’ ← bello ‘beautiful’), are called “emphatic”, probably because such patterns normally also convey at the same time particularly strong emotional overtones. But in general the term emphasis is reserved for other functions, as it should be (cf. already Mathesius 1939). Still a relatively close relationship exists between intensification proper and emphatic patterns which enhance the strength of the illocutionary force of a speech act. In the more familiar languages of Europe this kind of emphasis is normally expressed by adverbs or particles: It’s really true, etc. In some languages, however, it can be expressed morphologically. Abaza (Lomtatidze and Klychev 1989: 108), for example, has an emphatic suffix -ʒa- with the meaning ‘really’, added to the verb stem before tense/mood suffixes. The intimate relationship between truth-asseverating emphasis and intensification in our restrictive conception is obvious from at least two observations. First, it is well-known that “truth intensifiers”, as they are also called, are a major diachronic source for degree intensifiers (Peters 1993: 8, 39; cf. really good ), due to a straightforward pragmatic inference. And second, degree intensifiers are sometimes also extended in the opposite direction. Dressler and Merlini Barbaresi (1994: 499), for example, observe that the intensive Italian suffix -issimo in la stessa, la stessissima ‘the same, the very same’ (← stessa ‘same-FEM.’) “only produces emphasis interpretable as upgrading of strength of speaker’s commitment”. This truth-asseverating use was already foreshadowed in jocular

77. Intensification

1343

Latin examples such as ipsissimus ‘his own very self ’ (← ipse ‘himself ’). Another use of the term emphatic which is probably also best classified here is the one referring to the intensification of negation, as exemplified by the Japanese emphatic suffix -nanka ‘at all’: takaku-nanka nai ‘not high at all’ (← takaku ‘high’, nai ‘not’). What this suffix emphasizes is that the speaker wants the negation to be taken at face value. Although it is not generally called “emphatic”, the intensification of relative superlatives like German allerbester ‘very best, best of all’ (← aller ‘of all’, now felt to be a prefix, bester ‘best’) should probably also be classified here. While the kind of emphasis described in the previous paragraph comes close to intensification proper, this is not true for the following contrastive type, although it is also customary in the literature to speak of intensification (cf. Gast 2006). Many languages, in fact, have special emphatic patterns expressing contrast, be it on nouns (Konkani nátu ‘grandson’ → natúc ‘the grandson himself ’), possessives (Irish mo theach ‘my house’ → mo theachsa ‘MY house’), personal pronouns or deictic forms (Kokota ara ‘I’ → arahi, ade ‘here’ → adehi, where the suffix -hi in both cases adds contrastive emphasis). It would probably be preferable to stick to the term emphasis in these cases.

3. Scales There seems to be a wide consensus in the literature now that intensification, or expression of degree, consists in the assignment of some value on a “scale” associated with the intensified word. The ontological interpretation of this kind of metaphor, however, varies considerably according to theoretical orientation. Two of the most prominent schools of thought in the literature of the new millennium are, on the one hand, formal logic, and on the other, cognitive grammar. In their framework of extensional semantics, Kennedy and McNally (2005: 349), for example, define a scale as a “set of ordered degrees”, which in turn are characterized as “abstract representations of measurement”. Gradable adjectives are defined as “relations between individuals [denoted by the external argument of the adjectives; F. R.] and degrees”. What distinguishes gradable adjectives from each other is the dimension along which the ordering relation is defined (height in the case of tall and short, redness in the case of red, etc.). In cognitive terms, a scale is conceived of as an “image schema which provides a gradable dimension to a domain” (Croft and Cruse 2005: 65; cf. also Paradis 2008). The function of intensifiers consists in determining the value of the degree. More interestingly, research conducted on intensification over the last decades has shown that it is also necessary to take into account the structure of scales in order to properly describe the behaviour of intensifiers. The most important distinction to be made, as has long been recognized (cf., for example, Leech and Svartvik 1975: 103), is the one between open or unbounded scales and closed or bounded scales. A relatively reliable way of finding out whether a particular adjective has a closed scale is to test its compatibility with the proportional modifier half (Kennedy and McNally 2005: 352): half full/closed/invisible, etc. vs. *half tall/old/ expensive, etc. The choice of intensifier is heavily dependent on this open/closed distinction. Many intensifiers are restricted to open scale adjectives, for example the Latin prefix per-: peraltus ‘very high’ (← altus ‘high’), perarduus ‘very difficult’ (← arduus ‘difficult’), perdives ‘very rich’ (← dives ‘rich’), etc. Some are restricted to closed

1344

VII. Semantics and pragmatics in word-formation II: Special cases

scales, for example the Polish perfectivizing prefix do-: schnąć ‘to be drying’ → doschnąć ‘to get completely dry’, etc. Others are compatible with both types of scale, for example Turkish “emphatic” reduplication (Dhillon 2009), which operates not only over open scale adjectives, but also over closed scale ones: güpgüzel ‘very pretty’ (← güzel ‘pretty’), upuzun ‘very long’ (← uzun ‘long’), etc. vs. bomboş ‘completely empty’ (← boş ‘empty’), çırılçıplak ‘stark naked’ (← çıplak ‘naked’), etc. Kennedy and McNally (2005: 354−355) argue for the need to distinguish three kinds of closed scales, upper closed scales (pure/impure, safe/dangerous, etc.), lower closed scales (quiet/loud, unknown/famous, etc.) and scales closed on both sides (full/empty, open/closed, etc.). In order to justifiably call something pure, there can be no impurity, while for something to be called loud a minimum amount of noise is sufficient. According to Kennedy and McNally (2005: 369−378), this finer typology of scales is necessary in order to explain, for example, the distribution of the English intensifier much, which is said to select deverbal adjectives with lower closed scales (much criticized, etc.). The same restriction seems to hold for German adjectival compounds with a first element viel- ‘much’: vielbesucht ‘much visited’, vieldiskutiert ‘much discussed’, etc. Scale structure, of course, is not the only factor involved in the definition of the domain of intensifiers. The whole range of restrictions known from other word-formation processes is equally relevant here. Among the pan-European intensive prefixes extra-, hyper-, mega-, super- and ultra-, for example, each has its particular, synchronically unmotivated preferences concerning potential bases (extra- for commercial, ultra- for political terms, etc.). In this area, the availability of large corpora now permits a level of descriptive detail unimaginable two decades ago (cf. Rainer 2003 on Italian -issimo).

4. Intensifiable and non-intensifiable bases Words which have a scale associated with them are called intensifiable or gradable. In traditional grammar, gradability has often been conceived of as a stable feature of a subclass of adjectives, viz. those which allow the formation of a comparative. This view, however, is inadequate in at least two ways. One problematic point is that it seems to imply that the set of adjectives which allow the formation of a comparative is co-extensive with that of intensifiable adjectives in general. This, however, is not the case (Whittaker 2002). “Superlative” adjectives such as French abominable ‘id.’, for example, can hardly take ordinary intensifiers (cf. ?Son dernier livre est un peu/assez/très abominable ‘His last book is a bit/quite/very abominable’), though compatible with the comparative (Cf. Son dernier livre est encore plus abominable que ses livres précédents ‘His last book is even more abominable than his preceding books’). In Italian, this kind of adjective allows the booster suffix -issimo, as well as the comparative adverb più, but less naturally the more sober adverb molto ‘very’: splendidissimo/più splendido vs. ?molto splendido (← splendido ‘splendid’). On the other hand, the concept ‘dead’, which can hardly be combined with the comparative (?John is deader than Mary), is nevertheless intensifiable: stone dead, German mausetot ‘id.; lit. mouse dead’, etc. The use of the question mark instead of the asterisk in order to indicate dubious acceptability in these examples is deliberate, since acceptability judgements concerning

77. Intensification

1345

intensifiability are rarely hard and fast. Many words which look non-intensifiable at first sight can in fact be coerced into an intensifiable reading in a pertinent context (Leech and Svartvik 1975: 103; Paradis 2008: 328). An object, for example, would seem to be either spheric or not, but “in your first drawing the eye globe looks so very spheric” nevertheless sounds quite natural. Paradis (2008) aptly points out “[t]he flexibility of DEGREE in linguistic expressions” (p. 325), which she claims to be “neither a property of grammatical classes nor a property of individual words”, but “a configurational meaning structure” (p. 317). The second problem involved in the traditional view of gradability as a feature of a subclass of adjectives only is that there is now a general consensus that in reality it occurs in other lexical classes as well (Bolinger 1972; Os 1989; Doetjes 2008). Many researchers have also pointed out systematic correspondences among the different classes. The bounded/unbounded distinction, for example, cuts across categories: “Count nouns, noncontinuous verbs and [closed scale] adjectives are BOUNDED, while noncount nouns, continuous verbs and [open scale] adjectives are UNBOUNDED” (Paradis 2008: 331). Kennedy and McNally (2005: 361−367) describe regular correspondences between the aktionsart of verbs and the related deverbal adjectives in English. Doetjes (2008), furthermore, shows that many intensifiers extend over more than one lexical category. An extreme case in point is Warao (Vaquero 1965: 44, 51, 94−98), whose intensive suffixes may be added to all lexical categories except conjunctions and interjections: yakera ‘good’ → yakerawitu ‘very good’, ote ‘away’ → otewitu ‘far away’, ama ‘now’ → amawitu ‘right now’, era ‘much’ → erawitu ‘very much’, ubaya ‘he sleeps’ → ubayawitu ‘he sleeps like a log’, naria ‘I am off ’ → nariawitu ‘I am off-INT’, dima ‘father’ → dimawitu ‘real father’, yatu ‘you (pl.)’ → yatuwitu ‘you-INT’, etc. Adverbs behave very much like adjectives with respect to intensification. In Italian, the intensifying suffix -issimo can be added to underived adverbs associated with an open scale: bene ‘well’ → benissimo ‘very well’, presto ‘soon’ → prestissimo ‘very soon’, etc. While here the intensive suffix follows the adverb, in the case of deadjectival adverbs the intensive suffix precedes the adverbial suffix -mente: Italian serio ‘serious’ → serissimo ‘very serious’ → serissimamente ‘very seriously’; cf. also Polish cichy ‘quiet’ → cichuśki ‘very quiet’ → cichuśko ‘very quietly’. With some adverbs which would seem to be non-gradable from a strictly logical point of view, intensifiers − here the diminutive suffix -ito − can express precision: Spanish ahora ‘now’ → ahorita ‘right now’, etc. In the verbal domain, stative verbs are similar to adjectives with respect to intensification from a semantic point of view. This is particularly true for languages with few adjectives, which normally express states and even qualities by means of verbs: Jah-Hut num ‘to be ripe’ → raʔnum ‘to be very ripe’, hluk ‘to be heavy’ → sraʔluk ‘to be very heavy’, etc. Activity verbs, on the other hand, can normally only be “intensified” along dimensions such as duration, frequency/habituality, number of participants, or amount of result. It is customary to refer to this set of meanings as verbal plurality (Dressler 1968) or pluractionality (Wood 2007). That these semantic modifications form a family of related phenomena is suggested by the observation that in many languages one and the same formal pattern − very often reduplication − covers all or at least some subset of the semantic categories mentioned. Mathesius (1939: 408) has illustrated the phenomenon with the Czech prefix na-: dříti se ‘to struggle’ → nadříti se ‘to struggle hard’, čekati ‘to wait’ → načekati se ‘to wait long’, choditi ‘to walk’ → nachoditi se ‘to walk

1346

VII. Semantics and pragmatics in word-formation II: Special cases

a lot’, vařiti ‘to cook’ → navařiti ‘to cook a large quantity’, etc. Telic verbs, which involve an endpoint and are therefore bounded, have a close affinity to intensifiers with the meaning ‘completely’: German essen tr. ‘to eat (something)’ → aufessen ‘to eat completely; lit. to eat up’, etc. Nominal intensifiable bases are of different kinds. One substantial group is constituted by abstract nouns designating states, qualities or actions: German Heidenspaß ‘terrific fun’ (← Heide lit. ‘heathen’, Spaß ‘fun’), Vollrausch ‘drunken stupor’ (← voll ‘full’, Rausch ‘inebriation’), Riesenkraft ‘enormous strength’ (← Riese lit. ‘giant’, Kraft ‘strength’), Totalverzicht ‘complete renunciation’ (← total ‘complete’, Verzicht ‘renunciation’), etc. The other main group comprises intensifiable personal nouns: German Reaktionär ‘reactionary person’ → Erzreaktionär ‘ultrareactionary person’, Vollidiot ‘complete idiot’ (← voll ‘full’, Idiot ‘id.’), etc. If the noun is not intensifiable in itself, an intensifier picks out some intensifiable property closely associated with the noun: superdetective ‘finding out everything’, superglue ‘sticking well’, superprison ‘very safe’, supertanker ‘very large’, etc.

5. Degrees of intensity As already anticipated in the introduction, intensification as a linguistic term comprises not only a high degree, but all degrees of intensity. There can be no question here of reviewing the countless proposals in the literature concerning the classification of degrees of intensity (cf. Os 1989: 94−100 and Peters 1993: 3−9 for a partial review of the literature), nor the plethoric terminology. Instead, I must limit myself to presenting my own eclectic system based mainly on Os (1989: 99) and Paradis (2008: 321). Unfortunately, there are no typological studies available which would answer the interesting question whether these degrees are universal or, if not, on what points languages may differ. A category which is not always taken into account in the literature is that of e x c e s s and i n s u f f i c i e n c y, as we might call them: overambitious, overcareful, to overburden, to overcook, etc. vs. undercapitalized, underdeveloped, to underachieve, to underprice, etc. These intensifiers denote an upward or downward deviance with respect to some ideal norm. They easily turn into intensifiers denoting a high (cf. hyper- ‘too’ > ‘very’) or attenuative degree (Latin sub- ‘under’ > subacidus ‘slightly sour’ ← acidus ‘sour’, etc.). I will also set apart what Paradis (2008: 321) calls “totality modifiers”, i.e. m a x i m i z e r s (‘completely’) and a p p r o x i m a t o r s (‘almost’). What they have in common is that they require bases with a bounded scale. An example of a morphological approximator is Spanish cuasi-, as in cuasiperfecto ‘almost perfect’. In Urarina (Olawsky 2006: 414), the meaning ‘almost’ is expressed by a range of deverbal suffixes, which however function as moderators with stative verbs, turning, for example, ‘to be big’ into ‘to be quite big’. Examples of morphological maximizers have already been given at the beginning of section 4. Paradis’ “scalar modifiers” instead combine with bases associated with an unbounded scale. Beginning at the top, we must first mention the category of b o o s t e r s (‘very, extremely’). Though the distinction is often blurred in context, it is useful to distinguish

77. Intensification

1347

two subcategories of boosters according to the intensity conveyed. In Italian, for example, the suffix -issimo implies a stronger degree of intensity than the adverb molto ‘very’: molto ricco, anzi ricchissimo ‘very rich, even extremely rich’. Golin (Bunn 1974: 61) has two suffixes with the meaning ‘very’ and one suffix glossed as ‘very very’: taulé ‘few’ → taulé-ta, taulé-ga ‘very few’, taulé-waága ‘very very few’. Another means for boosting intensity found in many languages consists in the iteration of intensive patterns: German alt ‘old’ → uralt ‘very old’ → ururalt ‘very very old’ → urururalt ‘very very very old’, etc., Spanish caro ‘expensive’ → carísimo ‘very expensive’ → carisísimo ‘very very expensive’, Polish (Szymanek 2010: 215) mały ‘small’ → malusi ‘very small’ → malusieńki ‘very very small’ → malusienieczki ‘very very very small’, etc. A somewhat different case can be found in Bulgarian (Manova 2010: 283), where the iteration of a denominal diminutive suffix expresses intensity of the quality introduced by the first diminutive suffix: glava ‘head’ → glavica ‘small head’ → glavička ‘very small head’, etc. Stepping down the scale we come to the category of m o d e r a t o r s (‘quite, fairly’). In Latin, this degree could be expressed by the comparative: pulchrior ‘quite pretty’ (← pulcher ‘pretty’), etc. Sometimes the diminutive suffix -culus was added to the neuter form of the comparative, -ius: longus ‘long’ → longiusculus ‘fairly long’, etc. One step further down we find the category of d i m i n i s h e r s (‘slightly, somewhat’). Many languages have special patterns for this category, often called a t t e n u a t i v e or a p p r o x i m a t i v e : sickish, German dümmlich ‘somewhat silly’ (← dumm ‘silly’), Polish grubawy ‘somewhat fat’ (← gruby ‘fat’), etc. Some languages, interestingly, also resort to reduplication (cf. French bête ‘silly’ → bébête ‘somewhat silly’, etc.), a meaning which at first sight looks anti-iconic (cf. Abraham 2005), if we take into consideration the fact that reduplication more often denotes the high degree (cf. Jingulu kunumburra ‘fast’ → kunumburrakunumburra ‘very fast’, etc.). In Adyghe (Korotkova and Lander 2010: 308), the so-called “simulative” suffix -ŝw can be iterated with the effect of further weakening the degree of the quality expressed by the adjectival base: ʁw ež’ə-ŝw ‘yellowish’ → ʁw ež’ə-ŝwa-ŝw ‘paler than yellowish; lit. yellow-ish-ish’. Urarina (Olawsky 2006: 412−413) has two separate deverbal suffixes in the category of diminishers, -eri, glossed as ‘a little’, and -heriiri, glossed as ‘a tiny little bit’. As already mentioned in section 2.2, attenuation is also expressed in many languages by diminutive suffixes. Verbal diminishers (‘a little bit’) are also often diminutive suffixes and potentially cover the whole range of dimensions mentioned above with respect to verbal plurality: Italian saltare ‘to jump’ → salterellare ‘to perform a lot of small jumps’, Polish gwizdać ‘to whistle’ → pogwizdać ‘to whistle for a while’, etc. At the bottom of the scale, we first encounter a m i n i m a l degree, which in French is normally expressed by the adverb peu (as opposed to un peu) and in German by the adverb wenig (as opposed to ein wenig): peu convaincant/wenig überzeugend ‘rather unconvincing’, etc. With adjectives whose scale has no lower endpoint, the negative prefix un- expresses this minimal degree: unintelligent, etc. If, on the contrary, the adjective implies a lower endpoint, the negative prefix is taken at face value and expresses a n e g a t i v e d e g r e e : unidentified, etc. The difference between these two uses of un- is obvious also from the behaviour of the adjectives in the comparative construction: A is more/less unintelligent/ ?unidentified than B. It is often difficult to assign a particular intensifier to one of these classes with certainty because of a rhetorical tendency to use degree expressions for exaggeration, under-

1348

VII. Semantics and pragmatics in word-formation II: Special cases

statement or irony. Intensifiers therefore may have different functions depending on context and speaker intention.

6. The origin of intensifying patterns Intensifiers are a fruitful object of study for diachronic linguistics. The origin of degree adverbs has indeed already received much attention in the literature (cf. Peters 1993: 39−50; Claudi 2006), while there does not seem to exist a comprehensive study of the origin of morphological intensifiers (on booster prefixes in Old English, cf. Lenker 2008). This is why I must limit myself here to a brief sketch. The most important category quantitatively are boosters. This is because the high emotional involvement typical for boosting requires a constant renewal of the means of expression, since boosters inevitably lose their emotional force the more they are used. Among the source domains of boosters we can identify at least the following: 1) expressions of excess: hyper-, which now often simply means ‘very’ instead of ‘too’, its original meaning: hyper-beautiful, etc.; 2) locative morphemes with the meaning ‘beyond’: Latin trans- ‘beyond’ > French très ‘very’, which is now an adverb, but was long written with a hyphen like a prefix; 3) the relative superlative: Latin -issimus ‘the most’ > Italian -issimo ‘very’, etc.; 4) morphemes denoting the top rank in a hierarchy: German Oberbürgermeister ‘Lord Mayor’ (← ober ‘upper, senior’, Bürgermeister ‘Mayor’) and similar formations yielded derogatory formations such as Obertrottel ‘first-class idiot’ (← Trottel ‘idiot’); French archiconnu ‘well known’ (← connu ‘known’), which contains the same prefix etymologically as archbishop; the recent use of top as an intensifier (cf. top famous people) can also be classified here; 5) augmentatives: megastore > mega expensive, Italian omone ‘big man’ (← uomo ‘man’) > benone ‘very well’ (← bene ‘well’), etc.; 6) quantifiers: German vielgeliebt ‘loved a lot’ (← viel ‘much’, geliebt ‘loved’), etc.; 7) elements with a qualifying meaning: cf. the German bitter-compounds in 2.3; 8) a prefix expressing origin: German Urzeit ‘primeval times’ (← Zeit ‘time’; urultimately comes from uz ‘out’) → uralt ‘very old’ (← alt ‘old’), etc.; 9) metaphors, as first elements of compounds: German riesengroß ‘very tall’ (← Riese ‘giant’, groß ‘big’), etc.; 10) verbs expressing a consequence, as first elements of compounds: German bettelarm ‘very poor’ (← betteln ‘to beg’, arm ‘poor’), quietschvergnügt ‘happy as a sandboy’ (← quietschen ‘to squeak’, vergnügt ‘happy’), etc.; 11) taboo words: German scheißegal ‘absolutely indifferent’ (← scheißen ‘to shit’ or Scheiße ‘shit’, egal ‘indifferent’); 12) iconicity, through reduplication, possibly also gemination: Konkani sobit ‘pretty’ → sobitti ‘very pretty’, etc. This collection is certainly incomplete, but gives a sufficient idea of the multiplicity and heterogeneity of pathways which can lead to the coming into being of morphological boosters. A fact worth pointing out is that degree adverbs never seem to become bound intensifiers (Claudi 2006: 364), while the opposite evolution is well attested: Latin trans- > French très, super- > super adv. (cf. German das steht dir super ‘that suits you super well’), ur- > ur adv. in Viennese (cf. das taugt mir ur ‘I enjoy that very much’), etc. The second most frequent category are diminishers. As we have already mentioned, these are often difficult to distinguish from diminutives in many languages, which therefore constitute one important source domain. Another source domain are patterns ex-

77. Intensification

1349

pressing resemblance, like Latin -aster: filiaster ‘son-in-law’ (← filius ‘son’) > surdaster ‘somewhat deaf ’ (← surdus ‘deaf ’), English childish > bluish, German königlich ‘royal’ (← König ‘king’) > dümmlich ‘somewhat silly’ (← dumm ‘silly’), etc.

7. References Abraham, Werner 2005 Intensity and diminution triggered by reduplicating morphology: Janus-faced iconicity. In: Bernhard Hurch (ed.), Studies on Reduplication, 547−568. Berlin/New York: Mouton de Gruyter. Bauer, Laurie 2002 What you can do with derivational morphology. In: Sabrina Bendjaballah, Wolfgang U. Dressler, Oskar E. Pfeiffer and Maria D. Voeikova (eds.), Morphology 2000. Selected Papers from the 9 th Morphology Meeting, Vienna, 24−28 February 2000, 37−48. Amsterdam/Philadelphia: Benjamins. Bolinger, Dwight 1972 Degree words. The Hague: Mouton. Bunn, Gordon 1974 Golin Grammar. Ukarumpa: Summer Institute of Linguistics. Claudi, Ulrike 2006 Intensifiers of adjectives in German. Sprachtypologie und Universalienforschung 59: 350−369. Croft, William and D. Alan Cruse 2005 Cognitive Linguistics. Cambridge: Cambridge University Press. Delhay, Corinne 1996 Il était un “petit X”. Pour une approche nouvelle de la catégorisation dite diminutive. Paris: Larousse. Dhillon, Rajdip 2009 Turkish emphatic reduplication: Balancing productive and lexicalized forms. Grazer Linguistische Studien 71: 3−10. Doetjes, Jenny 2008 Adjectives and degree modification. In: Louise McNally and Christopher Kennedy (eds.), Adjectives and Adverbs. Syntax, Semantics, and Discourse, 123−155. Oxford: Oxford University Press. Dressler, Wolfgang U. 1968 Studien zur verbalen Pluralität. Iterativum, Distributivum, Durativum, Intensivum in der allgemeinen Grammatik, im Lateinischen und Hethitischen. Wien: Verlag der Österreichischen Akademie der Wissenschaften. Dressler, Wolfgang U. and Lavinia Merlini Barbaresi 1994 Morphopragmatics. Diminutives and Intensifiers in Italian, German, and Other Languages. Berlin/New York: Mouton de Gruyter. Gary, Edward N. 1979 Extent in English. A unified account of degree and quantity. Ph.D. dissertation, University of California, Los Angeles. Gast, Volker 2006 The Grammar of Identity. Intensifiers and Reflexives in Germanic Languages. London: Routledge. Kennedy, Christopher and Louise McNally 2005 Scale structure, degree modification, and the semantics of gradable predicates. Language 81: 345−381.

1350

VII. Semantics and pragmatics in word-formation II: Special cases

Kiesler, Reinhard 2000 Où en sont les études sur la mise en relief ? Le Français Moderne 68: 224−238. Kleiber, Georges 2007 Sur la sémantique de l’intensité. In: Juan Cuartero Otal and Martina Emsel (eds.), Vernetzungen. Bedeutung in Wort, Satz und Text. Festschrift für Gerd Wotjak zum 65. Geburtstag. Vol. 1, 249−261. Frankfurt/M.: Lang. Korotkova, Natalia and Yury Lander 2010 Deriving affix ordering in polysynthesis: Evidence from Adyghe. Morphology 20: 299− 319. Leech, Geoffrey N. and Jan Svartvik 1979 A Communicative Grammar of English. London: Longman. Lenker, Ursula 2008 Booster prefixes in Old English − an alternative view of the roots of ME forsooth. English Language and Linguistics 12: 245−265. Lomtatidze, Ketevan and Rauf Klychev 1989 Abaza. In: B. George Hewitt (ed.), The Indigenous Languages of the Caucasus. Vol. 2: The North West Caucasian Languages, 89−154. Delmar, NY: Caravan. Manova, Stela 2010 Suffix combinations in Bulgarian: Parsability and hierarchy-based ordering. Morphology 20: 267−296. Mathesius, Vilém 1939 Verstärkung und Emphase. In: Mélanges de linguistique offerts à Charles Bally, 407− 413. Geneva: Georg. Olawsky, Knut J. 2006 A Grammar of Urarina. Berlin/New York: Mouton de Gruyter. Os, Charles van 1989 Aspekte der Intensivierung im Deutschen. Tübingen: Narr. Paradis, Carita 2008 Configurations, construals and change: Expressions of DEGREE. English Language and Linguistics 12: 317−343. Peters, Hans 1993 Die englischen Gradadverbien der Kategorie booster. Tübingen: Narr. Petersen, Walter 1916 Latin diminution of adjectives. Classical Philology 11: 426−451. Rainer, Franz 1983 Intensivierung im Italienischen. Salzburg: Institut für Romanistik. Rainer, Franz 2003 Studying restrictions on patterns of word-formation by means of the Internet. Italian Journal of Linguistics 15: 131−140. Rainer, Franz 2010 Carmens Erwerb der deutschen Wortbildung. Vienna: Verlag der Österreichischen Akademie der Wissenschaften. Ruf, Birgit 1996 Augmentativbildungen mit Lehnpräfixen. Heidelberg: Winter. Rūķe-Draviņa, Velta 1953 Adjectival diminutives in Latvian. Slavic Review 31: 452−465. Sapir, Edward 1944 Grading: A study in semantics. Philosophy of Science 11: 93−116. Szymanek, Bogdan 2010 A Panorama of Polish Word-Formation. Lublin: Wydawnictwo KUL. Vaquero, Antonio 1965 Idioma Warao. Caracas: Sucre.

78. Negation

1351

Whittaker, Sunniva 2002 La notion de gradation. Applications aux adjectifs. Bern: Lang. Wierzbicka, Anna 1996 Semantics. Primes and Universals. Oxford: Oxford University Press. Wood, Esther Jane 2007 A semantic typology of pluractionality. Ph.D. dissertation, University of California, Berkeley.

Franz Rainer, Vienna (Austria)

78. Negation 1. 2. 3. 4. 5.

Introduction Semantic subcategories of negation Restrictions on negative prefixation The influence of negative prefixes on argument structure and further derivation References

Abstract As a universal concept negation is present in all languages and can be studied from different perspectives. The present article is dedicated more specifically to the expression of negation, privation, opposition, contrariety or other meanings closely related to negation by morphological means. The article also discusses the restrictions on negative prefixation, as well as the influence of negative prefixation on other aspects of grammar.

1. Introduction Negation is a concept that is expressed in all languages. Human beings view the world through the lens of contrasts such as good/bad or positive/negative. But negation is a phenomenon that extends far beyond language proper. Therefore, apart from linguistics, it is the object of study of disciplines as diverse as logic, psychology, philosophy, and mathematics. Rigorous treatments of negation from a general-linguistic perspective can be found in the works of Jespersen (1917), Klima (1964), Zimmer (1964), Ladusaw (1980), Attal and Müller (1984), Payne (1985), Brütsch, Nussbaumer and Sitta (1990), Laka (1994), Klosa (1996), Mollidor (1998), and Horn (2001). For Spanish, which will be at the center of this article, see Llorens (1929), Molho (1962), Gyurko (1971), Ibáñez (1972), Voigt (1979), Bosque (1980), Hernández Paricio (1985), Bustos (1986), Montero Curiel (1999), and Sánchez López (1999). Besides Spanish, occasional examples will also be

1352

VII. Semantics and pragmatics in word-formation II: Special cases

borrowed from the languages treated in the final two volumes of this handbook, where these are particularly suitable for the illustration of aspects of general interest. The mechanisms available for the expression of negation are manifold, and each language uses a different set thereof, be they syntactic, lexical, morphological or gestural. Italians, Germans, Spaniards, the French or the British express negation by moving the head along the horizontal axis, while other European people, such as the Bulgarians, move the head along the vertical axis, a movement that signals affirmation in most parts of the world. Among the grammatical processes used for expressing negation we may distinguish sentential negation, phrasal negation and word-level negation (cf. Sánchez López 1999: 2563−2565). Sentential negation contradicts the meaning of a sentence (e.g., They did not dance vs. They danced ); it is generally expressed in European languages by a negative adverb or particle, but can also be realized morphologically, for example, by the inflectional suffix -ma- in Chuvash (see article 190 on Chuvash, section 2): tašlamarĕ ‘(S/he) didn’t dancе’ ← tašă ‘dance, n.’ + -la- (verb-forming affix) + -ma- (negation) + -r- (past tense) + -ĕ (3rd person), or by the inflectional prefix da- in Aghul (see article 198, section 3.2), which must be added to the light verb (e.g., ʜazur daxas ‘not to get ready’ ← ʜazur xas ‘to get ready’). In Itsari (see article 202 on Dargwa, section 6) some verbs are negated via reduplication. In phrasal negation the negative element does not have scope over the whole sentence, but only over a phrase (e.g., Not all of them danced vs. All of them danced ). In word-level negation it is a word or, to be more precise, a lexeme, that is negated (e.g., immoral vs. moral ). Word-level negation does not render the sentence negative that contains the word, as can be inferred from the behavior of polarity items (e.g., This is not clear for anybody vs. *This is unclear for anybody). Finally, negation can also be embodied in simplex lexemes, as is the case with the negative auxiliary ńiisʲ in Nenets (see article 175, section 4.3.3). In the present article we will concentrate on word-level negation, which is normally expressed by prefixation in the languages of Europe. Note, however, that suffixes are commonly used to express the concept of privation, and some languages have suffixes deriving negative participles from verbs (e.g., Finnish rakasta-a ‘to love’ → rakastamaton ‘unloved’, see article 176, section 4.2; Turkish görül- ‘to be seen’ → görül-medik ‘unseen, unprecedented’, see article 184, section 4.2.3). In many languages, negative affixes present problems of delimitation with respect to particles and adverbs, from which they have often arisen via g r a m m a t i c a l i z a t i o n . In some languages, negative prefixes are still identical with the adverb or particle expressing negation in syntax, as in G. Nicht-Raucher, Hungarian nemdohányzó, Sp. no fumador, all meaning ‘non-smoker’. Even in these cases, however, syntactic criteria allow one to distinguish the prefix from the particle/adverb (e.g., Sp. Juan no es fumador ‘John is not a smoker’ vs. Juan es no fumador ‘John is a non-smoker’). Privative affixes are sometimes still transparently related to the preposition ‘without’ (e.g., Sp. sin, Russian bez). In the Digor dialect of Ossetic (see article 173, section 4.1.1) word order can distinguish between the use of ɐnɐ ‘without’ as adverb and prefix (e.g., ači ɐnɐzongɐ adɐjmag this NEG-known person ‘this unknown person’ vs. ɐnɐ ači zongɐ adɐjmag without this known person ‘without this known person’). Another source of privative affixes are adjectives expressing lack of something (e.g., E. -less, which is etymologically related to loose, lose and loss, as well as to G. -los). An early stage of such a construction can be observed in Lak (see article 201, section 3.2), where a negative adjective is

78. Negation

1353

formed by adding the negative copula -aq:a- ‘not being’ to the base (e.g., taχsir-baq:as:a ‘innocent; lit. guilt III.not-being’, æq’lu-baq:asa ‘insane, crazy; lit. mind III.notbeing’, k:ak:an χ:uj-baq:a-s:a ‘ugly; lit. looking good III.not-being’). A similar situation obtains in Botlikh (see article 204, section 3.2), where privative adjectives are formed by adding the negative auxiliary participle guč’a-b ‘not been/being, absent’ to a noun (e.g., míq’i guč’a-b ‘roadless’ ← míq’i ‘road’).

2. Semantic subcategories of negation At the outset, the term negative prefixation is in need of clarification. Under the label ‘negative’ many handbooks of morphology subsume concepts certainly close to that of negation, but not absolutely identical with it, such as opposition, contraposition, privation, separation, reversal, or even diminution. Furthermore, in the wake of Aristotle (cf. Hernández Paricio 1985: 131−135; Hamawand 2009), many authors make subtler distinctions within the concept ‘negation’ itself. In the case of so-called c o n t r a r y n e g a t i o n , both the positive and the negative predicate can be true at the same time: John is neither friendly nor unfriendly, for example, is not a contradictory statement. There is a middle ground on the scale of friendliness which is neither covered by friendly nor by unfriendly. In the case of c o n t r a d i c t o r y n e g a t i o n , on the other hand, this middle ground does not exist (philosophers speak of the law of excluded middle): John is neither married nor unmarried contains a contradiction. As can be seen from these two examples, one and the same affix can express both shades of meaning at the same time, but this need not necessarily be so. The prefix non-, for example, is decidedly oriented towards the contradictory pole (e.g., non-toxic, non-smoker). In the present article, a broad view of the subject will be adopted, including not only negation in the strict sense and privation, but also opposition, separation, reversal and diminution, since in these cases too some reality is negated. N e g a t i v e prefixes in the strict sense are those which, more or less forcefully, negate the content of the base to which they are attached, e.g., E. illegal vs. legal, unnecessary vs. necessary, discontinue vs. continue, G. missglücken ‘to be a failure’ vs. glücken ‘to be a success’, etc. Denominal formations such as Sp. desconfianza ‘distrust’ vs. confianza ‘trust’, It. disamore ‘disaffection’ vs. amore ‘love’, etc. are probably better viewed as privative, but the boundary is fuzzy. Negative prefixes are also occasionally added to pronouns, as in Chuvash (see article 190, section 4), where the negative pronouns are derived from the indefinite pronouns by prefixing ni- (e.g., nikam ‘nobody’ ← kam ‘who’, nihăš ‘no one’ ← hăš ‘which’, nimĕn ‘nothing’ ← mĕn ’what’), or Rutul (see article 195, section 4.3), where they are derived from the interrogative pronouns by adding the suffix -ni (e.g., vɨš-ni ‘no one’ ← vɨš ‘who’). This seems to be the central group in most languages, which is probably why the label ‘negative’ has come to be used as a cover term for related concepts as well. Karaim (see article 189, section 3.2) deserves special mention, since the antonymous adjective is formed in this language by compounding, adding the negative copula tiuviul’ ‘is not’ to the adjective (e.g., tatuvlu tiuviul’ ‘not tasty’ ← tatuvlu ‘tasty’). P r i v a t i v e affixes, which are also called c a r i t i v e , especially in Uralic linguistics, denote the loss or lack of something. From an etymological point of view, the first of

1354

VII. Semantics and pragmatics in word-formation II: Special cases

these terms is derived from the Latin transitive verb privare ‘to deprive’, while the base of caritive is intransitive carēre ‘to lack’. One might therefore expect the first to refer to deprivation or loss, while the second could simply refer to lack of something, without any implication that the entity of which the quality or state is predicated had possessed beforehand what it now lacks. But in reality these two terms are normally used as synonyms. The two concepts, by the way, are difficult to distinguish in many cases: A fatherless child, for example, can refer just as much to a child that has lost his or her father (loss), as to one who never knew him (lack). The boundary is equally blurred between privative/caritive meaning and negative meaning proper, which is why some authors prefer to speak of “negative-privative” affixes. Just think of a word like nonaggression. Is this a simple negation of aggression, or a lack thereof ? Privative/caritive affixes are very well represented in European languages, both as prefixes and as suffixes. The clearest cases are denominal adjectives such as the following: E. stressless, G. risikolos ‘riskless’, Sp. sinvergüenza ‘shameless’, Nenets ńe-sʲa ‘unmarried’ (← ńe ‘woman’), Finnish pilvetön ‘cloudless’, Hungarian felhőtlen ‘cloudless’, Turkish telsiz ‘wireless, walkietalkie’, Kabardian a:da-nʃa ‘fatherless’, Russian bez-lošad-nyj ‘horseless, carless; lit. without-horse-ADJ’, etc. In Estonian, the privative/caritive suffix can be preceded by the plural suffix: sõpra-de-tu ‘without friends; lit. friend-PL-PRIV’. A second series consists of denominal nouns such as the following: E. nonsense, Russian neporâdok ‘mess, minor malfunction; lit. not-order’, Ossetic ɐ-gad/ɐ-gadɐ ‘dishonor’ (← kad/kadɐ ‘glory, honor’), etc. Sometimes, r e v e r s a t i v e prefixes are grouped together with privative ones. Reversative prefixes are those that refer to the undoing of an action, or more precisely, of a state caused by the verb serving as the base, as in E. unfasten, untie, unwrap, disassemble, or G. entsperren ‘to unlock’. It would probably be preferable to reserve the term privative verb for those cases where the base is a noun of whose referent(s) the entity referred to by the object of the verb has been deprived, as in Sp. desinsectar ‘to clear of insects’ (← insecto ‘insect’; there is no verb *insectar). Reversative prefixes should better be grouped together with prefixes expressing o p p o s i t i o n . A prefix expressing opposition from the nominal domain is counter- (and its equivalents in other European languages, e.g., F. contre-, G. Gegen-): counterrevolution, for example, refers to a revolution whose aim is the reversal of a political system brought about by a previous revolution. In a similar vein, an antipope is opposed to a pope considered to be the legitimate pope by others. Note that in German, but not in English, the counterrevolution and the antipope share the same prefix: Gegenrevolution, Gegenpapst. A meaning closely related to that of reversative affixes is change of direction as expressed in countermarch ‘march in the reverse direction’, Sp. desandar (el camino) ‘to retrace one’s steps’ (← andar ‘to go’). Note that in German counter- must be translated differently in counterrevolution and countermarch: While the first one is translated as Gegenrevolution, the equivalent of the latter is Rückmarsch (← rück ‘back’, an allomorph of the adverb zurück used in compounds). Other meanings are still farther away from the prototypical negative meaning. A prefix like pseudo- ‘false, fake’, for example, contains a negative feature that clearly comes out in the paraphrase ‘not genuine’: pseudoscience is not genuine science, it only pretends to be science, like astrology or alchemy. In Yiddish, the prefix kmoi- serves the same function (e.g., kmoy-visnshaftlekh ‘pseudo-scientific’). There is also an overlap between temporal-locative and negative prefixes. The meaning ‘former’ of ex- (Hoppe

78. Negation

1355

1999; e.g., ex-husband ), for example, could also be counted among the marginally negative prefixes, since one plausible paraphrase would be ‘no longer’. And the locative prefix extra- invites negative inferences in some of its uses: an extracurricular activity is not part of the curriculum, and the defining characteristic of an extraparlamentary opposition is that it is not part of the parliament. D i m i n u t i o n or a t t e n u a t i o n (cf. Montero Curiel 2001), though not negative notions in the strict sense, also show some tenuous relationship with negation. They form one pole of contrasting pairs of words and often, though not necessarily, have implications of insufficiency, lesser importance, or at least non-prototypicality: hypoglycemia vs. hyperglycemia, underdeveloped vs. developed; minigolf vs. golf, microorganism vs. macro-organism; Vice-Chancellor vs. Chancellor; semi-vowel vs. vowel, etc. The great diversity of semantic values which we have just seen led Leal Cruz (1989− 90, fn. 2) to call the corresponding Spanish prefixes “non-positive”. No wonder then that, given such semantic diversity and fluidity, the classifications of negative prefixes in handbooks and specialized works vary widely. The issue is further complicated by the fact that many prefixes are polysemous and can express more than one meaning. Varela and Martín (2009: 76), for example, attribute to the Spanish prefix des- no less than seven different meanings: negation (e.g., desconocer ‘to not know’), reversion (e.g., desaparecer ‘to disappear’), privation (e.g., desplumar ‘to pluck’ ← pluma ‘feather’), separation (e.g., descarrilar ‘to be derailed’ ← carril ‘rail’), result (e.g., destrozar ‘to shatter’ ← trozo ‘piece’), instrument (e.g., despinzar ‘to remove with tongs’ ← pinza ‘tongs’) and intensity (e.g., deslavar ‘to wash superficially’). We therefore concur with Ibáñez (1972: 2), when he states: “The boundaries of ‘negativity’, i.e. the boundaries between what is negative and what is no longer negative, are fuzzy.”

3. Restrictions on negative prefixation As other prefixes, negative prefixes are generally attached in front of the base without further formal changes. The final vowel of vowel-final prefixes may coalesce with the first vowel of the base or be dropped (e.g., It. contrattacco ‘counterattack’ ← contro ‘counter’ + attacco ‘attack’), but not necessarily (e.g., Fr. contre-espionnage ‘counterespionage’). The same is true for consonant-final prefixes, which sometimes show allomorphs due to assimilation (e.g., illegal ← in- + legal ), and sometimes not (e.g., G. unlogisch ‘illogical’). The more loosely integrated negative prefixes can occasionally be factored out (e.g., pro- and anti-government protests). The rules governing factorization, however, vary from language to language. German, for example, tolerates un- oder amoralisch, while in- ou amoral is not possible in French. Negative prefixes can be added to already prefixed bases, as long as this makes sense from a semantic point of view (e.g., Sp. desencarecer ‘to make less expensive’ ← des‘un-’ + encarecer ‘to make more expensive’ ← en- (factitive) + caro ‘expensive’, -er (infinitive ending)). There are many language-specific restrictions on the combination of prefixes that need not concern us here. One general restriction, however, is worth mentioning: Though negation is recursive in logic, in natural languages a prefix expressing contrary negation hinders the attachment of another negative prefix (e.g., G. *ununhübsch ‘ununpretty’). This is all the more noteworthy since adjectives with negative

1356

VII. Semantics and pragmatics in word-formation II: Special cases

prefixes can be subjected again to phrasal negation in some languages, for example German (e.g., nicht unhübsch ‘quite pretty; lit. not unpretty’). This restriction probably is just an instance of the more general restriction postulated by Zimmer (1964) according to which negative affixes in the strict sense of the word must attach to a positive base: the oddness of, for example, G. *unhässlich ‘unugly’ is also accounted for by this general restriction. Prefixes expressing contradictory negation are not subject to this restriction and can be attached outside other negative prefixes (cf. Martín 1995: 475), for example no- outside of des- in the Spanish example un asunto no-desconocido ‘a not unknown affair’). The restriction does not apply either to negative prefixes expressing opposition (cf. anti-anti-missile system, counter-counter-revolution, etc.). Some morphologists hold that prefixes cannot change the word-class of the base. Most negative prefixes obey this restriction. Un- in unclear transforms an adjective into another adjective, It. dis- in disamore ‘disaffection’ transforms a noun into another noun, and un- in untie transforms a verb into another verb. Some patterns, however, seem to be difficult to reconcile with this position. One such pattern are Spanish formations of the type anti-caspa ‘anti-dandruff ’, as in champú anti-caspa ‘anti-dandruff shampoo’. Such formations show adjectival behavior to a certain extent. They can be placed after a nominal head, a position typically occupied by adjectives, and some can even be found in predicative position (e.g., soy decididamente anti-gobierno ‘I am decidedly anti-government’), though not very frequently. Since in Spanish the noun can be placed after the head noun in appositions, some scholars (cf. Martín 2005) prefer to interpret expressions of the anti-caspa type as appositions, thereby avoiding a split of anti- into a nominal and adjectival prefix. A further aspect that shows that expressions of the anti-caspa type have not yet attained the status of fullfledged adjectives is the fact that they do not show plural agreement with plural head nouns, as normal adjectives do (cf. champús anti-caspa/*anticaspas). It is probable that the awkwardness of the anti-caspa pattern on the background of Spanish word-formation is a consequence of its foreign − probably English − origin. A second pattern that seems to cause problems for the prohibition against word-classchanging prefixes are verbs of the type desinsectar ‘to clear of insects’. If we take insecto ‘insect’ to be the base of this verb, as the paraphrase seems to suggest, one could be led to claim that des- changes the word-class, since -ar is only an inflectional ending (infinitive). The answer to this question is highly theory-dependent. Suffice it to say here that some scholars have tried to preserve the non-category-changing nature of des- by assuming that desinsectar is not directly derived from insecto, but from a hypothetical verb *insectar, which happens not to be attested, while others see des- as only one part of a parasynthetic pattern consisting of the prefix and a zero suffix, which is made responsible for the change of word-class. Since the whole question is dealt with in detail in article 29 on parasynthesis in Romance, the reader interested in the intricacies of the problem is deferred to this article. Another property often ascribed to prefixes is that they tend to accept bases of different word-classes more freely than suffixes. At first sight, this also seems to be correct for negative prefixes. Sp. des-, for example, is found with verbs (e.g., desaparecer ‘to disappear’), nouns (e.g., desamor ‘coldness, indifference’ ← amor ‘love’) and adjectives (e.g., desigual ‘unequal, uneven’). Nevertheless, most prefixes appear to have a clear preference for one particular word-class: in Spanish at least, anti- prefers nouns, in- on the contrary adjectives. A further restriction ties this prefix to bases ending in the suffixes

78. Negation

1357

-ble (e.g., inacabable ‘never-ending’ ← acabable ‘achievable’) and -do (e.g., inacabado ‘unfinished’ ← acabado ‘finished’), which are by far the most productive subpatterns.

4. The influence of negative prefixes on argument structure and further derivation Negative prefixes in the strict sense do not normally affect the argument structure of the base, just like negation in general: That’s not / (un-)clear to me, The object is not / (un-)affected by acids, etc. In some cases, however, a negative prefix can also be seen to affect the argument structure of the derived verb. While Sp. aparecer ‘to appear’, for example, can only be followed by en ‘in’, desaparecer ‘to disappear’ tolerates both en ‘in’ and de ‘from’: El libro apareció en mi estudio ‘The book appeared in my study’ vs. El libro desapareció en/de mi estudio ‘The book disappeared in/from my study’. As the parallelism between English and Spanish suggests, this change in argument structure seems to be a straightforward consequence of the semantic change operated by the prefix. The new concept of disappearance contains a feature of ‘movement away’ which is absent from the concept of appearance, and it is this new feature which licenses the preposition de in Spanish and from in English. If the preposition en continues to be compatible with desaparecer this is because the ‘movement away’ implied by this verb can also take place in a certain space: en mi estudio/in my study, in this case, have the simple status of adjuncts. Another case in point is the effect of the negative prefix des- on the choice of mood in the subordinate clause. Since negative predicates require the subjunctive in Spanish, a verb with a negative prefix will have the effect of requiring the subjunctive (e.g., Pedro conoce que hay un nuevo reglamento ‘Peter knows that there are new regulations’ vs. Pedro desconoce qua haya un nuevo reglamento ‘Peter ignores that there are new regulations’). While in the example treated in the last paragraph the argument structure and the syntactic behavior of the verb can be seen to be automatic consequences of the conceptual structure of the derived verb, in other cases the change in argument structure and syntactic behavior seems to be rather arbitrary from a synchronic perspective. In the pairs esperar ‘to wait’ vs. desesperar ‘to drive mad’, mentir ‘to lie’ vs. desmentir ‘to deny’, caminar ‘to walk’ vs. descaminar ‘to misdirect’ the verb prefixed with des- is transitive, while the non-prefixed verb is intransitive. But there seems to be no way to explain the change from intransitive to transitive as a consequence of the attachment of des-. These pairs can no longer be related by a productive rule, their relationship is more or less lexicalized. The transitivity of the des-verbs must therefore be stated as a brute fact in the respective lexical entries. The same essentially applies to an intransitivereflexive pair such as vivir ‘to live’ vs. desvivirse (por algo) ‘to crave (sth.)’: While the choice of the preposition por by desvivir is predictable form the meaning ‘to crave’, the meaning ‘to crave’ cannot be derived in synchrony from ‘to live’ plus that of des-, nor can the change from intransitive to reflexive be accounted for in a regular manner. A negative prefix can also influence the further derivability of a derivative, since some patterns of word-formation are sensitive to the presence of a negative feature in the base. The Italian suffix -aggine (cf. Rainer 1989: 49), for example, forms quality

1358

VII. Semantics and pragmatics in word-formation II: Special cases

nouns from negative bases, which is why we sometimes find a derivative from the negative derivative, but not from its base (e.g., insensataggine ‘craziness’, but not *sensataggine ‘sensibleness’). In Spanish, but also in other Romance languages, the conversion of qualifying adjectives into personal nouns requires that the property be somehow noteworthy: So you can nominalize subnormal ‘mentally handicapped’, yielding un subnormal ‘a mentally handicapped person’, but hardly normal ‘normal’ (*un normal ). Since negative properties seem to be considered as more noteworthy than behavior corresponding to the norm, it follows that adjectives with negative prefixes are nominalized much more frequently than their bases (e.g., un infeliz ‘an unhappy person’, but *un feliz ‘a happy person’). The presence of the feature ‘negative’ can also be seen to have an impact on the formation on adverbs, with -mente in Spanish and -ly in English, but here the cause is less obvious: Why is inevitablemente/inevitably more frequent and natural than evitablemente/evitably, and why inolvidablemente/unforgettably so much more than olvidablemente/forgettably?

5. References Attal, Pierre and Claude Müller (eds.) 1984 La negation. Paris: Larousse. Bosque, Ignacio 1980 Sobre la negación. Madrid: Cátedra. Brütsch, Edgar, Markus Nussbaumer and Horst Sitta 1990 Negation. Heidelberg: Groos. Bustos, Eduardo 1986 Pragmática del Español. Negación, cuantificación y modo. Madrid: UNED. Gyurko, Lanin A. 1971 Affixal negation in Spanish. Romance Philology 25(2): 225−240. Hamawand, Zeki 2009 The Semantics of English Negative Prefixes. London: Equinox. Hernández Paricio, Francisco 1985 Aspectos de la negación. León: Universidad de León. Hoppe, Gabriele 1999 Das Präfix ex-. Beiträge zur Lehnwortbildung. Tübingen: Narr. Horn, Laurence R. 2001 A Natural History of Negation. 2nd ed. Chicago: University of Chicago Press. Ibáñez, Roberto 1972 Negation im Spanischen. München: Fink. Jespersen, Otto 1917 Negation in English and Other Languages. Köbenhavn: Höst. Klima, Edward S. 1964 Negation in English. In: Jerry Fodor and Jerrold Katz (eds.), The Structure of Language. Readings in the Philosophy of Language, 246−323. Englewood Cliffs: Prentice-Hall. Klosa, Annette 1996 Negierende Lehnpräfixe des Gegenwartsdeutschen. Heidelberg: Winter. Ladusaw, William A. 1980 Polarity Sensitivy as Inherent Scope Relations. Bloominton, IN: University Linguistic Club.

78. Negation

1359

Laka, Itziar 1994 Negation in Syntax. On the nature of functional categories and projections. New York: Garland. Leal Cruz, Pedro 1989−90 Incongruencias en la relación ‘positivo/no positivo’ en el español actual. Revista de Filología de la Universidad de La Laguna 8−9: 183−194. Llorens, Eduardo L. 1929 La negación en español antiguo con referencia a otros idiomas. Madrid: CSIC. Martín García, Josefa 1995 La creación de términos contrarios y contradictorios: Los prefijos negativos in-, des- y no- en español. In: Carlos Martín Vide (ed.), Lenguajes naturales y lenguajes formales XI, 471−477. Barcelona: Promociones y Publicaciones Universitarias. Molho, Maurice 1962 De la négation en espagnol. In: Maxime Chevalier, Robert Ricard and Noël Salomon (eds.), Mélanges offerts à Marcel Bataillon, 704−715. Bordeaux: Grassi. Mollidor, Jutta 1998 Negationspräfixe im heutigen Französisch. Tübingen: Niemeyer. Montero Curiel, María Luisa 1999 La prefijación negativa en español. Cáceres: Servicio de Publicaciones de la Universidad de Extremadura. Montero Curiel, María Luisa 2001 Prefijos aminorativos en español. Cáceres: Servicio de Publicaciones de la Universidad de Extremadura. Payne, John 1985 Negation. In: Timothy Shopen (ed.), Language Typology and Syntactic Description. Clause structure. Vol. 1, 197−242. Cambridge: Cambridge University Press. Rainer, Franz 1989 I nomi di qualità nell'italiano contemporaneo. Wien: Braumüller. Sánchez López, Cristina 1999 La negación. In: Ignacio Bosque and Violeta Demonte (eds.), Gramática descriptiva de la lengua española. Vol. 2, 2561−2634. Madrid: Espasa-Calpe. Varela, Soledad and Josefa Martín 1999 La prefijación. In: Ignacio Bosque and Violeta Demonte (eds.), Gramática descriptiva de la lengua española, 4993−5040. Madrid: Espasa-Calpe. Voigt, Burkhard 1979 Die Negation in der spanischen Gegenwartssprache. Analyse einer linguistischen Kategorie. Frankfurt/M.: Lang. Zimmer, Karl E. 1964 Affixal negation in English and other languages: An investigation of restricted productivity. Word 20(2): 1−105.

Marisa Montero Curiel, Cáceres (Spain)

1360

VII. Semantics and pragmatics in word-formation II: Special cases

79. Negation in the Slavic and Germanic languages 1. 2. 3. 4. 5. 6. 7. 8.

Introduction Nominal negatives Pronouns Verbal negation Adjectival negatives Adverbial negatives Negatives in synsemantic word classes References

Abstract Negatives are modificational derivatives formed mostly by prefixation, less frequently by suffixation in the West Slavic languages and their neighbouring West European languages which have many common features in the word-stock of negatives. They are represented by contrariety and contradictority, the relation of antonymy and other means of expression reflecting a scale of negation. The present article will concentrate on negation and word-formation in the Slavic and Germanic languages. The aim of this article is to demonstrate that negation as a phenomenon of word-formation is connected with all word classes.

1. Introduction The area of linguistic negation regularly leads the authors to the world of formal logic. The basic terminology of this area reveals its origin in ancient philosophy. From ancient times on, negation has been identified with the polarity of the positive and negative: affirmo and nego. On the other hand, we also grasp polarity in the ordinary experience of the sensual knowledge of the world around us: light and darkness by sight, noise and silence by hearing, sweet and salty by taste, pleasant and unpleasant odours by smell, hot and cold by touch. The world of ordinary experience contains numerous semiotic entities existing in polarity. The polarity of a language sign reflects extra-linguistic reality in more complicated ways, even if it is true that the elementary basis is also here the polarity of yes and no, the polarity of antonyms, the polarity of contrary and contradictory word meanings as well as the polarity of utterances with content affirmo and nego. But this basis can be modified in many ways. The basic structural problem of affirmation and negation lies in the fact that the presence of a negative exponent is not always an explicit signal of a negative content. In logic, a negative exponent corresponds to a negative content, also in the case of the double negative (the negation of a negation that cancels each other out). In the linguistic description of negation, the logic approach could be misleading. In some types of negation the meaning of negation can also be expressed by an “affirmative” exponent and vice versa, a negative exponent can express affirmative meaning.

79. Negation in the Slavic and Germanic languages

1361

Here we find the discrepancy between the grammatical and the semantic structure of the sentence. Linguistic polarity is thus not captured exclusively by the two extreme endpoints, but by a whole scale of means. Interestingly, this scale exists not only with respect to the means of expression, but also on the meaning level (which will be shown later, e.g., in the discussion of the so-called weak negation, see section 7). Language as a system has its own inner logic, but it is different from the rules of formal logic. Linguistic negation varies greatly in meaning according to which level of language it belongs. These are the semantic and formal subsystems of language and that is also why it is necessary to observe the semantic as well as the formal aspect when talking about negation. Negation has most often been investigated on the lexical and grammatical (syntactic) level. In the first case the subject-matter is the lexical unit, i.e. the isolated language item and the possibility of negating it. In the second case, the participation of negation in the grammatical and semantic sentence structure is investigated. In this article I will take the West Slavic languages Slovak (Slk.), Czech (Cz.), and Polish (Pol.) as a basis and compare them with the West European languages English (E.), French (Fr.), but mainly with German (G.) as a geographically and culturally neighbouring language. In some cases I will also mention the situation in Russian (Ru.), referring to Scheller (2008). Lexical negation constitutes a subsystem of the word-stock that is represented either by a negative meaning of some of its units or by the presence of a negative exponent (not taking into account the affirmativeness or negativity of meaning). Negative meaning is registered when the language reflects phenomena or attributes of extra-lingual reality as non-existing, absent, unreal, impossible, etc. In the Slavic languages, the universal exponent of negation is the morpheme ne-, but it has numerous synonymous formants (prefixes, prefixoids, very rarely suffixoids). Similarly in English, according to Mazzon (2004: 111), the most productive prefix is un-. In a small group of Slavic words, the prefixal morpheme ne- has become part of the base because over the course of time it has lost the semantic connection to its original function. Often only etymologists can reconstruct how the meanings of these words have changed and lexicalized in the process of their usage in context. Such are, e.g., the Slovak words nedeľa ‘Sunday’, nevesta ‘bride’, neduh ‘infirmity’, necudník ‘lecher’, nedopatrenie ‘inadvertency’, etc. InterSlavic comparison often provides clues to the emergence of such changes as, for instance, the differences that are visible in the word for Sunday: Slk. nedeľa, Cz. nedele, Pol. niedziela (< *ne dělati ‘to not do/work’) mean ‘the seventh day of the week’. (The Russian homophone nedelja has the meaning ‘period of seven days from Monday to Sunday’ since Sunday is expressed by voskresen’e lit. ‘resurrection’.) In these languages, the word denoting Sunday is considered to be positive even though it contains the morpheme ne-. After the isolation of the segment ne in this kind of word, the rest of the word often cannot stand alone, it does not carry any meaning. Many words are only formally derived from an affirmative word, but semantically they originate in a more complicated, often descriptive expression, in a similar way to, e.g., the German word Nichtraucher ‘non-smoker’ from nicht rauchen ‘to not smoke’, Nichtschwimmer ‘non-swimmer’ from nicht schwimmen können ‘to not be able to swim’, even if they stand in opposition to the corresponding nouns Raucher ‘smoker’, Schwimmer ‘swimmer’ (Wellmann 1998: 508). The difference between the genetic and synchronic motivation of negatives lies in the fact that even if some of them were originally motivated, in the present-day language

1362

VII. Semantics and pragmatics in word-formation II: Special cases

they are felt as if they were non-motivated and thus they are affirmative and may become bases of other derivatives: Slk.: nedeľa (noun) → nedeľný (adj.) → nedeľne (adv.) (they can even be further negated, e.g., nenedeľná nálada, i.e. ‘everyday mood; lit. nonsunday-ADJ mood’). This kind of negative belongs to the class of demotivated words. But in the vocabulary, there are also types of negatives that do not go back to demotivated words. Different attempts from the past to grasp them systematically (lexical negation, lexicalized negation, contrary and contradictory expressions, etc.) can be made sense of if we understand them from the perspective of word-formation. Negation is a phenomenon belonging above all to the word-formation system of language. This side of negation has often been overlooked. It is surely related to the fact that the negatives have not usually been recorded in dictionaries, that is, apart from the already mentioned small class of demotivated words as well as affirmative base words with negative meaning. The majority of negatives are nevertheless motivated, but these words belong to the invisible part of the vocabulary, not represented in dictionaries because their existence is assumed. This is, of course, a “bad habit” of lexicographers, but also a paper-saving necessity. That is the case at least in the dictionaries of Slavic languages. Why do I call this tradition of lexicographic practice a bad habit? The answer lies in the comparison of the dominant prefix ne- ‘non-, un-’ with other prefixes, e.g., Slk. pre- (Cz. pře-) ‘through’, Slk. and Cz. od- (Ru. ot-) ‘away’, etc. which have a notably higher representation in dictionaries. But in a certain sense they represent the same means of derivation as the denying element ne-. This word-formation element causes (cf., for example, the negation of verbs, which have usually entered in dictionaries only in their affirmative form) a change in the intentional structure of the new negated word. In the West Slavic languages, for instance, this change applies to genitive constructions competing with the accusative (after a negated verb); further, the negation of verbs requires the negation of pronominal quantifiers, but also of connectors, due to the requirement of negative concord. Changes in meaning are related to the external or internal positioning of the negative morpheme in the complex predicate (e.g., auxiliary/modal verb or modal predicative + infinitive), and thus to whether it belongs to the first or to the second component (Slk. nemožno ľúbit’ ‘impossible to love’, možno neľúbit’ ‘possible not to love’). Negation in word-formation deals not only with the process of deriving new negatives from already existing affirmative words, but also with the r e l a t i o n s between two or more words − the relations of negatives to affirmatives but also to other differently derived negatives. In the Slavic languages, the negative derivational morpheme (a prefix or, perhaps, a prefixoid) is regularly attached to the front of the word-formation base of the affirmative word. On the other hand, there are languages like, for example, German in which negative derivatives with suffixes or suffixoids exist. In the West Slavic languages, only a few derivatives with suffixoids are found. They are mostly calques that can be explained by the geographical and cultural contact of these languages with German, cf. adjectives with the suffixoids Slk. -prázdny, Cz. -prázdný , Slk. and Cz. -pustý ‘-less; lit. empty, free from, void of ’: Slk. bohaprázdny, Cz. bohaprázdný, Slk. and Cz. bohapustý ‘godless; lit. God-GEN-void of ’ − G. gottlos ‘id.’, Slk. duchaprázdny/duchapustý ‘spiritless; lit. spirit-GEN-void of ’ − G. geistlos ‘id.’, Slk. ľudoprázdny ‘deserted; lit. people-linking vowel-void of ’, Cz. liduprázdný ‘id.; lit. people-GEN-void of ’− G. menschenleer ‘id.’, Slk. vzduchoprázdny − G. luftleer ‘void of air’, etc. But the last two examples have synonyms both in Slovak and Czech: bezvzdušný lit. ‘without-air-SUFF.ADJ’, Slk. bezvodný, Cz. bezvodý ‘waterless, anhydrous’. Exam-

79. Negation in the Slavic and Germanic languages

1363

ples with the element boh- ‘god-’, duch- ‘spirit-’ were probably inspired by the exemplary translations of the Bible in the past. This word-formation principle (based on loan translation) is on rare occasion also used in the formation of poetisms of the type Slk. láskypustý ‘loveless; lit. love-GEN-void of ’ (cf. G. lieblos), Slk. kúzlaprostý ‘charmless; lit. charm-GEN-void of ’. In word-formation, negation originates in the combination of a negative formant (negator) with an affirmative word-formation base. This is a type of modification, by which the negative derivatives are based on the semantic change of the affirmative motivating word. Negation affects different word classes in different ways. The discussion of whether negation as a phenomenon belongs to grammar (syntax) or to word-formation is to some extent well-founded. When negation is effected by a suffix, the negator is a formant which is often subject to pattern-specific restrictions on the formation of new words that don’t apply in the case of a syntactic phrase. For example, Lenz (1995: 28) shows that adjectives that can only be used either attributively or predicatively cannot be combined with the negative prefix un- in German, cf. *unangeblich ‘*unalleged’ or *unquitt ‘*uneven’. No such conditions restrict the negated version of the sentential equivalent, cf. Wir sind (nicht) quitt ‘we are (not) even’, cf. also Lenz (1995: 9 and 141).

2. Nominal negatives Nominal negatives can be divided into contrary and contradictory negatives. The distinction between contrary and contradictory negation, according to Mollidor (1998: 11−12), goes back to Plato. In Slovak and Czech (and in other Slavic languages as well), both contrary and contradictory negatives contain the prefix ne-, but they differ semantically.

2.1. Contrary nominal negatives Contrary negatives occur when two semantic contents exclude each other, but individually they both represent something affirmative, a form of existence. Their mutual exclusion is limited by certain criteria (e.g., the same gender, the same individual, the same spatial or temporal substance), cf., e.g., Slk. priateľ ‘friend’ and the negative nepriateľ (Cz. nepřítel, Pol. nieprzyjaciel ) ‘enemy’: The test for the identification of a contrary negative must prove that it is a part of a synonymic chain, in the given case with the synonyms Slk. protivník ‘antagonist’, odporca ‘adversary’, opponent ‘opponent’, súper ‘contestant’, sok ‘challenger’, rival ‘rival’, opozičník ‘oppositionist’, protihráč ‘competitor’. Of course, the original negated meaning and its nuances are also important. Contrary negatives are not generated from all meanings of an affirmative word. Though if communicative necessity requires it (in spite of the common usage), the contrariety is usually indicated with emphasis on the negative morpheme, in writing by hyphenation, e.g., with Slk. priateľ ‘boyfriend’ meaning ‘partner, lover’ it would be expressed as ne-priateľ, and its most relevant synonym would be sok ‘rival’. The contrary relation between the affirmative and the negative within a given language does not always require an explicit expression of negation, as it can also be seen in the above mentioned words. That is also

1364

VII. Semantics and pragmatics in word-formation II: Special cases

why correspondence in related languages should not be expected. The Russian equivalent of this word does not contain a negator (vrag), and similarly the German words Feind, perhaps Gegner (e.g., in sports). The contrary noun to Slk./Cz. odborník ‘a professional, expert’ is the negative neodborník, with equivalents in other languages, such as Pol. niefachowiec, Ru. nespecialist, G. Nichtfachmann ‘a non-professional’. Interlingual symmetry in the exposition of the “n-negator” − in the above mentioned examples ne, nie, nicht − does not normally take place in concrete communication. A negative with an explicitly expressed negation is usually used as a notional instrument to express the meaning ‘who does not have professional knowledge about something’. New semantic and stylistic nuances are provided by synonyms like Slk. amatér ‘amateur’, laik ‘layperson’, ochotník ‘dabbler’, samouk ‘self-educated’, babrák ‘muff ’, diletant ‘dilettante’, nedouk ‘half-scholar’, fušer ‘butterfingers’, etc. The negative morpheme ne- is redundant in ancient words, such as Slk. nevraživosť ‘rancor, spite, grudge’, nevrlosť ‘moroseness, glumness’, neplecha ‘mischief ’, nesvár ‘discord, discordance’ which − just as the unprefixed base words − have negative meaning from an etymological point of view. Alongside contrary nouns with the negative prefix ne-, negatives with the prefix bezcan be also used. This form goes back to the preposition bez ‘without’ (in opposition to the prepositions s, so ‘with’). Slavic nominal derivatives can be formed especially from negative adjectives (with the prefix bez-) by addition of the suffix -osť, e.g., Slk. bezbolestnosť ‘painlessness’, bezbožnosť ‘godlessness’, bezdetnosť ‘childlessness’, bezhlavosť ‘headlessness’. As one can see from these examples, deadjectival nouns ending in -osť express mostly the reification of an absent, missing feature that is designated in the word-formation base. Some derivatives containing the prefix bez- are the result of univerbation (in Slavic word-formation defined as a combination of ellipsis and suffixation of a word-group or a descriptive expression), e.g., Slk. bezbožný človek → bezbožník ‘godless person’, druh hmyzu bez krídel ‘species of insects without wings’ → bezkrídlový (adj.) hmyz ‘wingless insect’ → bezkrídlovce (noun pl.) ‘wingless insects’. Generally, this group does not comprise a large number of negatives. More contrary negated nouns in -osť are derived from adjectives with the prefix ne-. The equivalent German derivatives ending in -keit, -heit have the prefixes Un- and Nicht-, cf. Slk. nepozornosť − G. Unachtsamkeit, Unaufmerksamkeit ‘inadvertancy, carelessness, inattention’, Slk. nepresnosť − G. Un-/Nichtexaktheit ‘inaccuracy’. Words of the type Slk. intolerancia (G. Intoleranz) ‘intolerance’, indiskrétnosť (G. Indiskretion) ‘indiscretion’, etc. are also commonly considered nominal negatives. Nominal, but more often adjectival internationalisms with the prefixes in-, non-, nec-, ab-, de-, dis-, ex-, marginally also prō- (prōmiscuus), re- (reprobus), per- (perfidiōsus) (E. promiscuous, reprobate, perfidious, Slk. promiskuitný, perfídny) represent loanwords borrowed long ago and negated already in Latin. That is why they should not be included into descriptions of the synchronic state of native languages (except for the case of analogical formations, including hybrid ones). Clarenz-Löhnert (2004: 61) regards these prefixes as “euromorphemes” that signal the convergence of (West)European languages.

79. Negation in the Slavic and Germanic languages

1365

2.2. Contradictory nominal negatives Contradictory negatives occur when two opposites exclude each other. The semantic contents of a contradictory negative and an affirmative exclude each other absolutely and without interposition of a third word (synonym). Contradictory negatives are in opposition to what is denoted in the affirmative counterpart. The negator (here: the prefix) denies the affirmative content, but it does not specify new features. The negated substance does not belong to that which is expressed in the base, e.g., Slk. člen ‘member’ − nečlen ‘non-member’, vojak ‘soldier’ − nevojak ‘non-soldier’, kresťan ‘Christian’ − nekresťan ‘non-Christian’. In other words: negation denies the original feature without specifying it (Lotko 1975: 23). In some cases, the new meaning that arises as a result of the negation can be divided into two meanings: 1. contradictory meaning (‘not that what is denoted in the wordformation base, something different’), and 2. shifted affirmative meaning ‘false, improper, weak, etc.’. Compare Slk. neherec 1. ‘not an actor, somebody else’, 2. ‘an actor, but a very bad one’, e.g.: Nie div, že z dnešného predstavenia nehercov diváci začali odchádzať ‘It is not surprising that the audience started to leave the performance of the NEGactors [i.e. actors with weak dramatic skills]’. This second meaning of the negative neherec is accompanied by irony and at the same time it obtains contrariety.

2.3. Negative occasionalisms, potential negatives and derivatives with a restrictive negative meaning Negative occasional words are bound to their context and may be semantically polarized − contrarily as well as contradictorily. The Slovak word nerodičia ‘NEG-parents’ in the communicative structure of the following announcement of a nursery negates its affirmative counterpart rodičia ‘parents’, i.e. those who have the right to take the children home: Nerodičia sa pri preberaní dieťaťa musia preukázať písomným splnomocnením ‘Non-parents should submit a written authorization when coming for the child/when picking up the child’. Even if the meaning ‘something other than what is denoted in the word-formation base’ is not specified, it is factually limited to ‘people who are interested in taking care of the respective child (in supporting his/her parents, etc.)’. The word nerodičia ‘non-parents’ implies a set of legally established properties identical with those of parents. Another kind of an occasional negative is existentially connected to d i s j u n c t i v e s y n t a c t i c c o n s t r u c t i o n s of the type Slk. Kopec − nekopec, ideme ďalej! ‘Hill or not a hill, we keep going!’ Zákaz − nezákaz, nalej! ‘Forbidden or not, pour!’. Potential negatives (i.e. potentially derivable, but not lexicalized antonyms) − “ c o n d e n s e r s ” (cf. Lotko 1975: 15) − arise for reasons of linguistic economy. To this class belong, e.g., transformations of a descriptive expression or of a whole (usually subordinate) sentence into one (negated) word. An appropriate Czech example goes back to the former Czechoslovak president Masaryk: Ovšem i v neválce, kdykoli jde o nejaký těžší úkol, musí být důstojník vzorem ‘Also in the times of no war, any time when there is some more difficult task, the commander should be an example for the others’ (Lotko 1975: 15). The author of this utterance could use the opposite to válka ‘war’, i.e. mír ‘peace’, but he chose a stylistically stronger expression neválka ‘non-war’.

1366

VII. Semantics and pragmatics in word-formation II: Special cases

Nominal derivatives with a restrictive negative meaning represent a rich group in which the content of the base word is modified in such a way that it does not reach the zero polarity, but tends toward or comes close to it. Native and also borrowed prefixes and prefixoids can be used in their formation. Weakly productive in Slovak word-formation is the native prefix pa- ‘pseudo-’. It modifies, restricts the affirmative meaning of the base without denying it completely, but the derivatives acquire a pejorative colouring: paveda ‘pseudoscience’, paumenie ‘pseudoart’. Just like formations with the foreign prefix pseudo- they belong to the semantic field ‘illusive, misleading, not real, not true’: Slk. pseudoproblém ‘pseudoproblem’, pseudodemokracia ‘pseudo-democracy’, pseudokultúra ‘pseudo-culture’, pseudoumenie, G. Pseudokunst ‘pseudo-art’, pseudoveda, G. Pseudowissenschaft ‘pseudo-science’, etc. The synonymous prefixoid Slk. kvázi-/E. quasi- (kváziodborník ‘quasiexpert’, kvázimetatext ‘quasi-metatext’) has some analogous native prefixoids in German: Scheinangriff ‘feigned attack’, Scheinbild ‘simulacrum’. Other derivatives with the meaning ‘incorrect, wrong, inappropriate’ are formed in German by widely differentiated means: Fehlzündung ‘failing, backfire’, Fehlurteil ‘false judgement, misjudgement’, Mistwetter, Dreckswetter ‘filthy weather; lit. dung/dirt-weather’; Scheissarbeit ‘shit work’. The German prefix Miss- is of special interest in this context: it is used to express pure negation, e.g., Missachtung ‘disrespect’, Missverhältnis ‘disproportion’, but also a restricted negative meaning, e.g., Misswirtschaft ‘mismanagement’. A t e m p o r a l relation between the motivating and the motivated word is expressed by the prefix ex- ‘former’: Slk. expriateľ, G. Exfreund ‘exfriend’; Slk. expremiér, G. Exkanzler ‘ex-prime minister’. Relations of d i r e c t i o n , p o s i t i o n and others are expressed by the prefix anti- and its native counterparts Slk. proti-, G. gegen-: Slk. antiglobalista ‘antiglobalist’, protitlak ‘back pressure, counter pressure’, G. Gegendruck ‘id.’.

3. Pronouns In negation, pronominal correlates of nouns, adjectives, numerals and adverbs represent a relatively closed group from the viewpoint of word-formation. Besides occasional nominalization (Slk. veľké nič, G. das große Nichts ‘the big nothing’; Slk. obyčajný nikto, G. ein Niemand ‘a mere nobody’), and derivation (Slk. niktoš ‘a nobody’ ← nikto ‘nobody’, ničnerobenie ‘doing nothing; lit. nothing-not-doing’, etc.), nothing noteworthy happens in this group. But in word-formation, one should not be interested only in the process-related side of word-formation, but also in the system of the results of derivation. Pronouns usually have more than one negative counterpart with the general meaning of a negated substance, characteristics, relation, number or amount, place or time. Slavic negative pronouns are typically formed by the prefix ni-. Slk. nik, nikto ‘nobody’; nič, pranič ‘nothing’; nijaký, niktorý, žiaden, žiadny ‘none, neither’; ničí (a negative relational-possessive pronoun) ‘nobody’s’, nikoľko ‘nil’; nikde, nikam ‘nowhere’; nikdy ‘never’; nijako ‘by no means’; nikade, nikadiaľ ‘no way’; odnikade, odnikadiaľ, znikade, znikadiaľ, znikiaľ ‘from nowhere’ take part in the formation of negative sentences as quantifiers and qualifiers (they reinforce the negative meaning of the proposition). For the most part, negative sentences in the Slavic languages have a double meaning. For instance, the Slovak sentence Neurobil to po prvý raz ‘He did not make it for the first time’ could

79. Negation in the Slavic and Germanic languages

1367

be interpreted in two ways: 1. ‘He did not make it and it happened for the first time’, 2. ‘He has made it more times already, not for the first time’. In this context we can differentiate complete and partial negation (according to Horn 1989: 204: nexal and special negation). The double meaning can be eliminated with the above mentioned pronouns (= negative total pronouns) and adverbs, called t o t a l i z e r s (cf. Horn 1989: 222). The genitive of negation, emphasis and other means by which negation can be intensified belong here, e.g., Slk. vôbec ‘neither’, vonkoncom ‘throughout’, zhola ‘entirely’, celkom ‘completely’, úplne ‘fully’, absolútne ‘absolutely’, naskrze ‘wholly’, etc. The Slavic languages adhere to the rule of negative concord in the use of pronominal negatives. This means that negative pronouns are used with a negative predicate in sentence actualization. The negative pronouns stand in opposition to universal and indefinite pronouns: Slk. nikto ‘nobody’ vs. každý, všetok ‘everybody’ or niekto ‘somebody’; nič ‘nothing’ vs. všetko ‘everything’ or niečo ‘something’; nikdy ‘never’ vs. vždy ‘always’ or niekedy ‘sometimes’. The exclusivity of these negatives is functionally redundant from the viewpoint of the expression of negation, because they alone do not express negation, but occur in accordance with the negation expressed in the predicate. That is why it is wrong to consider the negated sentential constructions in which they occur double negatives (on the contrary, double negation refers to cases, as in the negation of two verbs, that result in the elimination of negation: Nemôže nekonať lit. ‘He cannot not do something’ = Musí konať ‘He must do/has to do something’). The combination of a negative pronoun with a negative predicate does not eliminate the negation, but renders it clear, unambiguous and complete. Nor is the French type ne-pas (je ne suis pas) considered to be a double negative, but the intensification of negation (cf. Horálek 1955: 248). If a sentence which contains an accumulation of several exceptive expressions, is negated, all exceptive expressions must be negated, cf. Slk. Niekto niekde niekedy niečo videl lit. ‘Somebody somewhere somewhen something saw’ and Nikto nikde nikdy nič nevidel lit. ‘Nobody nowhere never nothing NEG-saw’. Pronominal negatives do not express the negation, but they accompany the operator of negation realized on the predicate. Only in the case of the ellipsis of the verb in the predicate, usually found in shortened dialogue, do they take on the function of expressing negation: Slk. Nikto neprišiel? ‘Nobody has come?’ − Nikto. ‘Nobody.’.

4. Verbal negation Verbs are the most “rewarding” word class from the viewpoint of negation because all of them can be negated. Dictionaries contain only the affirmative forms. In the Slavic languages, verbs are negated by means of the morpheme ne- (Polish nie-) ‘not’. In the linguistic literature, this element is traditionally described as a grammatical, more precisely, a morphological element, but its presence in inflected verb forms or sentence actualization does not essentially influence either the formal side of verbs or the concrete sentence structure. The element ne- adds the lexical-semantic value of non-existence of dynamic characteristics (action, state of being, relations between substances, qualities and abilities) to verbs expressed in the affirmative form. This is the reason why the morpheme ne- should be understood as an element of word-formation, and every verbal negative as a derivative.

1368

VII. Semantics and pragmatics in word-formation II: Special cases The following semantic elements can be distinguished in the verbal negatives:

1. c o n t r a r i e t y (a case of sentential, i.e. complete negation), and 2. c o n t r a d i c t o r i t y (a case of negation of components, i.e. partial negation). The semantic difference between these two classes may be inferred from their usage in context: Slk. Peter to neurobil ‘Peter did not do that’ 1. Subject Peter did not do the action in question; 2. The action was done, but not by Peter − compare a synonymous sentence Nie Peter to urobil, (niekto iný) ‘Not Peter has done it (but somebody else)’. In verbal predicates of the type “modal/auxiliary verb + infinitive”, the negative morpheme ne- is regularly placed e x t e r n a l l y, i.e. on the modal verb: Slk. nemusieť prísť ‘to not have to come’, nemôcť spievať ‘to cannot sing’, nesmieť fajčiť ‘to not must smoke’, nemať robiť ‘to not have to do’, nechcieť čítať ‘to not want to read’. Modal/ auxiliary verbs do not have a full lexical meaning in that they do not designate any action, but in the Slavic languages they do express predicative categories, which is why the negative modificative element is added to them. The semantic core of these predicates lies in the infinitive of the full verbs. Cases with an i n t e r n a l l y positioned negator are stylistically coloured and rare (cf. Nechápem, ako sme to vtedy mohli nevidieť lit. ‘I don’t understand how we could NEG-see it then’; Keďže pred Silvestrom mohli poniektorí nespievať s odôvodnením, že nepoznajú text, [...] lit. ‘If before New Year some could NEG-sing with the argument that they don’t know the text [...]’). The so-called a c t i v e n e g a t i o n is characteristic of the group of verbs that have an affirmative form and whose content may be paraphrased without the exponent of negation with negatives (Slk. nebyť ‘not to be’, nemôcť ‘cannot’, nechcieť ‘to not want’, nemať ‘to not have’, etc.). These verbs form affirmative predicates. Their negative meaning is limited by context and is usually typical of secondary meanings. The verb that is closest to a negative is Slk. chýbať ‘to miss, lack; to be absent’, which expresses the deficiency or absence of somebody or something. The distribution of this verb in Slovak and German, for instance, has analogous functions: Slk. chýbať na zápase − G. bei einem Wettkampf fehlen ‘to be absent from a contest’ and Slk. chýba mu odvaha − G. es fehlt ihm der Mut ‘he lacks the courage [to do it]; lit. it lacks him the courage’. The so-called s e p a r a t i v e v e r b s also belong to this group: Slk. prekážať ‘to restrain’, priečiť sa ‘to oppugn, to disagree’, hanbiť sa ‘to be ashamed’, vystríhať sa ‘to beware of ’, vzoprieť sa ‘to oppose’, zaklínať sa ‘to swear by’, zapovedať sa ‘to renounce sth.’, zdráhať sa ‘to balk’, zdržať sa ‘to hold back’, zostávať ‘to remain’, zunovať sa ‘to pall on’, odriecť sa ‘to disown’, odhodiť sa ‘to cast aside’, odprisahať sa ‘to swear off ’, odžehnať sa ‘to revoke’, okúňať sa ‘to shrink from’. These are the Slovak representatives which have the function of implicating negation like their equivalents in the other Slavic languages. Negative meaning is more evident in the passive forms of verbs that acquire the meaning of a characteristic or quality: Slk. roztok zbavený prímesí ‘a solution free [lit. ‘ridded’] of admixtures’, vynechaný člen ‘an omitted element’, vylúčená doprava ‘stopped traffic’, vyňatý z registra ‘taken out of register’. E n a n t i o s e m y is a rare phenomenon by which one word contains in itself both affirmative and negative meaning. This has different, very individual manifestations in different languages, compare Slk. obiť omietku z domu ‘to put the plaster down from the house’ − obiť strechu plechom ‘to build a metal roof ’, obieliť ‘to colour something in white’ − obieliť zemiaky ‘to skin the potatoes’; G. übersehen ‘to survey’ or ‘to overlook’;

79. Negation in the Slavic and Germanic languages

1369

F. replacer ‘to put in a different place’ or ‘to put back in the place where it was before’, E. wear ‘to last under use’ or ‘to erode under use’.

5. Adjectival negatives Adjectival negatives are derived by the prefixes ne- (in Polish nie-) and bez- ‘without’. The combination of negative prefixes is impossible. (There are, however, a few examples of Russian euphemistic adjectives such as, e.g., nebezynteresnyj ‘not uninteresting’.) The negation of derived negatives is possible only in sentence actualization in the form of the so-called partial negation: nie nesmrteľný ‘not immortal’. Similarly in German: nicht unsterblich (cf. Lenz 1995: 171). But in German, prefixal-suffixal negation is also possible: unzweifellos ‘undoubtless’ (Lenz 1995: 160). The latter seems to be an occasional example which could also be interpreted as contamination of G. zweifellos and unzweifelhaft both meaning ‘doubtless, undoubted’ (Lenz 1995: 150). The majority of adjectives can be negated. Negatives cannot be formed from the Slavic individual-possessive adjectives, e.g., Slk. Annin ‘Anna’s’, Jánov ‘Jan’s’, matkin ‘mother’s’, otcov ‘father’s’, mačkin ‘cat’s’, etc. There is no *Neannin ‘no(t)-Anna’s’, *Nejánov ‘no(t)-Jan’s’, etc. The possibility of negation is also limited in two- and multiword units of the type Slk. slepá ulička ‘blind end, dead end’, mastná pokuta ‘a stiff fine’, čestná funkcia ‘a honorary position’, výsostný štátny znak ‘sovereign national emblem’. Negatives of the type Slk. neposedný ‘restless’ are not derived by prefixation, but originate from the descriptive expressions of the type ‘who cannot rest/sit for a while’. Negation of adjectives produces (1) c o n t r a r y adjectival negatives which are the negative antonyms to their affirmative motivating words and usually synonymous with positive expressions: Slk. opitý ‘drunk’ − neopitý ‘not drunk’ (syn. triezvy ‘sober’), potrebný ‘necessary’ − nepotrebný ‘unnecessary’ (syn. zbytočný ‘redundant’), zložitý ‘complicated’ − nezložitý ‘uncomplicated’ (syn. jednoduchý ‘simple’). They represent a rich scale from a semantic point of view. The content of the affirmative motivating word is modified by (2) c o n t r a d i c t o r y adjectival negatives in such a way that the content of the affirmative adjective is negated and a negative component ‘other than that what is denoted by the base’ is added to it, but the new feature is not specified more closely. The polarity of adjectives in the relation of a n t o n y m y is interesting for two reasons: a. Antonyms can be two formally affirmative words (Slk. vysoký ‘tall’ − nízky ‘short’, dobrý ‘good’ − zlý ‘bad’, veľký ‘big’ − malý ‘little’) and yet they represent negatives of each other from a semantic point of view: there is mutual negation between them and at the same time mutual disjunction (they are mutually exclusive), interdependence and semantic contrast. According to Clarenz-Löhnert (2004: 64), formally affirmative lexemes such as G. ledig ‘unmarried’, Junggeselle ‘bachelor’ contain an implicit negation and have equivalents in explicitly negated words unverheiratet or nicht verheiratet ‘not married’.

1370

VII. Semantics and pragmatics in word-formation II: Special cases

b. In the negation of any member of an antonymic pair by means of the morpheme ne-, a so-called intermediate zone (= “neither a nor b”; named “Zwischenbereich” in Mollidor 1998: 26) comes into being which usually has a meliorative function in speech, e.g., Slk. nemladý ‘not young’ (= ‘neither young nor old’), nestarý ‘not old’ (= ‘not old, but not young’). The negation of a does not imply the assertion of the opposite (b). The semantic nuances of such negatives exceed the parallelism of the semantic structure of antonyms. (For the relation between sentence implication and the “intermediate zone”, see Mollidor (ibid.), analyzing sentences such as G. Das Kind ist nicht unglücklich ‘The child is not unhappy’; cf. also Clarenz-Löhnert 2004: 68−69.) Negatives with the prefix Slk. bez- ‘without’ often do not originate from their affirmative counterparts by means of a direct negation, but must be considered as c o n d e n s e r s of multi-word expressions (cf. also section 2.2), e.g., Slk. bezplatný ‘free, costless’ = ‘made without payment’. This can be seen by the fact that many of them do not have an affirmative adjectival counterpart because they are formed from expressions with the preposition bez ‘without’, e.g., Slk. bezsenný (G. schlaflos) ‘sleepless’, Slk. bezradný (G. ratlos) ‘helpless’, Slk. bezúhonný (G. fleckenlos, makellos) ‘blameless’. The German adjectives of this type are derived by a suffix whereas in the Slavic languages affixal negation is limited to prefixation. In the Slavic languages, the distribution of the negative morphemes within the contrary and contradictory adjectives and other word classes is not regular. One and the same prefix can express several types of semantic negatives, and derivatives with the same word-formation element can differ stylistically. The same is true also in the West European languages; cf. for German and French Mollidor (1998: 245−247). The concurrence of the prefixes ne- and bez- in Slavic word-formation provides the possibility of deriving a smaller group of negatives in parallel. In German, this wordformation parallelism is multiplied by the fact that negatives can be derived by prefixes (ungefettet ‘non-fatted’, entfettet ‘degreased’) as well as by suffixes (fettlos, fettfrei ‘fatless, fat-free, free from grease’), naturally, with semantic differences. For further information see Lenz (1995: 138). The prefix ent- is characterized by the possibility of producing a particular group of derivatives which express the reverse of the original state (cf. Mollidor 1998: 74). The se m a n t i c i n c o m p a t i b i l i t y of some negatives is manifested as the semantic distance of a negative form from the affirmative form, e.g., Slk. nezmyselný ‘having no sense’ − zmyselný ‘erotic, sexual’ (cf. Lotko 1975: 41). Lenz (1995: 93−94) shows a similar phenomenon in German: vergesslich ‘forgetful’ − unvergesslich ‘unforgettable’, verantwortlich ‘responsible’ − unverantwortlich ‘unjustifiable’, tröslich ‘comfortable, consoling’ − untröslich ‘disconsolate’. These cases are very rare.

6. Adverbial negatives Slovak adverbial negatives are derived from affirmative bases with the prefix ne-: vysoko ‘highly’ − nevysoko ‘not highly’, pekne ‘nicely’ − nepekne ‘not nicely’, vedecky ‘scientifically’ − nevedecky ‘unscientifically’, svedomito ‘consciously’ − nesvedomito ‘unconsciously’. The possibility of negating adverbs is limited to deadjectival adverbs. Adverbs based on the merger of prepositions and nouns, such as Slk. dodnes ‘until today’, nabok

79. Negation in the Slavic and Germanic languages

1371

‘to the side’, osamote ‘alone’, odkraja ‘from the corner’, sčasti ‘partially’, zvečera ‘in the evening’, vzadu ‘behind’, etc., cannot be negated. A special group of adverbs are the negative predicatives in impersonal sentences. (In Slovak linguistics they are classified as negative modal sentential adverbs (vetná príslovka); however, as can be seen by the examples below, this notion does not convey the content of the English term.) In Russian they are represented by the predicative neľzja ‘(it is) forbidden, must not-PRED [+V]’ or ‘cannot’ (in contemporary Russian only in the negated form, but cf. the Czech predicative lze ‘(it is) possible’), Ukr. ne možna lit. ‘not possible-PRED [+V]’ (cf. Ozerova 1978: 53). In the West Slavic languages negative predicatives express: a) Impossibility of realizing the action: Slk. nemožno ‘it is not possible; lit. not-possiblePRED’, Slk. neslobodno ‘it is not allowed to; lit. not-free-PRED’, Slk. and Cz. neradno ‘one should rather not; lit. not-advisable-PRED’. Slk. nedajboh and nedajbože express either impossibility or undesirability: ‘Heaven forbid (that); lit. not-give-IMP-GodNOM/not-give-IMP God-VOC’. b) Unnecessity, undesirability or inacceptability of realizing the action: Slk. netreba, Cz. netřeba ‘it is not necessary; lit. not-necessary-PRED’, Slk. nenačim ‘one should not; lit. not-on-what’, Slk. nehodno ‘it is not worthy; lit. not-worthy-PRED’. Cf. also non-modal predicatives such as Slk. nevidno ‘one does not see; not-visiblecf. cez stromy nevidno les ‘to not see the forest/wood for the trees (idiom.); lit. through the trees not visible-PRED forest-ACC’, Slk. nevedno ‘nobody knows; lit. notknow-PRED’, and Slk. nečudo ‘no wonder’ (cf. Pavlovič 2003: 63−65). PRED’,

7. Negatives in synsemantic word classes The b a s i c n e g a t i v e p a r t i c l e that stands in opposition to the affirmative particle Slk. áno, Cz. ano ‘yes’ is Slk. nie, Cz. ne, nikoli(v), cf. also Pol. nie, Ru. net, ne, G. nein, Fr. non ‘no’. The scale of e x p r e s s i v e n e g a t i v e p a r t i c l e s is rich: Slk. figu ‘my foot!’, čerta, paroma, hroma, bohchráň ‘God forbid!’, horkýže, kdeže ‘rubbish!’, etc. In colloquial speech, particles are also used which originate in vulgar taboo-words: hovno ‘shit’, prd ‘fart’, riť ‘ass’, etc. Due to this stylistic characteristic, negation in combination with a predicate or alone is not only expressed, but at the same time intensified. A typical i n t e n s i f i e r o f n e g a t i o n is the particle Slk. ani ‘neither’. In combination with nouns denoting small objects of the type vlások, chĺpok ‘hair’, náprstok ‘thumb-stall, thimble’, myš ‘mouse’, lístoček ‘leaf ’, mak ‘poppy seed’, the intensification of negation is very close to nullity (similarly in English “negative polarity items”: not one iota, not a tittle); cf. also idioms such as Slk. neskrivit’ (komu) ani vlások na hlave ‘to not harm a hair on sb.’s head’. The so-called w e a k n e g a t i o n is expressed by particles termed a p p r o x i m a t o r s that occur in negative contexts only. Clarenz-Löhnert (2004: 49) termed them (G.) “Negativoide” (negativoids). By the means of approximators like Slk. ťažko ‘hardly’, sotva ‘barely’, ledva ‘scarcely’, takmer, temer ‘almost’, in Czech also málem ‘nearly’, a high

1372

VII. Semantics and pragmatics in word-formation II: Special cases

degree of limitation of the affirmative propositional content is achieved, cf. Slk. Sotva sa z toho dostane ‘He will barely get out of it’ (= ‘He will probably not get out of it’), Ťažko nájsť ozajstného priateľa ‘One can hardly find a real friend’ (= ‘One will not find a real friend’). By the use of restrictive particles (e.g., Slk. iba ‘only’, len ‘alone’, jedine ‘merely’, výhradne ‘purely’), the validity of an utterance is limited to the reality expressed by the word following the particle. Due to the restriction of the valid content, the remaining content is polarized, e.g., Slk. Sústreďuje sa len na pitie ‘He thinks only of drinking’, Vlak premáva denne okrem nedele ‘The train runs every day except Sundays’. Direct and indirect (see below) restrictors are able to express weak negation corresponding to their lexical meaning. I n t e r j e c t i o n s as independent utterances are used above all in the expression of negation in dialogues with the communicative function of disapproval, averse, denial, disinterest and embitterment, e.g., Slk. fuj, kuš, huch, bŕ, no, nono, etc. Most of them can express several communicative functions; the newer loanword blabla, for instance, expresses still other evaluations besides disagreement or refusal. Implicative negation (indirect restriction) can be expressed by some p r e p o s i t i o n s which − on the basis of their lexical meaning − exclude objects from validity, e.g., Slk. mimo ‘outside’, okrem (Cz. kromě ) ‘except (for)’, s výnimkou ‘with exception of ’, až na ‘exclusive of ’, namiesto ‘instead of ’, etc., cf. Slk. Lopta je mimo hry ‘The ball is outside the playground’, Vlak premáva denne okrem nedele ‘The train runs daily except Sundays’.

8. References Clarenz-Löhnert, Hildegard 2004 Negationspräfixe im Deutschen, Französischen und Spanischen. Aachen: Shaker. Horálek, Karel 1955 Úvod do studia slovanských jazyků. Praha: Nakladatelství Československé akademie věd. Horn, Laurence R. 1989 A Natural History of Negation. Chicago: The University of Chicago Press. Horn, Laurence R. (ed.) 2010 The Expression of Negation. Berlin/New York: de Gruyter. Lenz, Barbara 1995 un-Affigierung. Unrealisierbare Argumente, unausweichliche Fragen, nicht unplausible Antworten. Tübingen: Narr. Lotko, Edvard 1973 Lexikální negace v současné češtině. Praha: Státní pedagogické nakladatelství. Mazzon, Gabriella 2004 A History of English Negation. Harlow: Pearson/Longman. Mollidor, Jutta 1998 Negationspräfixe im heutigen Französisch. Tübingen: Niemeyer. Ozerova, Nina G. 1978 Sredstva vyraženija otricanija v russkom i ukrainskom jazykach. Kyïv: Naukova dumka. Pavlovič, Jozef 2003 Negácia v jednoduchej vete. Bratislava: Slavistický kabinet SAV.

80. Spatial and temporal relations in German word-formation

1373

Scheller, Andrea 2008 Negationskonkordanz im Slavischen. Frankfurt/M.: Lang. Wellmann, Hans 1998 Die Wortbildung. In: Duden. Grammatik der deutschen Gegenwartssprache, 408−557. 6 th ed. Mannheim: Dudenverlag.

Jozef Pavlovič, Trnava (Slovakia)

80. Spatial and temporal relations in German word-formation 1. 2. 3. 4.

General remarks Options in word-formation Concluding remarks References

Abstract In the German language, there are two central ways of integrating spatial and temporal information by means of word-formation. Firstly, this type of information is typically located in the verbal phrase of sentences. As a consequence, it plays a major role in the area of word-formation of verbs too. The two major classes of such verbs found in German (“Partikelverben” and “Doppelpartikelverben”) are located in the transition zone between syntax and word-formation. The same adverbial relation is found in one type of nominal compounds (“Rektionskomposita”). On the other hand, space and time are prominent among the relations constituting the patterns of the prototypical type of noun compounds (“N+N-Komposita”). The integration of these relations into compounds involves some kind of functional interpretation.

1. General remarks 1.1. Space, time, and linguistic structures The construction and reconstruction of the spatial and temporal embedding is of central importance to the understanding of discursive worlds. Space and time are without doubt elements of a linguistic ontology, and the linguistic consequences have been discussed from different linguistic viewpoints (cf. Levinson 2003; for the interacting factors that have been discussed cf. Tversky 2005). In the last years, a number of papers have

1374

VII. Semantics and pragmatics in word-formation II: Special cases

appeared that deal with the question of the cognitive basis of these relations (e.g., Hickman and Robert 2006; Evans 2013). It is a consequence of their central role in linguistic categorizations that information about spatial and temporal conditions is systematically built into the grammatical and lexical systems of languages (for an overview see Filipović and Jaszczolt 2012, Hickmann and Robert 2006).

1.2. The case of German Elements of spatial as well as temporal classification are part of the same level of syntactic modification (cf. Zifonun, Hoffmann and Strecker 1997: 799−806): adverbial constructions of where (“Ortspezifikation (wo-Spezifikation)”) and when (“Zeitspezifikation (wann-Spezifikation)”) specify propositions in principle. In grammatical terms, this means that they function as supplements to the verb. With certain classes of verbs however, spatial and temporal information must be counted as part of the argument structure and therefore categorized as verbal complements (cf. Zifonun, Hoffmann and Strecker 1997: 1099−1104). The most important case in this respect is directional information (“Direktivkomplement”). Furthermore, there is a relevant group of verbs with local arguments (“Situativkomplement”) but only a few special cases where a temporal argument of a certain type (“Dilativkomplement”) is required. Spatial and temporal complements show some peculiarities and are therefore often seen as marginal cases of this category (“Halbkomplement”). On the other hand, the ability of configurations of verbal complements to constitute “families” of argument structures on their own plays a major role in the discussion about the nature of the interface between lexicon and grammar, and directionality is one of the relations in question (cf. Engelberg 2011: 93−96). The parts-of-speech classification corresponds with these syntactic facts. There are different types of local and temporal adverbs, and all classes of connecting particles (prepositions, “Konnektoren”; see Pasch et al. 2003) contain sub-types of these two categories. The continuum of temporal, aspectual, and modal classification finds its grammatical correspondence in German primarily in the inflectional morphology of verbs, and also has ramifications in different areas. There are no counterparts to this type of grammatical embedding on the spatial side.

1.3. Conceptualization The linguistic conceptualization of space has three points of reference (cf. for instance Levinson 2003: 64−74): The axes of Euclidian three-dimensional space, topological relations relative to objects and persons, position and status of the speaker-ego. Reference to space may concern position in space (local) and movement (directional). Movement can be divided with respect to its segments of origin, path and goal (for a critical discussion see Gambarotto and Muller 2003; for the cultural relativity of these abstract categories cf. Bohnemeyer and Tucker 2013: 639−642).

80. Spatial and temporal relations in German word-formation

1375

There is, up to a certain point, some analogy within the structuring of orientation in time. Points of reference are (for a discussion of these structures see Klein 2009: 25− 35, Evans 2013: 81−141): The concept of a timeline, the position relative to different points of time and the concept of experienced time.

2. Options in word-formation Word-formation as a means of realizing relations inside complex words without using specific means of morphology or syntax as far as the German language is concerned has two options of representing spatial or temporal information, namely, 1) an adverbial type, and 2) a categorizing type.

2.1. The adverbial type: verbs, particles, adverbs, and prepositions 2.1.1. General remarks Concerning the adverbial character of encoding techniques for spatial and temporal information, some kind of integration into the verbal lexeme seems a plausible option. And in fact, this type of word-formation, characterized by the incorporation of adverbial elements and local particles, plays a major role in German word-formation making German an example of a “satellite-framed” type of languages in the sense of Talmy (2000) (cf. Grinevald 2006: 30−37). The “polymorphematic constructions” produced by this type of word-formation are very near to instances of certain syntactic constructions. As these verbs obviously mark the transition between syntax and (lexical) morphology, they are classified in one or the other way, either as some kind of fixed syntactic construction (cf. Müller 2002) or as complex words following lexical rules (cf. Fleischer and Barz 2012: 91 f., 419 f.; Motsch 2004: 44−48). The semantic parallelism of particles with prepositions, adverbs, conjunctions and local particles is obvious. The structural ambivalence is still fostered by the fact that these complex verbs are separable, only in a subset of their structural configurations are they written as one word (for a systematic discussion cf. Jacobs 2011: 356−362 and article 34 on verbal pseudo-compounds in German). In recent years, different ways of description have been discussed that attempt to model the transitional character of these constructions in a systematic way (cf. Booij 2002 for Dutch; Felfe 2012).

2.1.2. Spatial schemata and lexical generalizations In addition to this, the conditions in German, e.g., in comparison with English (cf. Bauer, Lieber and Plag 2013: chapter 16: Locatives of time and space), are characterized by the existence of two types of separable verbs. The following texts contain examples of the two types: complex verbs with double particles (“Doppelpartikelverben”; cf. examples (2), (3), and (5) with hinein- and heraus- − hin- and her- are elements of some situational

1376

VII. Semantics and pragmatics in word-formation II: Special cases

specification) or with single particles (“Partikelverben”; cf. examples (1) and (4) with ein- and aus-): (1)

Als der Gast eingetreten war, schloss er die Tür. [After entering the room the guest closed the door.]

(2)

Als er in den dunklen Raum hineintrat, sah er zunächst nichts. [As he stepped into the dark room, he saw nothing at first.]

(3)

Eine andere Person war durch die Tür im Hintergrund hereingetreten. [Another person had entered through the door in the background.]

(4)

Viele Mitglieder sind aus den Parteien ausgetreten. [Many members have left the parties.]

(5)

Er wollte aus dem dunklen Raum heraustreten. [He wanted to step out of the dark room.]

(6)

Sie waren aufgeregt, als sie auf die Bühne hinaustraten. [They were excited when they stepped on the stage.]

In these examples, a rather general verb of motion (treten) is combined with particles or particle combinations concerning the arguably most basic topological relation, the one between inside and outside (ein and aus). But it is the combination with the pronominal adverb (hinein/hinaus, herein/heraus) by which the pure local meaning is encoded, while complex verbs containing the single local particle tend to have some specialized, idiomatic meaning (cf. Eichinger 2000: 231−242; similarly McIntyre 2002). So examples of a verb like eintreten ‘to enter’ in a large corpus reveal a strong co-occurrence with more abstract directional objects (with in) like Tagesordnung ‘agenda’, Mittagspause ‘lunch break’, Dialog ‘dialogue’, or like Gewerkschaft ‘trade union’. In the subject position, there is a preference for nouns like Situation ‘situation’, Gegenteil ‘opposite’, Schaden ‘damage’, Verbesserung ‘improvement’, Stille ‘quietness’. On the other hand, combinations containing adverbial double particles are, obviously, a principled means of incorporating the directional and situational information expressed by the directional adverbs (“Richtungsadverbien”). These adverbial particles in addition to their local part (aus in examples (5) and (6)), contain the elements hin- and her-. These elements relate the motion expressed in the verbal base (treten) to the position of the speaker. A motion towards the speaker is signaled by her-, a motion directed to another area (‘not the speaker’) by hin-. So in (5) the movement of the subject begins its path in an area distant form the author/speaker, and is aimed in the speaker’s direction (er ‘he’), in (6) the Bühne ‘stage’ as the goal or the motion is distant from the speakers place/starting point. So in principle these verbal constructions are of a syntactic type, they add an adverbial modification to the verbal base. But often they are embedded in more or less fixed constructions; for instance heraustreten ‘to come out of’ is highly correlated with aus dem Schatten ‘out of the shade’, and this construction has a type of metaphorical usage: (7)

Es gelang ihm, aus dem Schatten seines Vaters herauszutreten. [He was able to step out of his father’s shadow.]

80. Spatial and temporal relations in German word-formation

1377

And it is not only this use in constructions such as this that shows modifications of pure syntactic adverbial behavior. There are many cases showing some kind of idiomatic meaning that cannot be traced back to a syntactic construction, cf. for instance: (8)

Wir möchten uns bei allen bedanken, die hereingeschaut haben. [We would like to give thanks to all those who stopped by (lit. ‘looked in’).]

In comparison to English, in German there is an additional differentiation made possible by the use of the two types of particles, the “prepositional” simple particle and the “adverbial” complex one (cf. also articles 35–37 on particle verbs in Germanic, Romance and Hungarian). On the one hand, the semantic parallels between verbal particles, adverbs, and prepositions are obvious. But within this area of word-formation (“separable verbs”) the existence of two formal types allows for a functional difference. So the simple verbal particle in German may have an adverbial or a prepositional reading − in both cases connected with a strong tendency toward a specialized meaning −, while the non-specialized local reading is connected with the use of the adverbial “double particle” in the context of a dimensional reading (and with (clearly syntactic) prepositional phrases in the case of topological reference). (9)

Er hatte sich den Berg hinaufgewagt. [He had ventured up the mountain.]

(10) Auf den Gipfel hinauf hatte er sich nicht gewagt. [Up to the summit he had not dared to go.] (11) Wir möchten eine neue Mauer auf die alte aufsetzen. [We want to set up a new wall on the old one.] (12) Wir möchten eine neue Mauer auf die alte setzen. [We want to put a new wall on the old one.] So, as a consequence, the “adverbial” double-particle type is concerned with dimensional orientation, while only the single-particle type is able to encode topological (“prepositional”) information (but is open to dimensional orientation, as well). Obviously this type corresponds to some type of adverbial relation, which in German − in contrast to the English data − differs from the “prepositional” form of the single particle: (13) Der Politiker ist schnell aufgestiegen/nach oben gestiegen. [The politician has risen very quickly/risen to the top.] (14) Der Mann ist sehr hoch hinauf gestiegen. [The man has risen very high (lit. ‘up’).] The spatial information conveyed by the particle verb is regularly connected with some contextual restrictions, some kind of idiomatic or at least some metaphorical usage. This technique produces labels for certain types of fixed actions within some kind of frame, rather than the description of a dimensionally or topologically-oriented movement. This is indicated by (among other things) the fact that any part of a frame can be used as the base of such a verb, cf.:

1378

VII. Semantics and pragmatics in word-formation II: Special cases

(15) Zu Weihnachten wird gerne eine Gans aufgetischt. [At Christmas a goose often is served (lit. ‘uptabled’).] (16) Er wird ihnen noch ganz andere Schauermärchen auftischen. [He will tell (lit. ‘uptable’) them yet more horror stories.] In principle this is true for both types of spatial orientation, cf. the following “dimensional” example: (17) Segelboote werden vor dem Winter aufgekrant und gesäubert. [Sailing boats are craned (lit. ‘craned up’) and cleaned before winter.] (18) Zur Jagd werden die Greifvögel abgehaubt. [The hoods of birds of prey are freed for hunting. (lit. ‘The birds of prey are dehooded.’)] The prototypical case of verbs of the adverbial “double-particle” type represents a directional orientation, and therefore uses the ability of adverbials of the directional type to form a directional structure by itself, and thus leads to numerous opportunities for the modification of this movement by way of variation of the base. The simplest way to do this is, of course, the combination with a verbal lexeme of the movement type as a base of the complex verb. It is not very surprising that these general verbs tend to develop some specific kind of meaning in addition to their literal use. (19) Der Kellner sieht die Besucher zur Tür hereinkommen. [The waiter sees the visitors come in through the door.] (20) Das eingesetzte Geld muss wieder hereinkommen. [The money used must come in again.] As directionality is already given with the adverbial particle, this type of verb allows the expression of movement types by using different types of modifying aspects as the base of the complex verb: (21) Die Investoren sollten viel Geld hineinbuttern. [The investors should put in (lit. ‘butter into’) a lot of money.] (22) Wenn sich die Golden Globes zu einer Werbeveranstaltung herabdegradieren. [If the Golden Globes degrade down to a promotional event.] All of these types of verbs are suited to be used with respect to movements of the subject as well as the object.

2.1.3. Relations between actors and objects Another topological type is represented in connections with relevant prepositional particles and adverbs. These use other objects as points of reference for defining the respective movement:

80. Spatial and temporal relations in German word-formation

1379

(23) Ein Zug fährt durch. [A train passes through.] (24) Man sollte das Land mit dem Rad durchfahren. [One should ride through the country by bicycle.]

2.1.4. The temporal turn This last type of spatial ordering seems especially apt to be used for modeling actions in time (for the analogies and differences in the conceptualization of space and time see Tenbrink 2007: 19−29). Simply being somewhere in time seems to be a less important topic than the ordering and adjustment of temporal experience in relation to the relative position of the speaker, other elements, and elements of time constituted by this type of orientation. Behind all this structuring is the idea of time being organized in a linear order and having a natural back-front direction. This conceptualization is the basis for two types of moving entities. The sequencing of temporal sensations can refer to some subject moving along this line, or to the moving of time relative to the standpoint of the ego involved. It is events in time that move along the time-line, “relevant time”, as can be seen by the following examples. Regarding the subject or experiencer of these events, two interpretations are possible: there is a kind of “Deixis-am-Phantasma” [imagination-oriented] interpretation. In this case the person in question imagines himself moving with the elements on the time line. In the other interpretation, on his way along his own timeline − the row of events that have passed − meets the events coming upon to him from the opposite direction. A view fitting the modeling of all these events is that, once they have passed, they are behind him, so the preceding (vorausgehende) events are further in the past than the following ones (nachfolgende). Moreover, verbal particles of adverbial character are able to differentiate between sheer serialization of events (e.g., voraus-) and of intentional correlations (e.g., vorher-); in the latter case it is more likely to have a “person moving with the time” interpretation than in the first one (cf. Eichinger 1989: 346−350). (25) Vier Jahre intensiver Arbeit waren diesem Tag vorausgegangen. [Four years of hard work had preceded this day.] Just walking along the time-line with the time passing on in the same direction is obviously interesting, if some effort has to be made to reach the temporal position one has reached (cf. Eichinger 1989: 385−391): (26) Er hat eine schwere Zeit durchgemacht. [He’s been through a tough time.] On the other hand, time can be seen somehow from the outside: passing by (not without consequences and at a different speed; cf. Eichinger 1989: 393 ff.):

1380

VII. Semantics and pragmatics in word-formation II: Special cases

(27) Die Zeit ist auch an ihm nicht spurlos vorbeigegangen. [Time has left its mark on him also. (lit. ‘Time has not passed him by without trace.’)]

2.2. The categorizing type: spatial and temporal relations in composition and prefixation 2.2.1. Introductory remarks While the types of word-formation dealt with so far reflect the incorporation of adverbial relations of space, direction, and time, immediately − and word-formation of this type is near to some kind of fixed syntax − the compositional types follow their own rules. They rely heavily on the calculation of relations between parts of semantic frames indicated by two lexical items that are thought to represent structural elements of such a frame. The set of such elements can be read as a kind of communicative ontology. The structural power of such constellations is reflected in clusters of formations brought together by means of analogy and family resemblance. Whatever the granularity of the linguistic description of German composition may be − local and temporal relations form a relevant group in any case. Some types of compounds and prefixes categorize local and spatial situations in a more abstract way using the categories of topological orientation.

2.2.2. Spatial relations within nominal compounds A description like this fits the prototypical instance of nominal compounds. Such compounds are composed of two nominal and non-relational elements. How are we able to (re)construct the meaning of entities of such a structural type? First of all there are a large number of compounds that are a fixed part of the lexicon; whereas they may not undergo decomposition when accessed, nevertheless they may be used as a blueprint for analogous cases. Quite a number of compounds are part of the linguistic inventory of the average reader of German newspapers. This is true for words like Gebirgsdorf ‘mountain village’, Straßenschlacht ‘street battle’, Abendlicht ‘evening light’, Winterfahrplan ‘winter timetable’ and also for adjective-noun compounds like Frühstart ‘early start’ (cf. Eichinger 2006a, 2006b; we leave aside compounds with a second element with local (Tierpark ‘zoo; lit. animal park’) and temporal (Ferienmonat ‘holiday month’) meaning). Compounds that are not lexicalized to such an extent are to be read as an instruction to look for their specific meaning in the context they are used in, cf. a word like Gartenvogel in the following example (28): (28) Waldkauz und Schwarzspecht − wirklich nicht die typischen Gartenvögel. [Tawny Owl and Black Woodpecker − not really the typical garden birds.] Complex nouns like Gartenvogel describe certain objects or actions as positioned into spaces. The denotations of these spaces form the modifying first part of the composite.

80. Spatial and temporal relations in German word-formation

1381

And so at first sight only this relation of ‘being in a certain space’ can be deduced from the semantics of the two nominal elements. So far the information is similar to that given in a syntactic construction like in (29): (29) Es reicht, die Vögel im Garten eine Stunde lang zu zählen. [It is enough, to count the birds in the garden for an hour.] But there is a strong tendency for constructions like these to acquire some kind of idiomatic reading in addition to this “morphological” information. The interesting feature of this development is not that there is some more or less idiosyncratic case that has developed a meaning that is only loosely connected with the semantic information to be drawn from our knowledge of its constituents. The functional advantage of using a compound like, for instance, Landregen ‘steady rain; lit. land rain’ − and not a syntactic phrase like Regen auf dem Land ‘rain in the countryside’ − is that some functional instruction is connected with the topological information given as the motivational base. The function in this case is a terminological differentiation, cf. (30) Diese Arten von Wolken sind bei lang anhaltendem Regen, so genanntem Landregen, zu beobachten. [These types of clouds are observed during prolonged rain, so-called steady rain (lit. ‘land rain’).] There is obviously some loose association of some type of rain, which is typically experienced in the countryside (Land), but the definition − even the layman’s definition − of this meteorological event no longer relies on this element. This is shown by the usual paradigm of “rain”-compounds in German: (31) Intensivregen ‘intensive rain’, Nieselregen ‘drizzling rain’, Platzregen ‘torrential rain’, Sprühregen ‘spray rain’, etc. A single instance of the compound Stadtregen lit. ‘city rain’ found in the corpora shows the option for a stylistically motivated re-motivation of lexico-morphological constructions like Landregen: (32) Eine kleine Wolke ließ sanften Stadtregen über die Besucherscharen ab. [A small cloud sprayed gentle Stadtregen (lit. ‘city rain’) on the crowds of visitors.] This item obviously is an ad-hoc formation in analogy to the rather fixed form Landregen − this compound being associated with ampleness (ergiebig) of the precipitation: in contrast the Stadtregen is marked as an instance of slight rain (sanft). This example shows that the productivity of compounds of this type is guided by lexical analogies and family resemblances as well as by the contextual computation invested in introducing some kind of “uncommon” combination. So when we read the title “Die Himmelsbrücke” (‘heaven’s bridge’) in a recent literary text we may have some guess about a meaningful interpretation of the local relation implied, but it is only

1382

VII. Semantics and pragmatics in word-formation II: Special cases

two pages further, when we read the following lines that we realize what fixed interpretation applies to this complex formation: (33) Der weithin sichtbare Feuerschweif mußte den Augenzeugen […] als eine […] Brücke aus Licht erschienen sein, die Himmel und Erde für einen Atemzug verband. [The highly visible fire tail had to appear to the eyewitnesses […] as a […] bridge of light that united heaven and earth for a breath.] (Ransmayr: 43) And as the text goes on like (33) we are finally on safe lexical ground. (34) Auch wenn diese Himmelsbrücke […] [Even if this sky bridge […]] But obviously this is only one functional use of the element Himmel ‘heaven’ as a type of classifier of Brücke ‘bridge’, cf. (35) 1796 entstand die […] Bogenbrücke und bekam wegen ihrer steil aufsteigenden Form den Namen Himmelsbrücke. [1796 […] the arch bridge was built and because of its steep ascending form got the name sky bridge.] (URL: http://www.sohland.de/verzeichnis/ objekt.php?mandat= 64549) There is a wide range of options for compounds from fully lexicalized items to rather surprising singular constructions that require contextual help of different dimensions. German tends to use compounds consisting of two (not too complex) constituents as the base level for everyday classifications, numerous compounds can be found that use the functional localization of objects as motivation for their naming, e.g.: (36) Gartenfest ‘garden party’, Hotelzimmer ‘hotel room’, Gartenarbeiter ‘garden worker’, Stirnlampe ‘headlamp; lit. forehead lamp’, Kopfende ‘head (of the bed/ the table etc.); lit. head end’, Straßenschlacht ‘street battle’, Friedhofsblume ‘cemetery flower’ (all examples taken from Ransmayr) In compounds of this type, some kind of local relation is read as a functional relation of ‘being made for, being intended for’. A different type is represented in cases with a directional (and in certain cases local, see (41)) reading like in the following examples; these are compounds whose second part contains some relational element (“Rektionskomposita”): (37) Dieser Anblick hatte alle Italienfahrer der Zeit fasziniert und erschreckt. [This sight had fascinated and frightened all persons travelling to Italy at that time.]

80. Spatial and temporal relations in German word-formation

1383

(38) Die Umgebung profitiert von der Londoner Stadtflucht. [The area around London benefits from the urban exodus.] (39) Die Bevölkerungszahl in den Städten wird durch die Landflucht in die Höhe getrieben. [The population in the cities is increased by the rural exodus.] (40) Das Fahrzeug ist ideal für eine Wüstendurchquerung. [The vehicle is ideal for a desert crossing.] (41) Die Künstler genossen ihren Romaufenthalt. [The artists enjoyed their stay in Rome.] We find here systematically those types of compounds in which the second element incorporates a relation; the base of these words is a verbal lexeme with some derivational element incorporating the agent-complement (agent nouns: e.g., -fahrer ‘driver’) or referring to the procedure itself (action nouns: e.g., -flucht ‘flight/escape’). In the first element then some kind of local complement is realized. As can be seen by the examples, the different segments of movement can be encoded in this way.

2.2.3. Nominal compounds and temporal relations It is an obvious choice to have classifications for actions or events that make reference to the time(span) in which they typically happen. If one looks at compounds with Morgen ‘morning’ as the first element, any dictionary will give quite a number of examples: (42) Morgenbad ‘morning bath’, -bummel ‘stroll’, -feier ‘ceremony’, -konzert ‘concert’, -lektüre ‘reading’, -schicht ‘shift’ There are quite a lot of compounds, too, referring to states and conditions typical for this time of the day: (43) Morgendunst ‘morning mist’, -kälte ‘cold’, -licht ‘light’, -luft ‘air’, -nebel ‘mist’, -sonne ‘sun’, -stern ‘star’, -tau ‘dew’ And there are names for objects connected in a typical (functional) way with mornings, as well as names for persons taking their typical morning behavior as a motive for naming them: (44) Morgenappell ‘morning roll call’, -ausgabe ‘edition’, -crew ‘crew’, -kaffee ‘coffee’, -kreis ‘circle’, -magazin ‘show’, -mantel ‘dressing gown’, -programm ‘program’, -radio ‘radio’, -rock ‘robe’ (45) Morgenmoderator ‘presenter of the morning show’, -muffel ‘morning grouch’ While all these compounds seem to refer to morning as a point of time, reference can be made to the duration of certain entities by means of suitable lexemes as well:

1384

VII. Semantics and pragmatics in word-formation II: Special cases

(46) Wochenbericht ‘weekly report’, -bilanz ‘balance’, -pensum ‘workload’, -umsatz ‘turnover’, -verlauf ‘course’ And there are groups of words where the temporal relation is conceptualized as a series of intervals that can be addressed directly or concerning their frequency of occurrence: (47) Wochenanfang ‘beginning of the week’, -mitte ‘midweek’, -ende ‘weekend’, -hälfte ‘half of the week’; Wochenstunde ‘hour per week’, -lohn ‘pay for the week’ Temporal modification can use temporal relations in different ways: so, entities of temporal extension may be referred to as fixed units of measure in regard to their duration or as recurring items of the same type.

2.2.4. Adjectives with temporal and local reference Somehow compounds like Wochenlohn ‘weekly wage; lit. week wage’ are names for an object that in a more descriptive way could be referred to as wöchentlicher Lohn ‘weekly wage’; in a similar way one could speak of Morgenkälte ‘morning cold’ and of morgendliche Kälte, and it is even possible to differentiate between zweiwöchentliche Treffen ‘biweekly meeting’ (frequency) and zweiwöchige Treffen ‘two-week meeting’ (duration). These adjectives already have a kind of adverbial character, and there are special formative elements like -weise (stundenweise ‘by the hour’) or -lang (tagelang ‘for days’) that spell out certain adverbial aspects. An even stricter functional differentiation exists between compounds and comparable attributive constructions with adjectives derived from a ‘local’ stem. Nominal phrases containing such adjectives are usually (but there are a number of idiosyncratic restrictions) used to give reference to the territory where an object originates or where an action takes place or exists. Adjectives of this group are preferably formed with the suffix -isch, sometimes with -er, cf. (48) Es wurden Zölle auf chinesische Lieferungen eingeführt. [Tariffs on Chinese supplies have been introduced.] (49) Er versucht einen amerikanischen Wein von Weltklasse zu produzieren. [He is trying to produce an American world class wine.] (50) Die Münchner Biergärten feierten ihr 200-jähriges Bestehen. [The Munich beer gardens celebrated their 200th anniversary.] There are only marginal instances of comparable compounds to be found. To a certain extent adjectives derived from adverbs − e.g., dort ‘there’, gestern ‘yesterday’ − offer the opportunity to integrate some deictic reference into noun phrases: (51) die dortigen/gestrigen Ereignisse [the local/yesterday’s events]

80. Spatial and temporal relations in German word-formation

1385

2.3. Between two worlds: topological compounds and prefixes There is a continuum of forms referring to an intrinsic or deictic interpretation of topological relations. It is no surprise that the inside-outside relation is a preferred pattern for spatial orientation, back-front orientation for the orientation in time. (52) Das Restaurant bietet Plätze im Innen- und Außenbereich. [The restaurant offers seats in the inner zone and outdoors.] This differentiation between inside and outside can be used for categorizing different situations, e.g., for two sides of the same thing, like in: (53) Innenseite und Außenseite der Fenster [inner and outer surface of the windows] But also for items belonging to the outer and inner space of objects, like, for instance, houses: (54) die Außen- und Innenwände eines Hauses [the exterior and interior walls of a house] Sometimes only the reference to some kind of containment viewed from the inside or from the outside is emphasized: (55) der Innenhof eines Hauses [the courtyard of a house] (56) die Außenterrasse des Lokals [the outdoor terrace of the restaurant] Other relations and dimensions (for instance, Ober-/Unterdeck/Zwischendeck ‘upper/ lower deck/steerage/deck in-between’; Vorbau ‘porch; lit. pre-structure’; Hinterzimmer ‘backroom’) are less systematically used; cf. Motsch (2004: 417 f.). The in-out classification plays an important role with adjectives (and adverbs) too, but especially in an abstract usage; adjectives (and adverbs) of this type that are unambiguously used in a local sense are not very common: (57) außerweltlich ‘extramundane’, innerweltlich ‘intramundane’, aushäusig ‘out-ofhouse’, inhäusig ‘in-house’ (only used in very marked contexts), außerorts ‘out of town’, innerorts ‘in town’ The central use refers to some organizational demarcation like in: (58) außerehelich ‘extramarital’, außeruniversitär ‘non-university’, außertariflich ‘non-payscale’, außerunterrichtlich ‘extracurricular’ In these cases usually antonyms have the form of the base adjectives that somehow constitute the default option.

1386

VII. Semantics and pragmatics in word-formation II: Special cases

Many adjectives that are used quite a lot show a high degree of lexicalization, like for instance: (59) außerordentlich ‘extraordinary’, außergewöhnlich ‘exceptional’ With adjectives like these, one can doubt the relevance of a description that relates to local categories. The educated versions of this type of adjective operate with some type of GrecoLatin prefix and on the average they signal their character as members of some kind of specialized language (only extraterritorial, intravenös, intramuskulär are found in corpora of general written German; for the use of German intra-, extra-, inter-, etc. see Kühnhold, Putzer and Wellmann 1978: 216−230). What is striking with this kind of “spatial” encoding is that the multidimensionality of space, the choice of different options of reference to relations in space and their bidirectionality are the basis of a very variable set of encoding means. This is different with the expression of time at least as far as its linguistic modelling in analogy with spatial relationships is concerned (cf. Evans 2013: 165−166) and we are dealing only with those cases. It is the experience of an individual, somehow living on his time-line (cf. Klein 2009: 13; Eichinger 1989: 339−347) that is reflected in the dominant encoding on a topological or dimensional front-back line (cf. Evans 2013: 53− 54; 247−249). According to this, complex adjectives with prefixes that refer to this orientation play a certain role for the encoding of time in analogy to spatial concepts. The ordering of periods of time relative to certain fixed points (for instance cultural events) is pictured in a spatio-temporal serialization of before/in front of and after/behind, cf. (60) Die Städte sind vorweihnachtlich geschmückt. [The cities are seasonally (lit. ‘pre-Christmas’) decorated.] (61) Die nachweihnachtlichen Einkäufe übertrafen alle Erwartungen. [The post-Christmas purchases exceeded all expectations.] The vor-nach pattern is used for some kind of periodization: (62) die voraufklärerische/aufklärerische/nachaufklärerische Philosophie [the pre-enlightenment/enlightenment/post-enlightenment philosophy] But it extends beyond temporal classification, so the prefix vor- ‘pre-’ in many cases suggests an interpretation of a provisional and sometimes even incomplete state: (63) eine vortheoretische und vorwissenschaftliche Sicht der Probleme [a pre-theoretical and pre-scientific view of the problems] The use of the autochthonous German prefixes like vor- and nach- is supplemented by the learned prefixes prä- and post- (which are much more in the center of this semantic area in English; see Bauer, Lieber and Plag 2013: 332−352); but adjectives with these prefixes are at least more common than those with intra- and extra-. There are some prä-/post- pairs:

80. Spatial and temporal relations in German word-formation

1387

(64) neueste Technologien in der prä- und postoperativen Versorgung [the latest technologies in pre- and post-operative care] And we find a “mixing” of the two formal patterns: (65) die Produktion in vorindustrieller Zeit [the production in pre-industrial times] (66) in der postindustriellen Gesellschaft [in post-industrial society] There is a certain productivity of post- in public discourse (Kühnhold, Putzer and Wellmann 1978: 233 note the beginning of such a tendency), where post-, as in postmodern ‘postmodern’ assigns a preliminary status to the un-prefixed category (e.g., modern), cf.: (67) einige post- und neofaschistische Parteien [some post- and neo-fascist parties]

3. Concluding remarks Localization and movement are not only important categories for ongoing situations. There are syntactic means of introducing such information into utterances by means of adverbs, prepositional phrases, subordinate clauses, or similar sentential constructions. On the other hand, there are also techniques for integrating some kind of spatial information into a complex word. This is a way to reduce the amount of singular recurring situations by ordering them as a kind of localization or movement types. Verbs, with their ability to highlight the syntacto-semantic schemata they are to be integrated into, are the prototypical candidates for the attachment of elements of spatial information. And in this respect, it is not surprising that there is some connection between adverbs, prepositions, and similar connecting elements on the syntactic level and corresponding particles used in word-formation. That is why we spoke of an adverbial type of spatial information. German verbs with elements like herein- or hinein- are a case in point. Many of them represent names for movement types in a way similar to syntactic constructions. In contrast to this, particles like ein-, aus-, auf-, ab-, an-, and so on, generally show a kind of use that is different from that of comparable “regular” syntactic constructions with adverbs or prepositions: they systematically represent names of situational patterns characterized by a generalized interpretation of the relations between their parts. Their interpretation is guided by our knowledge of patterns that are presumably in some way analogous. A temporal reading of some of these relations is in some way the translation of our experience with time into certain spatial situation types. The categorizing way of using spatial information in one part uses verbal traits in some nouns to reduce the general information given to a situationally relevant level (so for instance using the valency of reisen in speaking of Amerikareisende ‘travelers to America’ instead of Reisende ‘travelers’. But the core of the categorizing techniques is represented by the combination of two nouns − the first with some local (or temporal) meaning − that represent relevant parts of some functional context (for instance, not

1388

VII. Semantics and pragmatics in word-formation II: Special cases

every animal we possibly could find in a house would be called a Haustier ‘pet’. It is our knowledge of such contexts and of analogous patterns of word-formation that helps us decode this information. Somehow a combination or the adverbial and the categorizing techniques is realized in some kind of left-modification (prefixes and the like) using dimensionally or deictic orientation tools for general classification purposes on a spatio-temporal base.

4. References Bauer, Laurie, Rochelle Lieber and Ingo Plag 2013 The Oxford Reference Guide to English Morphology. Oxford: Oxford University Press. Bohnemeyer, Jürgen and Randi Tucker 2013 Space in semantic typology: Object-centered geometries. In: Peter Auer, Martin Hilpert, Anja Stukenbrock and Benedikt Szmrecsanyi (eds.), Space in Language and Linguistics. Geographical, Interactional and Cognitive Perspectives, 637−666. Berlin/Boston: de Gruyter. Booij, Geert E. 2002 The Morphology of Dutch. Oxford: Oxford University Press. Eichinger, Ludwig M. 1989 Raum und Zeit im Verbwortschatz des Deutschen. Eine valenzgrammatische Studie. Tübingen: Niemeyer. Eichinger, Ludwig M. 2000 Deutsche Wortbildung. Eine Einführung. Tübingen: Narr. Eichinger, Ludwig M. 2006a Wortbildung − ein Haus mit drei Nachbarn. In: Kristel Proost and Edeltraud Winkler (eds.), Von Intentionalität zur Bedeutung konventionalisierter Zeichen, 179−196. Tübingen: Narr. Eichinger, Ludwig M. 2006b Dependenz in der Wortbildung. In: Vilmos Ágel, Ludwig M. Eichinger, Hans-Werner Eroms, Peter Hellwig, Hans Jürgen Heringer and Henning Lobin (eds.), Dependenz und Valenz. Ein internationales Handbuch der zeitgenössischen Forschung. Vol. 2, 1065− 1080. Berlin/New York: de Gruyter. Engelberg, Stefan 2011 Argumentstrukturmuster als Konstruktionen? Identität − Verwandtschaft − Idiosynkrasien. In: Stefan Engelberg, Anke Holler and Kristel Proost (eds.), Sprache zwischen Lexikon und Grammatik, 71−112. Berlin/New York: de Gruyter. Evans, Vyvyan 2013 Language and Time. A Cognitive Linguistic Approach. Cambridge: Cambridge University Press. Felfe, Marc 2012 Das System der Partikelverben mit “an”. Eine konstruktionsgrammatische Untersuchung. Berlin/New York: de Gruyter. Filipović, Luna and Kasia M. Jaszczolt (eds.) 2012 Space and Time in Languages and Cultures. 2 Vol. Amsterdam/Philadelphia: Benjamins. Fleischer, Wolfgang and Irmhild Barz 2012 Wortbildung der deutschen Gegenwartssprache. 4th ed. Berlin/Boston: de Gruyter. Gambarotto, Pierre and Philippe Muller 2003 Ontological problems for the semantics of spatial expressions in natural language. In: Emile van der Zee and Jon Slack (eds.), Representing Direction in Language and Space, 144−165. Oxford/New York: Oxford University Press.

80. Spatial and temporal relations in German word-formation

1389

Grinevald, Colette 2006 The expression of static location in a typological perspective. In: Maya Hickman and Stéphane Robert (eds.), Space in Languages. Linguistic Systems and Cognitive Categories, 29−58. Amsterdam/Philadelphia: Benjamins. Hickman, Maya and Stéphane Robert 2006 Space in Languages. Linguistic Systems and Cognitive Categories. Amsterdam/Philadelphia: Benjamins. Jacobs, Joachim 2011 Grammatik ohne Wörter. In: Stefan Engelberg, Anke Holler and Kristel Proost (eds.), Sprache zwischen Lexikon und Grammatik, 71−112. Berlin/New York: de Gruyter. Klein, Wolfgang 2009 How time is encoded. In: Wolfgang Klein and Ping Li (eds.), The Expression of Time, 39−81. Berlin/New York: de Gruyter. Kühnhold, Ingeburg, Oskar Putzer and Hans Wellmann 1978 Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Dritter Hauptteil. Das Adjektiv. Düsseldorf: Schwann. Levinson, Stephen C. 2003 Space in Language and Cognition. Explorations in Cognitive Diversity. Cambridge: Cambridge University Press. McIntyre, Andrew 2002 Idiosyncrasy in particle verbs. In: Nicole Dehé, Ray Jackendoff, Andrew McIntyre and Silke Urban (eds.), Verb-Particle Explorations, 95−119. Berlin/New York: Mouton de Gruyter. Motsch, Wolfgang 2004 Deutsche Wortbildung in Grundzügen. 2nd ed. Berlin/New York: de Gruyter. Müller, Stefan 2002 Syntax or morphology: German particle verbs revisited. In: Nicole Dehé, Ray Jackendoff, Andrew McIntyre and Silke Urban (eds.), Verb-Particle Explorations, 119−141. Berlin/New York: Mouton de Gruyter. Pasch, Renate, Ursula Brauße, Eva Breindl and Ulrich Hermann Waßner 2003 Handbuch der deutschen Konnektoren. Linguistische Grundlagen der Beschreibung und syntaktische Merkmale der deutschen Satzverknüpfer (Konjunktionen, Satzadverbien und Partikeln). Berlin/New York: de Gruyter. Talmy, Leonard 2000 Toward a Cognitve Semantics. Cambridge, MA: MIT Press. Tenbrink, Thora 2007 Space, Time, and Use of Language. An Investigation of Relationships. Berlin/New York: de Gruyter. Tversky, Barbara 2005 Visuspatial reasoning. In: Keith J. Holyoak and Robert G. Morrison (eds.), The Cambridge Handbook of Thinking and Reasoning, 209−241. Cambridge: Cambridge University Press. Zifonun, Gisela, Ludger Hoffmann and Bruno Strecker 1997 Grammatik der deutschen Sprache. 3 Vol. Berlin/New York: de Gruyter.

Ludwig M. Eichinger, Mannheim (Germany)

1390

VII. Semantics and pragmatics in word-formation II: Special cases

81. Adverbial categories 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Introduction: the heterogeneity of the notion of “adverb” The deadjectival manner adverb between inflection and derivation Issues of morpheme ordering: manner adverbs and degree Further meanings and uses for manner adverbializing morphemes Dedicated morphology for sentence adverbs Denominal adverbs De-numeral and time quantifying adverbs Place adverbs Diachronic sources for adverbializing morphology References

Abstract Adverbs are a very problematic category with respect to both their notional internal consistency and their status as an open lexical class on their own. After some considerations on the non-prototypical nature of adverb formation, notoriously on the border between inflection and derivation, the present article starts by characterizing the features of a central procedure within the domain, namely the deadjectival manner adverb. It then gives a cross-linguistic illustration of the main extensions related to this formation procedure and further proceeds to discuss other well-represented instances of productive adverbializing morphology, to end with some brief diachronic considerations.

1. Introduction: the heterogeneity of the notion of “adverb” Giving in a few pages an overview of the morphological expressions available for the “adverbial categories” is a nearly impossible task due to the well-known heterogeneity of the notion “adverb”. Fortunately, since in this handbook the focus is on regular morphological procedures only, we can restrict ourselves to a “core” notion of adverbs as a distinct, open word class which performs the function(s) of non-nominal modification (basically a quite traditional view, followed in several recent approaches like Hengeveld 1992: 37, Huddleston and Pullum 2002: 563 or Haser and Kortmann 2006: 68). Taking this simplifying perspective as a starting point, one is nevertheless confronted with two difficulties: (i) adverbs as a major word class on their own, freely extendable by productive language-internal means, are far from universal; (ii) the notion of “non-nominal modification” is inherently disjunctive, as it covers at least three quite different functions, namely predicate, sentence/clause, and adjective/adverb modification, and it is not obvious a priori that these three could reasonably be subsumed under a single word class from a cross-linguistic point of view. Space limitations prevent us from discussing the above topics in much detail. As for point (i), apparently no general typological survey has been attempted to determine the extent of adverbial categories in the languages of the world, as remarked by Payne, Huddleston and Pullum (2010: 69), and not a single map on adverb formation is found

81. Adverbial categories

1391

in WALS (Dryer and Haspelmath 2011). In the − admittedly limited − sample of Bauer (2002: 42) the adjective → adverb and noun → adverb formations (without any semantic specifications) are not ranked among the most frequent ones, and Bauer also remarks that no instances of such formations are recorded in the Southwest Asia, Oceania and North America portions of his sample. However, we can safely say that an open lexical class of adverbs is commonplace in Europe, although not necessarily present everywhere (see some discussion in section 2). As for point (ii), focusing on Europe is again helpful: among the languages of Europe, when a language has a productive morphological device for forming predicate modifiers distinct from nominal ones (as, for instance, -ly in E. he reacted cleverly vs. a clever reaction), the same items may often fulfill the other two roles of sentence modification (cleverly, he did not react) and adjective modification (a cleverly quiet reaction), as discussed in section 4. This looks like a strong argument in favour of the unity of the category, at least from a morphological perspective. It is much less straightforward to claim that -ly suffixation in English properly belongs to lexeme formation. The contrary view, which prefers to treat clever and cleverly as different word forms of the same lexeme, and consequently -ly as an inflectional marker, has several supporters (among others, Haspelmath 1996: 49−50, and, with some more caution, Plag 2003: 195−196; within the generative paradigm, e.g., Baker 2003: 230− 235). Taking the inflectional approach for -ly and its many cross-linguistic relatives would put much of the following pages outside of the topic of the present handbook. In section 2 some arguments in favour of the derivational option will be proposed (for a much wider discussion on English, which reaches the same conclusions, see Payne, Huddleston and Pullum 2010), while some morpheme-ordering issues raised in section 3 would rather point towards the inflectional option. As will become clear, at any rate, deadjectival adverbs offer a paramount example of the need for adopting a “fuzzy-border” approach to the evergreen issue of the distinction between inflection and derivation (see article 14 on the delimitation of derivation and inflection).

2. The deadjectival manner adverb between inflection and derivation The deadjectival manner adverbs, as E. slow-ly or It. lenta-mente ‘id.’, probably do not constitute the prototypical conceptual core for adverbs as a lexical word class, since adverbial concepts which most frequently occur in texts and appear to be lexicalized in most languages of the world, seem to belong to several semantically divergent, (semi-) closed subclasses: focalizers (‘also’), intensifiers (‘very’), quantity expressions (‘much’), phasal expressions (‘still’, ‘already’), connectors (‘however’, ‘therefore’), etc.; cf. Ramat and Ricca (1994: 315−316), Payne, Huddleston and Pullum (2010: 69−72). However, in languages which possess regular morphological devices to form adverbs, the rule ‘qualifying adjective’ → ‘manner adverb’ enjoys a wide diffusion and is definitely a good starting point for dealing with the languages of Europe. Its formal realization shows a great deal of variation, which also has an impact on evaluating the status of the output on the derivation-inflection continuum. A sort of prototype for the above-mentioned morphological procedure is given by cases like English -ly and Western Romance -ment(e): a “well-behaving”, easily seg-

1392

VII. Semantics and pragmatics in word-formation II: Special cases

mentable and highly productive suffix attaching to an adjectival base. Further instances include Albanian -isht (Buchholz and Fiedler 1987: 363−366), Finnish -sti (Sulkala and Karjalainen 1992: 349), Basque -ki (Hualde and Ortiz de Urbina 2003: 347−348); others will be met throughout the article. Of course, regular phonological adjustments or stem allomorphies often occur: for instance, in It., Sp. and Port. lentamente ‘slowly’, the adjectival base of first-class adjectives takes an -a- vowel which is etymologically connected to the inflectional feminine ending, but must be analyzed synchronically as a mere, semantically void stem formative (a “morphome” in Aronoff 1994’s terminology); Fin. -sti causes consonant gradation in the stem; Hungarian -ul/-ül, -an/-en and -lag/-leg display vowel harmony as well as does Turkish -ca/-ce/-ça/-çe (which also has a consonant alternation), and so on. In English and Romance respectively, -ly and -ment(e) are the only suffixes which apply productively to qualifying adjectives, but in other languages there is competition among different productive suffixes in the same domain. This is probably more in tune with the usual behaviour of derivational categories: although the division of labour between the competing affixes can be partially treated in terms of morphological/semantic restrictions, there is no strict complementary distribution (contrary to the case of the inflectional word classes). For instance, in Hungarian (cf. Kenesei, Vago and Fenyvesi 1998: 371−72) -an/-en is required after adjectives derived with -s (among others), and -ul/-ül after those derived with the negative suffixes -talan/-telen and -atlan/-etlen; a third suffix, -lag/-leg, is often selected by the -i adjectives. Since all the adjectivizing morphemes above are very productive, so are the adverbial derivations related to them. But there is also a good deal of overlap: both -an/-en and -ul/-ül are found with nonderived bases, giving also rise to doublets without a semantic distinction, and other kinds of derived bases can take more than one suffix. Latin has at least two productive suffixes for deadjectival manner adverbs, namely -ē and -iter, which in the classical language attach respectively to -o/-a stems (stultus, -a, -um → stultē ‘stupidly’) and -i or consonant stems (brevis, -e → breviter ‘briefly’). However, this interesting selection of competing morphemes according to the inflectional class of the base does not hold entirely either in Old or in Late Latin (with -iter extending also to several -o/-a adjectives, though not the opposite), and therefore it must be ascribed at least partially to the impact of the conscious literary norm (Rosén 1999: 56− 61; for a general treatment of Latin adverbs Pinkster 1972, Ricca 2010). Similarly, in Modern Greek -os and -a formations (-ōs was the productive device in Ancient Greek, although -a was already required with superlatives) are chiefly divided according to the adjectival inflectional classes, but some overlap is present, with or without semantic distinction; cf. both áδika and aδíkos ‘unjustly’ but télia ‘perfectly’ vs. telíos ‘completely’ (Holton, Mackridge and Philippaki-Warburton 1997: 90−92). The Lezgian suffixes -dakaz and -diz/-z are described by Haspelmath (1993: 113) as freely interchangeable, except for the adjectival bases borrowed from Russian, which take only -dakaz. DumTragut (2009: 667) lists all three Eastern Armenian suffixes -(a)pes, -(a)bar and -oren as productive. As said above, even suffixes like E. -ly are considered by many authors to be inflectional rather than derivational. The chief arguments for this position (many others, less relevant, are discussed and dismissed in Payne, Huddleston and Pullum 2010) are: (i) the high grade of productivity, generality and semantic transparency of the process, and (ii) the alleged full syntactic conditioning of the choice between the adjectival and the

81. Adverbial categories

1393

adverbial form: basically, it is argued, the difference between a slow walk/walker and he walks slowly resides uniquely in the non-nominal nature of the head in the latter phrase, which should determine obligatorily the occurrence of the -ly form within a single adjective/adverb paradigm. As for E. -ly, point (i) is undeniable, but its decisiveness for the inflectionalist claim can be disputed from two complementary points of view. On the one hand, while it is generally true that inflectional processes are tendentially more general and semantically less idiosyncratic than derivational ones, English itself offers instances of suffixes as productive and general as -ly, which have not often been proposed as candidates for inflectional status: -ness is a case in point (Payne, Huddleston and Pullum 2010: 62). On the other hand, -ly itself surely displays some amount of idiosyncrasy and semantic opacity as well as semantic restrictions. For instance, it is hard to apply -ly to colour adjectives (Huddleston and Pullum 2002: 566), even if paint the wall *yellowly should be a suitable syntactic and semantic context. More generally, adjectives expressing sensory properties often do not keep their primary meaning in the corresponding adverbs, but may be employed with metaphoric value only (cf. warmly, coldly, heavily, dryly). Finally, substantial and idiosyncratic semantic shifts occur in cases like hard/hardly, new/newly, etc. Such a relevant semantic mismatch is not to be expected from different word forms of the same lexeme, while it is obviously unproblematic in derivation, since the derived item may well undergo semantic evolution on its own, autonomously from its base (cf. Plag 2003: 98). For English, point (ii) has been treated at length by Payne, Huddleston and Pullum (2010), who deny a full syntactic complementarity between adjectives and adverbs. In particular, they stress that -ly formations can productively post-modify nouns in constructions like the impact environmentally, the presence locally, etc. Notice that for Italian -mente, whose behaviour is parallel to -ly in most respects, the complementarity argument becomes nearly void, given that an adjective may occur in co-predicative function, as in i treni corrono veloci ‘the trains run speedy:M:PL’, besides the adverb in i treni corrono velocemente ‘the trains run speedy-ADV’. The same applies in Spanish or Latin. A different syntactic property which can be put forward against the inflectional status of -ly concerns argument structure (see again Payne, Huddleston and Pullum 2010: 64). Several adjectives governing an argument do not transfer this capability to the corresponding -ly adverb, e.g., proud/*proudly of his daughter. Again, such a mismatch would not be expected for the different cases of the same adjectival inflectional paradigm, while it is quite common between a base lexeme and its derivations. The same sort of arguments can be raised pro and contra the derivational status of most adverbializing morphemes mentioned in this section, which are roughly equivalent to E. -ly both formally and semantically. Of course, a detailed language-specific analysis should be provided in each case, and it is likely that not all linguists would agree on the relevance of the different criteria. For instance, Haspelmath (1993: 113) discusses adjectival adverbs in Lezgian under the heading of “Adjectival inflection”. Although he has a separate chapter on (lexical) adverbs, he does not call the former deadjectival, consistently with his position on E. -ly. Sulkala and Karjalainen (1992: 349) argue for treating Finnish -sti as a case ending, and therefore for giving inflectional status to Finnish “adverbs”. As for Latin -ē/-iter, Pinkster (1972: 63−70) takes the derivationalist position (also preferable for Ricca 2010: 112−113), after discussing extensively the inflectionalist alternative.

1394

VII. Semantics and pragmatics in word-formation II: Special cases

Independently of their precise position between inflection and derivation, the adverbial markers mentioned so far are formally quite distinct from the (other) inflectional markers of the adjective (when they exist). In many other instances, however, the distinction is much less clear. Not rarely, a productive category of “adverb” is realized by a morpheme which formally coincides with some inflectional allomorph of the adjectival declension. For instance, Danish and Swedish -t is also the marker of the neuter singular indefinite, Modern Greek -a coincides with the neuter plural accusative, Estonian -lt with the ablative (Jänes 1972: 121). This introduces a further difficulty in assigning the process to inflection or derivation. However, the mere coincidence of the adverbial marker with some inflectional ending should not be a sufficient condition for assigning the process to inflection, although it most often reveals a diachronic connection between the two functions (see section 9). Synchronically, the semantic and syntactic criteria mentioned above should keep their validity. The extreme limit of this formal continuum is reached in the well known cases of German or Dutch, where the manner verb modifier coincides with the unmarked form of the adjective, which occurs in the predicative use, e.g., German Der Klavierspieler ist gut/spielt gut ‘The pianist is good/plays well’ vs. ein gut-er Klavierspieler ‘a good pianist’. The prevailing approach here is not to consider German gut as being both an adjective and an adverb, but rather to assign it to a single part of speech, usually simply labelled “adjective”. Another possible term is “attribute”, employed, for example, by Hummel (2014: 60, and elsewhere). This seems indeed preferable on economy grounds, but in principle one could speak equally well of adjective → adverb conversion, and consequently of a lexeme-forming process. The latter choice becomes more appealing for languages in which the phenomenon is less systematic and occurs in addition to a productive overt adverbializing rule. Italian or Spanish offer clear examples, in my opinion, of an adjective → adverb conversion process as formally distinct from both derivational -mente adverbs and co-predicative adjectives: a third version of ‘trains run fast’ in Italian can be I treni corrono veloce, where veloce is clearly distinguishable from both alternatives velocemente and veloci given above, being identical with the unmarked form of the adjective (the masculine singular) and showing no agreement with the plural subject treni. For a wider discussion, see Ricca (2004: 550−553). The same occurs in Spanish rápido and the like. However, Hummel (2014) prefers to include this Romance type among the polyfunctional “attributes” mentioned above. At any rate, an independent open class of adverbs must be posited for German as well, even assuming that gut and the like are not adverbs, not only to include lexical unanalyzable items like hier ‘here’, heute ‘today’, etc. but also productive derivational processes in other semantic domains (especially the -(er)weise sentence adverbs dealt with in section 5). Besides suffixation and conversion (the latter inextricably linked with the above issues on the adjective-adverb polyfunctionality), other kinds of morphological procedures are also attested in (deadjectival) adverb formation, although − as expected − they are much less frequent, at least in Europe. Instances of productive prefixing are hard to find. Welsh shows a border-line process between morphology and syntax, involving the prefix/particle yn: araf ‘slow’ → yn araf ‘slowly’ (King 1993: 238). English a- in aboard, afresh, etc. can be analyzed as a denominal and deadjectival prefix (Huddleston and Pullum 2002: 567), but is not productive.

81. Adverbial categories

1395

Reduplication is very productive in Afrikaans, admittedly not a European language: stil, for example, means both ‘quiet’ and ‘quietly’ (similarly to Dutch and German), but stil stil is only ‘quietly’ (Donaldson 1993: 447). In Turkish as well, reduplication of both adjective and nouns is a productive (although not unrestricted) strategy for forming adverbs, as in yavaş ‘slow/slowly’ → yavaş yavaş ‘slowly’, efendi ‘gentleman’ → efendi efendi ‘in a gentlemanly way’ (Lewis 1988: 193; Göksel and Kerslake 2005: 100), together and in competition with the -ca suffix and with the conversion strategy. Stress shift as an adverbializing device occurs, for example, in Slovene (lépo ‘beautiful:N:SG:NOM/ACC’ → lepó ‘beautifully’; Herrity 2000: 239) and in Modern Greek (véveos ‘certain:M:SG:NOM’ → vevéos ‘certainly’; Holton, Mackridge and PhilippakiWarburton 1997: 91). It is unclear, however, to what extent these are really productive processes: as for Greek, the stress shift is basically a remnant of the Ancient Greek phonological rule, conditioned by the long ō in the adverbial ending -ōs, while vowel quantity no longer plays a role in the modern language.

3. Issues of morpheme ordering: manner adverbs and degree A further, purely morphological, criterion for locating adverbs along the inflection-derivation continuum is given by morpheme ordering. It is well known that inflection tends to be external with respect to derivation (the famous universal 28 in Greenberg 1966: 93, which is not without exceptions, however; see article 14 on the delimitation of derivation and inflection). Thus an adverbializing morpheme which freely takes different and contrasting inflected forms as input could arguably be considered as inflectional itself; on the other hand, if the output of the rule freely feeds further derivation, the process can be more safely placed on the derivational side. The second condition is only a sufficient one, since derivational affixes can have “closing” properties anyway (cf. article 54 on closing suffixes). This criterion will be mentioned for some of the adverbial categories discussed in the following sections. However, its application is not unequivocal either, especially whenever other non-prototypical categories are involved. A paramount instance is given by the interaction of adverbializing morphemes with the category of degree. When the latter is expressed morphologically, it is mostly assigned to inflection, albeit a non-prototypical one (cf. Dressler 1989: 6, among many others). European languages expressing the category of degree on both adjectives and deadjectival adverbs behave quite inconsistently with respect to its ordering with the adverbializing markers. Although all grammatical descriptions speak of “comparative of the adverb” and the like, thus suggesting the idea that degree is external with respect to adverbialization (as should be expected for an inflectional vs. a derivational category), perhaps the only European language that matches this ordering precisely at the morphological level is Basque: errex ‘easy’, errex-ki ‘easi-ly’, errex-ago ‘easi-er’, errex-ki-ago ‘more easily’ (Allières 1979: 59, and similarly for the superlative -en, or the excessive -egi, cf. also Hualde and Ortiz de Urbina 2003: 195). The opposite strategy, with the adverbializing marker external to the degree marker, is more common, as in Hungarian: hangos ‘loud’, hangos-an ‘loud-ly’, hangos-abb ‘loud-er’, hangos-abb-an ‘more loudly’ (Kenesei, Vago and Fenyvesi 1998: 349). In other cases cumulation occurs, e.g., Lithuanian ger-as ‘good-M:SG:NOM’, ger-ai ‘good-ADV, well’, ger-esn-is ‘good-COMP-

1396

VII. Semantics and pragmatics in word-formation II: Special cases

M:SG:NOM, better (adj)’, but ger-iau ‘good-COMP:ADV, better (adv)’ (Ambrazas 1997: 138, 386). The cumulative solution is obviously irrelevant for the ordering criterion as such, but can be taken by itself as an argument in favour of the inflectional nature of the adverbializing morphology (assuming that the comparative is inflectional), due to the rarity of the instances of cumulation between inflection and derivation (Ricca 2005: 206). The picture is further troubled by the fact that many languages may have a neat contrast between adjective and adverb in the positive degree, which then fades out for other values of the category. This is obviously the case for English (slow vs. slowly, but slower and slowest serving both adjectival and adverbial function; cf. Zwicky 1989). Latin behaves in the same way in the comparative (fort-ius both ‘strongCOMP:N:SG:NOM/ACC’ and ‘more strongly’) but not in the superlative, which patterns like Hungarian (fort-issim-ē ‘strong-SUP-ADV’). In Latvian (Holst 2001: 124−127), the comparative of the adverb employs the same comparative suffix -āk- of the adjective, but followed by a zero suffix, which never occurs in the adjectival paradigm: lab-s ‘good-M:SG:NOM’, lab-i ‘good-ADV, well’, lab-āk-s ‘good-COMP-M:SG:NOM, better (adj)’, lab-āk ‘good-COMP, better (adv)’. Thus the comparative adverb does not display any overt adverbial marker, as in Latin or English; moreover, it is less marked than any comparative adjectival form. The examples of rather idiosyncratic behaviour of the adjective vs. adverb contrast across different values of the degree category could be multiplied, and they too are problematic for the derivationalist position: it is surely undesirable to think of the adverb as being derived from the adjective in the positive degree, but being the same lexeme in the comparative. On the whole, the interaction between adverbializing and degree morphology displays a fair amount of complexity cross-linguistically, but would seem more compatible with a model which assigns both categories to the same subdomain of morphology, perhaps with a preference for subsuming both under inflection, contrary to the suggestions of the syntactic and semantic criteria put forward in section 2.

4. Further meanings and uses for manner adverbializing morphemes As said in section 1, deadjectival manner adverbs, at least within Europe, display a rather systematic polyfunctionality, both syntactically (i.e. beyond predicate modification) and semantically (i.e. beyond manner). The most important syntactic extension is given by sentence adverbs (for some crosslinguistic work, see Ramat and Ricca 1998). A sentence like cleverly, he did not react, where the adverb cleverly is outside the scope of negation, neatly shows that it may modify syntactic units of higher order than the bare predicate. In this case, the semantics of the adverb is intact, but the manner evaluation regards the participant’s behaviour (‘acting cleverly, he …’). Similarly, evaluative sentence adverbs can refer to the event (strangely/fortunately, he did not react). In present-day English, nearly any deadjectival adverb with suitable semantics can be employed in the syntactic function outlined above: morphology does not distinguish sentence from predicate modification. More or less the same behaviour occurs for many nearly-equivalents of E. -ly throughout Europe, like

81. Adverbial categories

1397

Romance -ment(e), Albanian -isht (Bucholz and Fiedler 1987: 365−366), Modern Greek -a (Holton, Mackridge and Philippaki-Warburton 1997: 363), although it is not easy to ascertain if this syntactic shift is equally productive in the different languages. A different level of sentence modification is expressed in frankly/sincerely, you are a liar, where the manner evaluation applies to the illocutionary force of the utterance, qualifying the speaker’s commitment. Even in contemporary English, this function is much less widespread and generalizable, and diachronically, it appears much later in the history of the language: Swan (1988: 434−435) dates it around the 17 th century. The deadjectival modal sentence adverbs (E. possibly, probably, certainly, surely and the like), which assign a truth value to the proposition expressed by the sentence, may also develop from ordinary manner predicate adverbs. See for instance It. difficilmente, meaning both ‘with difficulty’ and ‘hardly’, or E. easily when used modally with the meaning of ‘probably’. Finally, a further kind of sentence adverb quite widespread and productive in many modern European languages is given by “point of view” or “domain” adverbs, as in E. technically it’s right, but politically it’s unfeasible. Romance -ment(e) adverbs display the same use (e.g., It. politicamente), and so do Albanian politikisht, Modern Greek politiká, etc. In this case, beyond the syntactic shift from predicate to sentence modification also a significant semantic change is involved. Manner deadjectival adverbs display another relevant syntactic extension: they usually also serve as adjective modifiers. This use is not limited to degree/intensification meanings: in English the whole semantic range displayed by -ly adverbs in their predicate and sentence functions, briefly illustrated above, may be transferred to adjective modification (pleasantly fresh, politically correct, locally famous, surprisingly frank, frankly surprising, etc). The same holds for Romance languages, Latin (cf. Ricca 2010: 160− 168), Albanian (Bucholz and Fiedler 1987: 340), Modern Greek (Holton, Mackridge and Philippaki-Warburton 1997: 360), Hungarian (Maria Grossmann, p.c.); clearly, this should not be taken for granted for every language in which a category “adverb” is identified. Among semantic extensions, an area of cross-linguistic variation in Europe concerns the availability of manner adverbs derived from ethnic adjectives with the meaning ‘in X language’. Compare E. speak, sing *Frenchly and It. parlare, cantare *francesemente with Lat. Gallice loqui, Turkish Fransızca söyledi ‘(s)he sang in French’, (Göksel and Kerslake 2005: 215), Hungarian franciaul beszélni ‘to speak (in) French’, Modern Greek miló γaliká ‘I speak (in) French’, etc. An interesting instance of a dedicated affix is Ancient Greek -istí (rōmaistí ‘in Latin’, dōristí ‘in Doric’), connected to the bases Rōmaîos, Dō´rios, via the derived verbs in -ízō (dōrízō ‘I speak Doric’), although the usual deadjectival derivation in -ōs from -iko- adjectives is also possible (rōma-ik-ôs, dōr-ikôs).

5. Dedicated morphology for sentence adverbs It is not frequent at all, at least in Europe, that the many functions of sentence modification discussed in section 4 are contrastively marked with respect to the function of a manner predicate modifier (Ramat and Ricca 1998: 203−206). The best known instance

1398

VII. Semantics and pragmatics in word-formation II: Special cases

is surely given by German -(er)weise and its cognates in Germanic languages other than English (Dutch -(er)wijs, Danish/Swedish/Norwegian -vis). In German the suffix is very productive and covers a wide semantic span of sentence adverbs: mostly participantoriented (klugerweise ‘cleverly’) and event-oriented (glücklicherweise ‘fortunately’) evaluatives, but also modals (möglicherweise ‘possibly’). Thus morphology alone is sufficient to contrast er hat mir klugerweise geantwortet ‘cleverly, he answered me’ with er hat mir klug geantwortet ‘he answered me cleverly’. Interestingly, however, German -(er)weise is not usually found with domain meaning, where the “short” polyfunctional adjective/adverb occurs: politisch/*politischerweise ist es Unsinn ‘politically, it’s nonsense’. So the border between the two strategies does not properly coincide with the one between predicate and sentence modification. In Scandinavian languages the affix -vis is probably not productive any more, but contrasts similar to German do occur, as in Danish Han talte helt naturligt ‘he spoke very naturally’ (predicate adverb) vs. Han var naturligvis ikke hjemme ‘he was of course not at home’ (Allan, Holmes and Lundskaer-Nielsen 1995: 337). In Swedish (Holmes and Hinchliffe 1994: 330), -vis is added to the -t marker: naturligtvis ‘naturally’ (with sentential meaning). Scandinavian languages also offer an interesting example of a developing grammaticalization process towards a new marker for sentence modifiers. Besides the limited occurrences of the -vis suffix, there is the option of postposing nok/nog ‘enough’ to a manner adverb (e.g., Swedish klokt ‘cleverly’) to allow its use as a sentence modifier: for instance Swedish *klokt/klokt nog svarade han inte mig ‘cleverly, he did not answer me’ (Ramat and Ricca 1998: 209). For Norwegian, Swan (1991: 423) goes further in taking nok as a productive derivational suffix for evaluative sentence adverbs. The same function is found with English enough (oddly enough, he did not come), but with a much lower level of obligatoriness/grammaticalization.

6. Denominal adverbs In some languages, the productive suffixes for deadjectival manner adverbs take also nouns as bases, with comparable semantics. This is clearly not the case for Romance -ment(e) or English -ly (barring very few exceptions, like partly). On the contrary, Armenian (Dum-Tragut 2009: 667) has erexa ‘child’ → erexa-ya-bar ‘childishly’ besides azniv ‘honest’ → aznv-a-bar ‘honestly’, bžišk ‘doctor’ → bžšk-a-pes ‘like a doctor’ besides xor ‘deep’ → xor-a-pes ‘deeply’. Similarly, Albanian -isht takes both adjectives and nouns as input (Buchholz and Fiedler 1987: 363−364). Turkish -ca (and allomorphs) has a very wide range of uses, but it is chiefly an adverbializing suffix, and it freely takes both adjectives and nouns in this function. As a denominal suffix it may have both manner and domain/viewpoint semantics: çocuk-ça ‘like a child, childishly’, but also yaş-ça ‘age-wise’ (Kornfilt 1997: 463). However, there are several instances of specialized de-nominal adverbializing suffixes, which seem to concentrate on the two meanings already seen for de-adjectival procedures, namely manner ‘as a N, like a N, in form of N’ and domain/viewpoint ‘with respect of N, concerning N’. E. -wise productively covers both functions: crosswise, food-wise (Plag 2003: 98). The German etymological cognate -weise (Drosdowski 1984:

81. Adverbial categories

1399

501) − to be kept separated, both in meaning and form, from the deadjectival -(er)weise discussed in section 5 − is not used for domain meanings, but besides manner meaning proper (e.g., kreuzweise ‘in form of a cross’), it often occurs in dimensional or distributive contexts (massenweise auswandern ‘to emigrate in masses’); the same holds for the Scandinavian cognate -vis. The distributive meaning seems to be quite common for denominal adverbs, cf. Basque -ka mentioned below, and Lat. -ātim (centuriātim ‘by centuries’, gregātim ‘in herds’, etc.), although -ātim is also deadjectival. A specialized distributive denominal suffix is Hungarian -nként (e.g., házanként ‘house by house’), considered a case marker by some authors (e.g., Kenesei, Vago and Fenyvesi 1998: 192), but an adverbial derivational suffix by others (e.g., Kiefer 1987: 98). Interestingly, Romance languages do not have productive devices for forming denominal adverbs, neither with manner nor with domain meanings. In many cases the gap is easily filled by applying the deadjectival suffix -ment(e) to all kinds of denominal relational adjectives: It. [[[stipendi ]N-al ]Adj -mente]Adv ‘wage-wise’, [[[kant]N-iana]Adj -mente]Adv ‘in the manner/style of Kant’, etc. Denominal adverbial formations raise even more issues with respect to the inflection/ derivation continuum than adjectival ones, because a denominal adverbializing suffix, if considered inflectional, could be more easily inserted in the declensional pattern simply as a manner case marker. Indeed, affixes with meanings like ‘as N’ are often described exactly so in languages with a rich case system, e.g., Hungarian -ként in turistaként ‘as a tourist’ may be labelled as an “essive-formal” case (Kenesei, Vago and Fenyvesi 1998: 192, 370−371). For a general discussion of the problematic delimitation between cases and derivational markers in Hungarian, see Kiefer (1987). Chiefly on the basis of the criterion of co-occurence with nominal modifiers, he includes Hungarian -ként among nominal cases, but considers the semantically similar -képpen (e.g., példaképpen ‘as an example’) as an adverbial derivational suffix (Kiefer 1987: 97−98). Notice that the Hungarian suffix -ként behaves derivationally in being further derivable by the adjectivalizer -i (Kenesei, Vago and Fenyvesi 1998: 371), e.g., turistakénti utazás ‘journey as a tourist’; however, since -i also freely attaches to postpositional phrases, its diagnostic value is not very high. Basque -ka is chiefly denominal, with distributive or manner meaning: talde ‘group’ → taldeka ‘in groups’, urte ‘year’→ urteka ‘yearly’, pelota ‘ball’ → pelotaka ‘playing ball’, but in this latter meaning it can also be deverbal: korri(tu) ‘to run’ → korrika ‘running’, thus approaching converbal − and therefore inflectional − status (Hualde and Ortiz de Urbina 2003: 348). A quite different semantics which also involves nominal bases is given by place adverbs, dealt with in section 8 below.

7. De-numeral and time quantifying adverbs Although not as pervasive as deadjectival manner adverbs, de-numeral adverbs as a productive morphological category are widespread in Europe. Probably the most typical semantic value is given by time quantifiers (frequency adverbs) derived from numerals with the meaning “X times”. A cross-linguistic study by Moreno Cabrera (1998: 162− 165) shows that a contiguous area in Central Europe displays such feature, including

1400

VII. Semantics and pragmatics in word-formation II: Special cases

German -mal, Dutch -maal, Slovene -krat, Czech -krát and Hungarian -szor/-szer/-ször. Further instances of the same procedure are Lezgian -ra in Eastern Europe (Haspelmath 1993: 234), and Welsh -(g)waith in the West (e.g., pedairgwaith ‘four times’, if the noun gwaith ‘time’ can be treated as a grammaticalized suffix in this instance, cf. King 1993: 123−124). In all these cases a dedicated adverbializing morpheme occurs. Dum-Tragut (2010: 668) lists even two such suffixes for Armenian: (a)-patik (hing ‘five’ → hng-a-patik ‘five times’) and -ic’s (erek’ ‘three’ → er-ic’s ‘three times’): the latter is claimed to be more productive. Other languages in Europe have similar suffixes which, however, are limited to some low cardinal numbers: an extreme case is English, which has only once and twice (with base allomorphy) plus the obsolete thrice. Interestingly, both Classical languages had productive devices to form de-numeral adverbs: Latin -iē(n)s and Ancient Greek -ákis, but the former did not survive at all in Romance languages, while the latter is limited to formal registers of Modern Greek. For most languages with a productive procedure for denumeral adverbs, the same affix extends to at least some non-numeral quantifiers; cf. Lat. tōtiēs ‘so many times’, Ancient Greek oligákis ‘few times’, German manchmal ‘sometimes’, Slovene kólikokrat ‘how often’ (Herrity 2000: 245), Czech mnohokrát ‘many times’, Hungarian néhányszor ‘sometimes’, Lezgian sa šumud-ra ‘several times’ (Haspelmath 1993: 234), Eastern Armenian both bazmic’s and bazmapatik ‘several times’ from bazum ‘many’. Again, the derivational status of this class of morphological procedures can be questioned, especially when they exhibit full generality. In Basque (e.g., mila-tan ‘a thousand times’), the formation seems to be clearly inflectional, as it coincides with the inessive (or locative) case -tan which equally applies to all nouns − indeed, phrases (Moreno Cabrera 1998: 162). As for Hungarian -szor/-szer/-ször, the label “multiplicative case” is also found (cf. Kenesei, Vago and Fenyvesi 1998: 191−192), but the suffix applies to numerals and quantifiers only. A pro-derivation argument can be raised when the adverbs so formed may be further derived, particularly to give adjectives with the meaning ‘repeated X times, X-fold’. This was not the case for Latin and Ancient Greek, but holds for German dreimal-ig, Slovene tríkrat-en (Herrity 2000: 139), Hungarian háromszor-i, all meaning ‘repeated three times’. Hungarian may make other kinds of deadverbial derivations (a háromszor-os bajnok ‘the three-times champion’, Kenesei, Vago and Fenyvesi 1998: 345), even with non-adjectival outputs (háromszor-oz ‘to triple’). Adverbs may also be derived productively from ordinal numbers. With the meaning ‘for the Xth time’, the same suffix may be used, as in Hungarian harmad-szor ‘for the third time’, or a different one, as in Slovene trétj-ič ‘id.’ (Herrity 2000: 148). Another frequent meaning for de-ordinal adverbs is their use as connectives, to list different textual units in an argumentative discourse: cf. E. thirdly, fifthly or German drittens, fünftens, etc. Since ordinal numbers often show more similarity with usual adjectives than cardinal ones, it is not surprising that the usual deadjectival adverbializing devices may extend to this domain, as is the case for E. -ly or Fr. cinquième-ment ‘fifthly’, etc.

8. Place adverbs Within the highly heterogeneous adverb word class, the adverbs of place are also peculiar from a syntactic point of view, as they often serve as predicate arguments instead of

81. Adverbial categories

1401

circumstantials. Since in these cases they could not be seen as modifiers proper, some authors (e.g., Huddleston and Pullum 2002: 564; Haser and Kortmann 2006: 67−68) even exclude them from the category, but note that also manner adverbs may serve as arguments, for instance of verbs like behave or treat. At any rate, adverbs of place seem at first sight to contribute little to derivation: many basic ones obviously have no structure at all (E. here, there, out, up, etc.), and many others are not the product of contextfree morphological rules but rather of idiosyncratic univerbation processes originating in the syntagmatic discourse chain, especially from PPs (E. inside, away, behind, etc.). However, there do exist instances of derivational procedures in this domain as well. A good example may be given by English -wards, and similarly German -wärts, Dutch -waarts, to indicate direction of motion. These suffixes take as input several basic place adverbs/prepositions (German aufwärts ‘upwards’), which is untypical for derivation; but also names for cardinal points (German ostwärts ‘eastwards’) and common nouns denoting usual reference points (German seewärts ‘seewards’, Dutch huiswaarts ‘homewards’, etc.). E. -wards may also take place names (heading Londonwards). Latin -itus (cf. Ricca 2010: 116) seems to have a similarly limited productivity: it has mainly ablative and sometimes also locative meaning, and is both deadjectival (antīquus ‘ancient’ → antīquitus ‘from/in ancient times’) and denominal (rādix ‘root’ → rādīcitus ‘from the roots’). The frequent instances of autonomous meaning evolution also suggest a lexeme forming process in this case. Ancient Greek displays a numerically reduced, but paradigmatically ordered set of items with locative/ablative/allative meanings, formed by dedicated suffixes: -thi/-si (locative), -then (mainly ablative) and − less frequently − -se (allative), which take as input several pronominal roots (e.g., from allo- ‘other’ → állothi ‘elsewhere’ / állothen ‘from elsewhere’ / állose ‘towards elsewhere’), but also place names (Ilióthi ‘in Troy’, Athē´nēthen ‘from Athens’) and some lexical bases as well: agróthi ‘in the country’, oíkothen ‘from home’, etc. Since the processes are not general enough, the outputs are usually taken as derivational rather than inflectional, given that they are systematic enough to posit a word-formation rule. Finally, a possible instance of a specialized derivational marker for temporal adverbs is Hungarian -kor (e.g., éjfélkor ‘at midnight’, Kiefer 1987: 98) which is again considered inflectional by other authors (e.g., Kenesei, Vago and Fenyvesi 1998: 192): at any rate, its exclusively temporal meaning would be quite uncommon for a case as well. Even from the few examples above, it should be clear that once again the label of word-formation for these kinds of processes is largely a matter of degree, between the two extremes of a fully productive inflectional process on the one end and a closed group of wholly idiosyncratic, hardly analyzable items on the other. Semantic criteria are probably less helpful in this domain than in those discussed in the preceding sections. On the other hand, Kiefer’s criterion mentioned in section 6 points consistently against inflectional status for all the items discussed above, since none of them can co-occur with nominal modifiers (e.g., *Great Londonwards).

9. Diachronic sources for adverbializing morphology In the survey above it has been repeatedly stressed how problematic adverbs are regarding their position along the continuum between inflection and derivation. This duplicity

1402

VII. Semantics and pragmatics in word-formation II: Special cases

is reflected in diachrony as well, since productive adverbializing morphology arises via two very different diachronic paths. On the one hand, some adverbializing suffixes are well traceable to a lexical source, often − but not necessarily − via the intermediate stage of a compounding procedure. This is the case of E. -ly, ultimately from Old E. lic ‘body’ via the -e adverb in -lice from the compound adjectives in -lic ‘with the shape of X’. German -weise (and cognates) is transparently related to Weise ‘manner’, -weg ‘way’ is productive as an adverbializing suffix in Afrikaans (Donaldson 1993: 445) and also in Dutch (Donaldson 1997: 130), at least in spoken registers. Of course, Romance -mente also belongs here, although the details of its grammaticalization from the ablative of Latin mens ‘mind, attitude’ are not so straightforward as it might be expected (see Karlsson 1981; Bauer 2003; Ricca 2010: 181−185; article 106 on the Romance adverbs in -mente). To give one instance outside manner adverbs, the German frequency adverb marker -mal is still practically identical with the noun Mal ‘time’. The case of Scandinavian nok/nog mentioned in section 5 is rather peculiar, as it is normally expected that the new grammatical morphemes come from constructions in which they fulfill the role of (syntactic) head (see, e.g., Haspelmath 1992: 81), while nok/nog ‘enough’ clearly starts its path as a modifier. The emergence of new productive derivational affixes via the grammaticalization path sketched above is a quite different thing, of course, from the instances of univerbation/ lexicalization by which single adverbial items like outside, indeed, etc., enter the lexicon: the latter process does not change the speakers’ morphological competence, while a new derivational procedure, once acquired, enables the speaker to form new lexemes at once and free from context. See also Ramat (2011: 506−508). On the other hand, several adverbializing markers do not come from a lexical source. In particular, most of the productive adverbial markers in ancient Indo-European languages can be traced to originally inflectional material: Latin -ē to an ancient ablative (or perhaps an instrumental), Ancient Greek -ōs possibly to an instrumental, Slavic -ě to a locative (Ricca 1998: 456; Cuzzolin, Putzu and Ramat 2006: 13). Even the conversionlike procedure of Modern German is the accidental byproduct of the loss of a final -e, still present in Middle High German, which goes back to Old High German -o and ultimately to an Indo-European inflectional marker (Braune and Eggers 1987 [1886]: 228). The new derivational function may co-exist with the old inflectional one, yielding many of the instances of synchronic coincidence between adverbializing and inflectional markers mentioned in section 2. Similarly to the preceding case, also this second diachronic path must be kept separated from the frequent occurrences of single lexical adverbs arising from “frozen” inflectional forms, either because the relevant inflectional category has been lost in the language (e.g., Latin vesperī ‘at evening’, Ancient Greek oíkoi ‘at home’, from old IndoEuropean locative forms, Modern Greek práγmati ‘really, in fact’, originally the (lost) dative of práγma ‘thing, fact’), or because the lexeme has disappeared as such, leaving only an isolated form of the paradigm (e.g., Lat. palam ‘overtly’, ōlim ‘once’ from the accusative of synchronically unattested lexemes). In the latter cases we can speak of idiosyncratic lexicalizations, because no productive word-forming device has arisen; on the contrary, Latin -ē or Ancient Greek -ōs left the inflectional paradigm but survived in the grammar as productive derivational devices. Such “de-inflectionalization” processes have been placed among the rare instances of degrammaticalization by Norde (2009: 152−185) − although she does not mention the

81. Adverbial categories

1403

case of de-inflectional adverbs − because she considers derivation as intermediate on a linear scale between lexicon and inflection; for other authors, however (e.g., Ricca 1998: 453; Haspelmath 2004: 32), the lexicon-inflection and lexicon-derivation chains are rather two independent grammaticalization paths, and consequently the transitions between the two domains − which are indeed attested in both directions − should not be relevant for the debated issue of (uni)directionality in grammaticalization.

10. References Allan, Robin, Philip Holmes and Tom Lundskaer-Nielsen 1995 Danish. A Comprehensive Grammar. London: Routledge. Allières, Jacques 1979 Manuel pratique de basque. Paris: Picard. Ambrazas, Vytautas (ed.) 1997 Lithuanian Grammar. Vilnius: Baltos Lankos. Aronoff, Mark 1994 Morphology by Itself. Stems and Inflectional Classes. Cambridge, MA: MIT Press. Baker, Mark C. 2003 Lexical Categories. Verbs, Nouns and Adjectives. Cambridge: Cambridge University Press. Bauer, Brigitte L. M. 2003 The adverbial formation in mente in Vulgar and Late Latin: A problem in grammaticalization. In: Heikki Solin, Martti Leiwo and Hilla Halla-aho (eds.), Latin vulgaire − Latin tardif VI, 439−457. Hildesheim: Olms. Bauer, Laurie 2002 What you can do with derivational morphology. In: Sabrina Bendjaballah, Wolfgang U. Dressler, Oskar E. Pfeiffer and Maria D. Voeikova (eds.), Morphology 2000. Selected papers from the 9 th Morphology Meeting, Vienna, 24−28 February 2000, 37−48. Amsterdam/Philadelphia: Benjamins. Braune, Wilhelm and Hans Eggers 1987 [1886] Althochdeutsche Grammatik. 14th ed. rev. by Hans Eggers. Tübingen: Niemeyer. Buchholz, Oda and Wilfried Fiedler 1987 Albanische Grammatik. Leipzig: Enzyklopädie. Cuzzolin, Pierluigi, Ignazio Putzu and Paolo Ramat 2006 The Indo-European adverb in diachronic and typological perspective. Indogermanische Forschungen 111: 1−38. Dressler, Wolfgang U. 1989 Prototypical differences between inflection and derivation. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 42: 3−10. Donaldson, Bruce C. 1993 A Grammar of Afrikaans. Berlin/New York: Mouton de Gruyter. Donaldson, Bruce C. 1997 Dutch. A Comprehensive Grammar. London: Routledge. Drosdowski, Günter (ed.) 1984 Grammatik der deutschen Gegenwartssprache. Mannheim: Bibliographisches Institut. Dryer, Matthew S. and Martin Haspelmath (eds.) 2011 The World Atlas of Language Structures Online. Munich: Max Planck Digital Library. Available online at http://wals.info [last access 14 Nov 2014]. Dum-Tragut, Jasmine 2009 Armenian. Modern Eastern Armenian. Amsterdam/Philadelphia: Benjamins.

1404

VII. Semantics and pragmatics in word-formation II: Special cases

Göksel, Aslı and Celia Kerslake 2005 Turkish. A Comprehensive Grammar. London: Routledge. Greenberg, Joseph 1966 [1963] Some universals of grammar with particular reference to the order of meaningful elements. In: Joseph Greenberg (ed.), Universals of Language. 2nd ed., 73−113. Cambridge, MA: MIT Press. Haser, Verena and Bernd Kortmann 2006 Adverbs. In: Keith Brown (ed.), Encyclopedia of Language and Linguistics. 2 nd ed. Vol. 1, 66−69. Amsterdam: Elsevier. Haspelmath, Martin 1992 Grammaticization theory and heads in morphology. In: Mark Aronoff (ed.), Morphology Now, 69−82. New York: SUNY Press. Haspelmath, Martin 1993 A Grammar of Lezgian. Berlin/New York: Mouton de Gruyter. Haspelmath, Martin 1996 Word-class-changing inflection and morphological theory. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1995, 43−66. Dordrecht: Kluwer. Haspelmath, Martin 2004 On directionality in language change with particular reference to grammaticalization. In: Olga Fischer, Muriel Norde and Harry Peridon (eds.), Up and down the Cline − The Nature of Grammaticalization, 17−44. Amsterdam/Philadelphia: Benjamins. Hengeveld, Kees 1992 Parts of speech. In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds.), Layered Structure and Reference in a Functional Perspective, 29−55. Amsterdam/Philadelphia: Benjamins. Herrity, Peter 2000 Slovene. A Comprehensive Grammar. London: Routledge. Holst, Jan Henrik 2001 Lettische Grammatik. Hamburg: Buske. Holmes, Philip and Ian Hinchliffe 1994 Swedish. A Comprehensive Grammar. London/New York: Routledge. Holton, David, Peter Mackridge and Irene Philippaki-Warburton 1997 Greek. A Comprehensive Grammar of the Modern Language. London: Routledge. Hualde, José Ignacio and Jon Ortiz de Urbina (eds.) 2003 A Grammar of Basque. Berlin/New York: Mouton de Gruyter. Huddleston, Rodney and Geoffrey K. Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Hummel, Martin 2014 The adjective-adverb interface in Romance and English. In: Petra Sleeman, Freek Van de Velde and Harry Perridon (eds.), Adjectives in Germanic and Romance, 35−71. Amsterdam/Philadelphia: Benjamins. Jänes, Henno 1972 Grammatik der estnischen Sprache. Malmö: Liber. Karlsson, Keith E. 1981 Syntax and Affixation. The Evolution of MENTE in Latin and Romance. Tübingen: Niemeyer. Kenesei, István, Robert M. Vago and Anna Fenyvesi 1998 Hungarian. London: Routledge. Kiefer, Ferenc 1987 The cases of Hungarian nouns. Acta linguistica Academiae Scientiarum Hungaricae 37: 93−101.

81. Adverbial categories

1405

King, Gareth 1993 Modern Welsh. A Comprehensive Grammar. London: Routledge. Kornfilt, Jaklin 1997 Turkish. London: Routledge. Lewis, Geoffrey L. 1988 [1967] Turkish Grammar. Oxford: Oxford University Press. Moreno Cabrera, Juan Carlos 1998 Adverbial quantification in the languages of Europe: Theory and typology. In: Johan van der Auwera (ed.), Adverbial Constructions in the Languages of Europe, 147−185. Berlin/New York: Mouton de Gruyter. Norde, Muriel 2009 Degrammaticalization. Oxford: Oxford University Press. Payne, John, Rodney Huddleston and Geoffrey K. Pullum 2010 The distribution and category status of adjectives and adverbs. Word Structure 3(1): 31− 81. Pinkster, Harm 1972 On Latin Adverbs. Amsterdam: North−Holland. Plag, Ingo 2003 Word-Formation in English. Cambridge: Cambridge University Press. Ramat, Paolo 2011 Adverbial grammaticalization. In: Heiko Narrog and Bernd Heine (eds.), The Oxford Handbook of Grammaticalization, 502−510. Oxford: Oxford University Press. Ramat, Paolo and Davide Ricca 1994 Prototypical adverbs: On the scalarity/radiality of the notion of ADVERB. Rivista di Linguistica 6: 289−326. Ramat, Paolo and Davide Ricca 1998 Sentence adverbs in the languages of Europe. In: Johan van der Auwera (ed.), Adverbial Constructions in the Languages of Europe, 187−275. Berlin/New York: Mouton de Gruyter. Ricca, Davide 1998 La morfologia avverbiale tra flessione e derivazione. In: Giuliano Bernini, Pierluigi Cuzzolin and Piera Molinelli (eds.), Ars linguistica. Studi offerti da colleghi ed allievi a Paolo Ramat in occasione del suo 60° compleanno, 447−466. Roma: Bulzoni. Ricca, Davide 2004 Conversione in avverbi. In: Maria Grossmann and Franz Rainer (eds.), La formazione delle parole in italiano, 550−553. Tübingen: Niemeyer. Ricca, Davide 2005 Cumulative exponence involving derivation. In: Wolfgang U. Dressler, Dieter Kastovsky, Oskar E. Pfeiffer and Franz Rainer (eds.), Morphology and Its Demarcations. Selected papers from the 11th Morphology Meeting, Vienna, February 2004, 161−182. Amsterdam/Philadelphia: Benjamins. Ricca, Davide 2010 Adverbs. In: Philip Baldi and Pierluigi Cuzzolin (eds.), New Perspectives on Historical Latin Syntax. Vol. 2: Constituent Syntax: Adverbial Phrases, Adverbs, Mood, Tense, 109−191. Berlin /New York: Mouton de Gruyter. Rosén, Hannah 1999 Latine loqui. Trends and Directions in the Crystallization of Classical Latin. München: Fink. Sulkala, Helena and Merja Karjalainen 1992 Finnish. London: Routledge.

1406

VII. Semantics and pragmatics in word-formation II: Special cases

Swan, Toril 1988 Sentence Adverbials in English. A Synchronic and Diachronic Investigation. Oslo: Novus. Swan, Toril 1991 Adverbial shifts: evidence from Norwegian and English. In: Dieter Kastovsky (ed.), Historical English Syntax, 409−438. Berlin/New York: Mouton de Gruyter. Zwicky, Arnold M. 1989 Quicker, more quickly, *quicklier. In: Geerd Booij and Jaap van Marle (eds.), Yearbook of Morphology 1989, 139−173. Dordrecht: Foris.

Davide Ricca, Turin (Italy)

82. Denominal verbs 1. 2. 3. 4.

Introduction The interpretation of denominal verbs Structural questions References

Abstract The article overviews (the literature on) (affixed and unaffixed) denominal verbs. Section 2 discusses semantic and pragmatic issues. Section 3 treats structural matters, namely (i) whether denominal verbs are best analysed using derivation from nouns, category underspecification or category-neutral roots, (ii) the various syntactic and non-syntactic approaches, and (iii) complex denominal verbs, where prefixes and particles co-occur with otherwise unacceptable denominal verbs.

1. Introduction When a journalist wrote that protestors had gnomed the Reserve Bank of Australia (i.e. adorned it with garden gnomes), when Shakespeare wrote It out-Herods Herod, and when a linguist was heard to say I George W. Bushed my way through my talk, they were exploiting a word-formation pattern which has inspired a voluminous literature. We will see that a correct analysis of the creation of verbs from (elements otherwise used as) nouns could teach us much about the interaction of semantics, pragmatics, syntax, morphology and the lexicon. This overview begins by reviewing the main interpretational issues discussed in the literature on denominal verbs. Section 2.1 illustrates the main semantic classes of denominal verbs and problems of semantic analysis (e.g., does saddle horses directly express

82. Denominal verbs

1407

motion of the saddle onto the horse or is this a grammatically irrelevant inference from a semantics corresponding more closely to provide horses with saddles). Section 2.2 deals with semantic and pragmatic constraints on their interpretation, e.g., why the verb file can mean ‘to put in a file’ only in contexts where the file is used in its intended storage function: file {documents/*paperclips}. Section 3 concerns morphological and syntactic aspects of denominal verbs, firstly treating the derivational source of (items called) denominal verbs (section 3.1): Is hammerV (i.e. the verb hammer) derived from hammerN (the noun hammer), or vice-versa, or are both derived from a root whose category is underspecified? We review the various (syntactic and non-syntactic/lexical) accounts of the structure of denominals (section 3.2). Finally, section 3.3 examines problems raised by complex denominal verbs, in which prefixes and particles occur with otherwise unacceptable denominal verbs (we pigged out vs. *we pigged). Notes on terminology: The term denominal (verb) refers to all verbs which are seen by at least some linguists as being derived from nouns. It applies to hammerV although not all theories derive it from hammerN, as well as to cases like profiteerV ← profiteerN (where profiteerN is uncontroversially the derivational source since -eer is otherwise a noun-forming suffix). The term denominal implies no analytic choice; it is used even by linguists who deny that hammerV is from hammerN. We speak of denominal verbs irrespectively of whether they are formally identical to nouns (hammer) or show formal differences such as consonant voicing (shelfN/shelveV), apophony/ablaut/umlaut (German Haut ‘skin’ → häut- ‘to skin’) or overt affixes (crystallize, derail ). Our area of interest thus partly overlaps with conversion/zero derivation, terms used here for any categorychanging derivation effected without overt category-determining affixation (see article 17 on conversion). A final terminological point is that we will refer to the noun which is the (putative) base of the denominal verb as the incorporated N, again without presupposing any particular analysis.

2. The interpretation of denominal verbs 2.1. Semantic classes of denominal verbs Semantic analyses of denominals often classify them into subgroups according to the relation of the incorporated N to the event they name. An illustration is the classification of English converted denominal verbs in (1)−(7), a simplification and partial modification of the classification in Clark and Clark (1979; henceforth C&C). The terms location and locatum in (1) and (2), now standard in research on denominals, refer respectively to the internal (ground) and external (figure, theme) arguments of prepositions. (1)

Location verbs: Direct object (intransitive subject) moves to/into/onto incorporated N: a. bottle the wine, {box/shelve/catalogue/blacklist} the books b. the share {peaked/bottomed out} at $1; they surfaced

(2)

Locatum verbs: Incorporated N moves to/onto/into the direct object (intransitive subject):

1408

VII. Semantics and pragmatics in word-formation II: Special cases a. {perfume/cream/powder/bandage} her face, {paper/paint/tile/soil} the wall, man a ship, {name/crown/arm} people, {address/stamp} a letter b. the window {iced over/fogged up}; the sky clouded over

(3)

Privative location verbs: Direct object is removed from incorporated N: mine the gold, quarry the marble, shell the peas

(4)

Privative locatum verbs: Incorporated N is removed from direct object: {gut/skin/milk} animals, dust shelves, stone the fruit, weed gardens

(5)

Instrument verbs: incorporated N is an instrument in the event V names: a. rollerskate/jet to London; ship/wheelbarrow them to another place b. microwave the food, knife people, mop/sandpaper/varnish/plane the floor

(6)

The incorporated N comes into existence: a. Cause object to become N: {orphan/outlaw/scapegoat} someone, cup one’s hands b. Become N: the clouds mushroomed, the truck jack-knifed c. Produce N: the animal foaled/calved, the child teethed

(7)

Agent verbs: Subject acts in the capacity named by incorporated N: a. {pilot/guard} aircraft, butcher animals, mother children, police the area b. they {pickpocketed/modelled for H&M/starred in films/fooled around}

This classification is one of several semantic classifications in the literature on denominal verbs in English and other languages (others are given in Gottfurcht 2008; Kaliuščenko 1988, 2000; Karius 1985; Leitner 1974; Marchand 1969; Rimell 2012). Since detailed comparison of these is beyond the scope of this article, we will confine ourselves to illustrating the most important problems in the semantic classification of denominals. First, it is often hard to determine which semantic-conceptual properties of a situation named by a denominal verb are referred to by the rule which created it. It is thus unclear if the rule that productively forms ‘locatum’ verbs like (2) specifically states that verbs can be formed from Ns naming objects that move to the direct object referents. Noticeably many ‘locatum’ verbs can equally well be analysed in terms of possession or some related notion expressible with with or have. Thus, WD-40 the chain would mean ‘provide it WITH WD-40 (a lubricant)’ rather than ‘put WD-40 on it’, the sky clouded over could mean that it became covered WITH clouds or came to HAVE clouds on it. Linguists have often assumed either this possessive analysis (Hale and Keyser 1993, 2002; Hirschbühler and Labelle 2009; Kiparsky 1997; Stiebels 1997; Volpe 2002) or the motion analysis (C&C, Gottfurcht 2008; Lieber 2004; Plag 1999) without real comparison of the two analyses. This requires detailed empirical study, but my preliminary tests favour the possessive analysis: Post-mortem/wound him entails that he HAD a post-mortem/ wound. Post-mortems and wounds do not move (except perhaps in a hard-to-verify metaphorical sense). Partitioning a room with a shelf entails providing the room WITH a partition (and not a physically impossible movement into the room of something which, prior to the event, was not a partition and may have already been in the room). The possessive analysis is dubious for they stoned him, but it is unclear if the motion of the stones is relevant here or if such examples are instances of the productive class of instrument verbs in (5).

82. Denominal verbs

1409

The lesson here is that obvious aspects of a situation, say the motion of the oil in I oiled the chain, could be mere entailments of the semantics of denominal verb formation without being intrinsically part of it. This problem of identifying grammatically relevant aspects of meaning applies to countless other issues, for instance, (i) whether the fact that many -ee-nouns like employee correspond to objects of the related verbs is specifically referred to in the job description of the affix or is a coincidental side-effect of the semantics of -ee (cf. article 52 on semantic restrictions on word-formation: the English suffix -ee), (ii) whether the non-head of car wrecker is grammatically represented as an argument of wreck or merely as an entity standing in some relation to wrecker (cf. articles 20 on composition, 33 on synthetic compounds in German) and (iii) whether to in give books to her characterises its complement as a goal of motion or as a recipient/ possessor (Rappaport Hovav and Levin 2008). Any polysemous word-formation device raises questions about the right level of generality: Is it empirically feasible to posit unified semantic descriptions for -er in writer of novels, toaster and prisoner (cf. article 74 on agent and instrument nouns) or for compounds of the types toymaker, toy box and toy car (cf. article 20 on composition)? Taxonomies like (1)−(7) likewise raise questions about the right compromise between (i) maximally specific descriptions which risk losing larger generalizations by trying to capture unsystematic, unproductive cases (the barstaff carded me ‘checked my ID card’ vs. *the police passported me), and (ii) maximally general (underspecified) semantic descriptions, which risk overgeneration, vagueness and untestability. To illustrate (ii), consider a claim that the classes in (1)−(7) are not explicitly sanctioned anywhere in grammar but are just specific manifestations of a general rule which states only that denominal verbs can be formed from nouns naming salient participants in an event (cf. Aronoff 1980). Such analyses are challenged by the differing degrees of productivity of the patterns in (1)−(7). Agent verbs like (7) are less productive than those in (1) and (2). Examples like *they actor/lawyer for a living or *he schoolmasters too often strike me as at best adventurous and metalinguistic. One might tender pragmatic explanations for such gaps (see Aronoff 1980: 753 on *surgeonV), but this risks excluding well-formed parallel verbs in German (schauspielern ‘actorN>V, to do acting’, schulmeistern ‘to act like a schoolmaster’). The classes in (1)−(7) are meant to illustrate problems of semantic analysis of denominal verbs using the example of English converted verbs, of course without implying that every denominal verb formation process in every language will yield the same semantic classes. Within English there has been comparison of the semantics of conversion and of overt affixes like those in hospitalize or classify, particularly with reference to whether the apparent greater semantic flexibility of the conversion mechanism points to a fundamental difference between conversion and affixation (see Baeskow 2006; Gottfurcht 2008; Lieber 2004; Plag 1999 and section 3.2). Apart from Kaliuščenko (2000), crosslinguistic variation in denominal verb formation has had little attention, despite the potential interest in knowing how classes like (1)−(7) are expressed with overt and unmarked denominal verb formation mechanisms in other languages, and which semantic classes tend towards having uniform exponents in other languages. Also illuminating would be a crosslinguistic study which asks which functions a marker can have in addition to denominal verb formation. Examples of such polyfunctional markers include -i/-j affixation in early Germanic languages, which formed deverbal and deadjectival causatives and various kinds of denominal verbs (van Gelderen

1410

VII. Semantics and pragmatics in word-formation II: Special cases

2011 and references), Indonesian -kan, which forms causative, benefactive, resultative and denominal verbs (see Son and Cole 2008, esp. footnote 12, p. 130 f.) or mediopassive/non-active morphology in Latin, which formed, e.g., passive, impersonal, reflexive, deadjectival and denominal verbs (Miller 1993: 225−229; Kallulli 2013).

2.2. Semantic drift and pragmatic questions Semantic analyses of denominal verbs must deal with various forms of semantic drift. Denominals often undergo metaphoric shifts (ideas can be shelved, ditched or spiced up) or idiosyncratic semantic specialization: for some speakers doctorV only has a metaphoric reading (doctor the evidence/*patient), and there is no (obvious) reason why trashV means ‘to lay waste to’ and rubbishV means ‘to criticise’, and not vice-versa (see C&C, p. 781 for other examples). Some denominals are synchronically unanalysable (and perhaps not perceived as denominal) because the extralinguistic factors that led to their creation are forgotten. Lynch and boycott are derived from the names of now littleknown historical figures. Dialing numbers or booking people need no longer involve dials and putting names in books. These shifts are nothing new in word-formation. Nominal compounds and complex verbs can also be metaphoric (wallflower, overturn governments), specialised (wheelchair ‘chair for the disabled’, overstep the {line/*objects on the floor}) or outlive the extralinguistic phenomena for which they were originally coined (watchmaker ‘repairer of clock-like devices’, ring off ‘to end a telephone conversation, no longer with concomitant ringing’). In much-discussed contrasts like (8) the acceptability of the parenthetic instrument phrases diagnoses the ability of a denominal verb to be used when the incorporated N does not strictly name the entity involved in the event. Such contrasts have often been taken to show that there are two semantically distinct rules at work, for instance differing in whether the input is a noun (as with tape) or a category-neutral root (hammer) (Kiparsky 1997: 485−491; Arad 2003; Don 2005). However, such formal approaches are not compelling. Driving in nails with shoes has a far greater resemblance to using hammers than inserting pins does to applying tape, and Dowd (2010) notes that analogous contrasts exist with use shoes as hammers vs. *use pins as tape, arguing that hammer, unlike tape, is defined more in terms of function than shape. It is thus superfluous to posit distinct denominal verb formation rules for contrasts like (8a, b). Harley and Haugen (2007) observe that (8c) is possible if the nails are twisted in in a screw-like manner. One can add that the unacceptable interpretations of (b) and (c) are blocked by the wellestablished verbs pin, zip and nail respectively. (8)

a. hammer nails in (with shoes), brush coats (with towels), shelve (books on tables) b. tape it to the wall (*with pins), button the coat up (*with a zip) c. #screw it to the wall with nails

Denominal verbs are sometimes subject to what has been called the canonical use constraint: they are often specialized to naming events in which the incorporated N is used in its canonical, designated function (Kiparsky 1997; C&C 785 f.). Thus, we find fileV

82. Denominal verbs

1411

the documents but not *file the paperclips, since files are made to function as containers for documents but not for paperclips. Similar examples are given in (9). (9)

hospitalise {patients/*nurses}, postcode {letters/*an address book}, drug {athletes/ *the medicine cabinet}, clothe {children/*washing lines}

The canonical use constraint is perhaps an instance of a more general phenomenon in which some word-formation processes are confined to nameworthy concepts. This notion is frequently invoked in the literature on incorporation (see sources in Rimell 2012: 186). An illustration is that English backformed noun-incorporating verbs tend to name activities which are sufficiently frequent and entrenched in Anglophone (sub)cultures to merit names (babysit, fundraise, head-bang, but *clock-scrub, *piano-burn, *frog-criticize). Similarly, AN compounds are confined to nameworthy concepts, cf. híghchair (well-known type of chair for children) and upside-dówn fridge (fridge with the freezer at the bottom), while non-nameworthy cases of elevated chairs and inverted fridges do not receive compound stress. Nameworthiness also constrains the interpretation of denominal verbs. Data like she pocketed her wallet (*in the trousers lying on the sofa) do not follow from the canonical use constraint, but make sense if pocketV lexicalises the familiar (and thus nameworthy) action of putting objects in pockets in clothes one is currently wearing. Privative verbs like weed gardens or dust the shelf likewise name memorized activities (and not canonical uses of the incorporated Ns) and the limited nameworthiness of the act of removing powder from objects makes it unlikely that powder the shelf can have this sense. Nameworthiness can also explain the effects of the canonical use constraint: the contrast crown {kings/*display cabinets} exists because coronations have more ritual significance than putting crowns on display. Finally, the nameworthiness of denominals can sometimes be confirmed by their having simplex translational equivalents in other languages: stone ‘to execute with stones’ matches French synchronically underived lapider; French balayer ‘to sweep’ (← balai ‘broom’) and poignarder ‘to stab’ (← poignard ‘dagger’) match underived English sweep and stab. The canonical use constraint and nameworthiness effects do not hold for all denominal verbs: The use of gnome ‘to adorn with gnomes’ mentioned at the start of this article is hardly a canonical, memorized use of gnomes. C&C (785 f.) give further examples, including attested uses of bottleV in the sense ‘to throw bottles at’. Such examples may involve spontaneously created concepts, but this is hard to test. Perhaps denominal verb formation occupies an intermediate position between word-formation processes like those seen above which are confined to nameworthy concepts, and processes showing no such constraints (e.g., nominals like clock-scrubber, piano burner, frog criticizer). For more on the canonical use constraint and nameworthiness, see Rimell (2012, ch. 5). Kiparsky (1997) contended that the canonical use constraint supports a lexical approach to denominals, but, building on Harley (2008), this wrongly commits us to lexical analyses of clearly syntactic structures. The PPs in (10) are syntactic entities, since they contain full DPs/NPs, but have canonical use interpretations. Specifically, (10a) entails that the workers went to (possibly different) banks as customers (an effect which disappears if the spatial preposition zu is used). Similarly, if (10b) has a weak definite interpretation in which the cleaners visited different pubs/doctors, then the cleaners must be interpreted as customers/patients of the pubs/doctors.

1412

VII. Semantics and pragmatics in word-formation II: Special cases

(10) a. Die Bauarbeiter gingen auf die Bank. The building workers went to the bank.

German

b. The cleaners both went to the local {pub/doctor}.

3. Structural questions 3.1. Are denominal verbs really denominal? Analyses of the structure of denominal verbs must decide whether they are denominal at all. In any given N-V pair, one must decide between the options in (11): (11) a. V is in no sense denominal, since N is derived from V. b. V is formed by a process specifically taking a noun as input. c. Non-directional analyses: Neither N nor V is derived from the other, and either i. N and V are both derived from category-neutral roots, or ii. A root is underspecified between N and V, and may thus be used in either noun-typical or verb-typical environments without any process which specifically turns it into a noun or verb (e.g., Farrell 2001). Option (c i) requires comment. Some linguists (e.g., Arad 2003, Borer 2014, Harley 2005) maintain that lexical roots do not intrinsically have categories like N or V, but obtain them by embedding in noun- or verb-specific contexts (e.g., her loveN vs. you will loveV it) or by merger with category-determining affixes (love-ableA). In such approaches derivatives like syllab-ifyV, syllab-icA need not be derived from the noun syllable by truncation but are derived from the root syllab- (as is syllab-leN). Neither element of the N-V pair syllable-syllabify is derived from the other. Similar remarks would hold for the denominals in (12). (12) colonize, fantasize, harmonize, notarize, prioritize, theorize, calcify, quantify, terrify Arguments for category-neutral roots extend beyond the mere desire to eliminate truncation from grammar. Borer (2014) describes several such arguments. We give one example. Many morphologically unmarked nominalizations in English do not support argument structure (the walk (*of the dog) vs. the walking of the dog). This correlation between lack of affixes and lack of arguments is explicable if we assume that (i) argument structure can only emerge when roots are embedded under functional heads that categorize roots as verbs, (ii) affixes like -ing can select projections of such functional heads, and thus allow argument realization, (iii) English has no zero affix that acts like -ing in this respect, (iv) walkN is a root that is nominalized by embedding under nominal functional heads like the. Thus, walkN was never a verb and thus never had the argument structure of walkV. Since converted denominal verbs involve no overt affixation, they also invite us to contemplate derivations involving category-neutral roots. While the argument from (8)

82. Denominal verbs

1413

that some denominal verbs are formed from category-neutral roots (Kiparsky 1997: 485− 491; Arad 2003; Don 2005) is not compelling (see section 2.2 and Borer 2014), this does not in itself refute the claim that at least some English denominal verbs are formed from category-neutral roots, and root-based derivations appear to be particularly plausible for Hebrew (Arad 2003). We now turn to the directional analyses in (11a, b), listing a selection of (alleged and genuine) arguments in their favour. The conclusion will be that at least some denominals are clearly from nouns, but that non-directional analyses like (11c) might be valid in other cases. A) Semantic evidence: intuitively personifyV is semantically more complex than personN. As in other denominals meaning roughly ‘to cause to become N’ (demonize/saint someone, cf. (6a)), personN names a concept which is part of the meaning of personifyV and can be conceptualized without reference to the verbal event-kind, but not vice-versa. These considerations (coupled with overt affixation on personify) exclude any derivation of personN from personify, so personify is derived either from personN or from a category-neutral root. B) Overt affixes on the input: Verbs containing noun-forming affixes constitute an apparently uncontroversial argument for a N → V analysis: leverageV, commissionV, referenceV, profiteerV; French règlement(er) ‘to regulate’ ← règlement ‘regulation’; clôture(r) ‘to close off ’ ← clôture ‘enclos.ure’, German gärtner(n) ‘to garden’ ← Gärtner ‘garden.er’. These French and German examples are given in their infinitive forms (the standard citation forms), with the infinitive affixes bracketed to distinguish them from the verb stems created by denominal verb formation rules. C) Regularization of inflection: In some N-V pairs V has regular (default) inflection and not the irregular inflection of a morphologically related verb. Thus, although slide is normally an irregular verb, cf. (13a), the use of slide in (13b) is regular. The regular use is, unlike that in (13a), intuitively “derived from” slideN as a productive instantiation of the well-attested location pattern in (1a). This is easily explained on a directional approach: slideN is derived from slideV in (13a) (it names an entity which is slid under a microscope). Denominal slideV cannot inherit information about irregular past tenses from the lexical entry of slideN, since nouns cannot inflect for tense. This, coupled with the low productivity of irregular inflection, means that slideV in (13b) can only receive default (regular) inflection. Variants of this account available in some theories include that the information about irregular inflection is too deeply embedded in the structure [V[N[V slide]]] to be accessible (cf. Pinker 2000: 184) or that slideV in (13b) is headed by an unpronounced affix, meaning that slide is not the head and thus unable to contribute information about inflection (cf. section 3.2). (13) a. I {slid/*slided} the sample under the microscope. b. I {slided/*slid} the sample. (‘put it on a slide’) Regularizations like (13b) are well-attested. With examples noted in Pinker (2000: 185), we can name they grandstanded ‘sought applause, as if in a grandstand’, he flied out ‘did a fly in baseball’ (he flew out would be used if speakers treat fly out as being directly derived from flyV or follow injunctions of untrained language critics, cf. Pinker 2000: 186 f.). German has regularly inflected beauftragen ‘to assign someone a task’, from Auftrag ‘task’, itself from irregularly inflected auftragen ‘to assign as a job’. Analogous

1414

VII. Semantics and pragmatics in word-formation II: Special cases

is haushalten ‘to keep house, budget’ ← Haushalt ‘household’ ← irregular halten ‘to hold’. The best cases of irregular inflection in denominal verbs I know are hamstrung (← hamstringN) and archaic clad (← clothe), shod (← shoeV). Here the events seem to be conceptually dependent on the entities named by the related nouns, excluding a V-N derivation (see point A above), so the verbs are either from nouns or category-neutral roots. Farrell (2001: 126−128) criticizes directionality-based accounts of inflectional regularization on the grounds that they cannot explain regularization phenomena such as computer mouses or baddest (superlative of bad in the sense ‘excellent’). For Farrell the explanation for regularization is always that semantic shifts have obscured the relation of the shifted lexeme to the source lexeme, so there would be no reason to apply the irregular inflection of the latter to the former. This is arguably supported by speakers who use highlighted but floodlit, since floodlightV is a hyponym of lightV while highlightV is not. Farrell’s account would need to be completed with an explanation for why irregular forms like underwent, understood, upset, overtook, withstood, babysat, midwives exist despite their at best tenuous semantic connections to the heads, and why there is no tendency towards regularization in any of the sense extensions of highly polysemous verbs like have, do, make. Does I took sick have a stronger connection to other senses of take than He flied out does to flyV? Future work would have to ask whether the irregular inflection in these cases is a memorized relic of stages where the structures had more compositional interpretations (e.g., midwives once meant ‘intermediate women’ and withstood ‘stood against’). D) Phonological facts can aid linguists in determining the direction of a derivation. We give three examples. First, Don (2005) observes that Dutch allows consonant clusters in syllable codas in nouns but not verbs, except denominal verbs (oogst- ‘to harvest’ ← oogst ‘harvest’). That denominals are exceptional in this regard follows if they are truly derived from nouns. Second, German has umlauted verbs like hämmern ‘to hammer’, häuten ‘to skin’ which are related to un-umlauted nouns (Hammer, Haut). (This is a relic of phonological conditioning by the Germanic -i-affix mentioned at the end of section 2.1.) Since some German morphological processes trigger umlaut (e.g., plural: Hämmer ‘hammers’) but no morphological rule de-umlauts its input, we cannot derive the un-umlauted nouns from umlauted verbs. Thus, either we derive the verbs from the nouns, or derive both the nouns and verbs from category-neutral roots. Third, English has dozens of N-V pairs where N has initial stress and V final stress, cf. remákeV vs. rémakeN, and analogous contrasts with the items in (14a−b). These stress-shift pairs do not involve N → V derivations, as the verbs allow irregular verbal inflection (remade, redid, rethought) and the prefix re- normally attaches to verbs. In triplets like (14c−e) the verbs on the right have the initial stress and idiosyncratic interpretation of the nouns, suggesting that these verbs are truly denominal (Kiparsky 1982: 13, Arad 2003: 759 f.). The relation between nouns and the finally stressed verbs like those in (14c−e) is less transparent semantically and the stress shift is unpredictable ((14f ) shows that stress shift probably cannot be predicted from independently needed principles like noun extrametricality). Kiparsky explained triplets like (14c−e) by assuming that V → N conversion is at level I and N → V conversion at level II (under the level ordering hypothesis, level I is prior to and more idiosyncratic than level II). Arad argued that the finally stressed verbs and the nouns in (14c−e) are both derived from

82. Denominal verbs

1415

category-neutral roots, while the initially stressed verbs are from nouns. The greater semantic and phonological idiosyncrasy of stress shifting derivations follows from the common assumption that processes operating directly on category-neutral roots are more prone to irregularity. (14) a. b. c. d.

conflict, produce, torment, present, transfer, invite, download, affix, permit retard, rethink, redo, rework, repaint digéstV → dígestN ‘summary’ → dígestV ‘to summarise’ discóuntV ‘to not count, to disregard’ → díscountN ‘rebate’ → díscountV ‘to sell at a rebate’ e. protéstV → prótestN ‘political demonstration’ → prótestV ‘to demonstrate’ f. mísprintN /mistrústN , díschargeN /disgústN , cónvertN /contrólN

Particularly from points B and D above, we can conclude that there are verbs which are unambiguously derived from nouns. However, this does not show that alternative derivations involving category-neutral roots or underspecification are not warranted for other N-V pairs, leaving us with much work to do in finding out which denominal verbs are truly denominal.

3.2. The structure of denominal verbs 3.2.1. Syntactic analyses We now describe the main structural approaches to denominals, starting with syntactic approaches. Many current syntactic approaches (e.g., Arad 2003; Hale and Keyser 1993, 2002; Harley 2005, 2008; Haugen 2009, Mateu 2001) form denominals by incorporation of nouns or category-neutral roots into syntactic heads which may be unpronounced. We present a simplified version of such analyses, based on (15). (See also article 22 on incorporation, section 4 for a more technical overview.) The underlined elements (either category-neutral roots or, as depicted here, bare nouns) are assumed to unite with the elements in bold type by head movement (see below) to form what we perceive as denominal verbs. (15) a. [VP they [V′ b. [VP they [V′ horses.) c. [VP they [V′ thief.) d. [VP they [V′

Vdo [N experiment]]] (VP in They experimented.) Vcause [VP horses [V′ Vbe [PP Pwith [N saddle]]]]]] (They saddled Vcause [VP the thief [V′ Vgo [PP Pin/to [N jail]]]]]] (They jailed the Vcause [VP him [V′ Vbecome [N outlaw]]]]] (They outlawed him.)

The bold-faced elements in (15) are unpronounced light verbs or prepositions which contribute closed-class elements of meaning corresponding to semantic primitives used in decompositional theories of semantics. They are marked with subscripts giving rough English glosses of their interpretations. Some such elements can appear overtly in other constructions. The silent P in (15c) is pronounced in imprison or put in jail. Vcause in

1416

VII. Semantics and pragmatics in word-formation II: Special cases

(15d) is arguably realized as an affix in colonize or demonize. Vdo in (15a) surfaces as a light verb in they did an experiment or as an affix -ier- in German experimentier‘experimentV’. Hale and Keyser (1993: 54 f.) appeal to the more widespread use in some languages of overt light verbs or affixes corresponding to the item labelled Vdo in (15a) in arguing that all unergative verbs like laugh, dance are underlyingly denominal. (This claim, if applied to irregularly inflected forms like I spoke/thought/spat, relativizes the claim reviewed in section 3.1.1 that denominals always inflect regularly. To uphold both claims, one would presumably have to argue that irregularly inflected unergatives like speak are derived by incorporation of category-neutral roots into Vdo , and that inflectional regularization occurs with verbs derived from nouns, but not (necessarily) with verbs derived from category-neutral roots.) In other cases explicit overt paraphrases are often either unacceptable (*they caused horses to be with saddles, cf. (15b)) or have the wrong interpretation (they caused the thief to go to jail expresses less direct causation than (15c)). This could have trivial causes such as the availability of more economical ways of expressing the same situations (including denominal verbs or single-verb structures like give the horse a saddle, put the thief in jail ), but may also suggest that the semantic analysis underlying the syntax needs adjustment. For instance, consensus is mounting that monoclausal causatives should not be decomposed with BECOME in addition to CAUSE (McIntyre 2010: 1254, and references), so one might replace (15d) with a structure closer to what is overtly visible in they made him an outlaw. In most syntactic approaches denominal verbs are formed by incorporation of the bare nominals or roots into the heads selecting them. Here incorporation takes the form of head movement, which can be defined as in (16). (16b) names two operations which have been employed for denominal conversion verbs (see Haugen 2009 for comparison). Strategy (16b i) is more appropriate for cases like those mentioned above in which the empty elements in (15) are pronounced. (16) Head movement is a situation where both (a) and (b) hold. a. The complement of a head X is (the projection of) another head Y. b. Either (i) Y adjoins to X to form a complex word [X° YX], or (ii) X has a defective phonological matrix and inherits Y’s phonological representation. (16) reflects the standard assumption that a head can only move to the head selecting it as a complement. This makes some empirical predictions which Hale and Keyser (1993: 60) saw as evidence that denominal verb formation involves syntax. The first prediction is that there is no incorporation of specifiers, which excludes incorporation of themesof-motion and of agents. An illustration of this might be that (15c) lacks paraphrases such as *The judge thiefed into jail (theme incorporation) or *It judged the thief into jail (agent incorporation). The ban on agent incorporation mirrors a crosslinguistic tendency against incorporating agents in noun incorporation constructions (see article 22 on incorporation). An open question here concerns data like (7). Perhaps modelV and motherV children derive from structures like be a model or be mother (of) children, and thus involve incorporation of (heads of) complements, which conforms to the Hale-Keyser approach, but this matter requires careful assessment. The ban on incorporating themes of motion is only justifiable if the incorporated Ns in they saddled the horses and other cases in (2) are not grammatically represented as moving to the direct objects, but merit

82. Denominal verbs

1417

analyses like (15b). Hale and Keyser assumed this analysis without providing independent motivation, but we saw in section 2.1 that it might be empirically superior to the motion analysis. A second consequence of (16a) is that jail in (15c) can only reach Vcause by moving through Pin and then Vgo. Longer movements would violate the widely accepted head movement constraint. Hale and Keyser (1993: 60) argued that this explains the unacceptability of cases with unincorporated prepositions like (17): to incorporate into V, bottle/ shelf would have to incorporate into in/on, which should then incorporate into V (this kind of derivation is observed in French embouteiller le vin ‘to in.bottle the wine’ or imprison the thief ). However, there are potential counterexamples like box/cage/fence them in, and German structures like (20c) below. If the incorporated Ns here truly start life as complements of the overt prepositions, then the Hale-Keyser argument is empirically faulty. If they do not, then there is a way of forming denominals other than head movement, which also undermines this argument. (17) *I bottled the wine in. *I shelved the books on. If any denominals are formed by head movement, it is clear that not all of them are. As Harley (2005: 60−65) and Rimell (2012: 25) note, head movement derivations are implausible for the very productive class of denominals naming instruments (hammerV and others in (5)), e.g., since instruments are otherwise adjuncts and not complements, and since complement positions are often occupied by other material (brush the coat, hammer the metal flat). Similar problems attend (less frequent) cases like she authored/ refereed a text and other (apparently) agent-naming denominals in (7). Head-moving the nominals out of the specifier positions where agents are introduced is undesirable for reasons seen above. Rimell (2012) notes other problems for head movement approaches, e.g., that English denominal verbs naming the patients or themes of non-motion events are far less productive than one would expect if incorporation of complements were a theoretical option. We do not find appleV ‘to eat an apple’ or novelV ‘to read a novel’, although such events involve canonical uses of apples/novels which are readily reconstructed in cases like I finished the apple/novel. Rimell argues that roots in denominal verbs are not incorporated out of argument positions but merge with null light verbs in a compounding-like operation (see also Hirschbühler and Labelle 2009). (18) illustrates this structural proposal (to save space Vcreate and Vgo replace the more sophisticated system of light verbs used by Rimell, which cannot be reviewed here). A leading idea here is that roots which normally name properties of entities are coerced into being event modifiers. (18) a. They [V° drill-Vcreate] a hole. b. They [V° author-Vcreate] a text. c. They [V° cycle-Vgo] to London.

1418

VII. Semantics and pragmatics in word-formation II: Special cases

3.2.2. Non-syntactic analyses There are numerous non-syntactic accounts which generate denominal verbs in a component of grammar often identified as the lexicon. Like most syntactic accounts, many non-syntactic accounts assume semantic decomposition approaches to verb meanings, but semantic primitives like GO or CAUSE are not present in syntax and indeed need not have any specific morphological realization. Such non-syntactic decompositional accounts have been applied to both affixed and affixless denominal verbs (Gottfurcht 2008; Kiparsky 1997; Lieber 2004; Plag 1999; Stiebels 1997; Wunderlich 1987). A simplified illustration of such approaches is (19) (a non-syntactic version of (15)). In each decomposition, y corresponds to the incorporated N, z to the subject and x to the direct object. (19) a. b. c. d.

(z, y) (z, CAUSE (z, CAUSE (z, DO

CAUSE

(they experimented/sermonized) (x, y)) (they motorized/named it) GO (x, TO (y))) (they hospitalized/jailed him) BECOME (x, y)) (they demonized/outlawed him) HAVE

Non-syntactic decomposition approaches differ regarding notation but also in more fundamental semantic assumptions. For instance, Plag (1999: 140) posits (a notational variant of) (19c) as a uniform semantics for all -ize verbs, except that the CAUSE component and its causer argument are absent in intransitives (it carbonized). Gottfurcht (2008) makes similar claims for all denominal verb formation processes in English. This localistic approach entails for instance that the incorporated N is a metaphorical goal in demonize someone and that the incorporated N in motorize a car realizes the x variable in (19c) (Plag and Gottfurcht do not discuss the possessive analysis in (19b) and section 2.1). This approach is less convincing in cases like (19a). Does theorizeV really mean ‘to cause theories to go somewhere’ or ‘to cause something to go to theories’, and if humans conceptualize theorizing in terms of such metaphors, what events could not be conceptualized in such terms? More promising is Lieber’s (2004: 86 f.) suggestion that these deviate from the core meaning of -ize (which for her is also roughly as in (19c)) due to extralinguistic pressure to coin names for such events. Other approaches like those of Kiparsky (1997), Stiebels (1997) and Wunderlich (1987) assume that the operations forming denominal verbs do not lead to unified semantic representations. A potential pitfall here is a proliferation of semantic representations with little in common, making one wonder why English -ize and conversion denominals have a largely similar range of uses. Perhaps these decompositions have in common that they instantiate particularly frequent skeletons of verbal meaning (“prototypical verbal concepts”, Wunderlich 1987: 314). The advantage is that the semantic representations are more precise and testable and court fewer potential overgeneration problems. Distinct semantic representation approaches also allow analysts to state generalizations on the semantic properties of the incorporated N. Kiparsky (1997) and Stiebels (1997: 273) suggest that only the least prominent entity in a semantic representation can incorporate (for instance only the goal in (19b) and only the possessed entity in (c)), replicating Hale and Keyer’s syntactic account of these effects. Proponents of unified representations would presumably appeal to pragmatic constraints like those in section 2.2 to explain these effects.

82. Denominal verbs

1419

Various assumptions about the structure of denominal verbs are available in nonsyntactic accounts. Denominal conversion verbs could be derived with the help of an unpronounced (zero) morpheme (Kiparsky 1982; Marchand 1969; Wunderlich 1987: 312−322), but the process of morphologically unmarked denominal verb formation might alternatively be seen as a lexical operation which has no formal correlate. Apart from one’s position on the cline between an item-and-arrangement view (all morphology involves concatenation of morphemes) and an item-and-process view (affixation has no privileged status and is one of several possible morphological processes), the choice hinges mainly on whether one sees substantial semantic or grammatical differences between overt affixation and conversion. Arguments against zero affixation include that the putative zero affix in (English) denominal verbs has a range of uses not found with overt affixes (Lieber 2004; Plag 1999), and positing multiple zero affixes incurs the cost of accidental homophony. Imaginable replies might invoke Gottfurcht’s (2008) arguments that zero and overtly affixed denominal verb formation devices have the same overall distribution of senses, and the fact that some morphologically unmarked processes are less productive than overtly affixed ones: zero-derived property-naming deadjectival nouns like goodA>N (a force for good) are marginal compared to overtly affixed ones like justice, fairness, scarcity. Moreover, the unmarked denominal verb formation mechanisms in current English have various diachronic sources, including French unmarked N/V pairs (armN/V, troubleN/V and others in Marchand 1969: 365) and formations that still had the overt affix -i in Old English (answarian ‘to answer’ ← answaru ‘answer’, endian ‘to end’ ← ende ‘end’). Zero affixation proponents might argue that any connection between lack of overt marking and extensive polysemy need not be encoded in synchronic grammar but is a reflex of the fact that zero exponence is a common direction of phonological change. These arguments are indirect. Direct evidence against objections to zero affixation would be an overt denominal-verb-forming affix in some language which functions as freely as English conversion. It would be interesting to know if such affixes exist. If zero affixes are rejected there are several ways of expressing a more direct connection between the semantic flexibility of denominal verb formation by conversion and the lack of formal marking. Lexicalists could express this idea using some variant of the underspecification hypothesis (recall (11c) above). An alternative defended by Lieber (most recently 2004: 91−93) is to view conversion as an unsystematic and semantically undetermined process of coining names for events by relabeling nouns as verbs. Relevant to these questions is the claim that overt markers that specifically form denominal verbs are crosslinguistically less frequent than affixes marking V → N derivations (see sources in Słodowicz 2011). If this observation is correct, how do we explain it? Does it show that verbs are formed with the mechanisms mentioned in the previous paragraph more often than nouns, and if so, why? Or does it have explanations that are compatible with zero affixation? For instance, do V → N derivations need more overt disambiguation than N → V ones? (A nominalizer with all the functions seen in the employment/employer/employee of John would perhaps be intolerably ambiguous, while ambiguities in denominals like dustV cause no problems for English speakers.) Or are there more diachronic sources for overt V → N markers than for N → V markers (excluding affixes used in other verbal functions like causative, see the last paragraph of section 2.1)?

1420

VII. Semantics and pragmatics in word-formation II: Special cases

3.3. Complex denominal verbs Special problems arise with complex denominal verbs (henceforth CDVs), i.e. denominals coexisting with preverbs (i.e. verb prefixes or particles). Many CDVs are unacceptable without preverbs. For instance they sexed up the theory ‘made it more appealing’ lacks a corresponding use of sexV without up. More cases of CDVs lacking (relevant) preverbless uses are seen in (20). (20) a. derail, behead, underline, outjockey, unhand, rejig, downsize, imprison, endanger, enthrone, enrage, entomb, enslave, encase, envision b. tone/scale down, size up, soldier on, pig out, tee off, clock in c. German aufbahr- ‘put on a stretcher; lit. on.stretcher’, anlein- ‘put on a leash’, eintüt- ‘put in a bag’, einsarg- ‘put in a coffin’, einkerker- ‘put in a dungeon’, einsack- ‘put in a sack’, einzäun- ‘fence in’ d. German entkalk- ‘decalcify’, versilber- ‘silver-plate’, erdolch- ‘stab with a dagger’ e. French emprisonn- ‘imprison’, embarque- ‘board a boat’, empoche- ‘pocket’ None of these verbs has an overt verbalizing suffix, but German has some denominal verbs which require a prefix and an overt suffix, consider benachrichtigen ‘to inform’ (← Nachricht ‘news’; cf. *nachrichten, *benachrichten, *nachrichtigen), and similarly with beherzigen ‘to take to heart‘, beerdigen ‘to bury‘. The closest English analogues are sporadic deadjectival verbs like embolden, enliven. The role of preverbs as apparent licensers of denominal verbs is more prominent in languages like German. While German unmarked noun-to-verb conversion is productive (cf. es müllerte ‘it Müller-ed, i.e. Thomas Müller scored a goal’), Kaliuščenko (1988: 112) notes that preverbs have become increasingly common with denominal verbs since Old High German times. My Germanophone students propose that the creative use of gnomeV mentioned at the start of this article should be translated with prefixed verbs like bezwergen, verzwergen (← Zwerg ‘gnome’). Some cases where preverbs occur with otherwise illicit denominal verbs may have trivial explanations like ambiguity avoidance (cf. the semantic contrasts unhand vs. handV, scale down vs. scaleV) and blocking (imprison blocks *prisonV, which is not intrinsically ill-formed, cf. jailV), but it is unclear if all cases allow such explanations. The ability of prefixes to co-occur with otherwise unacceptable denominal verbs like those in (20a, d, e) is sometimes explained by treating the prefixes as heads, making the complex verbs rare instances of left-headed structures (e.g., Williams 1981: 250; Lieber and Baayen 1993: 65−69). Williams suggested that English en- is independently the head because it can license otherwise illicit -ment-nominals (entrapment/*trapment; *(en)actment, *(en)listment). This argument predicts that the prefix in *(with)drawal is head of withdraw, although draw is its head judging by its irregular inflection (withdrew). Cases like *(en)listment can be explained without reference to headedness by assuming that -ment prefers Romance and/or morphologically complex input (most nouns with monosyllabic input like payment, placement, treatment, figment are synchronically memorized exceptions which entered the language as French borrowings, as can be verified by typing in *ment at www.onelook.com).

82. Denominal verbs

1421

Some preverbs in (20) are clearly not verbalizing heads. First, the particles in (20b, c) are not verbs. If anything they are prepositional (see article 23 on particle-verb formation), and they do not host verbal inflection: I pigged out/*pig outed. In German eingezäunt ‘in.ge.fence.d, i.e. fenced in’ the participial circumfix ge-…-t regards the nonparticle constituent as the head. Second, the phenomenon in (20) is not confined to preverbs. The denominal verb uses in (21) obligatorily occur with syntactic complements which one would not want to regard as verbalizing heads. Expletive it in (a), the PPs in (b, c) and the possessed NPs in (c, d) are not verbs. (21) a. stiff-upper-lip *(it) ‘adopt a stiff upper lip, i.e. not display emotion’ (C&C 768), leg *(it) ‘run’, lord *(it) over others ‘tell other people what to do’, cane *(it) ‘travel fast’ b. she padded *(down the stairs) ‘walked on the pads of her feet’ c. he wormed *(his way) *(out of responsibility) ‘tried to avoid it’ d. he craned *(his neck/his arm) ‘stretched it out, like a crane’ Third, particles like those in (20b) and prefixes like out-, re- only sporadically co-occur with items not otherwise used as verbs, and thus differ substantially from genuine verbalizers like -ize (idolize, Clintonize). This raises the question of how to handle preverbs which appear to combine systematically with nouns. Candidates include German/Dutch be- and ver-, but these also have various deverbal uses (see, e.g., Wunderlich 1987; Lieber and Baayen 1993; Stiebels 1997; Mateu 2001 for analyses). We will illustrate this problem with English/French en-/em-/im- in (20a, e), which more often combines with nouns than with verbs (enclose). It functions semantically like a bound variant of the preposition in/en than like a verb (imprison/emprisonner vs. in prison/en prison; enrage/enrager vs. in a rage/en rage; enslave vs. make into a slave). Productive German analogues like those in (20c) feature the prepositional particle ein- ‘in’ which is not a verbalizing head for reasons seen above. Ein- combines with verbs, but dozens of einverbs combine with otherwise illicit denominal verbs (this is rare with English in). On ein- see the articles in Olsen (1998). The paraphrases with in-PPs tempt one to treat the incorporated N as the ground (internal argument) of a prepositional in-relation expressed by en-. One could express this by treating en- as a bound P which merges (either in syntax or the lexicon) with a noun to form a constituent [P(P) en-N] which is obligatorily verbalized (with or without an empty V). Bound P+N constituents have a precedent in in-home entertainment, inpocket device, over-shoulder bag (cf. *it was in-home/in-pocket/over-shoulder). One question regarding such analyses is why [P(P) en-N] is only usable as part of a complex verb, and not, e.g., as a compound non-head in *the encage lions (‘the encaged lions’). Analyses of this type are also hard to defend for separable particle constructions like fence/box/cage them in and the German ein-verbs in (20c). An alternative is to assume that the incorporated N in encage or cage in is not grammatically represented as an argument of a preposition en- or in. Rather, cage forms a denominal verb by whatever mechanism one adopts for simplex denominal verbs like I caged the lion. We can still treat en- as prepositional, but it means ‘in(to) some (contextually identified) enclosure’, as it clearly does in deverbal cases like enclose or close in. In this analysis encage and cage in would mean ‘perform a cage-related act, causing the

1422

VII. Semantics and pragmatics in word-formation II: Special cases

object to go into an enclosure’, with the enclosure pragmatically identified as a cage (see Mateu 2001 for partly related proposals). The latter kind of analysis poses a paradox in any case where the preverb systematically combines with denominal verbs (as for instance with en-, which mostly co-occurs with roots otherwise used as nouns). Here linguistic theory must somehow accommodate complex operations, operations which obligatorily consist of two other independently attested operations (here: addition of prefix/particle plus conversion, however these operations are implemented theoretically). Precedents for this include Raffelsiefen’s (1992) composed functions, and others mentioned in article 29 on parasynthesis in Romance and article 33 on synthetic compounds in German. If, e.g., en-prefixation is an example of a complex operation in this sense, challenging questions arise about how it should be analysed: Is it, for instance, part of a lexically listed constructional template, or is it a bound preposition which merges with N and must incorporate into a light verb? This is yet another question which the present article cannot answer, but which deserves further attention.

4. References Arad, Maya 2003 Locality constraints on the interpretation of roots: The case of Hebrew denominal verbs. Natural Language and Linguistic Theory 21: 737−778. Aronoff, Mark 1980 Contextuals. Language 56: 744−758. Baeskow, Heike 2006 Reflections on noun-to-verb conversion in English. Zeitschrift für Sprachwissenschaft 25: 205−237. Borer, Hagit 2014 The category of roots. In: Artemis Alexiadou, Hagit Borer and Florian Schäfer (eds.), The Roots of Syntax, the Syntax of Roots, 91−121. Oxford: Oxford University Press. Clark, Eve and Herbert Clark 1979 When nouns surface as verbs. Language 55: 767−811. Don, Jan 2005 Roots, deverbal nouns and denominal verbs. In: Geert Booij, Emiliano Guevara, Angela Ralli and Sergio Scalise (eds.), Morphology and Linguistic Typology, 91−104. http:// morbo.lingue.unibo.it/mmm/proc-mmm4.php [last access 23 Jan 2015]. Dowd, Andrew 2010 More on instrumental denominal verbs. Snippets 21: 7−8. Farrell, Patrick 2001 Functional shift as category underspecification. English Language and Linguistics 5: 109−130. Gottfurcht, Carolyn 2008 Denominal verb formation in English. Ph.D. dissertation, Northwestern University. Hale, Kenneth and Samuel Jay Keyser 1993 On argument structure and the lexical expression of syntactic relations. In: Kenneth Hale and Samuel Jay Keyser (eds.), The View from Building 20, 53−109. Cambridge, MA: MIT Press. Hale, Kenneth and Samuel Jay Keyser 2002 Prolegomenon to a Theory of Argument Structure. Cambridge, MA: MIT Press.

82. Denominal verbs

1423

Harley, Heidi 2005 How do verbs get their names? Denominal verbs, manner incorporation, and the ontology of verb roots in English. In: Nomi Erteschik-Shir and Tova Rapoport (eds.), The Syntax of Aspect, 42−65. Oxford: Oxford University Press. Harley, Heidi 2008 Bare roots, conflation and the canonical use constraint. Handout. Talk presented at the NORMS Workshop on Argument Structure, University of Lund. Harley, Heidi and Jason D. Haugen 2007 Are there really two different classes of instrumental denominal verbs in English? Snippets 16: 9−10. Haugen, Jason 2009 Hyponymous objects and late insertion. Lingua 119: 242−262. Hirschbühler, Paul and Marie Labelle 2009 French locatum verbs and incorporation. Lingua 119: 263−279. Kaliuščenko, Vladimir 1988 Deutsche denominale Verben. Tübingen: Narr. Kaliuščenko, Vladimir 2000 Typologie denominaler Verben. Tübingen: Niemeyer. Kallulli, Dalina 2013 (Non-)canonical passives and reflexives. In: Artemis Alexiadou and Florian Schäfer (eds.), Non-canonical passives, 337−358. Amsterdam/Philadelphia: Benjamins. Karius, Ilse 1985 Die Ableitung der denominalen Verben mit Nullsuffix im Englischen. Tübingen: Niemeyer. Kiparsky, Paul 1982 Lexical morphology and phonology. In: Linguistic Society of Korea (ed.), Linguistics in the Morning Calm, 3−92. Seoul: Hanshin. Kiparsky, Paul 1997 Remarks on denominal verbs. In: Alex Alsina, Joan Bresnan and Peter Sells (eds.), Complex Predicates, 473−499. Stanford: CSLI. Leitner, Gerhard 1974 Denominale Verbalisierung im Englischen. Tübingen: Niemeyer. Lieber, Rochelle 2004 Morphology and Lexical Semantics. Cambridge: Cambridge University Press. Lieber, Rochelle and Harald Baayen 1993 Verbal prefixes in Dutch: A study in lexical conceptual structure. In: Geert E. Booij and Jaap van Marle (eds.), Yearbook of Morphology 1993, 51−78. Dordrecht: Kluwer. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. Mateu, Jaume 2001 Preverbs in complex denominal verbs. Catalan Working Papers in Linguistics 9: 37− 51. McIntyre, Andrew 2010 The BECOME=CAUSE hypothesis. Linguistics 50: 1251−1287. Miller, D. Gary 1993 Complex Verb Formation. Amsterdam/Philadelphia: Benjamins. Olsen, Susan (ed.) 1998 Semantische und konzeptuelle Aspekte der Partikelverbbildung mit ein-. Tübingen: Stauffenburg. Pinker, Steven 2000 Words and Rules. London: Phoenix.

1424

VII. Semantics and pragmatics in word-formation II: Special cases

Plag, Ingo 1999 Morphological Productivity. Berlin: Mouton de Gruyter. Rappaport Hovav, Malka and Beth Levin 2008 The English dative alternation. Journal of Linguistics 44: 129−167. Raffelsiefen, Renate 1992 A non-configurational approach to morphology. In: Mark Aronoff (ed.), Morphology Now. 133−162. Albany: State University of New York Press. Rimell, Laura 2012 Nominal roots as event predicates in English denominal conversion verbs. Ph.D. dissertation, New York University. Słodowicz, Szymon 2011 Defective denominal verbs in Polish. In: Piotr Bański, Beata Łukaszewicz, Monika Opalińska and Joanna Zaleska (eds.), Generative Investigations, 230−246. Cambridge: Cambridge Scholars Publishing. Son, Minjeong and Peter Cole 2008 An event-based account of -kan constructions in Standard Indonesian. Language 84: 120−160. Stiebels, Barbara 1997 Complex denominal verbs in German and the morphology-semantics interface. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1997, 265−302. Dordrecht: Kluwer. van Gelderen, Elly 2011 Valency changes in English. Journal of Historical Linguistics 1: 106−143. Volpe, Mark 2002 Locatum and location verbs in lexeme-morpheme base morphology. Lingua 112: 103− 119. Williams, Edwin 1981 On the notions “Lexically Related” and “Head of a Word”. Linguistic Inquiry 12: 245− 274. Wunderlich, Dieter 1987 An investigation of lexical composition: The case of German be-verbs. Linguistics 25: 283−331.

Andrew McIntyre, Berlin (Germany)

83. Valency-changing word-formation 1. 2. 3. 4. 5. 6. 7.

Introduction Transitivity alternations Argument hierarchy Valency-decreasing operations Valency-increasing operations Residual issues References

83. Valency-changing word-formation

1425

Abstract This article deals with valency change in general. Verbs denote states of affairs with a varying number of participants; when their valency is shifted, different participants are put into the foreground (realized as direct object or subject). Transitivity alternations represent the systematic shift between transitive and intransitive verbs. It is claimed that at each state of a verb there is a certain argument hierarchy that determines the grammatical realization of the arguments. A valency-changing operation either decreases or increases the number of arguments. Valency-decreasing includes several specialized operations such as passive, antipassive, reflexive, reciprocal and noun incorporation, as well as the more general middle or medio-passive. Valency-increasing includes causative and applicative. Valency-changing operations can be realized morphologically (where they contribute to word-formation) or periphrastically (e.g., by means of auxiliaries), or they are hidden and only visible by the actual number of complements. In terms of morphology, (i) affixes may be general or specialized intransitivizers/transitivizers, (ii) several cycles of valency-changing are possible, (iii) the order of affixation mirrors the semantic order of application, (iv) valency-changing precedes all modal, aspectual, temporal or personal specifications − it is therefore derivational. In addition, there are argument-alternating operations that do not change valency.

1. Introduction The notion of valency is used for categories that take a varying number of complements, such as verbs and verbal nouns. Semantically, valency refers to the number of arguments that make a proposition out of the verbal head, while in syntax valency refers to the number of complements that are necessary to form a full clause with this verb. Thus, a valency-changing operation is the same as an operation on argument structure (see Wunderlich 2012b). Arguments are distinct from adjuncts: if an argument is omitted, the utterance is felt to be incomplete, while an adjunct can be more easily omitted. Under certain contextual conditions, however, even arguments might be omitted without any interpretational effect, so that the valency doesn’t change. Sometimes it is useful to make a distinction between core arguments, realized by structural case (accusative, ergative, dative or nominative), and peripheric arguments, realized obliquely. There is no substantial difference between NP/DP and sentential complements concerning valency; both satisfy just one argument position. Infinitival complements, however, are different. Consider raising to object in (1), where the subject of the dependent proposition receives accusative from the matrix verb. Obviously, thought has three complements in (1b), while it has only two in (1a). Nevertheless, raising to object is not considered to be a valency-changing phenomenon; semantics is the determining factor. (1)

Raising to object a. I thought [(that) he was at home]. b. I thought [him] [(to be) at home].

1426

VII. Semantics and pragmatics in word-formation II: Special cases

“Valency change” presupposes that there is some valency to begin with. Not all linguists share this assumption. According to the radical Neo-Davidsonian approach (Parsons 1990; Schein 1993), syntax begins with a categorial constellation into which lexical roots are inserted. These roots, such as run, eat or give, have no arguments by themselves; however, when a root is inserted into V, it achieves an eventive argument role. By means of higher functional categories further argument roles can be introduced. Thus, give would become a three-place verbal entity by successive steps of valency-increasing, see (2a). (2)

Radical Neo-Davidsonian approach a. λe GIVE(x,y,z)(e) 5 λe [GIVE(e) & AGENT(e,x)] b. λe HAIL(e)

THEME(e,z)

&

RECIPIENT(e,y)

&

This approach intuitively captures the possibility of zero-place verbs quite well. An impersonal weather verb such as German hageln ‘to hail’ incorporates a root that is also used as a noun (historically it is related to a noun meaning ‘pebble’), with no further argument roles (2b). Now, surprisingly, even a zero-place verb can be used with further arguments, as shown in (3). (Note that hageln has an expletive subject es, so the syntactic valency is 1.) (3)

Weather verbs a. Es hagelte faustgroße Körner. ‘It hailed pebbles as big as your fist.’ No passive: *Faustgroße Körner wurden gehagelt. b. Es verhagelte den Weizen. it ver-hailed the wheat ‘The hailstorm ruined the wheat.’ Passive is possible: Der Weizen wurde (vom Sturm) verhagelt. c. Es hagelte uns faustgroße Körner aufs Dach. it hailed us fist-big pebbles on.the roof ‘It hailed pebbles as big as your fist on our roof.’ d. λz λy λx λe [HAIL(e) & PEBBLES(e,x) & BECOME LOC (x, ON(z)) & e. Wir kriegten faustdicke Körner aufs Dach gehagelt. we got fist-thick pebbles on.the roof hailed ‘We got pebbles as big as your fist hailed on the roof.’ f. Der Sturm/es behagelte unser Dach mit faustdicken Körnern. the storm/it be-hailed our roof with fist.thick pebbles Passive: Unser Dach wurde mit faustdicken Körnern behagelt.

POSS (y,z)]

(3a) shows an instance with a cognate object in the accusative, and (3b) an instance with the applicative prefix ver- (adding the result that something gets ruined). Interestingly, the applicative verb becomes truly transitive and can be passivized, differently from the verb with a cognate object. Since hailing is a weather-conditioned movement of little icy hailstones, the verb hail licences a directional PP as in (3c); in turn, the presence of a directional object licenses so-called possessor-raising in German, so that the dative possessor uns ‘us’ (a complement of the verb) is regularly interpreted as the possessor

83. Valency-changing word-formation

1427

of the roof in (3c) − thus, one arrives at the interpretation roughly given in (3d). The so-called kriegen-passive, then, can shift the dative possessor to a nominative subject, as shown in (3e). It is also possible to add the applicative prefix be-, shifting the goal to a direct object (which is known as locative alternation), as shown in (3f) − which in turn licenses passive. The examples in (3) illustrate in a nutshell that even a simple zero-place verb such as hageln ‘to hail’ can be the source of several valency-increasing and -decreasing operations. One also can see that some of these operations are morphologically marked, and others are not. Three of the valency-increasing operations observed in (3) are not marked morphologically in German: (i) the introduction of a cognate object, (ii) the introduction of a directional PP, and (iii) the introduction of a possessor-dative. In other words, these operations do not contribute to word-formation. Nevertheless, they allow the syntactic frame of a lexical item to be enriched. Most interestingly, the possessor-dative shown in (3c) (leading to the passive in (3e)) clearly increases the valency of the verb, which, however, is licensed only when the directional PP is added − this shows that syntax and lexicon can interact without any morphological mediation. Four other operations observed in (3) are marked morphologically: two applicatives (different prefixes) and two passives (the same participle formation, but different auxilaries): (i) the introduction of the result state ‘something becomes damaged’ by means of the prefix ver- (one of several alternatives of ver-, see Stiebels 1996: 302), (ii) the introduction of the goal ‘become located somewhere’ by means of the prefix be-, (iii) the kriegen-passive, and (iv) the werden-passive. This article clearly focusses on operations that are visible morphologically. If there were no languages that mark these operations, they would not come into consideration so easily; possibly one would just assume some flexibility (or ambiguity) at work. Interestingly, often only the existence of an operation is marked, leaving a certain amount of ambiguity for its outcome (see section 5). In the examples (3) above each applicative can be followed by the passive, so that the applied object becomes the subject of the clause. This indicates the functional advantage of having these operations sequentially applied: if a participant belongs to the type of events decribed by the verb but is usually not an argument of that verb, it can be made into an argument by means of the applicative operation, and ultimately can become the subject (or major topic) by means of passivization. As shown in (3), valency-increasing operations can even be based on zero-place verbal roots, which, however, are rare cross-linguistically. Much more often one finds verbal roots with one or two arguments required lexically. Every language seems to distinguish between at least two sorts of intransitive verbs (active and inactive ones, often also called “unergatives” vs. “unaccusatives”) on the one side, and transitive verbs, having both a more active and a more inactive participant (logical subject vs. logical object), on the other side. Many languages also have basic ditransitive verbs of the give or the put kind (give him the book, put the book on the desk). No language has any basic verbs with a valency more than 3, but many languages allow, by valency-increasing operations, up to 4 or even 5 arguments. Most valency change observed cross-linguistically concerns transitions between transitive and intransitive verbs in both directions (the so-called transitivity alternations); in other words, 2→1 and 1→2 changes are the most common ones. The same types of operations, however, can apply to verbs with one or even two arguments more, thus,

1428

VII. Semantics and pragmatics in word-formation II: Special cases

3→2 and 2→3 changes are possible as well. Totally unattested are 3→1 and 1→3 operations in a single step, which indicates that valency change is always a minimal change. Bigger changes must come about by iterated operations. The major valency-increasing operations are causative and applicative: causative adds a new subject, while applicative adds an object. Valency-decreasing operations include passive and antipassive: passive demotes the subject, while antipassive demotes the object (or one of the objects). These four operations are also included in the World Atlas of Language Structures (WALS 2011, chapter 107−111); the respective frequencies (computed on different but representative sets of languages) are shown in (4). As one can see, both applicative and passive occur in every second language, while antipassive is much rarer, and the presence of causative is a nearly universal feature of language. (4)

Frequency of visible valency changing operations in the languages of the world, according to WALS 2011 a. Passive is present in 43 % of languages (162 out of 373; Siewierska 2011) b. Antipassive is present in 25 % of languages (48 out of 194; Polinsky 2011a) c. Non-periphrastic causative is present in 92 % of languages (287 out of 310; Song 2011a, 2011b) d. Applicative is present in 45 % of languages (83 out of 183; Polinsky 2011b)

Indo-European (IE) originally realized both mediopassive and causative by suffixes (while antipassive and applicative were unmarked), whereas most of the modern IE languages developed periphrastic passives (auxiliary + participle) and causatives (a light verb ‘to make, let’ + infinitive). Some IE languages also developed applicative markers (prefixes, particles) from prepositions. A final remark concerns the theoretical model used in this article. Since one finds basic transitives in every language (and often also basic ditransitive verbs), the radical Neo-Davidsonian approach (illustrated in (2)) would not work. Whether a basic verb requires one, two or three arguments is always a lexical property. Moreover, Neo-Davidsonian representations have a flat structure; they do not show any ordering of arguments, including the order in which the arguments are added. Therefore, this approach is often accompanied with some sort of functional syntax, which has a more hierarchical structure, however, neither the syntax nor the semantics can explain why passive voice is a marked operation and active voice is not. I will use a moderate form of lexical decomposition, which makes the argument hierarchy and the operations that rely on it more transparent.

2. Transitivity alternations Let us for the time being assume that (prototypical) transitive verbs are asymmetric and could be described as {ACT(x) & AFF(y)}(e) ‘there is an event e in which x is active in some way & y is thereby affected in some way’. Then, all 2→1 and 1→2 changes have at least two alternatives. In case of valency-decreasing, either the active argument or the affected argument is demoted; the corresponding operations are called “passive”, “reflexive”‚ “anticausative” or “middle” if the active argument is demoted, and “antipas-

83. Valency-changing word-formation

1429

sive” or “noun incorporation” if the affected object is demoted. In case of valencyincreasing, either an active argument or an affected argument is added; these operations are called “causative” vs. “applicative”. Thus, valency-changing can form a fully symmetric structure built on the asymmetric conception of argument roles. Yucatec Maya is a language that demonstrates this idea most clearly because no verb can actually have more than two arguments (Krämer and Wunderlich 1999). Consider the following diagram (5). (5)

Transitivity alternation in Yucatec ACT(x) ei antipassive applicative

{ ACT(x)

&

AFF(y)

inherently imperfective

}

causative passive etc. ie AFF(y)

inherently perfective

There are mainly two classes of intransitive verbs in Yucatec; those that are inherently imperfective and are marked only for the perfect, and those that are inherently perfective and are marked only for the imperfect. In the imperfect, the argument is encoded by an ergative clitic attached to a preceding auxiliary, while in the perfect, the argument is encoded by a nominative (= absolutive) suffix; note that there is no 3rd person suffix. These two classes of intransitive verbs roughly correspond to active vs. non-active verbs, although some verbs to be arbitrarily classified. The transitivity alternations are illustrated in (6) and (7). Intransitivization is mostly done by stem alternation, while transitivization is done by suffixation. (INC = incompletive or habitual aspect.) (6)

Intransitivization in Yucatec a. k=in hek?-ik. INC=1sg break-IMPF ‘I am breaking it.’ b. k=u héek?-el. INC=3 break.MID-IMPF ‘It is breaking.’ c. hé?ek?. break.PASS ‘It has been broken.’

transitive

middle (inherently perfective)

passive (inherently perfective)

d. k=in hèek?. antipassive (inherently imperfective) INC=1sg break.ANTIPASS ‘I am breaking (something).’ (Bricker and Yah 1981: xi)

1430 (7)

VII. Semantics and pragmatics in word-formation II: Special cases Transitivization in Yucatec a. k=a luk?-ul. INC=2 leave-IMPF ‘You are leaving.’ b. k=a luk?-s-ik. INC=2 leave-CAUS-IMPF ‘You are removing it.’ c. k=a k?óoy. INC=2 dig ‘You are digging.’

intransitive (inherently perfective)

causative (transitive)

intransitive (inherently imperfective)

applicative (transitive) d. k=a k?óoy-t-ik. INC=2 dig-APPL-IMPF ‘You are digging it (up).’ (Straight 1976: 193) Causative and applicative can only apply on actually intransitive verbs. In order to express three-place concepts such as ‘teach’ or ‘send’, it is necessary first to intransitivize the root verb. In (8a), the transitive root is passivized before the causative is applied (and the result again passivized); in (8b), an object is incorporated before the applicative. Again, note the symmetry in this system: the active argument is demoted before a new active argument is introduced, and the affected argument is demoted before a new affected argument is introduced. (8)

Three-place concepts in Yucatec a. k=u ká?an-s- á?al. transitive INC=3 learn.PASS-CAUS-PASS.IMPF ‘It is being taught.’ (Bricker 1978: 22) b. k=u kon-lol-t-ik-etʃ. transitive INC=3 sell-flower-APPL-IMPF-2 ‘He is selling you flowers.’ (Krämer and Wunderlich 1999: 467)

These examples clearly show that multiple operations on argument structure are possible, and thus overcome the restriction to maximally 2 arguments. There is no reason for assuming that Yucatec is somehow handicapped in expressivity.

3. Argument hierarchy In the beginning of the preceding section I said that the asymmetric nature of transitive verbs might be captured by {ACT(x) & AFF(y)}(e). The two components of the event cannot be independent of each other because then, they wouldn’t form a single event, neither can AFF(y) temporally precede ACT(x) because then, there would be an external

83. Valency-changing word-formation

1431

source for y’s affectedness. In the minimal event combining the two components it is the activity of x that accounts for y being affected. Hence, the conjunction “&” must be understood as asymmetric: AFF(y) is determined in some way (e.g., caused) by ACT(x). However, the actual relationship does not need to be expressed explicitly (see Wunderlich 2012a). I take it for granted that for each individual verb (be it basic or derived) the arguments form a hierarchy. In the example considered here, the actor (agent or logical subject) ranks higher than the affected object, which in turn determines which argument is realized by ergative or accusative: the higher argument is realized by ergative (if available), and the lower argument is realized by accusative (if available). The grammatical realization of the arguments of a verb thus depends on argument hierarchy. Even if one does not have clear semantic arguments to determine which argument ranks higher (as in experiencer verbs such as see, like, fear, dimensional verbs such as limit, surround, surpass, etc.) − nevertheless, the verbs are clearly asymmetric, and no dispute arises as to which argument could possibly be ergative or accusative. Usually one simply writes SEE(x,y), LIKE(x,y), SURROUND(x,y), etc., indicating the argument hierarchy by the linear ordering (x,y), with x > y. The argument hierarchy in (9b) corresponds to the structure in (9a): the more an argument is structurally embedded the lower it is ranked. From the open proposition in (9a), the corresponding predicate is formed by stepwise λ-abstraction, first concerning the highest argument, and so on recursively (9c). If the complex predicate is applied to a specific argument expression, the lowest argument (corresponding to the leftmost λabstractor) is captured first, and so on recursively (9d, e). That is to say, λ-abstraction takes stepwise a negative copy, which is then unwound in the reverse order to saturate the open proposition. (9)

The computation of a transitive verb: semantic composition a. {ACT(x) (& AFF(y))}(e) b. Argument hierarchy: e > x > y c. Lambda abstraction: λy λx λe {ACT(x) & AFF(y)}(e) d. Lambda application, first step: λy λx λe {ACT(x) & AFF(y)}(e)(the_apple) = λx λe {ACT(x) & AFF(the_apple)}(e) e. Lambda application, second step: λx λe {ACT(x) & AFF(the_apple)}(e)(peter) = λe {ACT(peter) & AFF(the_apple)}(e) f. Aspect and tense specify, and mood binds the event variable e.

Following a proposal by Kiparsky (1992), argument hierarchy and the structural cases should be encoded by the same system of relational features. Lexical decompositional grammar (Wunderlich 1997a, 1997b, 2000, 2006) uses the features +hr = ‘there is a higher argument role’ (= ‘not the highest role’), and +lr = ‘there is a lower argument role’ (= ‘not the lowest role’); for reasons of markedness, these features differ slightly from those proposed by Kiparsky. Consequently, the highest argument is −hr, and the lowest argument is −lr, while any medial argument would be +hr,+lr. The structural cases are specified as follows: dative = [+hr,+lr] is compatible with the medial argument, accusative = [+hr] with a non-highest, ergative = [+lr] with a non-lowest, and nominative = [ ] (the

1432

VII. Semantics and pragmatics in word-formation II: Special cases

unspecified case) with any argument. For the transitive verb shown in (10a) four different case patterns are possible, while all other combinations are ruled out (10c). Which of these patterns is selected depends on further conditions. Note that in a split language like Hindi all four patterns are possible, depending on aspect, animacy and definiteness (Wunderlich 2006: 101). (10) The computation of a transitive verb: association with structural case a. λy λx λe {ACT(x) & AFF(y)}(e) −lr +lr +hr −hr b.

= [+hr,+lr] = [+hr] ERG = [+lr] NOM = [ ] DAT

ACC

c. Possible case patterns for {x, y}: {ERG, ACC}, {ERG, NOM}, {NOM, ACC}, {NOM, NOM}. Selection according to the constraint ranking of the particular language Similarly, a ditransitive verb such as zeigen ‘to show’ has three arguments (Ich zeigte ihm den Orion ‘I showed him the Orion’); here the ranking x > y > z is best realized by the case pattern {NOM, DAT, ACC} in German. Lexical marking is an option that overrides the default features in (10). For instance, a dative-subject verb such as German gefallen or Icelandic likar (both ‘to like’) assigns [+hr] to the highest argument. Valency-changing operations operate on a verb stem; their effects can easily be described in terms of argument hierarchy. Passive binds the highest argument (−hr) existentially, while antipassive binds the lowest argument (−lr) existentially, and the reflexive identifies a lower argument with the highest one. Causative introduces a new highest argument (−hr), so that the former subject automatically becomes [+hr], while applicatives introduce a non-highest argument (+hr).

4. Valency-decreasing operations 4.1. Passive and antipassive Valency-decreasing operations apply directly on the sequence of λ-abstractors. (11) shows the passive, which binds the highest argument existentially, so that it remains unexpressed. (The event argument is irrelevant here and therefore ignored.) (11)

PASS

[... λx VERB(x, ... )] = ... dx −hr

VERB(x,

...)

Some languages only allow passivization of transitive verbs, while other languages also include some subclasses of intransitive verbs. The class of verbs that can be passivized

83. Valency-changing word-formation

1433

is often restricted to agentive verbs, but certain non-agentive verbs can be included as well (e.g., The garden is surrounded by a fence). Existential binding causes the passivized n-place verb to be realized with at most n-1 morphosyntactic complements; thus, a transitive verb is detransitivized, and an intransitive verb becomes impersonal. As a consequence, another argument is realized by nominative (the default case) and thus becomes morphosyntactic subject. (12) and (13) show that in the passive of ditransitive verbs there are different options regarding the choice of object that becomes nominative. In the two languages illustrated here (the Uto-Aztecan language Yaqui, and Georgian), both objects are marked by accusative in the active. If the recipient shifts to nominative in the passive (as in Yaqui) it is said to be the primary object, while if the theme shifts to nominative (as in Georgian) it is said to be the direct object (Dryer 1986). (12) Double accusative and passive in Yaqui a. Joan Peo-ta ʔuka vaci-ta miika-k. Juan Pedro-ACC DET.ACC corn-ACC give-PERF ‘Juan gave Pedro the corn.’ b. Peo ʔuka vaci-ta miik-wa-k. Pedro DET.ACC corn-ACC give-PASS-PERF ‘Pedro was given the corn.’ c. *Uʔu vaci Peo-ta miik-wa-k. DET.NOM corn Pedro-ACC give-PASS-PERF ‘The corn was given to Pedro.’ (Van Valin 2007) (13) Double accusative and passive in the present series of Georgian a. Ketino Eka-s xalitša-s s-čukni-s. Ketino Eka-ACC carpet-ACC 3D-present-PRES.3N ‘Ketino presents Eka with a carpet.’ b. xalitša e-čuk-eb-a Eka-s. carpet PASS-present-TH-PRES.3N Eka-ACC ‘The carpet is presented to Eka.’ c. *Eka e-čuk-eb-a xalitša-s. Eka PASS-present-TH-PRES.3N carpet-ACC ‘Eka is presented with a carpet.’ (Joppen-Hellwig 2001: 50, 130) Thus, passivization not only is a test for subjecthood (Which argument is demoted in the passive?) but also for objecthood, differentiating between two types of double objects: primary vs. secondary object on the one hand, and direct vs. indirect object on the other (Dryer 1986; Wunderlich 2006: 136). Languages with symmetric objects (Bresnan and Moshi 1990) allow both alternatives: either the recipient or the theme becomes the syntactic subject in the passive. These differences clearly indicate that the promotion to nominative is not part of the passive operation, but a subsequent effect dependent on typological factors.

1434

VII. Semantics and pragmatics in word-formation II: Special cases

Passive is an operation even found in languages where it is not overtly marked (such as Basque). It reflects a ubiquitous salience shift. If an argument other than the subject is the actual topic, definite/specific, or a speech act participant, it might be more salient than the subject; in that case, passive can shift it to a higher argument position, thus making its high salience visible. (Aissen 1999 discusses subject choice in the framework of optimality theory.) Passive can be marked by a verbal affix, a stem alternation, a particle, or a periphrastic construction with an auxiliary and a participial form; it can also be made visible only by a shift in the complement pattern. Some languages allow the so-called personal passive, in which the highest argument of a passivized verb is expressed by an oblique instrument, source, or agent phrase; such a phrase is best seen as an adjunct whose free argument is coindexed with the existentially bound argument. Assume that the sentence John was kissed by Ann is represented as de {dx KISS(x,John)(e) & AGENT(e,Ann)}, then x=Ann is a contextual default for the value of x. Various theories of the passive have been proposed, among them the voice hypothesis (Kratzer 1996), claiming that verbs have a basic form without agent, and only if they are integrated into a voice phrase, is an agent either added (in the active voice) or not (in the passive voice). This hypothesis suggests that active voice is the more marked variant of a transitive verb, which, however, is only rarely observed cross-linguistically (e.g., in languages of the Austronesian family). Moreover, it does not describe the semantic effect of passive as existential binding but rather as the absence of an argument. In view of examples such as those in (14), Keenan (1980) and Dowty (1982) have argued that English passive must operate on transitive verb phrases rather than verbs, e.g., on the VP think cancer to be unlikely to be caused by hot dogs in (14a). However, already Bresnan (1982) showed that a lexical rule of passive is able to handle these more complex instances, too. (14) Passive of raising and control predicates a. Cancer is now thought to be unlikely to be caused by hot dogs. b. Intruders are now forced to be prepared to be attacked by dogs. (Bresnan 1982: 65) Let us assume that (15a) represents the passive of think and (15b) the embedded complex (itself being passivized), then (15c) results through functional composition as the approximate representation of (14a). This clearly shows that the most internal argument is shifted to the subject of the whole complex by means of two passive operations. (Note that in a polysynthetic language such as Greenlandic all the higher predicates are affixes, thus the operations are necessarily word-internal rather than affecting VPs.) (15) Analysis of raising + passive a. λp dx THINK(x,p) b. λz dy UNLIKELY(CAUSE(y,z)) c. λz dx THINK(x, dy UNLIKELY(CAUSE(y,z)))

raising

Similarly, with the three pieces in (16a) one gets (16b) in the first step, and (16c) in the second step, representing (14b). Here, the most internal argument is stepwise identified with the subject of be prepared and the object of force; again, two passives are involved.

83. Valency-changing word-formation (16) Analysis of control + passive a. λz du ATTACK(u,z) λP λy PREPARED(y,P(y)) λQ λx dv FORCE(v, x, Q(x)) b. λy PREPARED(y, du ATTACK(u,y)) c. λx dv FORCE(v, x, PREPARED(x, du

1435

subject control object control ATTACK(u,x)))

Antipassive is the mirror-image of the passive; it binds the lowest (rather than the highest) argument existentially, as shown in (17). (17)

ANTIPASS

[λz ... VERB( ... , z )] = dz ... VERB(... , z) −lr

While the passive is triggered by a particularly high salience status of the lower argument, the antipassive is triggered by a particularly low status of that argument. It is therefore expected to be a less universal operation than passive. Whereas a canonical NOM-ACC verb turns into 0̸-NOM by passivization, a canonical ERG-NOM verb turns into NOM-0̸ by antipassivization. However, similarly to an agent phrase in the passive, the object can be expressed obliquely. Note the aspectual and definiteness shift in the Inuit antipassive. (18) Antipassive in South Baffin Inuit a. Anguti-up nanuq quqir-jaa. man-ERG polar.bear.NOM shoot-PART.3sg>3sg ‘The man shot the polar bear.’ b. Anguti quqir-si-juq nanu-mik. man.NOM shoot-ANTIP-PART.3sg polar.bear-OBL ‘The man is shooting/shot a polar bear.’ (Spreng 2006) Antipassive also makes an argument accessible to relevant grammatical processes. In the ergative language Chukchi, spoken in Northeast Siberia, relativization is only possible with nominative arguments. If a transitive subject is involved, the verb must be antipassivized in order to make that argument accessible to relativization, consider (19d, b), whereas (19c), based on the active verb, is ungrammatical. (19) Antipassive and relativization in Chukchi a. tumg-e ŋinqey rəyegtetew-nin. friend-ERG boy.NOM save-AOR.3sg.3sg ‘The friend saved the boy.’ b. tumgətum ŋinqey-ək ine-nyegtele-gʔi. friend.NOM boy-LOC ANTIP-save-AOR.3sg ‘The friend saved the boy.’ c. *[ŋinqey rəyagtala-lʔ-ən] tumgətum boy.NOM save-PTCP-NOM friend (‘the friend that saved the boy’)

antipassive

1436

VII. Semantics and pragmatics in word-formation II: Special cases d. [ŋinqey-ək ine-nyegtelewə-lʔ-ən] tumgətum antipassive boy-LOC ANTIP-save-PTCP-NOM friend ‘the friend that saved the boy’ (Polinsky 1994, 2011a)

Although antipassive is found particularly often in ergative languages, it is not restricted to this type of language, just in the same way as passive is not restricted to accusative languages. A language can show both passive and antipassive by means of marked morphemes; an example is given in (20) (from Zoque, spoken in Chiapas, Eastern Central Mexico). One often finds the combination of causative with a preceding antipassive, as in (20c), a kind of mirror-image of the combination of applicative and passive, seen in (3f, g) in the introduction. (20) Passive, antipassive, and causative+antipassive in Zoque a. huʔc-ʔəm-wə bi wakaš. stab-PASS-COMPL DEF cow ‘(They) killed the cow.’ b. behča cəm-ʔoy-pa. horse carry-ANTIP-INCOMPL ‘The horses will carry (it).’ c. miš-yak-keš-ʔoy-wə-ʔam dey. 2>1-CAUS-eat-ANTIP-COMPL-NOW now ‘Now you have already fed me.’ (Johnson 2000: 144, 150, 136)

4.2. Recent developments in the passive of North Germanic languages In this section, three interesting and still-ongoing innovations regarding passive are briefly discussed: (i) the so-called “new passive” of Icelandic, in which the demotion of the subject is not accompanied by any change of case, (ii) the reinvention of a pure morphological passive in mainland Scandinavian, particularly in Swedish, and (iii) the complex passive in Norwegian and Danish, in which passive simultaneously affects a matrix control verb and its dependent verb. These three developments have nothing in common, but together, they nicely illustrate the possible variation within passive constructions.

4.2.1. Emergence of a new passive in Icelandic Icelandic exhibits a canonical passive as well as several sorts of impersonal passive. In the canonical passive, the object of the verb becomes the subject. Accusative objects are shifted to nominative, and the passive participle agrees with the subject (21a). Objects that are lexically marked for dative or genitive retain their case; here, the participle takes the neuter singular form (21b). The same distribution of case and agreement is found if

83. Valency-changing word-formation

1437

the passive clause begins with an expletive það as in (21c); in this case, however, a definiteness effect takes place, that is, the non-initial subject must be indefinite. (21) Canonical passive a. maður-inn var barinn af þjófi-num. man.NOM.msg-DEF was beaten.NOM.msg by thief-DEF ‘The man was beaten by the thief.’ b. Honum var hrint. he.DAT was pushed.NEUT ‘He was pushed.’ c. Það var strákur barinn. / barinn strákur. it was boy.NOM.msg beaten.NOM.msg / beaten boy ‘There was a boy beaten.’ The impersonal passive always needs an expletive það; it is possible with intransitive verbs (22a), with inherent reflexives (22b), and with various kinds of non-inherent reflexives. This construction did not emerge before the mid-19th century (Árnadóttir, Eythórsson and Sigurðsson 2011). (22) Impersonal passive a. Það var dansað alla nóttina. it was danced all night ‘There was dancing all night.’

intransitive verb

b. Það var drifið sig á ball. it was hurried REFL.ACC to dance ‘There was hurrying off to the dance.’

itr. reflexive verb

c. Það var farið heim til sín. it was gone home to REFL.GEN ‘There was going home to oneself.’

itr. verb + reflexive adverbial

The latest innovation is the new passive, beginning in the mid-20th century. It always uses the expletive það, but differs from the canonical or the impersonal passive in at least three properties: (i) Not only the lexically marked dative and genitive objects, but also the regular accusative objects retain their case; (ii) It is not possible to add an agentive by-phrase; (iii) There is no definiteness effect, that is, definite and indefinite objects are likewise possible. (23) New passive a. Það var barið strák / strák-inn. it was beaten.NEUT.sg boy.ACC / boy.ACC-DEF ‘A boy/the boy was beaten.’ b. Það var skammað mig. it was scolded me.ACC ‘There was scolding me.’

1438

VII. Semantics and pragmatics in word-formation II: Special cases c. Það var fengið sér öllara. it was got REFL.DAT beer.ACC ‘There was having oneself a beer.’

Note that because of case syncretism ambiguities between the canonical and the new passive can arise. In minimally contrasting examples like the following ones it is the definite dative phrase that shows that (24b) is an instance of the new passive. (24) Minimal contrast a. Það var hrint stelpu. it was pushed girl.DAT ‘A girl was pushed.’

canonical passive

b. Það var hrint stelpu-nni. it was pushed girl.DAT-DEF ‘The girl was pushed.’

new passive

Several proposals have been made about the nature of the new passive and the constructions that served as a model for the change. Maling and Sigurjónsdóttir (2002) assume that the new construction, although it uses passive morphology, is nevertheless an active impersonal with a null thematic subject (pro), which developed by extending reflexive impersonal passives to transitive verbs. On the other hand, Eythórsson (2008) and Gísladóttir (2007) argue that the new construction is a “true passive”, modeled after those passives that retain lexical case. In any case, it should be clear that some sort of þaðpassive has been extended. Cross-linguistically, both options are attested. The deponent verbs of Latin show that verbs with passive morphology can become active (loqu-or [1sg.PASS] ‘I speak’). On the other hand, Ukrainian impersonal passive is an example where accusative case on the object is preserved (Sobin 1985; Lavine 2005). (Note that Ukrainian -no/-to, which differs from the neuter singular -ne/-te, is really non-agreeing.) (25) Ukrainian impersonal passive of transitive verbs a. Kulju rozirvano (cvjaxom). balloon.ACC pierced[−AGR] (nail.INSTR) ‘The balloon was pierced (by a nail).’ b. Vazu bulo rozbyto vitrom. vase.ACC was broken[−AGR] wind.INSTR ‘The vase was broken by the wind.’ (Lavine 2010) Under what conditions impersonal passives of transitive verbs in languages such as Ukrainian and contemporary Icelandic are preferred to personal passives has still to be explored.

83. Valency-changing word-formation

1439

4.2.2. Emergence of a dominant morphological passive in Swedish The mainland Scandinavian languages (Norwegian, Danish, Swedish) have two productive ways of forming the passive, a morphological way using the suffix -s, and a periphrastic way using the auxiliary bli ‘to become’ together with a passive (perfect) participle. (The ordering verb-PAST-PASS in Swedish is unexpected because passive, a valency operation, usually precedes inflection; on the other hand, since -s is an innovation resulting from a former reflexive pronoun, it is expected to be word-final.) Morphological and periphrastic passives in Norwegian, Danish and Swedish utover. (26) a. Dør-en åpne-s door-DEF open-PASS outward lit. ‘The door is opened outward.’

Norwegian

utover. b. *Dør-en blir åpne-t door-DEF is open-PART outward ikke mere dansk i Skåne. (27) a. Der tale-s there speak-PASS no longer Danish in Scania ‘Danish is no longer spoken in Scania.’

Danish

dansk i Skåne. b. Der bliver (ofte) tal-t there is (often) speak-PART Danish in Scania ‘Danish is (often) spoken in Scania.’ franska vid hov-ed. (28) På den tiden tala-de-s at that time speak-PAST-PASS French at court-DEF ‘At that time French was spoken at the court.’ (Engdahl 1999)

Swedish

In Norwegian and Danish, the s-passive is restricted to the uninflected verb form (infinitive or present), while Swedish has extended the use of s-passive to all contexts during the last century. The overall frequency of the s-passive is about 50 % in Norwegian and Danish, but more than 90 % in Swedish. Engdahl (2006) argues that the periphrastic passive in Swedish is now the marked form, which requires that the subject has some control over a change of state.

4.2.3. The complex passive in Norwegian and Danish In the double or complex passive shown in (29a), an object of the lower verb is raised to the subject of the higher verb. The matrix verb is a control verb requiring an infinitive complement; so if it is passivized, the dependent subject is targeted, too. The double passive seems to preserve the control relation between the two verbs. The construction is recursive, as shown in (29b, c).

1440

VII. Semantics and pragmatics in word-formation II: Special cases

(29) Complex passive in Norwegian skut-t. a. Jon ble forsøk-t Jon was attempt-P.PART shoot-P.PART ‘It was made an attempt to shoot Jon.’ forsøk-t skut-t. b. Jon ble love-t Jon was promised attempted shot ‘It was promised that there would be an attempt to shoot Jon.’ c. Jon ble påståt-t love-t forsøk-t skut-t. Jon was claimed promised attempted shot ‘It was claimed that it was promised that there would be an attempt to shoot Jon.’ (Hellan 2001) The initial verb can be in the bli passive or s-passive, all later verbs, however, must have the participial form with -t. Interestingly, the final verb also can be an active participle, undistinguishable from the passive participle (30a). Moreover, the final active participle also can go together with an auxiliary, as in (30b). (30) Complex passive based on active participles a. Jon påstå-s frykte-t omkomme-t. Jon claim-PASS fear-P.PART perish-A.PART ‘It is claimed that it is feared that Jon has perished.’ b. Jon påstå-s frykte-t å ha lest bok-en. Jon claim-PASS feared to have read book-DEF ‘It is claimed that it is feared that Jon has read the book.’ (Hellan 2001) In every case, the complex passive construction has a common argument that raises successively from the lowest (the last) verb into the matrix clause. Notice that the raised argument expression can also end as the object (rather than subject) of the matrix verb participle (31). Moreover, the participle phrase can be used as an attribute of a noun, where the raised argument is coreferenced with the head noun. (31) Complex passive in a subject-to object raising context Jeg har ofte sett ham forsøk-t introduser-t for direktøren. I have often seen him attempt-P.PART introduce-P.PART for the director ‘I have often seen attempts at introducing him for the director.’ (32) Complex passive in an attribute phrase en ofte forsøk-t skut-t bjørn an often attempted shot bear Danish has the complex passive construction, too (with slightly different restrictions that cannot be discussed here).

83. Valency-changing word-formation

1441

(33) Complex passive in Danish a. Bil-en forsøge-s reparere-t. car-DEF try-PASS repair-P.PART b. Bil-en blev forsøg-t reparere-t. car-DEF was tried repaired c. Bil-en blev love-t forsøg-t reparere-t. the.car was promised tried repaired ‘A promise was made to try to repair the car.’ (Ørsnes 2006) The following example shows that the complex passive is also possible with subject-toobject raising verbs. (34) Double participles with subject-to-object raising verbs in Danish a. Man forventer forslag-et vedtage-t. you expect proposal-DEF adopt-P.PART ‘You expect the proposal to be adopted.’ b. Forslag-et forvente-s vedtage-t. proposal-DEF expect-PASS adopt-P.PART lit. ‘The proposal is expected to be adopted.’ (Ørsnes 2006) The complex passive construction can be compared with the remote passive found with German control verbs, as in (35). (35) Remote passive in German weil der Wagen oft zu reparieren versuch-t wurde because the.NOM car often to repair try-P.PART was ‘because many attempts were made to repair the car’ (Müller 2003) Here, the matrix verb is passivized, and simultaneously, the object of the embedded verb is advanced to subject, as if this verb were passivized. Müller (2003) argues that this effect is the consequence of complex predicate formation, consider (36b), a reformulation of Müller’s HPSG-analysis. Assume that (36a) represents the lexical entry for a subjectcontrol verb, and (36b) represents the complex verb formed with REPAIR, whereby the embedded object becomes the object of the complex verb. The passive morphology only attaches to the matrix verb, and the object shifts to the subject of the whole complex. Note that coindexation of controller and controllee in the lexical entry guarantees that both subjects simultaneously are bound existentially. (36) (Remote) passive of a complex verb in German a. λV λx TRY(x,V(x)) subject-control b. λy λx TRY(x, REPAIR(x,y)) complex verb formation c. PASS [λy λx TRY(x, REPAIR(x,y))] = passive λy dx TRY(x, REPAIR(x,y))

1442

VII. Semantics and pragmatics in word-formation II: Special cases

This account is not possible for the complex passive in the Scandinavian languages because it gives no motivation for the distribution of the passive morphology to both verbs. Let us instead assume that the control condition is separately stated, which blocks the formation of a complex verb by a simple application, see (37a). In the next step, the control verb is applied to REPAIR, still with an additional identity condition for the two subjects (37b). Then, if the passive is applied, it must be distributive; otherwise, the identity condition might be violated. (Intuitively, z=x requires that everything that happens to x should also happen to z.) In a medial step, the embedded object is allowed to raise into the matrix. The final result, shown in (37c), is identical with what one gets in the German case. (37) Distributed passive for a subject-control verb in Norwegian and Danish a. λV λx TRY(x, λzV(z)) & z=x subject control composition b. {λx TRY(x, λy λz REPAIR(z,y)) & z=x} passive c. PASS {λx TRY(x, λy λz REPAIR(z,y)) & z=x} = PASS {λx TRY(x, pass λy λz REPAIR(z,y)}) & z=x} distribution of passive application of passive {dx TRY(x, λy dz REPAIR(z,y)) & z=x} = raising {λy dx TRY(x, dz REPAIR(z,y)) & z=x} = λy dx TRY(x, REPAIR(x,y)) The complex passive is also possible with object-control verbs such as ‘to ask for’, ‘to order’, ‘to forbid’. (38) Complex passive with an object-control verb in Danish Den gule stjer-ne blev af nazister-ne påbud-t båre-t af alle the yellow star-DEF was by nazis-DEF order-P.PART bear-P.PART by all jøder. jews lit. ‘The yellow star was ordered by the nazis to be born by all jews.’ (Ørsnes 2006: 32) If one represents these verbs as in (39a) and distributes the passive on both verbs, one finally gets the result shown in (39b). (39) Distributed passive for an object-control verb a. λV λx ORDER(x, y, λz V(z)) & z=y b. PASS {λx ORDER(x, y, λu λz BEAR(z,u)) & z=y} = PASS {λx ORDER(x, y, pass {λu λz BEAR(z,u)}) & z=y} = λu dx ORDER(x, y, dz BEAR(z,u)) & z=y In this case the two subjects are independent of each other, as also evidenced by the two agent phrases in (38). There is therefore less reason for distributing the passive morphology to both verbs. Perhaps one might argue that the complex passive is the only way to realize the embedded object as the topic of the construction. Note that German does not allow the remote passive with an object-control verb.

83. Valency-changing word-formation

1443

(40) Passive of an object-control verb in German a. *Der gelbe Stern wurde zu tragen befohlen. ordered the.NOM yellow star was zu bear b. Den gelben Stern zu tragen wurde befohlen. the.ACC yellow star to bear was ordered.

4.3. Reflexive, reciprocal, and noun incorporation Another valency-decreasing operation is the lexical reflexive, using an affix. It establishes an anaphoric relationship by identifying a lower argument with the highest one. Such an operation is distinct from using a reflexive anaphora (such as themselves) in the syntax (see Reinhart and Siloni 2005). (41)

REFL

[λz ... λx VERB(x, ... , z )] = λx VERB(x, ... , x) +lr −lr −lr

The lexical operation also allows to opt for a possessor reflexive. (42a) shows the canonical case, in which a transitive verb is detransitivized. In (42b) however, the verb remains transitive; in this case a possessor is added to the core arguments (which makes the verb ditransitive) and is then identified with the highest argument. In general, reciprocal functions similarly, but has a more complex semantics: the antecedent X must be plural and must receive a distributed interpretation, and any x2X is paired with some (or every) y2X, where y≠x. Interestingly, (42c) illustrates a case in which the reciprocal morpheme must be combined with the reflexive in order to unfold its full semantics. (42) Reflexive and Reciprocal in Bolivian Quechua a. Pedru maylla-ku-n. P wash-REFL-3sg ‘Pedro washes himself.’ REFL[λy λx WASH(x,y)] = λx WASH(x,x) b. Pedru uya-n-ta maylla-ku-n. P face-3sg-ACC wash-REFL-3sg ‘Pedro washes his (own) face.’ REFL[λz λy λx {WASH(x,z) & POSS(y,z)}] = λz λx {WASH(x,z) &

POSS(x,z)}

c. maylla-na-ku-yku. wash-REC-REFL-1pl ‘We wash each other.’ (van de Kerke 1996: 160, 146) Incorporation is quite a different type of valency-reducing; in this case, an argument is realized by a morphologically integrated nominal predicate. For instance, a noun can be prefixed to the verb stem (as in klavierspielen ‘piano-playing’), so that this noun predicates of the lowest argument of the verb. Van Geenhoven (1998) analysed incorporated nouns as predicative indefinites. Formally, one can assume an operation that takes two

1444

VII. Semantics and pragmatics in word-formation II: Special cases

elements in the input, a noun and a verb, and produces a coherent verb reading by argument identification, see (43). (43)

INCORP

< λv NOUN(v), λz ... VERB( ... , z )> = dz ... {VERB( ... , z) & NOUN(z)} −lr

This analysis suggests that noun incorporation always leads to a general or unspecific reading; however, some languages also allow a specific reading of the incorporated noun, as in (44a), where a demonstrative is “stranded”. Its referent has to be identified with the entity that the N-V complex is predicating of, see (44b). (44) Noun incorporation with definite reading in Southern Tiwa a. Yede a-seuan-mu-ban. that 2sg-man-see-PAST ‘You saw that man.’ b. R(‘that’) = ιz {SEE(you,z) & MAN(z)} (Baker 1988: 93) Only the lowest argument can be incorporated − probably because canonical λ-application takes place, affecting the lowest θ-role first. Thus, a ditransitive verb incorporates the theme, not the recipient. (45) ‘give’ in Southern Tiwa Ka-’u’u-wia-ban. 1sg>2sg-baby-give-PAST ‘I gave you the baby.’ λy λx λe dz {{ACT(x) &CAUS

BEC POSS(y,z)}

& baby(z)}(e) (Baker 1988: 110)

Noun incorporation cannot be iterated directly, but it is possible to have more than one cycle of operations, such as incorporation + causative in (46). This generates the complex meaning given in (46b). (46) Cyclic incorporation in Halkomelem (Salish) a. niʔ ʔə č sq-əəlcəp-st=ənəq-stəxʷ? AUX Q 2SUBJ cut-wood-CAUS-person-CAUS ‘Did you have him have people cut firewood?’ b. λu λv λe {ACT(u) & dx {ACT(v) & dy {CUT(x,y) & PERSON(x)}}(e) (Gerdts 2004: 215)

WOOD(y)}

&

Noun incorporation creates a configuration where other arguments can function as the object or subject: an instrument (47a), a possessor (47b, c), or a goal (47d). A transitive verb is detransitivized in (47a, b), and then undergoes instrumental or possessor applicative (see next section), becoming again transitive. Even an intransitive verb can incorporate (47c, d), and becomes again intransitive when it undergoes possessor or goal appli-

83. Valency-changing word-formation

1445

cative. (The bracketings in the semantic representations show the ordering of operations. It is open to discussion, whether the possessor is introduced by an applicative or inherited from the incorporated possessed noun.) (47) Noun incorporation, followed by an invisible applicative a. Kua ta fakatino he tama e malala. PERF-draw-picture ERG-child NOM-charcoal ‘The child has been drawing pictures with a charcoal.’ λz λx λe {dy {DRAW(x,y) & picture(y)} & INST(z)}(e) b. Wa-hi-nuhs-ahni:nu: John. PAST-1sg>3m-house-buy John ‘I bought John’s house.’ λz λx λe {dy {BUY(x,y) & (house(y)} & c. Hrao-nuhs-rakv ne sawatis. 3m-house-white DET John ‘John’s house is white.’ λy λe {dx {WHITE(x) & (house(x)} &

Niue (Polynesian)

Oneida (Iroquoian)

POSS(z,y)}(e)

Mohawk (Iroquoian)

POSS(y,x)}(e)

d. Am-seuan-wan-ban liora-n. Southern Tiwa (Kiowa-Tanoan) 3pl-man-come-PAST lady-pl ‘The man came to the ladies.’ λy λe {dx {COME(x) & man(x)} & GOAL(y)}(e) (Baker 1988: 128, 96, 107) Note that noun incorporation is also productive in Swedish, where it allows to introduce another participant as the direct object: Läkaren hjärt-opererade patienten ‘The doctor heart-operated the patient’ (Mellenius 1996).

5. Valency-increasing operations 5.1. Causative and similar operations Valency-decreasing always binds an existing argument, so the semantic core remains unaffected. By contrast, valency-increasing adds an argument as well as a licensing predicate, so the semantic core itself is enlarged. Either a higher predicate together with a higher argument is added, or a lower predicate together with a lower argument. A prototypical instance of the former type of operations is the causative, whereas various sorts of applicative are characteristic for the latter type of operations. The causative adds a causer, who instigates the event expressed by the basic verb, either by direct coercion, or more indirectly by giving an order or admitting a certain course of affairs. Some version of causative is found in nearly every language, and many languages have more than one type of causative (differing morphosyntactically and often also in their finer semantic aspects). It is disputed in the literature whether the causative has to be represented explicitly by the predicate CAUSE (and whether this CAUSE is a relation between two events or

1446

VII. Semantics and pragmatics in word-formation II: Special cases

between an entity and an event), or whether the causal relationship can be inferred from the lexical combination of an action predicate with another, simpler predicate (see, e.g., the different views advocated by Bierwisch 2002 vs. Wunderlich 2000, 2012a). For the purposes of this article, I use the notion &CAUSE, taken as a contextually-induced reading of the connector AND. (Notice that &CAUSE is asymmetric, just like & itself, when used in a lexical decomposition from which argument hierarchy is derived.) Moreover, the causative usually gets a factive reading, which is expressed by existential binding of the verb’s original event variable. (48) caus [λe′

VERB(

... )(e′)] = … λx λe {ACT(x) &CAUSE de′

VERB(

... )(e′)}(e)

In a typical causative formed from a transitive verb the causee becomes the medial argument; it is marked dative in an accusative language such as Japanese (see below (54a)), as well as in an ergative language such as Basque (49). (49) Causative in Basque Ama-k haurr-a-ri zopa jan-eraz-i dio. mother-ERG child-DET-DAT soup.NOM eat-CAUS-PERF have.3N.3sgD.3sgE ‘Mother let the child eat the soup.’ λz λy λx λe {ACT(x) &CAUSE de’ EAT(y,z)(e’)}(e) (Joppen and Wunderlich 1995: 145) In a double object construction the causee is realized as the primary object (which can become the subject under passive, can be co-indexed with an object affix, etc.). Besides this unmarked option, illustrated in (50a), there is also a marked option, in which the causee is obliquely realized and does not function as a structural object (50b). Such a marked option is found in various languages, even in those that otherwise have a dative; in Hungarian, for instance, it can be captured by the assumption that the causative morpheme lexically assigns instrumental case (51b). (50) Causative variation in Bantu: Chimwiini (a) vs. Chichewa-A (b) a. Mwa:limu 0̸-wa-andik-ish-ize wa:na xati. teacher SU-OB-write-CAUS-ASP children letter ‘The teacher made the children write a letter.’ b. Anyani a-na-wa-meny-ets-a ana kwa buluzi. baboons SU-PAST-OB-hit-CAUS-ASP children to lizard ‘The baboons made the lizard hit the children.’ (Baker 1988: 183, 163) (51) Medial arguments in Hungarian a. Anna Péter-nek adott egy könyv-et. A. P.-DAT gave a book-ACC ‘Anna gave a book to Peter.’ b. Anna könyv-et olvas-tat Péter-rel. A. book-ACC read-CAUS P.-INST ‘Anna has Peter read a book.’ (Wunderlich 2002)

83. Valency-changing word-formation

1447

The periphrastic causative is a construction formed with an object control verb, such as force, make, or let (force him to go, make him go, let him go). Such a verb adds two arguments, whereby it identifies the object with the subject of a dependent (infinitive) clause (52a). It would, therefore, be unexpected that a single morphological operation produces an object control configuration, as proposed by Alsina 1992 (see also Matsumoto 1998). (52) Periphrastic vs. morphological causative a. λP λy λx FORCE(x,y,P(y)) b. λp λx {ACT(x) &CAUSE p} A few languages have operations that add a highest argument in a function distinct from causer. One such operation is the assistive in Quechua. (53) shows that causative and assistive are structurally alike in Quechua: either a causer or a helper is added as the highest argument. Note that a helper does not necessarily contribute an additional event because she is involved in the same type of action as the helpee. (Quechua lacks a dative, therefore all objects are realized as accusative; object agreement on the verb refers to the highest object, which is the causee or helpee in these cases.) (53) Causative and Assistive in Bolivian Quechua a. mama-y Maria-ta maylla-chi-wa-rqa. mother-1sg Mary-ACC wash-CAUS-1A-PAST ‘My mother made me wash Maria.’ λz λy λx λe {ACT(x) &CAUSE WASH(y,z)}(e) b. mama-y Maria-ta maylla-ysi-wa-rqa. mother-1sg Mary-ACC wash-ASS-1A-PAST ‘My mother helped me to wash Maria.’ λz λy λx λe {HELPER(x) & WASH(y,z)}(e) (van de Kerke 1996: 153, 157) Another operation that adds a highest argument is the affective in Japanese. That causative and affective are structurally alike is shown in (54a, b). Although the affective is formed with the same suffix (-are) as the passive and is therefore traditionally called “indirect passive”, its argument structure is clearly distinct from that of a passive (54c). (Note that -ni functions both as dative and as adverbial postposition.) (54) Causative, affective and passive in Japanese a. John-ga Mary-ni tokei-o nusum-ase-ta. John-NOM Mary-DAT watch-ACC steal-CAUS-PAST ‘John let Mary steal a watch.’ λy λx λu λe {ACT(u) &CAUSE STEAL(x,y)}(e) b. John-ga Mary-ni tokei-o nusum-are-ta. John-NOM Mary-DAT watch-ACC steal-AFF-PAST ‘John had a watch stolen by Mary.’ = ‘John was affected by Mary stealing (his) watch.’ λy λx λu λe {AFF(u) & STEAL(x,y)} (e)

1448

VII. Semantics and pragmatics in word-formation II: Special cases c. Tokei-ga Mary-ni nusum-are-ta. watch-NOM Mary-BY steal-PASS-PAST ‘The watch was stolen by Mary.’ λy λe dx STEAL(x,y) (e) (Washio 1995: 6)

In Korean, the morphological causative and passive use the same suffix -i (or -hi, -li, -ki, depending on the phonological context), so (55) is ambiguous; here, the passive reading is only possible with an inalienable possession. (Note that Korean also has a light-verb passive and a periphrastic passive.) (55) Ambiguity between Korean causative and passive. John-i/nun Mary-eykey meli-ul kkakk-i-ess-ta. J.-NOM/TOP M.-DAT hair-ACC cut-CAUS/PASS-PAST-DECL (i) ‘John had Mary cut the hair.’ (John’s or someone else’s hair) (ii) ‘John was cut his hair by Mary.’ (Kim and Pires 2003)

Causative Passive

5.2. Applicative For operations that add a non-highest argument, the term applicative is used as a collective name; the added argument can be a recipient, possessor or beneficiary, a location or source, an instrument, a companion or some other participant of an event. In some languages, a single morpheme encodes all these semantic roles, while other languages have several distinct morphemes. The general scheme of applicatives when applied to a transitive verb is given in (56). (Whether BECOME is present or not depends on further circumstances, especially on the dynamics of the verb.) (56)

APPL [VERB(x,y)] = VERB(x,y)

& & &

POSS(z,y)

‘z is (or becomes) a possessor of y’ z) ‘y is (or becomes) located at z’ INST(z,y) ‘z operates as an instrument on y’ LOC(y AT

In principle, the operation is possible with ditransitive verbs, too. Some applicatives also apply to intransitive verbs; they then characterize a relation with the intransitive subject, or just a further participant of the event. Applied objects can also stand in a manner, comitative or sociative relation. In any case, the subject remains the same. The most prototypical instance of applicative is the benefactive alternation, shown in (57); here, the transitive verb ‘to buy’ becomes ditransitive by means of the applicative suffix -kan in (57b). Following Baker (1988), one might say that the preposition ‘for’ is incorporated into the verb, so that the prepositional object becomes a direct argument of the verb. However, since untuk ‘for’ and the applicative kan are quite distinct morphemes, “incorporation” would have to be understood in a rather abstract sense. Conceptually it is more convenient to consider the applicative as a way of expressing further participants, independently of whether corresponding prepositional means exist. Thus, the relationship between (57a) and (57b) is purely semantical, not generative; the appli-

83. Valency-changing word-formation

1449

cative applies to the verb, not to a syntactic construction. The predicate variable ‘P’ in (57a) serves as a placeholder for the prepositional phrase, which could also be ‘with’, ‘in’, etc. Even if one concedes that the applicative could have a comitative rather than a benefactive meaning, the verb in (57b) is more specific than that in (57a) because it has a third structural argument. (57) Benefactive alternation in Bahasa Indonesia a. Ali memi telefisi untuk ibu-nja. Ali TR.buy televison for mother-his ‘Ali bought a televison for his mother.’ λP λz λx λe{BUY(x,z) & P(z)}(e) b. Ali mem-beli-kan ibu-nja Ali TR-buy-APPL mother-his ‘Ali bought his mother a televison.’ λz λy λx λe{BUY(x,z) & BECOME POSS(y,z)}(e) (Chung 1976) Note that some German prepositions have developed to prefixes, which seems to support Baker’s incorporation hypothesis (er fuhr um den Park − er umfuhr den Park ‘he drove around the park’), see also (3b, f ) above. Prefixes and particles are widely used as applicative markers in the Germanic and Slavic languages (see Stiebels 1996; Olsen 1997; Rojina 2004), however, the benefactive expressed by a dative or primary object is not marked on the Indo-European verb. To illustrate how widely applicative constructions vary cross-linguistically, I will briefly discuss examples from three non-related languages: Maasai (Maa) (an Eastern Nilotic language spoken in southern Kenya and northern Tanzania), Halkomelem (a Salish language spoken in southwestern British-Columbia), and the Bantu-language Kinyarwanda (spoken in Rwanda). (The Maasai people had contact with Bantus for centuries, but taken their military strength in the past as well as their general mental attitude, it is unlikely that they have borrowed from Bantu.) Maasai has two general applicative morphemes: (i) the dative applicative -aki(n) (together with some allomorphs in the perfective), which adds a benefactive or goal argument, and (ii) the instrumental applicative -íé(k), which adds an instrument, associative, locative or agent role. Both can be attached to intransitive, transitive and ditransitive verbs, and appear also in the combination DAT-INSTR. The additional argument is always a core argument, due to the following tests: It can pronominally be indicated by one of the verb prefixes that encode a transitive relation; as an NP it appears with the accusative tonal pattern; in the passive/middle it appears as subject (marked by a pronominal prefix or with nominative tonal pattern). Obligatory for the Maasai verb is only the pronominal prefix marking the subject; with a 1sg or 2sg object a so-called inverse prefix marks both subject and object: áá- for 1sg>2sg, áa- for 3>1sg, and kí- for 2>1sg or 3>2sg. All objects can be realized by an accusative NP, regardless of whether they belong to the root or are added by an applicative; in principle, either one of them can become nominative in the passive/middle. It is possible to have four explicit NP arguments. Oblique arguments are introduced with the preposition tɛ.

1450

VII. Semantics and pragmatics in word-formation II: Special cases

(58) Applicatives in Maasai a. e-ton-íék-ì. 3sg-sit-INST-MED ‘It will be used to sit.’ b. áa-iger-óki ɛm-páláí. 3>1sg-write-DAT fsg-letter.ACC ‘He will write a letter to/for me.’ c. áá-ishɔ-ɔkí kanisa. 1sg>2sg-give-DAT church.ACC ‘I will give it to the church for you.’ d. ɛ-súj-ákin-íé ɛn-kɛráí ɛn-kítɛŋ e-ŋúdì en-tím. 3-follow-DAT-INST fsg-child.NOM fsg-cow.ACC fsg-stick.ACC fsg-forest.ACC ‘The child will use the stick to follow the cow into the bush.’ (Lamoureaux 2004: 64, 40, 41, 54) All three objects of a DAT-INST combination can potentially be middle subjects (59b−d); thus Maasai is a language with symmetric objects (see Bresnan and Moshi 1990). (59) Symmetric middle in Maasai a. e-duŋ-ókín-yìè in-kírí ɔl-álɛm en-kitók. 3-cut-DAT-INST fpl-meat.ACC msg-knife.ACC fsg-woman.ACC ‘He will cut meat with the knife for the woman.’ b. e-duŋ-ókín-óré in-kírí ɔl-álɛm en-kítok. 3-cut-DAT-INST.MID fpl-meat.ACC msg-knife.ACC fsg-woman.NOM ‘The woman will cut meat for herself with the knife.’ c. e-duŋ-ókín-óré in-kírí ɔl-alɛm en-kitók. 3-cut-DAT-INST.MID fpl-meat.ACC msg-knife.NOM fsg-woman.ACC ‘The knife is used to cut meat for the woman.’ d. e-duŋ-ókín-óré in-kirí ɔl-álɛm en-kitók. 3-cut-DAT-INST.MID fpl-meat.NOM msg-knife.ACC fsg-woman.ACC ‘The meat is cut with the knife for the woman.’ (adapted from Lamoureaux 2004: 95) Interestingly, the instrumental applicative also serves as causative; in this function it introduces a dependent agent role rather than the causing agent. This can easily be seen from the examples below: a machete is a good instrument for felling a tree (60a), while children achieve the instrumental role when they are told to do the action (60b). Thus, the Maasai causative does not introduce the causer but the causee. This is even true where other languages have a lexical causative; compare ‘look.at’ and ‘show’ in (60c, d). (60e) shows nicely the kind of ambiguity that can arise.

83. Valency-changing word-formation

1451

(60) Instrumental vs. causative use of the applicative a. e-ur-íé ɔl-pánkà ɔl-catá. 3-make.fall-INST msg.machete.ACC msg-tree.ACC ‘He will use the machete to fell the tree.’ b. e-ur-íé iŋ-kɛrà il-paék. 3sg-make.fall-INST fpl-children.ACC mpl-corn.ACC ‘He will make the children bend the corn.’ c. á-íŋɔr-ìè ɔl-ŋátúny ɔl-tɔrɔbíní. 1sg-look.at-INST msg-lion.ACC msg-binoculars.ACC ‘I will look at the lions with the binoculars.’ d. á-íŋɔr-ìè ɔl-ŋátúny ɔl-payíán. 1sg-look.at-INST msg-lion.ACC msg-man.ACC ‘I will show the man the lions.’ e. á-ínɣàŋ-un-ié ɛn-kítɛŋ. 1sg-buy-VENT-INST fsg-cow.ACC (i) ‘I will use something (e.g. money) to buy a cow.’ (ii) ‘I will make him/her buy a cow.’ (Lamoureaux 2004: 80, 74; the ventive expresses a motion towards the point of reference) Quite different from Maasai is Salish, a family of 23 languages spoken in British Columbia and northwest USA. Only up to two core arguments are possible, all others are marked as oblique. All Salish languages distinguish at least two (classes of ) applicative morphemes: APPL1 (called relational), adding an object to an intransitive verb, and APPL2 (called redirectional), adding an object to a transitive verb, whereby the former object becomes oblique (Kiyosowa 2006; Kiyosowa and Gerdts 2010). For instance, psych predicates are usually intransitive, and become transitive either by APPL1 (with the applied object as stimulus) or by the causative, as shown in (61). (The suffix -t is here a transitivity marker (TR), not a valency-changing morpheme.) (61) Intransitive psych verbs in Halkomelem a. cəqʔ-meʔ-t č ceʔ kʷθə nəcəwməxʷ. surprise-APPL1-TR 2sgSUBJ FUT DET visitor ‘You will be surprised at the visitors.’ b. niʔ cən siʔsiʔ-meʔ-t kʷθə sqʷəmey. AUX 1sgSUBJ frighten-APPL1-TR DET dog ‘I was frigthened at the dog.’ c. niʔ cən siʔsiʔ-stəxʷ kʷθə sməyəθ. AUX 1sgSUBJ frighten-CAUS DET deer ‘I frigthened the deer.’ (Gerdts and Kiyosawa 2005: 334, 339)

1452

VII. Semantics and pragmatics in word-formation II: Special cases

‘Give’ is basically a 2-place verb, the recipient is only added by means of the theme argument is demoted.

APPL2,

thus,

(62) Applicative with transitive verbs in Halkomelem a. ni ʔám- əs-t-əs kʷθə John ʔə kʷθə púkʷ. AUX give-APPL2-TR-3SUB DET John OBL DET book ‘He gave John the book.’ b. ni

ʔám- əs-t-əm kʷθə John ʔəƛ Mary ʔə kʷθə púkʷ. give-APPL2-TR-PASS DET John OBL Mary OBL DET book ‘John was given the book by Mary.’ (Gerdts 1988: 101, 233)

AUX

The applicative can also follow the causative, cf. (63), where ‘to eat’ is basically intransitive. In any case, the number of core arguments cannot exceed two, at any stage of derivation. (63) Causative + applicative in Comox ʔiltən-st-aʔam-θi tθəm ʔə tə čuy. eat-CAUS-APPL2-TR.2sgOBJ 1sgSUBJ.FUT OBL DET child ‘I will feed the child for you.’ (Watanabe 2003: 250, cited in Kiyosawa 2006: 295) The applicatives in the Bantu languages, particularly those in Kinyarwanda, are since Kimenyi (1980) the topic of various theoretical discussion (Baker 1988; Marantz 1984, 1993; Bresnan and Moshi 1990; Pylkkänen 2008; McGinnis and Gerdts 2003; Mc Ginnis 2005, 2008). Some applicative variants of Kinyarwanda are illustrated in (64a−c); here, benefactive, possessor-raising, and instrumental applicative are marked by different suffixes. As can be seen, -iish is ambiguous; similarly to Maasai it either marks instrumental applicative or causative (64c, d). The reading depends on the sortal properties of the complements: usually a child but not a piece of soap is washed, while soap but not a child can be an instrument of washing. (64) Applicatives in Kinyarwanda a. umugóre y-a-som-e-ye umwáana igitabo book woman 3sg-PAST-read-APPL-PERF child ‘The woman read the book to the child.’ (benefactive appl) b. umugabo a-ra-kikir-ir-a umugóre umwáana. man 3sg-PRES-hold-APPL-IMPF woman child ‘The man is holding the woman’s child.’ (possessor-raising appl) c. umugóre y-0̸-uhag-iish-ije umwáana isábune. soap woman 3sg-PAST-wash-APPL-PERF child ‘The woman washed the child with soap.’ (instrumental appl) d. umugóre y-0̸-uhag-iish-ije umukoóbwa umwáana. child woman 3sg-PAST-wash-CAUS-PERF girl ‘The woman made the girl wash the child.’ (causative) (Polinsky and Kozinsky 1992)

83. Valency-changing word-formation

1453

The different sorts of applicative behave differently in Kinyarwanda: benefactives are symmetric, while locatives are asymmetric. More precisely, in the benefactive applicative either object can become subject under passive or be realized with an object marker on the verb, while in the locative applicative only the applied object (the respective location) can become subject under passive or be realized with an object marker (65, 66). The object markers are similar to the class markers CL for the subject. Note that also the morpheme ordering in the locative differs from that in the benefactive in that the locative marker -ho follows the aspect marker; however, I don’t think that this is relevant for the different grammatical potential. (The data are reported by Kimenyi 1980; McGinnis 2001; Jeong 2006; and many others.) (65) Passive of Kinyarwanda applicatives Benefactives a. Umukoôbwa a-ra-andik-ir-w-a íbárúwa n’ûmuhuûngu. girl 3sg-PRES-write-APPL-PASS-ASP letter by boy ‘The girl is written the letter for by the boy.’ b. Íbárúwa i-ra-andik-ir-w-a umukoôbwa n’ûmuhuûngu. letter CL-PRES-write-APPL-PASS-ASP girl by boy ‘The letter is written for the girl by the boy.’ Locatives c. Ishuûri ry-oohere-j-w-é-ho igitabo n’úúmwáalímu. school CL-send-ASP-PASS-ASP-LOC book by teacher ‘The school was sent the book by the teacher.’ d. *Igitab cy-oohere-j-w-é-ho ishuûri n’úúmwáalímu. book CL-send-ASP-PASS-ASP-LOC school by teacher ‘The book was sent to school by the teacher.’ (66) Object markers in Kinyarwanda applicatives Benefactives a. Umugóre a-rá-mu-he-er-a ímbwa ibíryo. food woman 3sg-PRES-OBJ-give-APPL-ASP dog ‘The woman is giving food to the dog for him.’ b. Umugóre a-rá-bi-he-er-a umugabo ímbwa. dog woman 3sg-PRES-OBJ-give-APPL-ASP man ‘The woman is giving it to the dog for the man.’ Locatives c. Úmwáalímu y-a-ry-oohere-jé-ho igitabo. teacher 3sg-PAST-OBJ-send-ASP-LOC book ‘The teacher sent the book to it.’ d. *Úmwáalímu y-a-cy-oohere-jé-ho ishuûri. teacher 3sg-PAST-OBJ-send-ASP-LOC school ‘The teacher sent it to school.’ A possible explanation for this different behavior might go as follows. Let us assume that a locative semantically always adds LOC(y,z) to the event, where z is the location

1454

VII. Semantics and pragmatics in word-formation II: Special cases

and y another argument of the verb; thus, the added argument is always the lowest one. By contrast, what the benefactive adds might be cognized in two different ways: (i) it simply adds a benefactive to the event, i.e. BEN(z), where the added argument is again the lowest one, or (ii) it adds a possessor-relation to another argument of the event, i.e. POSS(z,y); here, the added argument is not the lowest one. Let us further assume that only the lowest argument can become subject under passive or indicated with an object prefix; then, the distribution in (65) and (66) automatically follows. In other words, in the symmetric benefactives two different analyses are possible. Pylkkänen (2008) distinguishes between high and low applicatives: low applicatives establish a semantic (possessor) relation between two individual arguments, such as a recipient and a theme, while high applicatives establish a semantic relation between an individual and an event; syntactically, the low applicative is contained within the verbal phase (VP), whereas the high applicative operates above of it. In the above-used notation, ‘& POSS(z,y)’ characterizes a low applicative, and ‘& BEN(e,z)’ a high applicative semantically. Consequently, when an applicative is low, it can only operate on transitive verbs, whereas when it is high it can also operate on intransitive verbs. In this respect, all the Maasai applicatives have to be classified as high. In the Salish languages, APPL1 must be high, whereas APPL2 might be classified as low. Indeed, all the documented semantic roles of APPL2 (see Kiyosawa 2006), such as recipients, sources, possessors, raised possessors and even delegatives (‘instead of’), are based on a twoplace relation between individuals. What is described as benefactive vs. malefactive could be differentiated by pragmatic inference. The delegative reading, however, is problematic for a syntactic account because it relates the applied object to the subject, which is VP-external. (67) Delegatives in Interior Salish a. Coeur d’Alene (Doak 1997: 157) níč-ši -t-s-es xʷe pilí. cut-APPL2-TR-1sgOBJ-3SUBJ DET Felix (i) ‘Felix cut (wood) instead of me.’ or (ii) ‘Felix cut (wood) for me.’ b. Okanagan (Mattina 1993: 272) kʷu qʷəlqʷíl-x-t-s. 1sgOBJ talk-APPL2-TR-3SUBJ ‘He talked for me (in my stead).’ (Kiyosowa 2006: 183) McGinnis and Gerdts (2003) study the position of objects as well as quantifier binding between them in several applicatives for determining their relative ordering. They also compare various multiple applicatives in Kinyarwanda, looking for the objects that can become subject under passive. They come to the conclusion that there is the ranking BEN > LOC > [VP transitive THEME > INST]. Since all these applicatives are possible with intransitives, they should be high, which constitutes a puzzle. One of the inherent problems of their account is that high/low is associated with several other contrasts: intransitive vs. transitive base, symmetric vs. asymmetric objects, phasal vs. non-phasal interpretation, interpreted as (z,y) vs. (e,z) relation. McGinnis (2005) admits that there might be some mismatching between the structure in which an applicative merges syntactically and the structure in which it is interpreted; for her, instrumentals merge low, but are high

83. Valency-changing word-formation

1455

semantically. However, note that both instrumentals and benefactives are marked nearer to the verbal root than locatives, thus, in a sense, they both are “lower” than locatives morphologically.

5.3. Multifunctional affixes In the preceding sections, several instances of affix ambiguity have been observed. Both the Maasai and the Kinyarwanda instrumental applicative can also be used as a causative (60b, 64d), the Korean causative can also be used as a passive (55), and the Japanese passive can also be used for introducing an affected argument in the highest position (54). From the viewpoint of their semantic effect, these affixes are clearly ambiguous: either a higher or a lower argument is added, and a highest argument either is introduced or existentially bound. Such an ambiguity can mostly be resolved, taking the specific lexical information of verb and complements into acccount. Still the question arises as to why this can happen. There seems to be some economy principle at work: it might be more costly to have a separate affix for each specific operation than having a single affix that is contextually specified. The Pama-Nyungan languages of Australia often have a single affix as a general transitivizer. The Kalkatunga suffix -nti adds a causer if it is combined with an inchoative or stative verb (68a), but it adds a beneficiary, instrument, or location if it is combined with an agentive verb (68b). (68) Transitivization in Kalkatunga a. iti ‘to return’ iti-nti ‘to send/bring back’ nguyi ‘fall’ nguyi-nti ‘push over’ b. nuu ‘to lie’ wani ‘to play’

nu-nti ‘to lie on (something)’ wani-nti ‘to play with (something)’ (Austin 1997)

Most transitive verbs of Salish have a transitivity marker which indicates transitivity, regardless of whether another transitivizing affix is present, recall (61, 62). Since the verbal roots of Salish are usually intransitive (Davis and Matthewson 2009), the presence of the transitivity marker is informative in general. In the reverse direction, there can be a general detransitivizer, usually called middle. According to the diagram (69), general transitivizers and general detransitivizers might be distinguished from more specific operations. (69) Transitivity alternations

transitive causative

applicative intransitive

antipassive

reflexive, reciprocal

(anticausative)

passive

1456

VII. Semantics and pragmatics in word-formation II: Special cases

Chukchi, a paleosibirian language, has the general detransitivizer -tku/-tko, comprising antipassive, reflexive, reciprocal, and anticausative. (This suffix has further functions not considered here.) (70) Detransitivization in Chukchi ʔ a. ʔəttʔ-e juu-nin əttʔ-ən nə-jγu-tku-qin ‘the dog bit him’ ‘the dog bites’

antipassive

b. tewla-nen ‘he shook it off’

tewla-tko-γʔe ‘he shook himself’

reflexive

c. ommačajpə-nen ‘he hugged him’

ommačajpə-tko-γʔat ‘they hugged each other’

reciprocal

d. ejpə-nin ‘he closed it’

ejpə-tku-γʔi ‘it closed’

anticausative (Nedjalkov 2006: 222)

A middle voice, often fused with tense or person, is found in many unrelated languages (e.g., Classical Greek, Fula, Munda). In some modern Indo-European languages (Spanish, Icelandic, Russian), the middle is derived by means of a reflexive marker. (Note that the German reflexive allows the same readings as Spanish except the passive one, see Steinbach 2002.) (71) Spanish reflexives a. Juan se lava. b. La cuerda se rompe. c. El libro se publicó en 1952. d. Se vive bien aqui. e. Estas frutas se comen.

‘Juan washes himself.’ ‘The rope splits.’ ‘The book was published in 1952.’

reflexive anticausative passive

‘People live well here.’ ‘These fruits are edible.’

impersonal modal

(Kaufmann 2004: 191) The anticausative is the odd one out among these specific detransitivizing operations. Unlike passive, it does not allow any reference to the agent of a transitive verb (using a by-phrase or ‘willingly’). Some part of the transitive verb’s meaning is really cut off. But that conflicts with the principle of monotonicity, stating that no semantic information is deleted in the course of derivation. Therefore it is hard to imagine that an affix could have emerged with a pure anticausative function; it would have been blocked by the monotonicity principle. What one indeed finds are morphemes with a broader function (such as the middle) including the anticausative reading as a special, contextually determined case. Japanese has the suffix -e, which either transitivizes or detransivizes. Comrie (2006) lists 57 inchoative/causative pairs where -e derives the causative verb (72a), and 36 pairs where it functions in the opposite manner, i.e. derives the inchoative verb from the transitive one (72b). This shows that the anticausative reading is restricted to a minor set of lexical items. Moreover, note that in addition to -e, Japanese has other means of deriving the transitive or intransitive variant, see (72c).

83. Valency-changing word-formation

1457

(72) Causatives vs. anticausatives in Japanese a. ak-u ‘to open’ ak-e-ru ‘to open (tr.)’ itam-u ‘to hurt’ tam-e-ru ‘to injure’ tat-u ‘to stand’ tat-e-ru ‘to raise’ b. nuk-u or-u tuka-u

‘to remove’ ‘to break (tr.)’ ‘to use’

c. kowa-s-u ‘to destroy’

nuk-e-ru or-e-ru tuka-e-ru

‘to come off’ ‘to break’ ‘to be usable’

kowa-re-ru ‘to be destroyed’ (Comrie 2006)

There is a lively debate as to whether anticausatives can preserve the reference to a causer (Koontz-Garboden 2009; Schäfer 2009) or not (Haspelmath 1993; Horvath and Siloni 2011), and also an intensive study into the historical development of anticausative constructions (Ottosson 2009; Cennamo, Eythórsson and Barðdal 2011). Multifunctional affixes are also interesting under the viewpoint of markedness. Usually the presence of a particular semantic feature (such as +plural, +2nd person, etc.) is morphologically marked. One could think of a very general affix that marks the nondefault value for each such dimension. In this respect it is interesting to note what Watters et al. (2006) report about a particular harmonic mutation in the verbs in Kusunda, a language isolate of Nepal. This mutation marks the semantically more complex category in various dimensions: it marks causative in the transitivity dimension, irrealis in the modality dimension, negation in the polarity dimension, and dependent in the dependency dimension. Thus, there is a possible scale of explicity: affixes that represent a specific operation on argument structure > those that represent a specific outcome of an operation (such as valency) > those that represent the non-canonical variant in a dimension (for instance, a deviation from what a lexical item means under usual circumstances).

6. Residual issues In this section some further issues involved in morphological valency-change are addressed.

6.1. Order of affixation Valency-changing operations can be combined cyclically, so that the output of a first operation serves as the input for a further operation. In particular, valency-increasing and valency-decreasing operations often alternate. Yucatec Maya illustrates a type of language in which no more than two structural arguments are possible; in a certain state of affairs a verb can either be transitivized or detransitivized. More precisely, causative (suffixation with -s) or applicative (suffixation with -t) applies to verbs in an intransitive state, whereas passive, antipassive or noun incorporation applies to verbs in a transitive state. These operations can easily be combined. (73a) shows the ordering V-CAUS-PASS,

1458

VII. Semantics and pragmatics in word-formation II: Special cases

(73b) V-APPL-PASS, (73c) V-PASS-CAUS-PASS, and finally (73d) V-INCORP-APPL. The last line in each example indicates the semantic representation of the complex verb. (73) Cyclic transitivization and intransitivization in Yucatec Maya a. kíin-s-áʔabʔ-en tumèen leti. die-CAUS-PASS.PERF-1 PREP PRON.3.SG ‘I was killed by him.’ λy λe dx {ACT(x) &CAUS DIE(y)}(e) kʔóoy-t-áʔal. dig -APPL-PASS.IMPF ‘It gets dug (up).’ λy λe dx {DIG(x) & AFFECTED(y)}(e)

b. k=a

INCOMPL=2

c. k=u

káʔan-s-áʔal. learn.PASS-CAUS-PASS.IMPF ‘It is being taught’ λz λe dx {ACT(x) & CAUS dy LEARN(y,z)}(e)

INCOMPL=3

d. taan=u

kon-lol -t -ik -et. sell-flower -APPL -IMPF -2 ‘He’s selling you flowers.’ (lit. ‘he’s flower-selling you’) λy λx λe {dz (SELL(x,z) & FLOWER(z)) & AFFECTED(y)}(e) (Krämer and Wunderlich 1999: 463−467)

INCOMPL=3

Several other languages supply evidence that valency-changing operations can be combined freely, though they may be subject to some sequential constraints, see Muysken (1986) on Quechua. (74a) is an example from Chichewa (Bantu) with the ordering VAPPL-REC-CAUS. (74b) shows the stepwise semantic interpretation; the symbol 4 indicates a reciprocal relationship between the two occurrences of a variable. (74) Interaction of argument changing operations in Chichewa a. M-lenje a-na-mang-ir -an -its -a a-tsikana nkhuni. 1-hunter 1-PAST-tie-APPL -REC -CAUS -FV 2-girl firewood ‘The hunter caused the girls to tie firewood for each other.’ b. mang: mang-ir: mang-ir-an: mang-ir-an-its:

λy λx λe TIE(x,y)(e) λy λz λx λe {TIE(x,y) & benefactive BEC POSS(z,y)}(e) λy λx4 λe {TIE(x,y) & reciprocal BEC POSS(x,y)}(e) λy λx4 λu λe {ACT(u) & causative CAUSE de’{TIE(x,y) & BEC POSS(x,y)}(e’)}(e) (Hyman 2003)

It is easy to imagine that any other order of affixes would yield a different interpretation: − −

REC-CAUS-APPL: CAUS-REC-APPL:

‘The hunter caused the girls to tie each other at the firewood’; ‘The hunters caused each other to tie firewood for the girls’.

83. Valency-changing word-formation

1459

However, some surface orders of suffixes are forbidden. According to Hyman and Mchombo (1992), the causative suffix -its may not appear after the applicative suffix -ir or the passive suffix -idw. In other words, the affix ordering is fixed, independently of semantic scope. (75) Surface alignment constraints in Chichewa a. *-ir-its (-APPL-CAUS) b. *-idw-its (-PASS-CAUS) The resulting ambiguity, however, can normally be resolved. This is illustrated in (76) for the first constraint. In (76a), the applied instrument sticks clearly relates to making cry and not to cry, whereas in (76b) the applied instrument spoon relates to the lower verb stir. (Interestingly, the passive of (76a) requires sticks as the subject, and the passive of (76b) requires woman as the subject, thereby reflecting the different semantic scope.) (76) Chichewa CAUS-APPL representing both scopes a. alenjé a-ku-líl-íts-il-a waná ndodo. hunters 3pl-PROG-cry-CAUS-APPL-FV child sticks ‘The hunters are making the child cry with sticks.’

(INST (CAUSE (CRY)))

b. alenjé a-ku-tákás-íts-il-a mkází mthíko. hunters 3pl-PROG-stir-CAUS-APPL-FV woman spoon ‘The hunters are making the woman stir with a spoon.’ (CAUSE (INST (STIR))) (Hyman 2003) According to Stiebels (2003), the affix order is transparent in (76a), but opaque in (76b) because of the language-specific constraint that violates the (revised) mirror principle. (77) Mirror Principle a. Morphological derivations must directly reflect syntactic derivations (and vice versa). (Baker 1985: 375) b. Revised: The affix order must mirror semantic composition. (Stiebels 2003: 292) The revised mirror principle claims a correspondence between morphology and semantics rather than between morphology and syntax.

6.2. Other types of argument alternation From the literature (especially Levin 1993 on English), many types of argument alternations are known, in which two arguments shift their place, such as wipe-alternation, locative alternation, or dative shift.

1460

VII. Semantics and pragmatics in word-formation II: Special cases

(78) wipe-alternation a. She wiped the crumbs from the table. b. She wiped the table. (79) Locative alternation a. Max loaded the luggage into the car. b. Max loaded the car (with luggage). (80) Dative shift a. Anna gave the photos to Max. b. Anna gave Max the photos. In the wipe- as well in the locative alternation, the respective a-sentence adds a locative (source or goal) PP to the transitive verb, while the verb in the b-sentences already incorporates that information (e.g., & dy BECOME¬LOC(y,ON(table)) in (78b). These alternations, thus, nicely illustrate that the locative applicative is possible even if it is not marked; the DP next to the verb realizes the applied object (78b, 79b). The dative shift is a bit different, because the DP next to the verb realizes a recipient (a possessor) (80b). Here, the ditransitive construction might be considered as an instance of the possessor applicative. It has often been observed that there are finer semantic differences between the respective a- and b-sentences, which, however, follow from the different status of the arguments involved. For instance, (78a) entails that all the crumbs are wiped off, whereas (78b) entails that the whole table is wiped, but not that all the crumbs are wiped off. If a language allows only two structural arguments, the base object must be realized obliquely when APPL2 is applied (see above). English, however, accepts a double-object construction − why not in the wipe- and the locative alternation? The answer is: the applied object in these constructions is not a medial argument, and the base argument is “moved” into a position, where it cannot be expressed as structural argument (see Wunderlich 1997a, b). Moreover, the crumbs must remain unexpressed in (78b) because an appropriate preposition in English doesn’t exist.

6.3. The place of valency-changing morphology in the grammar Valency is a basic property of lexical items, therefore, a morpheme that changes valency is expected to attach directly to the root or stem. A valency shift can affect the aspectual or temporal properties of the stem, and thus determines the expected association with tense-aspect (as seen in the Inuit antipassive, (18b)). Furthermore, valency shift is expected to preceed person-number inflection, relating to the arguments actually present. Valency-decreasing noun incorporation enables a further cycle of valency-changing operations, as shown in (46). Moreover, nominalization can follow a lexical reflexive like in Hungarian mos-akod-ás (wash-REFL-NOML ‘self-washing’; Reinhart and Siloni 2005: 411). In sum, valency-change belongs to the derivational component of morphology, which usually preceeds inflection. Both causative and passive are often periphrastic, while applicative and antipassive are not − why? Auxiliaries as well as light verbs operate as higher verbs, and thus can

83. Valency-changing word-formation

1461

only affect the highest argument. By contrast, applicative constructions alternate with oblique objects or low adverbials, which do not make use of a verb. From the viewpoint of interaction between morphosyntax and semantics, three general types have been identified along this article: (i) A valency-changing operation can be invisible on the verb. (ii) A valency-changing operation can be realized by a general transitivizer or detransitivizer, so that the actual meaning depends on further context. (iii) A valency-changing operation can be realized by a specific morpheme or construction (such as passive or causative). It is this order in which the operations vary between less and more explicity.

7. References Aissen, Judith 1999 Markedness and subject choice in optimality theory. Natural Language and Linguistic Theory 17: 673−711. Alsina, Alex 1992 On the argument structure of causatives. Linguistic Inquiry 23: 517−555. Árnadóttir, Hlíf, Thórhallur Eythórsson and Einar Freyr Sigurðsson 2011 The passive of reflexive verbs in Icelandic. In: Tania E. Strahan (ed.), Nordlyd 37. Relating to Reflexives, 39−97. Tromsø: Universitetet i Tromsø. http://www.ub.uit.no/ munin/nordlyd [last access 10 Oct 2014]. Austin, Peter 1997 Causatives and applicatives in Australian aboriginal languages. In: Kazuto Matsumara and Tooru Hayasi (eds.), The Dative and Related Phenomena, 165−225. Tokyo: Hituzi Syobo. Baker, Mark 1985 The mirror principle and morphosyntactic explanation. Linguistic Inquiry 16: 373−415. Baker, Mark 1988 Incorporation. A theory of grammatical function changing. Chicago: University of Chicago Press. Bierwisch, Manfred 2002 A case for cause. In: Ingrid Kaufmann and Barbara Stiebels (eds.), More than Words, 327−353. Berlin: Akademie Verlag. Bresnan, Joan 1982 The passive in lexical theory. In: Joan Bresnan (ed.), The Mental Representation of Grammatical Relations, 3−86. Cambridge, MA: MIT Press. Bresnan, Joan and Lioba Moshi 1990 Object asymmetries in comparative Bantu syntax. Linguistic Inquiry 21: 147−185. Bricker, Victoria R. 1978 Antipassive constructions in Yucatec Maya. In: Nora C. England, Collete G. Craig and Louanna Furbee-Losee (eds.), Papers in Mayan Linguistics, 3−23. Columbia: Museum of Anthropology, University of Missouri. Bricker, Victoria R. and Eleuterio Po’ot Yah 1981 Yucatec Maya Verbs (Hocaba dialect). New Orleans: Latin American Studies Center, Tulane University.

1462

VII. Semantics and pragmatics in word-formation II: Special cases

Cennamo, Michela, Thórhallur Eythórsson and Jóhanna Barðdal forthc. Constructions in competition: Anticausativity in Latin and Germanic. In: Leonid Kulikov and Nikolaos Lavidas (eds.), Typology of Labile Verbs. Focus on Diachrony [Special issue of Linguistics]. Chung, Sandra 1976 An object-creating rule in Bahasa Indonesia. Linguistic Inquiry 7: 41−87. Comrie, Bernard 2006 Transitivity pairs, markedness, and diachronic stability. Linguistics 44: 303−318. Davis, Henry and Lisa Matthewson 2009 Issues in Salish syntax and semantics. Language and Linguistics Compass 3(4): 1097− 1166. Doak, Ivy Grace 1997 Coeur d’Alene grammatical relations. Ph.D. dissertation, University of Texas at Austin. Dowty, David R. 1982 Grammatical relations and Montague grammar. In: Pauline Jacobson and Geoffrey K. Pullum (eds.), The Nature of Syntactic Representations, 79−130. Dordrecht: Reidel. Dryer, Matthew S. 1986 Primary objects, secondary objects, and antidative. Language 62: 808−845. Engdahl, Elisabet 1999 The choice between bli-passive and s-passive in Danish, Norwegian and Swedish. Ms. Göteborg University. Engdahl, Elisabet 2006 Semantic and syntactic patterns in Swedish passives. In: Benjamin Lyngfelt and Torgrim Solstad (eds.), Demoting the Agent. Passive, Middle and Other Voice Phenomena, 21− 45. Amsterdam/Philadelphia: Benjamins. Eythórsson, Thórhallur 2008 The new passive in Icelandic really is a passive. In: Thórhallur Eythórsson (ed.), Grammatical Change and Linguistic Theory. The Rosendal Papers, 173−219. Amsterdam/ Philadelphia: Benjamins. Gerdts, Donna B. 1988 Object and Absolutive in Halkomelem Salish. New York: Garland. Gerdts, Donna B. 2004 Combinatory conditions on Halkomelem causatives. Linguistics 42: 767−789. Gerdts, Donna B. and Kaoru Kiyosawa 2005 Halkomelem psych applicatives. Studies in Language 29: 329−362. Gísladóttir, Rósa Signý 2007 The new passive in Icelandic. A dynamic syntax perspective. M.Sc. thesis, King’s College, University of London. Haspelmath, Martin 1993 More on the typology of inchoative/causative verb alternations. In: Bernard Comrie and Maria Polinsky (eds.), Causatives and Transitivity, 87−120. Amsterdam/Philadelphia: Benjamins. Hellan, Lars 2001 On complex passives in Norwegian. In: Elena Anagnostoupolou and Marc van Oostendorp (eds.), Progress in Grammar. Amsterdam: Meertens Institute Electronic Publications in Linguistics. http://www.meertens.knaw.nl/books/progressinggrammar [last access 10 Oct 2014]. Horvath, Julia and Tal Siloni 2011 Anticausatives: Against reflexivization. Lingua 121: 2176−2186. Hyman, Larry M. 2003 Suffix ordering in Bantu: A morphocentric approach. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 2002, 245−281. Dordrecht: Kluwer.

83. Valency-changing word-formation

1463

Hyman, Larry M. and Sam Mchombo 1992 Morphotactic constraints in the Chichewa verb stem. Berkeley Linguistic Society 18: 350−364. Jeong, Youngmi 2006 The landscape of applicatives. Ph.D. dissertation, University of Maryland, College Park. Johnson, Heidi 2000 A grammar of San Miguel Chimalapa Zoque. Ph.D. dissertation, University of Texas at Austin. Joppen, Sandra and Dieter Wunderlich 1995 Argument linking in Basque. Lingua 97: 123−169. Joppen-Hellwig, Sandra 2001 Verbklassen und Argumentlinking. Tübingen: Niemeyer. Kaufmann, Ingrid 2004 Medium und Reflexiv. Eine Studie zur Verbsemantik. Tübingen: Niemeyer. Keenan, Edward L. 1980 Passive is phrasal (not sentential or lexical). In: Teun Hoekstra, Harry van der Hulst and Michael Moortgat (eds.), Lexical Grammar, 181−213. Dordrecht: Foris. Kim, Hee-Soo and Acrisio Pires 2003 Ambiguity in the Korean morphological causative/passive. In: William McClure (ed.), Japanese/Korean Linguistics 12, 255−266. Stanford: CSLI. Kimenyi, Alexandre 1980 A Relational Grammar of Kinyarwanda. Berkeley: University of California Press. Kiparsky, Paul 1992 Structural case. Ms. Berlin: Institute for Advanced Studies. Kiyosawa, Kaoru 2006 Applicatives in Salish languages. Ph.D. dissertation, Simon Fraser University, Burnaby, BC, Canada. Kiyosawa, Kaoru and Donna B. Gerdts 2010 Salish Applicatives. Leiden: Brill. Koontz-Garboden, Andrew 2009 Anticausativization. Natural Language & Linguistic Theory 27: 77−138. Krämer, Martin and Dieter Wunderlich 1999 Transitivity alternations in Yucatec, and the correlation between aspect and argument roles. Linguistics 37: 431−479. Kratzer, Angelika 1996 Severing the external argument from its verb. In: Johan Rooryck and Laurie Ann Zaring (eds.), Phrase Structure and the Lexicon, 109−137. Dordrecht: Kluwer. Lamoureaux, Siri van Dorn 2004 Applicative constructions in Maasai. M.A. thesis, University of Oregon. Lavine, James E. 2005 The morphosyntax of Polish and Ukrainian -no/-to. Journal of Slavic Linguistics 13: 75−117. Lavine, James E. 2010 Case and events in transitive impersonals. Journal of Slavic Linguistics 18: 101−130. Levin, Beth 1993 English Verb Classes and Alternations. A preliminary investigation. Chicago: University of Chicago Press. Maling, Joan and Sigríður Sigurjónsdóttir 2002 The ‘new impersonal’ construction in Icelandic. Journal of Comparative Germanic Linguistics 5: 97−142. Marantz, Alec P. 1984 On the Nature of Grammatical Relations. Cambridge, MA: MIT Press.

1464

VII. Semantics and pragmatics in word-formation II: Special cases

Marantz, Alec P. 1993 Implications of asymmetries in double object constructions. In: Sam A. Mchombo (ed.), Theoretical Aspects of Bantu Grammar, 113−150. Stanford: CSLI. Matsumoto, Yo 1998 A reexamination of the cross-linguistic parameterization of causative predicates: Japanese perspectives. In: Miriam Butt and Tracy Holloway King (eds.), Proceedings of the LFG98 Conference. Stanford: CSLI. http://web.stanford.edu/group/cslipublications/ cslipublications/LFG/LFG3-1998/lfg98-toc.html [last access 10 Oct 2014]. Mattina, Nancy 1993 Some lexical properties of Colville-Okanagan ditransitives. In: William Seaburg (ed.), Working Papers of the 28th International Conference on Salish and Neighbouring Languages, 265−284. Seattle, WA: University of Washington. McGinnis, Martha 2001 Variation in the phase structure of applicatives. Linguistic Variation Yearbook 1: 105−146. McGinnis, Martha 2005 UTAH at Merge: Evidence from multiple applicatives. MIT Working Papers in Linguistics 49: 183−200. McGinnis, Martha 2008 Applicatives. Language and Linguistic Compass 2: 1225−1245. McGinnis, Martha and Donna B. Gerdts 2003 A phase-theoretic analysis of Kinyarwanda multiple applicatives. In: Sophie Burelle and Stanca Somesfalean (eds.), Proceedings of the 2003 Annual Conference of the Canadian Linguistic Association, 154−165. Montréal: Université du Québec. Mellenius, Ingmarie 1996 On noun-verb compounding in Swedish. Lund Working Papers in Linguistics 45: 133−149. Müller, Stefan 2003 Object-to-subject-raising and lexical rule: An analysis of the German passive. In: Stefan Müller (ed.), Proceedings of the HPSG-​2003 Conference, Michigan State University, East Lansing, 278−297. Stanford: CSLI. Muysken, Pieter 1986 Approaches to affix order. Linguistics 24: 629−643. Nedjalkov, Vladimir P. 2006 Chukchi reciprocals. In: Tasaku Tsunoda and Taro Kageyama (eds.), Voice and Grammatical Relations. In Honor of Masayoshi Shibatani, 217−246. Amsterdam/Philadelphia: Benjamins. Ørsnes, Bjarne 2006 Creating raising verbs: An LFG-analysis of the complex passive in Danish. In: Miriam Butt and Tracy Holloway King (eds.), Proceedings of the LFG06 Conference, Universität Konstanz, 386−405. Stanford: CSLI. Olsen, Susan 1997 Über den lexikalischen Status englischer Partikelverben. In: Gisa Rauh and Elisabeth Löbel (eds.), Lexikalische Kategorien und Merkmale, 45−71. Tübingen: Niemeyer. Ottosson, Kjartan 2009 The anticausative and related categories in the Old Germanic languages. Ms. University of Oslo. Parsons, Terence 1990 Events in the Semantics of English. A Study in Subatomic Semantics. Cambridge, MA: MIT Press. Polinsky, Maria 1994 Relativization in Chukchee. In: Howard Aronson (ed.), Linguistic Studies in the NonSlavic Languages of the Commonwealth of Independent States and the Baltic Republics, 241−262. Chicago: Chicago Linguistic Society.

83. Valency-changing word-formation

1465

Polinsky, Maria 2011a Antipassive constructions. In: Matthew S. Dryer and Martin Haspelmath (eds.), The World Atlas of Language Structures Online. München: Max Planck Digital Library, chapter 108. Available online at http://wals.info/chapter/108 [last access 9 Nov 2011]. Polinsky, Maria 2011b Applicative constructions. In: Matthew S. Dryer and Martin Haspelmath (eds.), The World Atlas of Language Structures Online. München: Max Planck Digital Library, chapter 109. Available online at http://wals.info/chapter/109 [last access 9 Nov 2011]. Polinsky, Maria and Isaac Kozinsky 1992 Ditransitive constructions in Kinyarwanda: Coding conflict or syntactic doubling? Papers from the 28th Annual Meeting of the Chicago Linguistic Society, 426−442. Chicago: Chicago Linguistic Society. Pylkkänen, Liina 2008 Introducing Arguments. Cambridge, MA: MIT Press. Reinhart, Tanya and Tal Siloni 2005 The lexicon-syntax parameter: Reflexivization and other arity operations. Linguistic Inquiry 36: 389−436. Rojina, Nina 2004 English particles, Russian prefixes, and prepositional phrases. M.A. thesis, University of Tromsø. Schäfer, Florian 2009 The causative alternation. Language and Linguistics Compass 3: 641−681. Schein, Barry 1993 Plurals and Events. Cambridge, MA: MIT Press. Siewierska, Anna 2011 Passive constructions. In: Matthew S. Dryer and Martin Haspelmath (eds.), The World Atlas of Language Structures Online. München: Max Planck Digital Library, chapter 107. Available online at http://wals.info/chapter/107 [last access 9 Nov 2011]. Sobin, Nicholas J. 1985 Case assignment in Ukrainian morphological passive constructions. Linguistic Inquiry 16: 649−662. Song, Jau Jung 2011a Nonperiphrastic causative constructions. In: Matthew S. Dryer and Martin Haspelmath (eds.), The World Atlas of Language Structures Online. München: Max Planck Digital Library, chapter 111. Available online at http://wals.info/chapter/111 [last access 9 Nov 2011]. Song, Jau Jung 2011b Periphrastic causative constructions. In: Matthew S. Dryer and Martin Haspelmath (eds.), The World Atlas of Language Structures Online. München: Max Planck Digital Library, chapter 110. Available online at http://wals.info/chapter/110 [last access 9 Nov 2011]. Spreng, Bettina 2006 Antipassive morphology and case assignment in Inuktitut. In: Alana Johns, Diane Massam and Juvenal Ndayiragije (eds.), Ergativity. Emerging Issues, 247−270. Dordrecht: Springer. Steinbach, Markus 2002 Middle Voice. A comparative study of the syntax-semantics interface of German. Amsterdam/Philadelphia: Benjamins. Stiebels, Barbara 1996 Lexikalische Argumente und Adjunkte. Zum semantischen Beitrag von verbalen Präfixen und Partikeln. Berlin: Akademie Verlag.

1466

VII. Semantics and pragmatics in word-formation II: Special cases

Stiebels, Barbara 2003 Transparent, restricted and opaque affix orders. In: Uwe Junghanns and Luka Szucsich (eds.), Syntactic Structures and Morphological Information, 283−315. Berlin: Mouton de Gruyter. Straight, H. Stephen 1976 Decompositional structure in Yucatec verbs. In: Marlys McClaran (ed.), Mayan linguistics. Vol. 1, 189−201. Los Angeles: UCLA and American Indian Studies Center. van de Kerke, Simon 1996 Affix order and interpretation in Bolivian Quechua. Ph.D. dissertation, Universiteit van Amsterdam. van Geenhoven, Veerle 1998 Semantic Incorporation and Indefinite Description. Semantic and Syntactic Aspects of Noun Incorporation in West Greenlandic. Stanford: CSLI. Van Valin, Robert D. 2007 The role and reference grammar analysis of three-place predicates. Suvremena Lingvistika 33: 31−64. WALS = Dryer, Matthew S. and Martin Haspelmath (eds.) 2011 The World Atlas of Language Structures Online. München: Max Planck Digital Library. Available online at http://wals.info/ [last access 10 Oct 2011]. Washio, Ryuchi 1995 Interpreting Voice. A case study in lexical semantics. Tokyo: Kaitakusha. Watters, David, Madhav Pokharel, Balaram Prasain and Yogendra P. Yadava 2006 Notes on Kusunda grammar: A language isolate of Nepal. Himalayan Linguistics Archive 3: 1−182. http://www.linguistics.ucsb.edu/HimalayanLinguistics/grammars/HLA03.html [last access 10 Oct 2011]. Wunderlich, Dieter 1997a Cause and the structure of verbs. Linguistic Inquiry 28: 27−68. Wunderlich, Dieter 1997b Argument extension by lexical adjunction. Journal of Semantics 14: 95−142. Wunderlich, Dieter 2000 Predicate composition and argument extension as general options. In: Barbara Stiebels and Dieter Wunderlich (eds.), Lexicon in Focus, 247−270. Berlin: Akademie Verlag. Wunderlich, Dieter 2002 On the nature of dative in Hungarian. Approaches to Hungarian 8: 161−184. Wunderlich, Dieter 2006 Towards a structural typology of verb classes. In: Dieter Wunderlich (ed.), Advances in the Theory of the Lexicon, 58−166. Berlin: Mouton de Gruyter. Wunderlich, Dieter 2012a Lexical decomposition in grammar. In: Markus Werning, Wolfram Hinzen and Edouard Machery (eds.), The Oxford Handbook of Compositionality, 307−327. Oxford: Oxford University Press. Wunderlich, Dieter 2012b Operations on argument structure. In: Claudia Maienborn, Klaus von Heusinger and Paul Portner (eds.), Semantics. An International Handbook of Natural Language Meaning. Vol. 3, 2224−2259. Berlin/Boston: De Gruyter Mouton.

Dieter Wunderlich, Berlin (Germany)

84. Word-formation and lexical aspect: deverbal verbs in Italian

1467

84. Word-formation and lexical aspect: deverbal verbs in Italian 1. 2. 3. 4. 5. 6. 7. 8.

Deverbal verbs in Italian The data Direction of derivation Classification of the data Restrictions on the input of word-formation rules: the role of actionality Restrictions on the output of word-formation rules: the role of aspect and tense Italian dialects References

Abstract Evaluative morphology has been widely described in the literature of the field. However, scholars have always neglected evaluative constructions with a verb as the base. Actually, evaluative verbs are cross-linguistically less widespread than evaluative nouns and adjectives. Moreover, even in languages in which they display a satisfactory degree of productivity (as in Italian), their occurrences are far from being homogeneous: their formation is constrained by a set of restrictions that depend on the actional characterization of the base, while their distribution is highly constrained by the context of occurrence, that is by the tense of the verb.

1. Deverbal verbs in Italian In Italian, the only way to derive complex verbs by suffixation from simple verbs is to use suffixes that show many similarities with affixes traditionally labelled as evaluative: (1)

nominal affixes armadi(o)-etto wardrobe-DIM pastor(e)-ello shepherd-DIM govern(o)-icchio government-DIM

verbal affixes fischi(are)-ettare to whistle-EVAL salt(are)-ellare to jump-EVAL mord(ere)-icchiare to bite-EVAL

In order to interpret the Italian data correctly, a brief overview of Italian morphology is needed. Italian is a fusional language, with a rich inflectional and derivational morphology. As for verbs, the infinitive form, ending in -are, -ere, or -ire, is conventionally used to represent the whole paradigm (so, in a suffix as -ettare, -are is the inflectional ending of the infinitive). In nominal morphology, the singular is the headword: the inflectional ending is -a for the feminine, and -o for the masculine (with possible exceptions, of course). A third class is widely productive, that of nouns ending in -e in the singular: it

1468

VII. Semantics and pragmatics in word-formation II: Special cases

includes both masculine and feminine nouns (as pastore in (1) and volpe ‘fox’ respectively). Going back to the data in (1), while most nominal evaluative suffixes can be univocally classified with respect to their meaning (i.e. diminutive, augmentative, and pejorative), verbal evaluative suffixes cannot be classified in a unique way, since their meaning is often a “cluster” of different semantic nuances (attenuation, iteration, superficiality, etc.; and this explains why I use a generic abbreviation such as EVAL ‘evaluative suffix’ in the glosses above). Verbal evaluative constructions have been almost systematically ignored by scholars. This situation is probably a consequence of their being cross-linguistically less diffused than nominal (and adjectival) evaluative constructions. It also has to do with the fact that even in languages which make quite extensive use of verbal evaluative affixes (Romance languages above all), these are considerably less widespread and productive than nominal ones. The issue is even more pressing if one considers that verbal evaluative affixes are often identical to nominal ones, as shown by the data in (1). The reason for this disparity in frequency and productivity is the quite complex set of restrictions that constrain the application of verbal suffixes and their occurrence. As the data I present in this article will demonstrate, the restrictions on the input and the output of the wordformation processes under examination are largely based on actionality (which, in this article, is used as a synonym for lexical aspect and aktionsart) and aspect, respectively.

2. The data The research on which this article is based covers a sample of approximately 300 complex verbs in Italian (that is, verbs containing an “evaluative” suffix) that can be traced back to simple verbs and that have been extracted from two dictionaries of contemporary Italian (GRADIT edited by Tullio De Mauro and DISC edited by Francesco Sabatini and Vittorio Coletti). The data presented in this article are based on the analysis of a representative subset of approximately 150 complex verbs and 80 simple verbs, listed below: (2)

Base verbs: ammontare ‘to amount (to)’ annaspare ‘to grope about’ arrangiare ‘to arrange, to manage’ avvolgere ‘to wind, to roll up’ baciare ‘to kiss’ balzare ‘to jump’ beccare ‘to peck’ beffare ‘to mock, to trick’ bere ‘to drink’

braccare ‘to hunt’

Deverbal verbs: ammonticchiare annaspicare arrangicchiare avvoltolare baciucchiare, sbaciucchiare balzellare, sbalzellare becchettare, beccolare, beccuzzare, sbecchettare beffeggiare, sbeffeggiare bevacchiare, bevazzare, bevicchiare, bevucchiare, sbevacchiare, sbevazzare, sbevicchiare, sbevucchiare braccheggiare

84. Word-formation and lexical aspect: deverbal verbs in Italian bruciare ‘to burn’ bucare ‘to hole’ cagare/cacare ‘to shit’ campare ‘lo live, to scrape along’ cantare ‘to sing’ comprare ‘to buy’ copiare ‘to copy’ costare ‘to cost’ dormire ‘to sleep’ fischiare ‘to whistle’ forare ‘to make a hole in’ frugare ‘to rummage’ fumare ‘to smoke’

giocare ‘to play’ girare ‘to ramble, to wander’ gridare ‘to shout’ guadagnare ‘to earn’ guaire ‘to yelp’ guardare ‘to look’ guidare ‘to drive’ imparare ‘to learn’ inciampare ‘to trip’ insegnare ‘to teach’ lavare ‘to wash’ lavorare ‘to work’ leccare ‘to lick’ leggere ‘to read’ macchiare ‘to stain’ mangiare ‘to eat’ mordere ‘to bite’ palpare ‘to feel, to touch’ parlare ‘to speak’ pelare ‘to peel, remove hair/fur from’ pennellare ‘to paint’ piacere ‘to like’ piangere ‘to cry’ picchiare ‘to beat’ piegare ‘to bend, to fold’ piovere ‘to rain’

abbruciacchiare, bruciacchiare, sbruciacchiare bucherellare scacazzare, scagazzare campicchiare canterellare, canticchiare compricchiare scopiazzare costicchiare, costucchiare dormicchiare, sdormicchiare fischiettare foracchiare, sforacchiare frugacchiare, frughicchiare, frugolare fumacchiare, fumazzare, fumeggiare, fumicchiare, sfumacchiare, sfumazzare, sfumicchiare giocherellare, giochicchiare girellare, gironzolare gridacchiare guadagnucchiare guaiolare guardicchiare, guarducchiare, sguardicchiare, sguarducchiare guidacchiare imparacchiare, imparicchiare, imparucchiare inciampicare insegnicchiare, insegnucchiare lavicchiare lavoracchiare, lavoricchiare, slavoricchiare sleccazzare leggicchiare, leggiucchiare macchiettare mangicchiare, mangiucchiare, smangiucchiare mordicchiare palpeggiare parlacchiare, parlicchiare, parlottare, parlucchiare spelacchiare pennelleggiare piacicchiare, piaciucchiare piagnucolare, piangiucchiare picchierellare, picchiettare pieghettare, spiegazzare piovicchiare, pioviccicare, piovigginare

1469

1470

VII. Semantics and pragmatics in word-formation II: Special cases puzzare ‘to stink’ ridere ‘to laugh’ rigare ‘to rule, to scratch’ riposare ‘to rest’ rodere ‘to gnaw’ rubare ‘to steal’ saltare ‘to jump’ sballottare ‘to jolt (about)’ sbavare ‘to drool’ scherzare ‘to joke’ sciupare ‘to spoil, to waste’ scoppiare ‘to explode’ scrivere ‘to write’ sfottere ‘to make fun of ’ sparare ‘to shoot’ spelare ‘to remove the hair/fur from’ spendere ‘to spend’ spennare ‘to pluck’ spezzare ‘to break’ sputare ‘to spit’ stentare ‘to be hard up’ stirare ‘to stretch’ studiare ‘to study’ sudare ‘to sweat’ suonare ‘to play’ tagliare ‘to cut’ tastare ‘to feel, to touch’ tossire ‘to cough’ tremare ‘to tremble’ trottare ‘to trot’ vendere ‘to sell’ vivere ‘to live’ volare ‘to fly’

puzzacchiare, puzzicchiare ridacchiare righettare riposicchiare rosicchiare rubacchiare, rubicchiare saltellare, salterellare, salticchiare sballottolare sbavazzare scherzeggiare sciupacchiare scoppiettare scribacchiare, scrivacchiare, scrivicchiare, scrivucchiare sfotticchiare sparacchiare spelacchiare spendacchiare, spendacciare, spendicchiare, spenducchiare spennacchiare spezzettare sputacchiare, sputazzare stentacchiare stiracchiare studiacchiare, studicchiare sudacchiare suonacchiare, suonicchiare tagliuzzare tasteggiare tossicchiare tremolare trotterellare vendicchiare vivacchiare, vivicchiare, vivucchiare volacchiare, volicchiare, svolacchiare, svolazzare

The list of all the verbs taken from the two dictionaries mentioned above includes very frequent forms like mangiucchiare ‘to nibble’ or saltellare ‘to trip, to hop’ as well as very uncommon and out-of-date forms like ammalazzarsi ‘to be sickly’. In the subsample of 150 verbs only the more frequent forms are included. Of course, it is necessary to explain clearly the criteria according to which this selection has been made. As it is well-known, dictionaries always outline a static image of a language. So, in order to gather the real, actual behaviour of evaluative verbs in contemporary Italian, a sample of occurrences of these verbs has been selected at a first stage from the corpus CORIS (cf. Rossini Favretti 2000), a representative corpus of contemporary written Italian. However, the written variety of a language does not represent a suitable environment for

84. Word-formation and lexical aspect: deverbal verbs in Italian

1471

evaluative verbs, since they still remain confined to a few spoken varieties of the language usually used in informal situations. Therefore, occurrences taken from the corpus CORIS have been complemented with occurrences from the web, accessed by using the search engine Google. The web represents the simplest way to have access to a large number of texts that, even in written form, reproduce a linguistic use close to the orality pole of the diamesic (written vs. spoken) dimension (cf. Hathout, Montermini and Tanguy 2008). However, the use of the web as a source of data for linguistic investigations is not free of problems. For a survey of the main drawbacks, see Baroni and Bernardini (2006), Baroni and Ueyama (2006) and Hathout, Montermini and Tanguy (2008). One of these problems regards the well-known “changeableness” and “instability” of the web. Is it well-known that databases on which commercial search engines rely are constantly updated, always according to criteria that do not correspond to what linguists may require in their investigations. Moreover, criteria that are used in order to update databases are not always made explicit. As a result, linguistic tendencies that emerge from a survey of the web can be trusted only if they are confirmed or strengthened in an adequate lapse of time. Actually, one can easily predict that the more time that passes between two queries, the bigger the difference between the results will be, since the update of search engines’ databases is constant, but slow. The most logical solution to this problem is to repeat the query after an adequate period of time and to compare the results. The data to be discussed in the following sections are the outcome of a three-year research project, so the query has been repeated three times in three years: in September 2004; in October 2005 and in November 2006. In the first query, I collected and checked the occurrences of all 300 evaluative verbs taken from dictionaries and deleted from the database all verbs with less than 50 occurrences on the web. Then, I repeated the queries twice for a sub-sample of evaluative verbs, in order to understand whether the tendencies revealed by the first query were confirmed or not: as we will see in the next sections, the answer to this question is affirmative. Going back to the verbs extracted from the two dictionaries and then selected according to their frequency in the corpus CORIS and on the web, we obtain the following “evaluative” suffixes: (3)

-acchiare -ecchiare -icchiare -ucchiare -azzare -ezzare -uzzare -eggiare -(er)ellare -ettare -ottare -icare -igginare -olare -ucolare

1472

VII. Semantics and pragmatics in word-formation II: Special cases

They represent the inventory of verbal deverbal suffixes attested in contemporary Italian with varying degrees of productivity and frequency. Before analysing the data in detail, and before facing the issue of restrictions on the input and the output of these wordformation processes, it is necessary to deal with two problems posed by the data.

3. Direction of derivation The derivational history of some of the complex verbs under examination is far from unambiguous. Consider the case of fischiettare ‘to whistle thoughtlessly, in a happy-golucky manner’: according to GRADIT, it derives from the verb fischiare ‘to whistle’ by adding the verbal evaluative suffix -ettare: fischi(are)-ettare. However, it can also be analysed in two other ways: fischiettare might be the effect either of a conversion from the diminutive noun fischietto (fischi(o)-etto ‘small whistle’) or of the addition of the suffix -ettare to the noun fischio ‘whistle’. On the other hand, a verb like sputacchiare ‘to spit out, to splutter’ is described by GRADIT as a conversion from the diminutive/ pejorative noun sputacchio (‘spittle’); but we cannot exclude that it contains a suffix -acchiare joined to the noun sputo ‘spit’ (sput(o)-acchiare) or to the verb sputare ‘to spit’ (sput(are)-acchiare). Upon this background, one cannot avoid wondering what the actual base of these forms and of the other complex verbs is. In order to answer this question, it is necessary to make a brief remark about the history of Italian verbal evaluative suffixes. Most of them derive from nominal derivatives in Latin which underwent a process of conversion: (4)

TUSSIS]N

‘cough’ The form (5)

→

TUSS+ĬCŬLA]N DIM

‘little cough’

TUSSĬCULARE

TUSSĬCŬL+ARE

→

TUSSĬCŬL+ARE]V

‘to give a little cough’

was then reanalysed as follows:

> TUSS+ĬCŬLARE

giving rise to the new verbal evaluative suffix -ĬCŬLARE (> It. -icchiare). Verbs formed by conversion from evaluative nouns are widely diffused in the history of Italian morphology, uninterruptedly and with a constant productivity (cf. a very recent formation as spintaN ‘push’ → spintoneN AUG ‘shove’ → spintonareV ‘to shove’). Conversion, along with the reanalysis that often accompanies it, has the consequence of weakening the boundary between evaluative nouns and the corresponding verbs, of forming strong pairs of suffixes such as -icchio/-icchiare, -etto/-ettare, etc., and, finally, of making a verbal counterpart virtually available for most evaluative nouns. But, on the other hand, a nominal counterpart becomes virtually available for most evaluative verbs too (through backformation). Thus, this process produces the side effect of obscuring the direction of derivation and of creating a split between the formal and semantic component of these word-formation rules. Consider the emblematic case of ballettare, the base of which, according to GRADIT, is the noun balletto ‘ballet’ (that is: ballett(o)-are by conversion), and not the verb ballare ‘to dance’ (that is: ball(are)-ettare). Nevertheless, the meaning of this complex

84. Word-formation and lexical aspect: deverbal verbs in Italian

1473

Tab. 84.1: Formal representation of the Italian verb-formation process Category of the base X

N V

Suffix etto

are

ettare

Category of derived word

Meaning

V

‘to use Xetto, to make Xetto, to produce Xetto …’

V

‘to do X in a superficial, careless manner …’

verb is ‘to dance in an awkward, clumsy manner’, and not ‘to perform ballet (without grace and elegance)’, as we would have expected it to mean if its base had been balletto. In other words, there are no traces of the meaning of balletto in the meaning of ballettare. So, although from a diachronic perspective ballettare is probably the effect of a conversion from balletto, in a synchronic perspective it is always re-interpreted and used as if it were derived from ballare. More specifically, ballettare has undergone a formal and semantic reanalysis: ballett-are > ball-ettare. So it is evident that speakers feel the strong link between the members of pairs of suffixes as -etto/-ettare, and, as a consequence, apply reanalysis whenever it is formally possible. This situation can be represented as in Table 84.1 above. The crucial point of the issue is that Italian speakers easily move from one derivational structure to the other, irrespective of the original attestation. If these are the premises, it is almost impossible, even pointless, to answer the question posed above, since there is not a single answer, but many answers that could be applied to each and every ambiguous case. Furthermore, these answers would be interesting for a statistical or historical investigation, but meaningless when the intention is to understand the rule that is synchronically perceived by speakers. In this picture, the most convincing solution is probably to assert that the formation of Italian deverbal verbs is largely conditioned by analogy, that is by abstract schemas that speakers gather from sets of actual words such as ballo ‘dance’ / ballare ‘to dance’ / balletto ‘ballet’ / ballettare ‘to perform ballet without grace and elegance / to dance in an awkward, clumsy manner’: N / V / EVAL N / EVAL V. It is plausible that speakers perceive each set of words as a whole, since the course of the individual derivational paths is tarnished by reanalysis processes. Consequently, I will not draw a distinction between ambiguous forms like ballettare and clear cases like mangiucchiare ‘to nibble’ or svolazzare ‘to flutter’. In other words, my analysis will cover every complex verb in whose structure a simple verb is recognizable (irrespective of the etymology recorded in dictionaries of contemporary Italian), and that can be replaced in all occurrences (or in most occurrences) by the simple verb without compromising the grammaticality of the sentence.

4. Classification of the data Basing my analysis on the occurrences taken from the corpus CORIS and the web, the 150 verbs composing the sub-sample have been classified according to their argument structure following a schema proposed by Jezek (2003), as in (6). Then, they have been arranged in tables as shown in 84.2.

1474

VII. Semantics and pragmatics in word-formation II: Special cases

Tab. 84.2: Classification of data V TR

V INTR AV

picchiare ‘to beat’

X

X

picchierellare ‘to tap’

X

X

picchiettare ‘to tap’

X

X

suonare ‘to play’

X

X

suonacchiare ‘to play without much thought’

X

X

suonicchiare ‘to play without much thought’

X

X

(6)

V INTR ES

V INTR PRON

X

transitive verbs (V TR), intransitive verbs with auxiliary avere ‘to have’ (V INTR AV), intransitive verbs with auxiliary essere ‘to be’ (V INTR ES), intransitive pronominal verbs (V INTR PRON − PRON is usually a reflexive pronoun);

These tables serve as the basis for my analysis, the general results of which are presented in Grandi (2007, 2008). In the next sections of this article I will focus on two specific issues: the interaction between verbal evaluative suffixes and actionality (concerning the restrictions on the input of word-formation processes under examination) and the interaction between verbal evaluative suffixes and aspect (concerning the output of the wordformation processes under consideration).

5. Restrictions on the input of word-formation rules: the role of actionality According to Jezek (2003), Italian verbs are distributed among 15 classes with respect to their argument structures: (7)

1 V TR (e.g., abolire ‘to abolish’) 2 V INTR AV (e.g., russare ‘to snore’) 3 V INTR ES (e.g., cadere ‘to fall’) 4 V INTR PRON (e.g., pentirsi ‘to regret’) 5 V INTR AV and INTR ES (e.g., squillare ‘to ring’) 6 V INTR AV and INTR PRON (e.g., approfittare ‘to take advantage of ’) 7 V INTR ES and INTR PRON (e.g., ammuffire ‘to mould’) 8 V INTR AV and INTR ES and INTR PRON (e.g., sedimentare ‘to sediment’) 9 V TR and INTR AV (e.g., mangiare ‘to eat’) 10 V TR and INTR ES (e.g., affondare ‘to sink’) 11 V TR and INTR PRON (e.g., alzare ‘to lift’)

84. Word-formation and lexical aspect: deverbal verbs in Italian 12 13 14 15

V V V V

1475

TR and INTR AV and INTR ES (e.g., continuare ‘to continue’) TR and INTR AV and INTR PRON (e.g., chiudere ‘to close’) TR and INTR ES and INTR PRON (e.g., ingiallire ‘to turn yellow’) TR and INTR AV and INTR ES and INTR PRON (e.g., bruciare ‘to burn’)

If one compares the argument structures of bases and derived words, one may first note that in the selection of their domain, verbal evaluative suffixes show a strong preference for verbs which are both transitive and intransitive and which select avere as auxiliary (that is, verbs which alternate between a transitive and an unergative use; baciare ‘to kiss’, scrivere ‘to write’, etc.; about 50 % of the base verbs are members of this class); and, secondly, that an evaluative suffix often causes an increase in the class of intransitive verbs with the auxiliary avere (cf. line 2 in Table 84.3); and, finally, that almost any intransitive verb with essere as an auxiliary is formed by means of an evaluative suffix. It is necessary to point out that in verbs belonging to class 9 (that is, verbs in which the transitive value and the intransitive value with the auxiliary avere alternate), the intransitive value often predominates. If one refers to Table 84.3, it becomes apparent that approximately 72 % of Italian evaluative verbs are intransitive with avere, often irrespective of the argument structure of the base. Therefore, we can lay a first cornerstone asserting that evaluative suffixes show a strong preference for intransitive verbs with avere. This issue deserves to be investigated more deeply, since it can provide

Tab. 84.3: Argument structures of base and derived verbs Argument structures

Base verbs

Derived verbs

1

V TR

7.3 % (6)

10.7 % (16)

2

V INTR AV

7.3 % (6)

21.5 % (32)

3

V INTR ES

1.2 % (1)

1.3 % (2)

4

V INTR PRON

0%

1.3 % (2)

5

V INTR AV and INTR ES

8.5 % (7)

6 % (9)

6

V INTR AV and INTR PRON

0 % (0)

0 % (0)

7

V INTR ES and INTR PRON

0 % (0)

0 % (0)

8

V INTR AV and INTR ES and INTR PRON

1.2 % (1)

0 % (0)

9

V TR and INTR AV

48.8 % (40)

50.3 % (75)

10

V TR and INTR ES

1.2 % (1)

1.3 % (2)

11

V TR and INTR PRON

4.9 % (4)

4 % (6)

12

V TR and INTR AV and INTR ES

4.9 % (4)

0.7 % (1)

13

V TR and INTR AV and INTR PRON

12.2 % (10)

0.7 % (1)

14

V TR and INTR ES and INTR PRON

0 % (0)

0 % (0)

15

V TR and INTR AV and INTR ES and INTR PRON

2.4 % (2)

2 % (3)

1476

VII. Semantics and pragmatics in word-formation II: Special cases

promising hints concerning the restrictions on verbal evaluative suffixes. Upon this background, we should turn our attention to actionality. It is not easy to draw a uniform “actional” description of base verbs belonging to classes 2 and 9 in (7) (V INTR AV and V TR / V INTR AV), but a clear tendency emerges: very often they designate events with an atelic, dynamic or durative nuance. Therefore, the following picture can be outlined: (8)

Actional characteristic Durative verbs

Punctual verbs Dynamic verbs

Stative verbs Atelic verbs

Telic verbs

Evaluative morphology Example: Yes dormire ‘to sleep’ → dormicchiare ‘to snooze, to doze’ No esplodere ‘to explode’ Yes trottare ‘to trot’ → trotterellare ‘to scamper, to toddle’ No credere ‘to believe’ Yes cantare ‘to sing’ → canticchiare ‘to sing softly, to hum’ No morire ‘to die’

Thus, the application of verbal evaluative suffixes seems to be constrained by a set of semantic restrictions that reduce their domain of application considerably. In conclusion, I claim that the issue of restrictions on evaluative verbal suffixes can be best approached by adopting a method which integrates the syntactic level and the actional-semantic considerations. If this perspective is adopted, then an interesting tendency emerges: on formal-syntactic grounds, the evaluative suffixes exhibit a clear preference for unergative constructions, while on semantic grounds they prefer atelic, durative and dynamic verbs.

6. Restrictions on the output of word-formation rules: the role of aspect and tense The restrictions referring to the unergativity and to the atelic, durative and dynamic character of the base verbs have the effect of reducing the domain of verbal evaluative suffixes, but they do not explain why even evaluative verbs with an unergative and dynamic/atelic/durative base often show a low and heterogeneous diffusion. In fact, what is really surprising is that the degree of the acceptability of evaluative verbs exhibits a considerable range of variation according to the syntactic context, as, for example, when the tense of the verb changes, as confirmed by data in (9), which are the outcome of the first web and corpus search (Sept. 2004 − the same holds for data in 10; the third person plural has been chosen just because it usually does not show instances of homography; this does not mean that there is a particular correlation between the use of evaluative verbs and the third person):

84. Word-formation and lexical aspect: deverbal verbs in Italian (9)

1477 Occurrences

canticchiare ‘to hum, to sing softly’ canticchiano 3rd PERSON PLURAL − PRESENT canticchiarono 3rd PERSON PLURAL − REMOTE PAST hanno canticchiato 3rd PERSON PLURAL − PRESENT PERFECT canticchiavano 3rd PERSON PLURAL − IMPERFECT

543 62 210 476

mangiucchiare ‘to nibble, to have a bit’ mangiucchiano 3rd PERSON PLURAL − PRESENT mangiucchiarono 3rd PERSON PLURAL − REMOTE PAST hanno mangiucchiato 3rd PERSON PLURAL − PRESENT PERFECT mangiucchiavano 3rd PERSON PLURAL − IMPERFECT

834 15 36 255

parlottare ‘to whisper, to mutter’ parlottano 3rd PERSON PLURAL − PRESENT parlottarono 3rd PERSON PLURAL − REMOTE PAST hanno parlottato 3rd PERSON PLURAL − PRESENT PERFECT parlottavano 3rd PERSON PLURAL − IMPERFECT

776 79 139 1,060

saltellare ‘to trip, to skip, to hop’ saltellano 3rd PERSON PLURAL − PRESENT saltellarono 3rd PERSON PLURAL − REMOTE PAST hanno saltellato 3rd PERSON PLURAL − PRESENT PERFECT (and sono saltellati saltellavano 3rd PERSON PLURAL − IMPERFECT

15,700 294 75 16) 1674

The presence of an evaluative suffix seems to be more acceptable in some verb forms than in others. More specifically, the imperfect seems to be the most suitable environment for using a verb containing an evaluative suffix. On the other hand, the remote past and the present perfect show a strong dispreference for them. Of course, such statements must be supported by a comparison with the occurrences of the base verbs: (10)

Occurrences cantare ‘to sing’ cantano 3rd PERSON PLURAL − PRESENT cantarono 3rd PERSON PLURAL − REMOTE PAST hanno cantato 3rd PERSON PLURAL − PRESENT PERFECT cantavano 3rd PERSON PLURAL − IMPERFECT

403,000 20,400 57,800 138,000

mangiare ‘to eat’ mangiano 3rd PERSON PLURAL − PRESENT mangiarono 3rd PERSON PLURAL − REMOTE PAST hanno mangiato 3rd PERSON PLURAL − PRESENT PERFECT mangiavano 3rd PERSON PLURAL − IMPERFECT

526,000 33,900 38,500 111,000

parlare ‘to speak’ parlano 3rd PERSON PLURAL − PRESENT parlarono 3rd PERSON PLURAL − REMOTE PAST hanno parlato 3rd PERSON PLURAL − PRESENT PERFECT parlavano 3rd PERSON PLURAL − IMPERFECT

1,880,000 73,700 347,000 386,000

1478

VII. Semantics and pragmatics in word-formation II: Special cases saltare ‘to jump’ saltano 3rd PERSON PLURAL − PRESENT saltarono 3rd PERSON PLURAL − REMOTE PAST hanno saltato 3rd PERSON PLURAL − PRESENT PERFECT (and sono saltati saltavano 3rd PERSON PLURAL − IMPERFECT

274,000 19,300 845 16,800) 31,900

However, if one checks the distribution of the occurrences of the third person plural in the present, remote past and present perfect of base verbs, no clear tendency emerges. On the other hand, if we focus on the imperfect and present perfect of evaluative verbs (the two more productive past tenses in contemporary Italian; for a survey of the Italian verbal system cf. Bertinetto 1986), the proportion between them is unambiguous: on average, we observe a ratio of one present perfect to more than 16 forms of imperfect. (11)

canticchiare 1 / 7,7

vs.

p.p / impf.

cantare 1 / 2,4

mangiucchiare 1 / 7,08

vs.

p.p / impf.

mangiare 1 / 2,3

parlottare 1 / 7,6

vs.

p.p / impf.

parlare 1 / 1,11

saltellare 1 / 18,4

vs.

p.p / impf.

saltare 1 / 1,8

Therefore, what seems to emerge is a clear preference of evaluative verbs for tenses with an imperfective characterization. Of course, in order to confirm such a tendency, it is indispensable to test it over an extended period of time, repeating the query, since, as stated above, this is the only way to cushion the effects of the well-known instability of the web, which suffers from a continuous updating of the search engines’ databases. As to this point, it is clear that the tendency pointed out above would be strengthened further if it were carried out by means of two different queries with an adequate lapse of time between them. The claims presented above would be corroborated if, in these circumstances, the same result is obtained, or if the result of the second query strengthens the position of imperfect vis-àvis present perfect and remote past. As I stated above, the data discussed in this article have been collected over the course of a three-year research project: the investigation using Google has been repeated three times (in September 2004; in October 2005 and in November 2006). A comparison between the results of the first and the third query reveal that the preference for tenses with an imperfective nuance exhibited by evaluative verbs is gaining force: (12) canticchiano canticchiarono hanno canticchiato canticchiavano

Nov. 2006 5,360 193 196 1,810

Sept. 2004 543 62 210 476

84. Word-formation and lexical aspect: deverbal verbs in Italian

1479

mangiucchiano mangiucchiarono hanno mangiucchiato mangiucchiavano

1,170 38 63 295

834 15 36 255

parlottano parlottarono hanno parlottato parlottavano

6,050 551 135 3,740

776 79 139 1,060

19,100 1,040 118 6 6,180

15,700 294 75 16) 1,674

saltellano saltellarono hanno saltellato (and sono saltellati saltellavano

The data confirm the conclusion presented above: in the space of three years there has been a remarkable growth of imperfect and a corresponding decrease in both the past perfect and remote past tenses with evaluative suffixes. To pick up the thread of the argument again, the reasons behind the correlation between imperfective tenses and verbal evaluative suffixes must be sought in their meaning. A survey of the semantic readings of the 150 verbs under consideration listed in (2) reveals that some features appear to be constant: superficiality, habituality, iteration, and attenuation. That is, evaluative suffixes usually indicate an action which is performed with superficiality, which recurs iteratively, the effects of which are attenuated, etc. Some of these features (above all iteration and habituality) are very close to some values often associated with the imperfective aspect (in boldface in the following schema): (13)

Classification of aspectual oppositions (Comrie 1976: 25)

Perfective

Imperfective

Habitual

Non-progressive

Continuous

Progressive

Therefore, the tendency pointed out above is not surprising: because of the semantic closeness between imperfective aspect and the meaning of the so-called verbal evaluative suffixes, one can easily foresee that the “acceptability rate” of evaluative suffixes will be higher with verbal forms usually associated with imperfective aspect than with verbs forms that have a perfective nuance.

1480

VII. Semantics and pragmatics in word-formation II: Special cases

7. Italian dialects As it is well known, even if in the last 150 years the number of native speakers of Standard Italian has progressively increased, Italy still displays a high degree of linguistic diversity. Dialects are currently used, especially, but not exclusively in the Southern regions of Italy, even if their occurrence is strictly limited to informal situations (within the family, the peer group, etc.) and to the spoken variety. As mentioned above, these are just the situations in which evaluative affixes tend to occur with the highest frequency. So, one can wonder whether the situation observed for Standard Italian can be confirmed by the dialects or not. The problem is that dialects are not as well documented as Standard Italian: there are no representative corpora or databases available. And the dialect atlases usually do not cover the area of derivational morphology. So, data can be collected only by means of questionnaires submitted to native speakers. In the last section of this article I present some provisional results of a survey of verbal evaluative morphology in some dialects of Italy. The data have not been collected in a systematic way, creating a representative sample of these languages. Rather, they have been picked out depending on the availability of native speakers. So, the results presented in this section cannot represent an exhaustive sketch of the situation. Nevertheless, the data seem to suggest two generalizations. First, while nominal and adjectival evaluative morphology seem to occur frequently in almost all Italian dialects, verbal evaluative morphology is on average very rare. Second, most deverbal verbs seem to be found in Northern dialects. I’ll focus on two representative cases. Milanese displays a considerable number of deverbal verbs: (14) toccà ‘to touch’ bev ‘to drink’ saltà ‘to jump’ spuà ‘to spit’

→ → → →

toccascià ‘to touch insistently’ bevascià ‘to booze, to tipple’ saltascià ‘to skip, to hop’ spuascià ‘to spit out, to sputter’

The suffix -ascià seems to be the most productive. Quite widespread is also the suffix -attà: sgorà ‘to fly’ → sgorattà ‘to flutter, to flit’. As in Italian (cf. the cases of bere, parlare, fumare among others in (2)), different suffixes can be attached to the same base word: (15) sbragà ‘to shout’ → sbragalà, sbragagnà, sbragascià ‘to shout repeatedly and in a chaotic way’ The formal and semantic behaviour of these forms do not exhibit significant differences from Standard Italian verbs: the restriction to atelic, durative and dynamic verbs seems to be confirmed. Of course much work is needed in order to confirm also the restrictions on the output (that is, the preference of deverbal verbs towards tenses with an imperfective nuance).

84. Word-formation and lexical aspect: deverbal verbs in Italian

1481

All these verbs can be literally translated in Italian by deverbal verb: (16) Milanese toccascià bevascià saltascià sgorattà

Italian stoccazzare sbevazzare saltellare svolazzare

As already mentioned, the diffusion and the productivity of these affixes gradually decrease moving from Northern to Southern Italy. In Central and Southern dialects these forms are almost absent. What is interesting is that some verbs that apparently display evaluative affixes are attested. In some Abruzzese dialects, for example, we find zumbettejä/zumbéttià (the double spelling evidently depends on the lack of a common rule in writing dialects) ‘to trip, to hop’, pentecchiä ‘to spot, to speckle’, pezzettejä ‘to divide into small pieces’. But these forms are derived from “evaluative nouns” by conversion: zumbâtte ‘small jump’, pentâcchie ‘small mark, spot’ and pezzâtte ‘small piece’. So, the formation of an “evaluative verb” must be mediated by an evaluative noun; pairs V / EVAL V such as Italian mangiare ‘to eat’ / mangiucchiare ‘to nibble’ are not attested. As shown above, actual evaluative verbal suffixes are usually the outcome of the reanalysis of verbs formed by conversion from an evaluative noun (cf. the Latin forms in (4) and (5)) and of the consequent analogical extension of the suffix. We can assert that in Central and Southern dialects this process has not yet taken place, even if the necessary conditions are present. On the contrary, in some Northern dialects, as in Standard Italian, this process has already come to its end, giving rise to autonomous verbal suffixes. Since the data on this issue are still provisional and have not been collected through a systematic investigation, it is not possible to put forward an explanation for the irregular distribution of evaluative verbs in the dialects. In Central and Southern dialects, when the “evaluative form” of a verb is needed, usually an evaluative suffix is attached to the corresponding action or event noun. So, for example, if a sentence like non ho mangiato, ho mangiucchiato ‘I did not eat, I nibbled’ were to be translated into an Abruzzese dialect, a sentence such as n’ zo magnatə, so fattə na magnatellə (lit. ‘I did not eat, I just had a small feed’) would be used. The expression so fattə na magnatellə (lit. ‘I had a small feed’) is equivalent to Italian ho mangiucchiato ‘I nibbled’ and can be considered as a light-verb construction, where the scope of the nominal affix (-ella in magnatella) extends over the verb.

8. References Baroni, Marco and Silvia Bernardini (eds.) 2006 Wacky! Working Papers on the Web as Corpus. Bologna: Gedit. Baroni, Marco and Motoko Ueyama 2006 Building general- and special-purpose corpora by Web crawling. In: Proceedings of the 13 th NIJL International Symposium, Language Corpora. Their Compilation and Application, 31−40. Tokyo: NIJL. Bertinetto, Pier Marco 1986 Tempo, aspetto e azione nel verbo italiano. Il sistema dell’indicativo. Firenze: Accademia della Crusca.

1482

VII. Semantics and pragmatics in word-formation II: Special cases

Comrie, Bernard 1976 Aspect. Cambridge: Cambridge University Press. DISC = Sabatini, Francesco and Vittorio Coletti (eds.) 1999 DISC − Dizionario italiano Sabatini Coletti. 2nd ed. Firenze: Giunti. GRADIT = De Mauro, Tullio (ed.) 1999 Grande dizionario italiano dell’uso. Torino: Utet. Grandi, Nicola 2007 I verbi valutativi in italiano tra azione e aspetto. Studi di Grammatica Italiana 24: 153− 188. Grandi, Nicola 2008 I verbi deverbali suffissati in italiano. Dai dizionari al Web. Cesena: Caissa Italia. Hathout, Nabil, Fabio Montermini and Ludovic Tanguy 2008 Extensive data for morphology: Using the World Wide Web. Journal of French Language Studies 18(1): 67−85. Jezek, Elisabetta 2003 Classi di verbi tra semantica e sintassi. Pisa: Edizioni ETS. Rossini Favretti, Rema 2000 Progettazione e costruzione di un corpus di italiano scritto: CORIS/CODIS. In: Rema Rossini Favretti (ed.), Linguistica e informatica. Multimedialità, corpora e percorsi di apprendimento, 39−56. Roma: Bulzoni.

Nicola Grandi, Bologna (Italy)

85. Word-formation and aspect in Samoyedic 1. 2. 3. 4. 5.

Introduction Aspectuality Tense and aspectuality in the Samoyedic languages Conclusion References

Abstract In this article I discuss the relation between aspect, tense, and derivation in the Samoyedic languages. A typical feature of these idioms is that they lack the morphological category of present tense; instead, they display a morphologically unmarked neutral tense. It is the aspectual quality of the morphologically unmarked verb that determines the actual temporal meaning of the verb and thus, of the sentence in which it occurs. Telic verbs denote events and actions of the past, whereas atelic verbs refer to present events and actions. Modifying the aspectual quality of a verb (e.g., by changing the aktionsart with derivational means) also affects the temporal reference of the verb in the neutral tense, as well as of the sentence.

85. Word-formation and aspect in Samoyedic

1483

1. Introduction The Samoyedic languages form a sub-branch of the Uralic languages. On the basis of their geographical distribution they are commonly divided into the northern and southern branch. The northern branch comprises Nenets, Enets and Nganasan. Within this branch, Nenets and Enets are (mutually) more closely related, whereas Nganasan displays several distinctive features. Of the southern Samoyedic languages nowadays only Selkup is spoken. Mator became extinct in the first half of the 19th century. This idiom is scarcely documented; therefore, I will exclude it from the description below. The other extinct language of the southern branch, Kamas, was spoken until the second half of the 20th century. The temporal and aspectual system of Kamas displays essential differences compared to the four Samoyedic languages that are still spoken today, which can be explained by the significant influence of Turkic on the Kamas language. In the following discussion the focus will be on Nenets which is spoken on the European side of the Ural and in Northern Siberia, but at certain points I will also illustrate some features of its linguistic relatives.

2. Aspectuality Aspectuality is a complex phenomenon. It is commonly known that aspect and aktionsart are interrelated, and likewise, aspectuality cannot be treated apart from the category of tense (for a detailed discussion of the topic see Kiefer 2006 and Boogart 2004). However, before examining how aspectuality is related to derivation in the Samoyedic languages, it seems expedient to recall the essential differences between tense and aspect. Comrie (1985: 9) regards tense as being related to the external temporal structure of a state of affairs, and thus, the tense system of a particular language can be regarded as a deictic system. Aspect on the contrary is not deictic, but related to the internal temporal structure of the state of affairs (see also Comrie 1976). Hence, the aspectuality of a sentence refers to its internal temporal structure and therefore, it is a category of clausal semantics. Also, we speak of the aspectual quality of verbs, which is the “contribution of the verb semantics to the internal temporal structure of the sentence” (Kiefer 2006: 26, my translation). It is typical of the Samoyedic languages that − as we will see below − the aspectuality of the sentence is highly dependent on the aspectual quality of the verb. In this sense, the Samoyedic languages can be regarded as aspect languages. A slightly more complex issue in relation to aspectuality is the distinction between aspect and aktionsart. These categories may be discerned by semantic factors on the one hand and by formal factors on the other. Aktionsart is a semantically objective category, insofar as it expresses the internal temporal intervals of a particular event or action. From a formal point of view aktionsart is always marked by morphological means. Aspect, again, reflects the individual perspective of the speaker and can be regarded as a grammatical category that may combine with aktionsart. Changing the aktionsart by derivational means is only relevant from the point of view of aspect if the temporal structure of the basic verb and thus, its aspectual quality, has been modified.

1484

VII. Semantics and pragmatics in word-formation II: Special cases

3. Tense and aspectuality in the Samoyedic languages Until present, a number of studies have been dedicated to the relationship between tense and aspect as well as to the modes of expressing aspect in particular Uralic and Samoyedic languages. The most comprehensive descriptions have been provided by Kiefer (1996, 2006), Honti (1994), and Szeverényi (2007). The majority of the papers (as, e.g., Honti 1994; Rédei 1984, etc.) mainly follow a diachronic approach seeking to answer the (main research) question of whether the Uralic proto-language should be reconstructed as a tense language or an aspect language. In this research topic the Samoyedic languages are deemed to play a key role (see, e.g., Janhunen 1982: 36). Based on data from Samoyedic languages several scholars (e.g., Bartens 1993; Janhunen 1982) have concluded that in the Uralic proto-language the marking of aspectuality was prominent at the expense of the expression of tense. Honti (1994: 60−61), on the contrary, argues that the category of tense was primary in Proto-Uralic. With respect to data from certain Uralic languages, I agree with Kiefer (2006: 197) in that all Uralic languages display the lexical category of aspect and aktionsart. At the same time Kiefer notes that this research area is scarcely investigated in Uralistics, and therefore, on the evidence base available, I can only draw some cautious conclusions. Notwithstanding this fact I can ascertain that in the majority of contemporary Uralic languages the morphological category of aspect plays a rather marginal role; these languages can be rather classified as tense languages. It is only in the Samoyedic languages that verbal aspect preponderates in the expression of temporality compared to the morphological category of tense. This, however, does not imply that in some Samoyedic languages (as, e.g., in Nganasan) there is no differentiated tense system, including (several) past tenses and a future tense (for more details see Helimski 1998 and WagnerNagy 2002).

3.1. Tense and aspect In Samoyedic languages, aspect plays a prominent role in the expression of temporal relations, which means that both systems, i.e. aspect and tense, are closely interrelated. The tense system in these languages is not complex; each of the Samoyedic languages exhibits a morphologically marked past and future tense. However, it has to be stressed that in Nenets, as well as in Enets, the future tense can be regarded as an inflectional rather than a derivational, category. In both languages, the tense marker developed from a continuative derivational suffix. A strong argument in favour of the derivational character of the suffix is the fact that while the past tense marker, as a rule, appears on the auxiliary, the future tense marker (which is the durative-continuative marker -ŋku, -ta/ -da in Nenets, and -da/-ta in Enets) is always attached to the main verb. Of course, derivational suffixes are often reanalysed as tense markers, but for the durative suffix in Nenets this process is, in my opinion, not yet accomplished. As opposed to Nenets and Enets, in Nganasan and Selkup, the future tense is an inflectional category. The examples from Tundra Nenets below illustrate a negative construction of past tense (1), and future tense (2), respectively:

85. Word-formation and aspect in Samoyedic (1)

mań sʲaxartʔ tarcʲa xawna-na sarʲo-mʔ ńi-dam-cʲ manesʔ! I never such apart.from-LOC rain-ACC NEGAUX-1sg-PST see-CN ‘I have never seen such rain before.’ (Nenâng 2005: 129)

(2)

ŋabťe-d’iʔ, ńanand’iʔ ńi-dmʔ jilʲe-ŋku-ʔ! stink-2DU.R 2DU.LOC NEGAUX-1sg live-CONT-CN ‘You are stinky; I will not live with you.’ (Ângasova 2001: 28)

1485

Example (2) clearly demonstrates that the tense category does not appear on the auxiliary but is attached to the main verb. A highly interesting temporal category in the Samoyedic languages is the morphologically unmarked (and temporally ambivalent) aorist, the so-called neutral tense. This category cannot be regarded as referring to the present tense, but rather to the moment that precedes the utterance. The actual temporal reference of the aorist (or neutral tense) is determined by the aspectual content of the verb that occurs in the utterance. The following section describes how this is manifested in the Samoyedic languages. In these languages, the basic verbs, i.e. the verbs that do not display an aktionsart suffix, as expressing aspectuality, can be grouped into the two categories of imperfective and perfective verbs. The majority of the verbs belong to the perfective group. This group comprises all action verbs (verbs describing change of state), as, e.g., Nenets to-sʲ ‘to come’, ťu-sʲ ‘to enter, go in’, and also telic verbs that express an action that has just been finished, as, e.g., sʲerta-sʲ ‘to do’, xo-sʲ ‘to find’. The imperfective group typically comprises verbs referring to a static state, comprising atelic verbs as, e.g., Nenets jilʲesʲ ‘to live’, manzara-sʲ ‘to work’, pʲina-sʲ ‘to fear, be scared’. What is the consequence of the aspectual quality (i.e. imperfectiveness and perfectiveness, respectively) for the expression of tense? In case the verbs feature one of the aforementioned tense markers (i.e. the marker of the past or of the future tense) aspectuality doesn’t play any role in the expression of tense. In the aorist tense, however, aspect is decisive, in that the actual temporal reference of the verb will be determined depending on its aspectual content. The perfective verbs (that is, mainly the telic verbs) refer to the action or event accomplished (just) before the utterance time and thus, they express past tense. Imperfective verbs, again, express present tense. This also means that no deictic temporal adverb can co-occur with the aorist form of a verb, since the aorist itself is not deictic either. The examples from Nenets below demonstrate the dichotomy: (3)

atelic verb = present jilʲe-dmɁ ‘I live.’ manzara-dmɁ ‘I work.’ pʲina-dmɁ ‘I am scared.’

telic verb = past to-dmɁ ‘I came.’ ťu-dmɁ ‘I entered (went in).’ sʲerta-dmɁ ‘I have done it.’ xo-dmɁ ‘I found.’

As previously mentioned, tense markers neutralize the lexically inherent aspectual meaning: If the verb is marked by the past tense marker (Nenets -sʲ, -zʲ, -cʲ), this triggers the meaning of a general past. In Nenets and Enets, typically, the tense is marked wordfinally, e.g., Nenets manzara-dam-zʲ ‘I worked’, xaja-dam-zʲ ‘I left’. In Selkup and Nga-

1486

VII. Semantics and pragmatics in word-formation II: Special cases

nasan, the past tense marker occurs, as a rule, preceding the personal suffixes. The past tense marker in Nganasan has the form -suə (-sɨə, -d’üə, -d’iə). It appears before the personal ending: ńilɨ-d’iə-m ‘I lived’, kotu-d’üə-ŋ ‘You killed him/her’, ŋəjbəu-d’üə-m ‘I worked’, büü-d’üə-m ‘I left’. In Nganasan, the aorist is slightly more complex. Here too, we observe the above mentioned dichotomy of the aspectual quality, but a small set of verbs display a dual behaviour: they can be both perfective and imperfective. This category is named “biaspectual verbs”. In Nganasan personal suffixes are not attached to the stem directly in the aorist tense. A so-called coaffix (an aorist marker or linking element) is attached to the stem and personal suffixes follow this coaffix. The form of the coaffix depends on the aspect of the verb. Perfective verbal stems are followed by the coaffix -Ɂə (-Ɂa, -Ɂi), and imperfective verbal stems by the coaffix -ntɨ (-ntu, -tɨ, -tu, etc.). (4)

atelic verb = ńilɨ-tɨ-m ŋojbəu-tu-m hilʲi-tɨ-m

present ‘I live.’ ‘I work.’ ‘I am scared.’

telic verb tuu-Ɂə-m ťii-Ɂə-m mɨɨ-Ɂə-m ŋəði-Ɂəm

= past ‘I came.’ ‘I entered.’ ‘I have done it.’ ‘I found.’

These are underived verbs in Nganasan and can be followed by either of the two coaffixes. One of these coaffixes implies that the verb in the aorist tense has present meaning, while in the other case the meaning is past. Thus, in this case the aorist marker actually has tense meaning, while the verbal stem has no inherent aspectual meaning. (5)

hu͡aŋku-ďa ‘to be drunk’ → hu͡aŋka-Ɂa-m ‘I have got drunk.’ hu͡aŋku-tu-m ‘I am drunk.’

(6)

sojbu-sa ‘to be heard’ → sojbu-Ɂə ‘It has been heard.’ sojbu-tu ‘It is heard.’

The above examples clearly demonstrate that aspect plays a crucial role in the expression of tense in the aorist.

3.2. Derivation and aspect In section 3.1 I discussed the interrelation of tense expression and aspect. The examples presented may suggest that in the Samoyedic languages perfective verbs do not occur with present meaning. This indeed is the case, though it does not mean that in these languages the meanings ‘come’ or ‘make’ cannot be expressed in present tense. In order to change the tense of the sentence, however, it is necessary to modify the aspectual quality of the verb. Let’s take a look at the following sentences from Nenets: (7)

xan-mɁ sʲerta-dmɁ sledge-ACC make-1sg ‘I made a/the sledge.’

85. Word-formation and aspect in Samoyedic (8)

1487

xan-mɁ sʲerta-bi-dmɁ sledge-ACC make-CONT-1sg ‘I am making a/the sledge.’ (Tereŝenko 1947: 185)

The verb in example (8) displays the -mba (-ba, -bi) durative-continuative suffix. The aspectual quality of the verb base, and thus, that of the sentence, has been changed by the suffix. While example (7) displays a telic, perfective verbal construction, in example (8) we find a verbal construction with a verb of progressive aspectual quality. In Nenets, not only the above mentioned derivational suffix may be applied in this function, but also the derivational suffixes -ŋgo (-ŋgu) and -ta (-da), which also modify the aspectual quality of the basic verb. Next I will describe the behaviour of the verbs when the derivational suffix -ŋgu is attached to them. According to Tereŝenko (1947: 186) this suffix expresses that the action or event described by the basic verb has not yet been accomplished: tola ‘(s)he read, has read sth.’, but tolaŋgu ‘(s)he is reading now’. As was explained in section 3.1 above, it is this suffix that has emerged in the formation of the future tense and is in the process of being grammaticalized from a derivational to a tense suffix. But this suffix has still preserved its quality as a derivational suffix which can be seen from the fact that it can combine with the suffix of the past tense demonstrated in the examples below: (9)

a. ťiki-m sʲerta jaɁma-w this-ACC make.CN not.can-1sg.DET ‘I can’t do this.’

Nenets

b. Ivanko sʲiɁmʲi ńada-ŋgu. Ivanko myself.ACC help-CONT.3sg ‘Ivanko (will) help me.’ c. ńenesʲanda ŋod Ivanko ńada-ŋgu-sʲ indeed too Ivanko help-CONT.3sg-PST ‘And indeed, Ivanko helped.’ (Tereŝenko 1947: 187) While (9b) can be interpreted as an utterance of future tense, the temporal reference of example (9c) is obviously past, while its aspectual quality is durative-continuative. This holds for the suffix -ta (-da) as well, which combines with some verb stems in the function of future tense marker instead of the suffix -ŋgu. To these verbs belong, e.g., ŋam-zʲ ‘to eat’ and ta-sʲ ‘to give’. The meaning of the form ŋam-da is, consequently, ‘(s)he will eat’ or ‘(s)he is eating right now’, while the form ta-ta-dmɁ stands for ‘I will give’, ‘I am giving right now’. In the Western dialects of Nenets, however, the future tense form of the last verb displays the suffix -ŋgu (see Kupriânova, Barmič and Homič 1985: 146). In Nganasan as well, there is a derivational suffix that plays an important role in expressing aspectual quality: -ntə (-ndə, -tə). This aspect marker is the most frequently used and thus the most productive verb-forming suffix in Nganasan. The derived verb is always imperfective. The verbs derived by this suffix, however, never express future

1488

VII. Semantics and pragmatics in word-formation II: Special cases

tense in Nganasan, e.g., mi-sʲi ‘to give (perfective)’ vs. mi-tə-d’i ‘to give (imperfective)’; čii-d’i ‘to go away, leave’ vs. čii-tə-sʲi ‘to go’, etc. Let us observe the following sentences: (10) a. ... nɨ-rə kolɨ-ə bɨɁə hiri-Ɂə ... woman-2sg fish-ADJ.ACC soup.ACC cook-CO.3sg ‘Your wife cooked the fish soup.’ (Tereŝenko 1979: 86)

Nganasan

b. sʲimbia hiri-tə-sɨə kolɨ Simbia cook-CONT-PST.3sg fish.ACC ‘Simbia cooked fish.’ (Data from the author’s own collection: NDCh, 2006) c. sɨtɨ kolɨ hiri-d’iə (s)he fish.ACC cook-PST.3sg ‘(S)he has cooked the fish.’ (Data from the author’s own collection: NDCh, 2006) Examples (10a) and (10c) both display a perfective verb. Consequently, both sentences reveal a past meaning, but they refer to slightly different time periods. While example (10a) refers to the near past, (10c) refers to the past in general. The temporal reference of sentence (10b) is also the general past, but its aspectual quality is imperfective.

3.3. Aktionsart The Samoyedic languages display a rich stock of verbal derivational suffixes. As I have demonstrated in the previous sections, certain derivational suffixes play an important role in the expression of tense and of aspectuality. Again, there are other suffixes with the primary function of changing the aktionsart of the verb and not its aspectual quality. Certainly, in some cases, the modification of the aktionsart triggers the modification of the aspectual quality of the verb construction, and thus, of the whole utterance as well. Anytime a perfective verb (as I mentioned earlier, this is the case with the majority of verbs) is derived by a frequentative, habituative, iterative, etc., suffix, the verb loses its original perfective meaning and gains imperfective meaning. In Nenets, Enets and Selkup this change has no further morphological consequences. In Nganasan however this operation triggers a different coaffix (linking element) that precedes the personal suffixes, i.e. the morpheme -ntu instead of -Ɂə). The following examples demonstrate this with the case of the frequentative suffix -Ɂnar. Verbs displaying this suffix describe events that are repeated within irregular intervals during a certain period of time. (11)

d’esɨ-mə büü-Ɂnar-u gərədə-gətə Nganasan father-1sg go.away-FREQ-CO.3sg town-EL ‘My father sometimes goes away from the town.’ (Data from the author’s own collection: KNT, 1994)

85. Word-formation and aspect in Samoyedic

1489

The 3Sg form of the verb büü-ďa ‘to go away’ in the neutral tense is, regularly, biiɁia. However from example (11) it can be seen that it is not the linking element -Ɂə which occurs between the verbal stem and the personal suffix, but -u, which is the allomorph of -ntu occurring after the consonant -r. This fact clearly indicates that the aspectual quality of the basic verb has changed. As for the frequentative suffix, a morpheme can be observed in Nenets as well as a linking element between the verbal stem and the personal suffix. This morpheme however doesn’t have the function of indicating the aspectual quality; instead, its occurrence can be explained by phonological reasons (for a detailed discussion see Salminen 1997 or Hajdú 1989). In Nenets the modification of the aspectual quality can only be stated on the basis of the changed aspectual quality of the utterance itself. Let’s regard the following examples: (12) toxodanna-Ɂ klassa-xad tarpo-r-ŋa-Ɂ Nenets classroom-EL exit-FREQ-CO-3pl pupil-PL ‘The pupils go out from the classroom.’ (Kupriânova, Barmič and Homič 1961: 144) (13) toxodanna-Ɂ klassa-xad tarpɨ-dɁ classroom-EL go.out-3pl.R pupil-PL ‘The pupils went out from the classroom.’ (Kupriânova, Barmič and Homič 1961: 144) While the aspectual quality of example (13) is perfective, that of example (12) is imperfective by reason of the occurrence of the derivational suffix. Not only perfective verbs can be transformed into imperfective ones, but the opposite direction is also possible, i.e. imperfective verbs can be transformed into perfective verbs. The derivational suffixes of the momentaneous verbs are applied in this function, and among them, mainly the inchoative and resultative suffixes. As we have seen in section 3.1, the aspectual quality of Nenets jilʲe-sʲ, Nganasan ńilɨ-sɨ ‘to live’ is imperfective. If we attach to the verbal base an inchoative (ingressive) suffix (Nenets -lɁ, -l(a), Nganasan -lə, -lʲi), the aspectual quality of the verb will change. This can be seen on one hand by the temporal meaning of the utterance, and on the other, by the choice of the linking element (coaffix) in Nganasan. (14) ŋerm jand’er-Ɂ ńesej-wana jilʲe-l-jadɁ Nenets north(ern) inhabitant-pl new-PROL live-INCH-3pl.R ‘The inhabitants of the North started to live in a new manner.’ (Kupriânova, Barmič and Homič 1961: 148) (15) təsʲiəðə bənd’e-Ɂ ńaagə-iɁ ńilɨ-lʲi-Ɂə ŋuəčənu now all-pl new-pl.GEN live-INCH-3pl altogether ‘And now, they all started to live well together.’ (Tereŝenko 1979: 313)

Nganasan

4. Conclusion In this article I have argued that in the Samoyedic languages aspect is primarily comparable to morphological tense markers in expressing temporal relations. The aspectual

1490

VII. Semantics and pragmatics in word-formation II: Special cases

quality of a verb can be modified by derivation. The majority of the derivational suffixes are aktionsart suffixes which modify the aspectual quality of the verb by changing the internal temporal structure of the verb semantics. Only a small sub-group of verbal suffixes modify the aspectual quality of the verb; most of the suffixes also modify the aktionsart of the verb base. I also demonstrated that Nenets and Enets display derivational suffixes that play a role in expressing tense, although they have not yet become completely grammaticalized in this function but still display some features typical of derivational suffixes.

Abbreviations CO CONT CN DET DU EL

coaffix continuative connegative determinative conjugation dual elative

FREQ INCH LOC NEGNUX PROL

r

frequentative inchoative locative negative auxiliary prolative reflexive conjugation

5. References Ângasova, Nejko Maksimovna 2001 Neneckie skazki i neneckie pesni sudbabc, ârabc. Tomsk: Izdatel’stvo Tomskogo Universiteta. Bartens, Raija 1993 Suomalais-ugrilaisten kielten tempuksista. In: Sirkka Saarinenn, Jorma Luutonen and Eeva Herrala (eds.), Systeemi ja poikkeama. Juhlakirja Alho Alhoniemen 60-vuotispäiväksi 14. 5. 1993, 21−37. Turku: Department of Finnish and General Linguistics of the University of Turku. Boogaart, Ronny 2004 Aspect and Aktionsart. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-formation. Vol. 2, 1165−1180. Berlin/New York: Mouton de Gruyter. Comrie, Bernard 1976 Aspect. An Introduction to the Study of Verbal Aspect and Related Problems. Cambridge: Cambridge University Press. Comrie, Bernard 1985 Tense. Cambridge: Cambridge University Press. Décsy, Gyula 1966 Yurak Chrestomathy. Bloomington: Indiana University Press. Fancsaly, Éva 2009 Bemerkungen zur deverbalen Verbbildung im Nenzischen. In: Éva Fancsaly (ed.), Írások Györke József és Hajdú Péter tiszteletére 2002−2007, 43−51. Pécs: Dialóg Campus. Hajdú, Péter 1989 Chrestomathia Samoiedica. 3rd ed. Budapest: Tankönyvkiadó. Helimski, Eugen 1998 Nganasan. In: Daniel Abondolo (ed.), The Uralic Languages, 480−515. London/New York: Routledge.

85. Word-formation and aspect in Samoyedic

1491

Honti, László 1994 Zur Frage nach dem Aspekt und Tempus in der uralischen Grundsprache. In: Ago Künnap (ed.), Minor Uralic Languages. Structure and Development, 48−65. Tartu: University of Tartu. Janhunen, Juha 1982 On the structure of Proto-Uralic. Finnisch-Ugrische Forschungen 44: 23−42. Kiefer, Ferenc 1996 Az igeaspektus areális-tipológiai szempontból. Magyar nyelv 92: 257−268. Kiefer, Ferenc 2006 Aspektus és akcióminőség. Különös tekintettel a magyar nyelvre. Budapest: Akadémiai Kiadó. Kiefer, Ferenc 2010 Areal-typological aspects of word-formation: The case of aktionsart-formation in German, Hungarian, Slavic, Baltic, Romani and Yiddish. In: Franz Rainer, Wolfgang U. Dressler, Dieter Kastovsky and Hans Christian Luschützky (eds.), Variation and Change in Morphology. Selected papers from the 13 th International Morphology Meeting, Vienna, February 2008, 129−148. Amsterdam/Philadelphia: Benjamins. Kiefer, Ferenc and László Honti 2003 Verbal prefixation in the Uralic languages. Acta Linguistica Hungarica 50: 137−153. Kupriânova, Zinaida, Maria Âkovlevna Barmič and Lûdmila Vasiľevna Homič 1961 Neneckij âzyk. Leningrad: Gos. Učebno-pedagogičeskoe izdatel’stvo. Kupriânova, Zinaida, Maria Âkovlevna Barmič and Lûdmila Vasiľevna Homič 1985 Neneckij âzyk. Leningrad: Prosveŝenie. Lehtisalo, Toivo 1956 Juraksamojedisches Wörterbuch. Helsinki: Suomalais-Ugrilainen Seura. Márk, Tamás 1990 Tempus und Aspekt im Samojedischen. Specimina Sibirica 3: 137−141. Nenâng, M. A. 2005 Russko-neneckij razgovornik. Sankt-Peterburg: Drofa. Pusztay, János 1977 Bemerkungen zur Frage der Aspekte in den uralischen Sprachen. Finnisch-Ugrische Mitteilungen 1: 153−159. Rédei, Károly 1984 A neutrális idő nyomai egyes uráli nyelvekben. Nyelvtudományi Közlemények 86: 113− 117. Salminen, Tapani 1997 Tundra Nenets Inflection. Helsinki: Suomalais-Ugrilainen Seura. Salminen, Tapani 1998 A morphological Dictionary of Tundra Nenets. Helsinki: Suomalais-Ugrilainen Seura. Szeverényi, Sándor 2007 Igeidő és aspektus. In: István Kozmács and Katalin Sipőcz (eds.), Uralisztika. Fejezetek az uráli nyelvészetből, 102−122. Szeged: Szegedi Egyetemi Kiadó. Tereŝenko, Natal’â Mitrofanovna 1947 Оčerk grammatiki neneckogo (ûrako-samoedskogo) âzyka. Leningrad: Učpedgiz. Tereŝenko, Natal’â Mitrofanovna 1965 Nenecko-russkij slovar’. Moskva: Sovestkaâ ènciklopediâ. Tereŝenko, Natal’â Mitrofanovna 1979 Nganasanskij âzyk. Leningrad: Nauka. Wagner-Nagy, Beáta 2002 Alaktan. In: Beáta Wagner-Nagy (ed.), Chrestomathia Nganasanica, 71−126. Szeged/ Budapest: SzTE − MTA Nyelvtudományi Intézet.

Beáta Wagner-Nagy, Hamburg (Germany)

1492

VII. Semantics and pragmatics in word-formation II: Special cases

86. Verbal prefixation in Slavic: a minimalist approach 1. 2. 3. 4. 5. 6.

Introduction Effects of prefixation Prefixes are incorporated prepositions The syntactic and semantic derivation of prefixed verbs Conclusion References

Abstract Verbal prefixation has various effects in Slavic languages; it brings about perfectivity and telicity, it affects case assignment properties of the verb, it can change the meaning and the argument structure of the verb, etc. For these phenomena we propose an analysis couched in a morphosyntactic minimalist approach according to which prefixes are incorporated prepositions which project a result predicate in the complement position of the verb. We show how the particular syntactic and semantic properties of the incorporated preposition derive the various derivational effects.

1. Introduction Traditionally the prefixes in Slavic are treated in two different linguistic disciplines with different approaches or models: qualifying prefixes which build up new lexemes are treated in word-formation; modifying prefixes which exhibit salient semantic modification of the corresponding base verbs, i.e. inchoative, distributive, deliminative, etc., aktionsart, are in the focus of interest of aspectology. The Academic Grammars of Slavic languages (see, for instance, GSRLJa 1970; RG 1980; MČ 1986; GWJP 1984), dealing with the meaning of the prefixes in the chapters on word-formation, proceed from the assumption that the meaning of the prefix is added to the meaning of the base verb. Therefore their aim is to list all the submeanings of a given prefix which results in homonymy and synonymy of all prefixes. In the 1970s structural approaches (see, for instance, Flier 1975; Gallant 1979) tried to represent the semantic unity of a given prefix by means of invariant semantic features. But these invariant meanings were criticized as too abstract and not providing enough specific information to describe the semantics of individual instantiations of a prefix (cf. Janda 1986: 215). Therefore, beginning in the 1980s, cognitive semantic approaches have tried to explain the meaning of a prefix, not by means of invariant semantic features, but by means of prototypes (Janda 1986) or semantic networks (Krongauz 1998). In the 1990s the semantic approaches/descriptions were confronted with morphosyntactic approaches which aimed at explaining the syntactic and semantic properties of prefixation by means of morphosyntactic derivation.

86. Verbal prefixation in Slavic: a minimalist approach

1493

2. Effects of prefixation Verbal prefixation induces various effects. They concern perfectivity, telicity, case assignment, the argument structure of the verb, boundedness. Before discussing these topics, it is necessary to say a few words about verbal prefixes themselves.

2.1. Lexical vs. superlexical prefixes It has been argued in recent years that there are two types of prefixes in Slavic: lexical prefixes (also called internal or resultative) and superlexical prefixes (also called external ); see, for instance, Babko-Malaya (1999), Ramchand (2004), Romanova (2004, 2006), Svenonius (2004), Di Sciullo and Slabakova (2005), Richardson (2007), Gehrke (2008). Explicit formulation of this distinction with some criteria goes back at least to Isačenko (1962), who replaces the traditional distinction between lexical and empty/ grammatical prefixes with the distinction between qualifying (that is, lexical) and modifying (superlexical) prefixes. Let us now briefly overview the differences between the two classes that are often hypothesized in the literature. In contrast to lexical prefixes, superlexical prefixes do not affect the argument structure of the verb to which they attach. Superlexical prefixes mostly do not derive secondary imperfectives, in contrast to lexical prefixes. They also do not change the aspectual class of the base verb, in contrast to lexical prefixes. Whereas lexical prefixes have a compositional (spatial) or idiosyncratic meaning, superlexical prefixes can only have a compositional meaning. To give a few examples, superlexical prefixes are, for instance, the inceptive prefix za- ‘behind’ in the Russian verb zabolet’ ‘to become ill’, the repetitive prze- ‘over’ in the Polish przerobić ‘to rework’, the delimitative po- ‘along’ in the Serbo-Croatian pogledati ‘to take a look’, the completive do- ‘to’ in the Slovak dopracovat’ ‘to finish working’, and the saturative/cumulative na- ‘on’ in the Russian verb nabegat’sja, in the Slovenian nalaufati se and in the Czech naběhat se, all meaning ‘to come to have one’s fill of running’. Non-compositional lexical prefixes are, for instance, u- ‘at’ in the Bulgarian ugovorja ‘to arrange’, s- ‘from’ in the Russian sžit’ ‘to expel’ and ode- ‘away’ in the Czech odečíst ‘to subtract’. Compositional (spatial) lexical prefixes are cases like na- ‘on’ in the Croatian nabàcati ‘to throw’, nad- ‘above’ in the Slovak nadpísať ‘to superscribe, entitle’ and pod- ‘under’ in the Russian podpisat’ ‘to sign’. We observe that the superlexical and compositional lexical prefixes change the verbal meaning − more or less − in a predictable way, in accordance with the prepositional/prefixal meaning, and that the changes induced by the non-compositional lexical prefixes are unpredictable. It has also been argued that superlexical prefixes can stack in contrast to lexical prefixes and that the superlexical prefix must precede the lexical prefix if they co-occur. Furthermore, superlexical prefixes do not form idioms and certain adjectival participles in contrast to lexical prefixes. According to the most popular view in the recent literature, these differences are based on different base positions of the two types of prefixes: lexical prefixes are merged in a vP/VP-internal position, whereas superlexical prefixes are merged in a vP/VP-external position. There are also approaches questioning such a distinction or arguing that such a distinction is too rough and that a more fine-grained analysis is necessary; see Biskup

1494

VII. Semantics and pragmatics in word-formation II: Special cases

(2007, 2012, forthc.), Tatevosov (2008), Žaucer (2009, 2012). For instance, Žaucer (2009) discusses several superlexical prefixes and argues that the cumulative and saturative na- ‘on’, the Slovenian perdurative pre-, and its Russian counterpart pro- ‘through’ behave like resultative, that is, lexical prefixes. Tatevosov (2008) shows that the twoway distinction is not fine enough and argues that a third class − namely, intermediate prefixes − is necessary. Biskup (2012, forthc.) shows that many of the differences discussed above are rather a tendency than a clear-cut distinction and that superlexical prefixes behave like lexical prefixes in many respects. They also affect the argument structure, case assignment properties and selectional restrictions of verbs to which they attach and they also induce telicity. Moreover, he shows that prefixes do not behave consistently with respect to the diagnostics (the differences discussed above), which would be unexpected if the differences between the two classes of prefixes were based on their different syntactic positions. Because of lack of space, we will not go into further details and will turn now to the effects of prefixation.

2.2. Perfectivity and telicity Prefixation of simplex verbs − which are imperfective in the vast majority − induces perfectivity, as is obvious from the examples in the preceding section, where all the prefixed verbs are perfective. Prefixed verbs can be imperfectivized by the secondary imperfective suffix, as shown by the following examples based on the verbs from the previous section. (1)

a. za-bol-e-va-t’ behind-pain-TH-SI-INF ‘to be becoming ill’

b. na-bac-íva-ti on-throw-SI-INF ‘to be throwing’

c. ode-čít-a-t away-read-SI-INF ‘to be subtracting’

Because of this behaviour, it has been proposed that the secondary imperfective suffix represents the aspectual head Asp; see Pereltsvaig (2004), Ramchand (2004), Svenonius (2004), Gehrke (2008), Tatevosov (2008), among others. Given that the secondary imperfective morpheme is closer to the root √ than the tense/infinitive morphology and that the morpheme expressing the verbalizing head v (the conjugation class) occurs closer to the root than the aspectual morpheme, the morphosyntactic structure of Slavic verbs like in (1) looks like (2), abstracting away from prefixes for a moment. The mirror order of morphemes is achieved via head movement, as we will see below. (2)

[TP T [AspP Asp [vP v [√P √ ]]]]

Verbs imperfectivized by the secondary imperfective suffix can be turned into perfective again by a superlexical prefix; consider example (3), containing verbs from (1), prefixed with the distributive prefix po-. (3)

a. po-za-bol-e-va-t’ along-behind-pain-TH-SI-INF ‘to become ill one after another’

86. Verbal prefixation in Slavic: a minimalist approach

1495

b. po-na-bac-íva-ti along-on-throw-SI-INF ‘to throw one after another’ c. po-ode-čít-a-t along-away-read-SI-INF ‘to subtract one after another’ The fact that superlexical prefixes can take scope over the secondary imperfective suffix is one of the reasons why some researchers propose that superlexical prefixes are generated in a position in/above the aspectual phrase. However, it is not necessary to merge the prefix directly in such a high structural position; it could, for instance, be moved there from a lower position inside vP/VP. The “superlexical” scope could also be dissociated from the prefix itself, which could be a pure overt marker agreeing with the appropriate functional head bearing the superlexical semantics. It is a well-known fact that it is necessary to distinguish between the lexical aspect and the morphological aspect; see Smith (1991), Schoorlemmer (1995), Filip (1999, 2003), Borik (2002). Consequently, the question arises whether prefixes generally function as telicizers, in addition to their perfectivizing function. Taking into consideration prefixes in example (1), it seems that prefixes are indeed telic: the inceptive superlexical prefix za- contributes an initial boundary to the atelic predicate bolet’ ‘to be ill’, the non-compositional lexical prefix ode- contributes the result state to the activity verb číst ‘to read’ and the compositional lexical prefix na- specifies the result state of the telic verb bàcati ‘to throw’. In a similar vein, in the examples in the preceding section, for instance, the saturative/cumulative na- adds a result state to the Czech activity verb běhat ‘to run’, the non-compositional prefix s- adds a result state to the Russian atelic verb žit’ ‘to live’ and the compositional prefix nad- adds a result to the Slovak activity verb písať ‘to write’. There is no clear consensus on this matter in the literature. On the one hand, Filip (1999) claims that verbal prefixation induces lexical aspect shift. Similarly, Arsenijević (2006) argues that all Slavic prefixes are predicates of the result subevent, that is, that they are telic. And according to Piñón (1994) and van Hout (2008), Polish and Russian prefixes make verbs telic. On the other hand, Filip (2003), Romanova (2006) and Gehrke (2008) argue that prefixes do not function in all of their uses as telicity modifiers. In this respect, superlexical prefixes are more problematic than lexical prefixes. For instance, Romanova (2006) argues that superlexical prefixes do not always affect the lexical aspect, in contrast to lexical prefixes, which are always telic. The problematic cases, which are almost always cited in the literature, are the delimitative prefix po- and the perdurative prefix pro-. Although these prefixes make verbs perfective, they are compatible only with for-adverbials, which is taken to show that they do not induce telicity. This issue, however, is in need of more investigation because in certain contexts the delimitative po- behaves like a telicizer; consider the following Czech example. (4)

a. Jan (po-)vařil špenát pět minut / *za pět minut. Jan along-cooked spinach five minutes / in five minutes ‘Jan cooked the spinach for 5 minutes.’

1496

VII. Semantics and pragmatics in word-formation II: Special cases b. po-vařený špenát along-cooked spinach ‘the spinach is in the state of being cooked for a while’

Sentence (4a) shows the compatibility of the unprefixed verb and the po-verb with the for-adverbial (note that the in-adverbial is not meant to measure the time interval between a reference time and the beginning of the cooking event). Example (4b) then demonstrates that the po-verb can derive the past passive participle. Since past passive participles are standardly considered to express a result state, (4b) suggests − in contrast to (4a) − that the delimitative po- is telic (see Biskup forthc. for more on this point). As to the perdurative pro-, consider the Czech example (5). (5)

a. Jan ve vězení pro-seděl dva roky / *za dva roky. Jan in prison through-sat two years in two years ‘Jan was in prison for two years.’ b. Jan ve vězení pro-seděl židli *dva roky / za dva roky. Jan in prison through-sat chair two years in two years ‘Jan destroyed a chair by sitting in the prison in two years.’

According to the adverbial test in (5a), prosedět is atelic. However, the verb prosedět requires an accusative complement (in contrast to the unprefixed sedět), which suggests that the adverbial in (5a) is forced to bear accusative because of its complement status. Example (5b) supports this analysis because if an accusative complement is added to sentence (5a), the adverbial must be in the in-form. This in turn suggests that the perdurative pro- is in fact a telicizer. Since the presence of unselected objects indicates resultative predication, the fact that pro- licenses the unselected accusative object shows that the prefix is resultative, hence telic. For a more detailed discussion of the perdurative pro- and the syntactic status of the adverbial, see Schoorlemmer (1995), Borik (2002), Ramchand (2008), Žaucer (2009, 2012). Having said this, we prefer the view that prefixes contribute telicity.

2.3. Case assignment properties of prefixed verbs In some Slavic languages, there is a relation between the form of objective case and aspectual properties of the verb; see, for instance, Jakobson (1936), Paducheva (1998) for Russian and Wierzbicka (1967), Rozwadowska and Willim (2004) for Polish. In both languages, the partitive genitive is restricted to the perfective aspect, as demonstrated by (6), from Paducheva (1998: 80). (6)

a. *Ja pju vod-y. I.NOM drink.IMPF water-PART b. Ja vy-pil vod-y. I.NOM out-drank.PF water-PART ‘I drank (some) water.’

86. Verbal prefixation in Slavic: a minimalist approach

1497

Example (6a) shows that the unprefixed imperfective verb cannot co-occur with the object bearing partitive case. In contrast, in (6b) the prefix makes the verb perfective; hence the object can be marked with partitive. Regarding the meaning of the partitive argument, it has the partitive (‘part of ’, ‘some’) interpretation. The partitive case alternates with accusative, which expresses the total quantity of the noun. Superlexical prefixes, too, can affect case assignment properties of the verb to which they attach; consider the Polish example (7), containing the cumulative na- ‘on’. (7)

a. Agnieszka piekła bułk-i / *bułek. Agnieszka baked.IMPF roll-ACC.PL roll.GEN.PL ‘Agnieszka was baking rolls.’ b. Agnieszka na-piekła bułek / *bułk-i. Agnieszka on-baked.PF roll.GEN.PL roll-ACC.PL ‘Agnieszka baked a lot of rolls.’

The unprefixed verb takes the direct object marked with accusative, as shown in (7a), but if the verb is prefixed with the cumulative prefix, the plural object must bear partitive genitive, as demonstrated in (7b). For discussion of the cumulative na- mainly from the syntactic point of view, see Pereltsvaig (2006), Romanova (2006), Žaucer (2009), Biskup (2012), and for the semantic point of view, see Piñón (1994), Filip (2000, 2005) and Tatevosov (2007).

2.4. Boundedness effects The next effect of verbal prefixation concerns boundedness (definiteness). The Czech example in (8) demonstrates that there are two types of structural accusative, which differ with respect to extraction. (8)

a.

psal [článek t1 ]? [O čem]1 Jan about what Jan.NOM wrote article.ACC ‘About what was Jan writing a/the article?’

b. ?*[O čem]1 Jan do-psal [článek t1 ]? about what Jan.NOM to-wrote article.ACC c.

Jan do-psal článek o opicích. Jan.NOM to-wrote article.ACC about monkeys ‘Jan wrote a/the article about monkeys.’

Example (8a) shows that extraction of the prepositional phrase o čem ‘about what’ out of the accusative direct object is grammatical when it happens in a sentence with the unprefixed verb. In contrast, when the prepositional phrase is moved out of the accusative object in a sentence with the prefixed predicate, the extraction is strongly degraded; consider (8b). The control example (8c) demonstrates that the problem is indeed due to the movement of the prepositional phrase. In other words, the perfective accusative blocks extraction in contrast to the accusative assigned by the imperfective (unprefixed)

1498

VII. Semantics and pragmatics in word-formation II: Special cases

verb. This is not surprising in light of the claims that in Slavic languages the aspect properties of the verb affect the interpretation of the object, cf., e.g., Krifka (1989, 1992), Piñón (1995), Filip (1999). In Bulgarian, the presence of the prefix on the verb correlates with the presence of the definite article on the mass noun. Consider example (9), taken from Verkuyl (1999: 129), originally Guentcheva (1990: 36), which shows that if the prefixed verb co-occurs with an object without the definite article, the sentence is ungrammatical. (9) a. *Az iz-pix kafe. I out-drank coffee b.

Az iz-pix kafe-to. I out-drank coffee-the ‘I drank the coffee.’

The boundedness effect induced by verbal prefixes can be found in other domains as well, as demonstrated by the prefixed adverbs in the following Slovak example. (10) a. kedy b. od-kedy c. do-kedy when away-when to-when ‘when’ ‘since when’ ‘till when’ When the time adverb kedy is unprefixed, as in (10a), the time interval (or the time axis) is unbounded at both ends. However, when it is prefixed, as in (10b) and (10c), the time interval is bounded: at the beginning and at the end, respectively. Other boundedness effects can be found in the aspectual domain. They were already discussed in section 2.2. Verbal prefixation also has an effect on the argument structure. This will be discussed in the following section in connection with the identity between prefixes and prepositions.

3. Prefixes are incorporated prepositions Following Biskup (2007, 2012), we argue that verbal prefixes are incorporated prepositions. There are several arguments supporting this view. The first one is based on the argument structure effects of verbal prefixation. Schoorlemmer (1997) argues with respect to compositional telicity that perfective paired verbs, that is, prefixed verbs deriving the secondary imperfective like za-bolet’/za-bolevat’ ‘to become ill’, pere-mërznut’/ pere-merzat’ ‘all to freeze’, pod-rasti/pod-rastat’ ‘to grow up’, always have an internal argument in Russian. This means that intransitive paired verbs must be unaccusative. Concerning prefixed verbs derived from unergative verbs, Schoorlemmer argues that derivation of such paired verbs always involves transitivization; consider the following predicates: raz-igrat’/raz-igryvat’ ‘to raffle’, o-plakat’/o-plakivat’ ‘to bewail’, na-guljat’/ na-gulivat’ ‘to walk a lot’. Similar facts can be found in Czech, as shown by the examples below, containing the perfective/secondary imperfective pairs. (11) a. na-mrznout/na-mrzat on-freeze ‘to freeze onto sth’

b. s-růst/s-růstat with-grow ‘to grow together’

86. Verbal prefixation in Slavic: a minimalist approach c. do-hořet/do-hořívat to-burn ‘to burn out’

d. za-hloubat se/za-hloubávat se behind-pore.over self ‘to pore over’

e. vy-pracovat/vy-pracovávat out-work ‘to work out’

f.

g. při-dělat/při-dělávat at-do ‘to fix’

h. pode-psat/pode-pisovat under-write ‘to sign’

i.

1499

pře-křičet/pře-křikovat over-shout ‘to howl down’

při-hřát/při-hřívat at-warm ‘to warm a little’

Examples in (11a, b, c) illustrate that if prefixes are attached to an unaccusative verb, then the verb remains unaccusative. The examples in (11d, e, f ) then show that if prefixes are attached to an unergative verb, the verb is transitivized. And if prefixes are attached to a transitive verb, then the verb remains transitive, as demonstrated in (11g, h, i). The data also show that this holds for both lexical (11a, b, d, e, g, h) and superlexical prefixes (11c, f, i). The generalizations suggest that the prefixes are generated in the complement position of the root. They are prepositions projecting a phrase that competes for the complement position with the nominal complement of the unprefixed verb. Later the preposition incorporates into the root. Having this in mind, we can explain the argument structure generalizations in the following way: In the case of unaccusative verbs, as in (11a−c), the root is merged with the prefix (preposition, for more details on the syntactic structure, see section 4.1), which introduces an argument and projects a prepositional phrase. Since the prepositional phrase is in complementary distribution with the nominal complement of the root (the base verb), the argument structure is not augmented and the verb remains unaccusative. In the case of base unergative verbs, as in (11d−f ), the preposition again introduces an argument or arguments in its prepositional phrase. However, since the root is selected by the verbal head introducing an agentive argument in this case, this results in augmentation of the argument structure of the base verb. In this way, we account for Schoorlemmer’s (1997) observation that intransitive paired verbs are unaccusative. Regarding base transitive verbs, as in (11g−i), the prepositional phrase with its argument(s) replaces the nominal complement of the root; hence the prefixed verbs remain transitive. The following Russian example demonstrates that the argument of the verb prefixed with the cumulative na- or the distributive po- cannot be a singular noun. (12) a. Gruš-a padala na zemlju. on ground pear-NOM.SG fell ‘The pear was falling to the ground.’ b. *Gruš-a na-/po-padala. pear-NOM.SG on/along-fell c. Gruš-i na-/po-padali. pear-NOM.PL on/along-fell ‘Pears fell down in a certain quantity/one after another.’

1500

VII. Semantics and pragmatics in word-formation II: Special cases

Given that selectional relations are very local − standardly, it is the head-complement or the head-specifier relation −, selectional restrictions imposed on arguments by superlexical prefixes also support the proposed analysis. Specifically, if the appropriate argument in cases like (12c) is merged in the prepositional phrase projected by the prefix, as proposed above, then such a close selectional relation between the prefix and the argument exists. If prefixes are prepositions that incorporate into the root, then one expects that this movement will be subject to movement constraints. As demonstrated by the Russian example (13), this expectation is correct. (13) a. Popugaj v-letel v komnat-u. parrot.NOM in-flew in room-ACC ‘The parrot flew into the room.’ b. Popugaj v-letel na stol. parrot.NOM in-flew on table.ACC ‘The parrot flew onto the table.’ c. Popugaj v-letel v komnat-u na stol. parrot.NOM in-flew in room-ACC on table.ACC ‘The parrot flew into the room, onto the table.’ d. *Popugaj v-letel na stol v komnat-u. parrot.NOM in-flew on table.ACC in room-ACC (Inga Žirkova, p.c.) This example shows that prepositional phrases projected by a preposition that does not fit the verbal prefix cannot intervene between the homophonous prefix and preposition. Concretely, sentence (13a) demonstrates that the verb vletel can co-occur with the prepositional phrase headed by the preposition v, and (13b) shows that vletel can co-occur with the prepositional phrase projected by na. What is important is that vletel can cooccur with both prepositional phrases if naPP follows vPP, as shown in (13c). Crucially, (13d) is ungrammatical because na blocks the local relation between the incorporated preposition v in vletel and its copy in v komnatu. If it is correct that syntactically, verbal prefixes are just another copy of a preposition, then it predicts that all prefixes have a prepositional counterpart. This is not problematic from the diachronic point of view because prefixes are historically derived from prepositions or both categories have a common ancestor. This holds not only for Slavic languages; it has also been argued, for instance, that verbal prefixes and particles in the Germanic languages belong to the category preposition, see Jackendoff (1973), Emonds (1985), den Dikken (1995), Stiebels (1996, 1998), Zeller (2001), McIntyre (2007, and article 23 on particle-verb formation). Let us turn back to Slavic languages and consider, for instance, Russian and Czech. They have almost twenty verbal prefixes and only three of them do not have a prepositional counterpart: vz- ‘up’, vy- ‘out’ and raz/roz- ‘apart’. This is the modern state of affairs. In Old Czech as well as Old Russian, vz was used as a preposition. Vz is also present in today’s Serbo-Croatian (as uz/uza) and in certain Macedonian dialects (as voz); see Kopečný (1973). There is a relation between vz and vy because they are derived

86. Verbal prefixation in Slavic: a minimalist approach

1501

from the Indo-European uds ‘up, out’ and ud ‘up, out’, respectively; see Kopečný (1973) and Vasmer (1976). As far as raz/roz is concerned, it is derived from the Proto-Slavic form orz and the preposition raz can be found, for instance, in Slovenian (Kopečný 1973; Rejzek 2001). A possible synchronic analysis of these phenomena can be based on the necessity of prepositional incorporation in certain cases. For instance, the preposition vz needs to incorporate into the verb in modern Russian and modern Czech, whereas in Old Russian, Old Czech, Serbo-Croatian and certain Macedonian dialects, the preposition does not have to incorporate into the verb. Another possibility is to assume that the appropriate prefix and preposition are not etymologically related. Note that allomorphs generally do not have to be of the same form; see discussion of prepositional incorporation in Baker (1988) and Biskup and Putnam (2012), who argue that the German verbal prefix ent- is a spell-out of the incorporated preposition aus ‘out’. In contrast to the German case, Slavic languages spell out two copies of the prepositional chain: the head of the chain, that is, the prefix, and the tail of the chain, that is, the preposition selecting the complement. Given economy of grammar, the two copies must be licensed. As shown in Biskup (2012), the prefix copy is necessary because of the morphological aspect, and the prepositional copy is necessary because of the semantics of the preposition connected to the assigned case (cf. Nunes 2004 for the claim that spell-out can spell out more than one copy of the chain and Yadroff and Franks 2001 for the fact that in colloquial Russian even prepositions themselves can be multiplied). The next argument for the proposal that verbal prefixes are incorporated prepositions is based on the parallelism between extraction from the object marked with the perfective structural accusative and extraction from prepositional phrases. In section 2.4 we saw that extraction out of the object with the perfective structural accusative is bad, in contrast to the imperfective structural accusative. Recall that the perfective structural accusative is assigned by the prefixed verb, in contrast to the imperfective accusative, which is assigned by the unprefixed verb. Now consider the following Russian examples. (14) a. Popugaj v-letel v komnat-u. parrot.NOM in-flew in room-ACC ‘The parrot flew into the room.’ v-letel v t1? b. *Čto1 popugaj what parrot.NOM in-flew in c. Popugaj v-letel v komnat-u so stolom. parrot.NOM in-flew in room-ACC with table ‘The parrot flew into the room with the table.’ d. *[S čem]1 popugaj v-letel v komnat-u t1? with what parrot.NOM in-flew in room-ACC Examples (14a) and (14c) are the unmarked cases. Sentence (14b) shows that extraction of the prepositional complement is ungrammatical, and example (14d) shows the same for the prepositional subconstituent. The generalization drawn from (14) is that prepositions block extraction. Taken together, the fact that verbal prefixes and prepositions behave in the same way with respect to extraction supports the view that they are identical elements.

1502

VII. Semantics and pragmatics in word-formation II: Special cases

Verbal prefixes and prepositions also have very similar phonological properties; therefore it has been argued that they belong to the same category (see Matushansky 2002 for Russian). Even if they do not always behave in the same way in phonological terms, it does not necessarily mean that they are not identical elements. Their different behaviour can be induced by the different morphosyntactic contexts in which they occur. For instance, Biskup, Putnam and Smith (2011) argue that the different behaviour of German prefixes and particles/prepositions with respect to the stress can be derived from the fact of whether or not the appropriate preposition incorporates into the verb and consequently occurs in the same phonological domain as the verb. In certain cases, the bound-free morpheme distinction is even lost and prepositions behave like prefixes, as shown by the Slovak example below. (15) a. do-ň (= do neho) to-it ‘to it’

b. za-ň (= za neho) behind-it ‘behind it’

The final argument for the proposed analysis is based on the semantic properties of prepositions and verbal prefixes. A comparison of the prefixed verbs with the prepositional phrases in example (16) shows that the lexicosemantic import of the prefixes and prepositions is almost identical. This is demonstrated by the Polish lexical prefix and preposition w- in (16a, b), by the Russian lexical prefix and preposition na- in (16c, d), and by the Slovenian superlexical prefix and preposition od- in (16e, f ). (16) a. w-jechać in-drive ‘to drive in’

b. wdomu in house ‘in the house’

c. na-kleit’ on-glue ‘to glue on’

d. na stole on table ‘on the table’

e. od-peti away-sing ‘to finish singing’

f. od pekača away baking.sheet ‘away from the baking sheet’

4. The syntactic and semantic derivation of prefixed verbs The previous sections provide data and background information for our analysis, which will be presented in this section. We have shown that prefixation has various effects on the verb. In what follows, we present a syntactic and semantic derivation of the Russian example (17b), containing the compositional lexical prefix na- ‘on’. In connection with this example, we show how the various effects of prefixation can be analyzed in a minimalist framework. (17) a. Oleg risoval ptic-u na sten-e. Oleg.NOM drew bird-ACC on wall-LOC ‘Oleg was drawing the bird on the wall.’

86. Verbal prefixation in Slavic: a minimalist approach

1503

b. Oleg na-risoval ptic-u na sten-e. Oleg.NOM on-drew bird-ACC on wall-LOC ‘Oleg drew the bird on the wall.’ In contrast to the analyses proposed in the literature − which mostly show prefix derivations only in part − our proposal provides a complete syntactic and semantic derivation of a sentence containing the prefixed verb. The syntactic analysis in (18) is couched in a morphosyntactic minimalist approach and the semantic analysis in (20) in the framework of formal semantics, using lambda calculus. Since our proposal is formalized, it has the advantage that an observant reader can well control how the derivation of prefixed verbs works and how the truth-conditional meaning of the sentence is derived. Our analysis treats morphological processes as part of the syntactic computation, in line with recent developments in generative grammar; see, e.g., the distributed morphology framework. Such an approach has the advantage that there is only one generative component in the grammar − namely the (morpho)syntax − that combines particular elements together (for empirical arguments, see Baker 1988; Lieber 1992 and article 7 on word-formation in generative grammar; Biskup and Putnam 2012, among others). The same also holds for the semantic derivation; there is no need to assume semantic operations in the lexicon. The syntax provides a compositional apparatus and the derived constructions are then interpreted at the semantic interface. Now let us turn to the syntactic derivation.

4.1. The syntactic derivation The syntactic derivation of example (17b) proceeds as follows. (18)

CP 3 C TP 3 Oleg DP T’ [v -F,vT-F] 3 l T AspP 3 [u -F,vT-F] Asp vP 3 [u -F,uT-F] pticu DP vP 3 [v -F,vT-F] Oleg DP v’ 3 [v -F,uT-F] P ova v 3 PrefixP ris 3 Prefix PP 3 pticu DP P’ [v -F,uT-F] 3 na P DP [u -F, vT-F]

stene

[v -F,uT-F]

1504

VII. Semantics and pragmatics in word-formation II: Special cases

In the first step, the preposition na ‘on’ merges with the ground argument (relatum) stene ‘wall’. We follow Biskup (2007, 2009), who analyzes prepositional cases as an unvalued tense feature on the head D. Biskup extends Pesetsky and Torrego (2004, 2006) and proposes that prepositions bear unvalued φ-features and a valued tense feature. DPs bear valued φ-features because of their inherent properties and in addition, the unvalued tense feature. This means that all cases are treated uniformly, concretely, as a result of the operation Agree between φ-features and tense features of the probe and goal. In this respect, we depart from the standard approach, which assumes that the prepositional case assignment differs from the structural case assignment in that prepositional cases are assigned in connection with θ-roles and are of semantic nature. Note that structural cases also have semantic effects; consider, for instance, the accusative-partitive genitive alternation in Russian and Polish, discussed in section 2.3. With respect to the problematic connection between prepositional cases and θ-roles, consider Chomsky (2008), who proposes that the verbal head V inherits φ-features of the head v and assigns structural accusative to its object. Since V also gives a θ-role to its object, this means that certain structural cases are also licensed in connection with θ-roles. Returning to the derivation of (17b), the unvalued φ-features of the preposition na are valued by the valued φ-features of stene and the unvalued tense feature of stene is valued by the valued tense feature of na. This results in locative case on the noun stene. Although there is no overt agreement morphology on prepositions in Slavic languages, there are languages with overt prepositional agreement; see Baker (2008). Concerning the tense feature on prepositions, there are also languages that have tensed prepositions; see Bowern and Aygen-Tosun (2000), Harlow (2007). Next, P’ merges with the figure argument (locatum) pticu ‘bird’. Since pticu is marked with structural accusative in (17b), it is obvious that its unvalued tense feature is not valued by the tense feature on the head P. Assuming the activation condition (Chomsky 2000, 2004), which states that the probe and goal are active only if they have an unvalued feature, the unvalued tense feature on pticu cannot be valued in the prepositional phrase. In order not to violate the case filter, pticu must receive case somewhere higher in the derivation. The prepositional phrase is selected by the head Prefix and the preposition na incorporates into it. We assume that the prepositional head movement is triggered by an unvalued feature of the greedy type present on na. The head Prefix makes a prefix from the preposition and glues the prepositional phrase to the verb, as discussed in the semantic derivation below. The presence of the prefix phrase in the complement position of the verb (root) − in fact, the presence of the prepositional arguments − can bring about the argument structure augmentation, as discussed in section 3. In the case of example (17b), there is no effect visible because the unprefixed verb risovat’ also takes the direct object (with the prepositional phrase). The difference between examples (17a) and (17b) is that in (17a) the preposition na stays in PP and does not incorporate into the verb. In the next step, PrefixP merges with the root ris-, projecting the root phrase. The head composed of the preposition na and the head Prefix incorporates into the root, in accordance with our proposal in section 3 that verbal prefixes are incorporated prepositions. The head adjunction always happens to the left, which derives the correct order of morphemes, as shown by the complex head in (19).

86. Verbal prefixation in Slavic: a minimalist approach (19)

1505

T 3 Asp T l 3 v Asp 3 v ova 3 ris Prefix 3 na P Prefix

Since the suffix -ova- determines the conjugation class of narisovat’, a natural assumption is that it represents the verbal head v. (Note that other morphemes can turn the root into a different category; see, for instance, the Russian noun risunok ‘drawing’ and the Slovak noun nárys ‘sketch’.) Thus, the root phrase is selected by the verbal head -ova-, which has agentive properties, and the complex head root incorporates into it. The argument Oleg is merged in the specifier position of vP, with the consequence that it is interpreted as the agent. Given that vPs are phases and that the prepositional external argument pticu occurs in the phase complement, it must move to the edge of vP to escape the consequences of the phase impenetrability condition (Chomsky 2000). Recall that pticu still bears the unvalued tense feature, which will induce a derivational crash if pticu remains in PP. Thus, pticu moves to the outer specifier of vP, observing the extension condition, as demonstrated in derivation (18). We saw in section 2 that prefixes perfectivize verbs. Given this, there must be a relation between verbal prefixes and the aspectual head. Recall that prefixes, that is, incorporated prepositions bear a valued tense feature. The aspectual head is usually assumed to express a relation between times. One concludes that the relation between the aspectual head and prefixes can be based on tense features. This is another function of the prepositional tense feature. Hence we assume that the aspectual head bears an unvalued tense feature. When it merges with the verbal phrase, its unvalued tense feature is valued by the valued tense feature of the preposition incorporated into the head v. Following Biskup’s proposal (2009), the tense feature on prepositions has the value bounded. Consequently, the aspectual head receives the perfective aspectual interpretation. In cases where the preposition does not incorporate into the verb and is spelled out in the phase complement of the head v, as in (17a), the tense feature on the aspectual head receives the default value and Asp gets the imperfective interpretation (another possibility would be to assume that verbs bear an unbounded tense feature, which represents the default value when it co-occurs with the prepositional tense feature). That is, sentence (17a) does not say whether or not the image of the bird on the wall is finished, in contrast to the perfective (17b). In addition to the tense feature, the aspectual head also bears unvalued φ-features, which are responsible − together with the tense feature − for assigning objective case. Now we will present a couple of arguments for the view that objective case is assigned by the aspectual head in Slavic. If cases are unvalued tense features and if the aspectual head expresses a relation between times, then it is natural that objective case is assigned by the aspectual head. There is a relation between the form of objective case and aspectual

1506

VII. Semantics and pragmatics in word-formation II: Special cases

properties of the verb. For instance, the Russian partitive genitive on the object is restricted to the perfective aspect, as we saw in example (6). The same also holds for the partitive genitive in Polish partitive constructions. Furthermore, in both languages, prefixes like the cumulative na- affect the case properties of the direct object; see example (7) again. Since we know that there is a relation between prefixes and the aspectual head, this suggests that the aspectual head mediates between the prefix and the form of objective case (for the link between case and aspect in Russian and other Slavic languages, see, e.g., discussion in Babko-Malaya 1999; Pereltsvaig 2000; Borer 2005; Richardson 2007). Such phenomena are not restricted to Slavic languages; consider, for instance, the well-known accusativepartitive alternation in Finnish, connected to aspectual properties of the predicate or aspect split languages like Hindi, where ergative is restricted to the perfective aspect. Going back to our derivation, the φ-features on the aspectual head probe and find pticu, which is closer to them than Oleg occurring in the inner specifier of v. Thus, the φ-features of the head Asp are valued by the valued φ-features on pticu and the unvalued tense feature on pticu is valued by the valued tense feature on the aspectual head (which has been valued by the incorporated preposition na). This results in the bounded structural accusative on pticu. Recall from section 2.4 that the structural accusative assigned by a perfective verb differs from the structural accusative assigned by an imperfective verb with respect to the possibility of extraction. In languages like Bulgarian, the presence of the bounded tense feature on the object DP, which has been valued by the bounded tense feature of the aspectual head, can be manifested through the presence of the definite article on the object, as we saw in example (9). Data like (8) can be accounted for by Chomsky’s (1964) A-over-A principle applied to bounded tense features. In (8a) the prepositional phrase o čem ‘about what’, which bears the bounded tense feature, is extracted from the more inclusive category článek ‘paper’ whose tense feature is unbounded (its structural accusative comes from the imperfective Asp, having the unbounded tense feature). Therefore, the A-over-A principle is not violated and there is no problem with the extraction. In contrast, in example (8b), the prepositional phrase with its bounded tense feature is extracted from the argument also bearing the bounded tense feature, (its structural accusative is assigned by the perfective Asp, that is, by the bounded tense feature). This violates the A-over-A principle, hence the extraction is degraded. The fact that prepositional phrases are islands for extraction, as demonstrated in example (14), can be analyzed in the same way. The tense feature of the prepositional complement komnatu/čto is valued by the bounded tense feature of the preposition v ‘in’. Therefore, extraction of the argument from the dominating category bearing the bounded tense feature violates the A-over-A principle. The same also holds for extraction of the prepositional subconstituent from PP; consider example (14d) again. The question arises as to how case assignment works in the case of case alternations, for instance, in the case of the cumulative prefix na-, where the aspectual head assigns partitive genitive instead of accusative. Biskup (2012, forthc.) argues that superlexical prefixes can be analyzed in the same way as lexical prefixes. That is, they can also merge in the complement position of the root, where they project PP, and then they incorporate into the verb. He assumes that the preposition na can optionally bear the cumulative feature, which is responsible for the cumulative interpretation by checking the interpretable feature of the cumulative head. This feature is also responsible for the partitive genitive on the direct object; as in (7b). Concretely, when the preposition with the cumulative feature incorporates into the case assigning head − here Asp −, it changes case assignment properties of the head from accusative to partitive genitive.

86. Verbal prefixation in Slavic: a minimalist approach

1507

The particular selectional properties imposed on the prepositional external argument by superlexical prefixes (prepositions) like the cumulative na- or the distributive po-, discussed in connection with example (12), can also be analyzed through the cumulative feature (distributive feature). Specifically, the appropriate feature places additional selectional properties on the preposition so that, for instance, na requires its external argument to be a plural or mass term. In addition to the uniform analysis of cases and deriving perfectivity, using tense features has the following advantages. It is known that there is a relation between the lexical aspect (telicity) and the morphological aspect. This relation can be based on the presence of the bounded tense feature on incorporated prepositions since prepositions simultaneously contribute to the lexical aspect (projecting PP in the verbal domain) and to the morphological aspect (valuing the tense feature on Asp). The bounded tense feature on prepositions is also responsible for various boundedness effects: perfectivity, islandhood of PPs and islandhood of the argument bearing the perfective structural accusative, as discussed above. The bounded tense feature can also derive boundedness effects in the case of prefixed adverbs, assuming a noun with an unvalued tense feature in the complex morphological structure of the adverbs. In the case where the adverb is prefixed, that is, where there is a tense feature with the value bounded, as in (10b) and (10c), the path/interval is interpreted as bounded. In contrast, if there is no prefix present, that is, there is no bounded tense feature, the path or time interval is interpreted as unbounded, as in (10a). In the next step of our derivation, the aspectual phrase merges with the tense head and the complex aspectual head adjoins to it. The head T bears unvalued φ-features and the valued tense feature. The φ-features of the tense head probe and agree with Oleg, which is the closest active element. And the valued tense feature of the tense head values the tense feature on Oleg as nominative. Assuming that the tense head bears the EPPfeature, Oleg moves to the specifier position of the head T, which derives the right word order of the sentence under discussion. Finally, the tense phrase merges with the complementizer. Since the head C is of the declarative type, it has no overt syntactic effects on the derivation. It is obvious from example (17b) that two copies of the preposition must be spelled out. Specifically, the upper copy is spelled out as the prefix in the complex head T, as shown in (19), and the lowest copy is spelled out as the preposition selecting the argument stene. As already discussed in section 3, the upper copy is necessary because of the aspect and the lowest copy because of the semantics of the preposition connected to the assigned case.

4.2. The semantic derivation In this section, we present the semantic derivation of sentence (17b); consider (20). For ease of exposition, movements are omitted in the tree. First, the preposition na combines with the DP stene. The preposition bears the standard locative meaning, locating the external argument x with respect to the internal argument y, and stene is a definite DP of the type , derived by the iota operator. By means of functional application, we receive the meaning of P’: the external argument x is in the state of being on the wall. P’ applies to the definite DP pticu, derived by the

1508

VII. Semantics and pragmatics in word-formation II: Special cases

iota operator applied to the predicate NP. This gives the meaning of the prepositional phrase, that is, the bird is in the state of being on the wall. In the next step, the meaning of PP combines with the head Prefix, which makes a prefix from the preposition. Since the prefix na- behaves like a compositional lexical prefix in the sentence, that is, it does not induce unpredictable lexical changes in the verb risovat’ ‘to draw’, the head Prefix bears the general meaning of compositional prefixes, as proposed in Biskup (2012). The first conjunct in the meaning of Prefix stands for the meaning of PP; this represents the result state brought about by prefixation; see discussion of telicity in section 2.2. The second conjunct introduces an event that has properties of the root√. This means that the Prefix head works as glue between the prepositional phrase and the root (verb). The third conjunct of Prefix expresses the fact that it is prefixation that brings about a causative relation between the result state and the other subevent. Using functional application, the meaning of the prepositional phrase replaces the predicate variable P in Prefix. In this way, we derive the prepositional nature of prefixes, the identical lexicosemantic properties of prepositions and verbal prefixes, as discussed in the end of section 3. (20)

86. Verbal prefixation in Slavic: a minimalist approach

1509

Concerning the meaning of non-compositional prefixes, see Biskup (2011, forthc.), who proposes the following meaning: λ√λxλsλe[P.√(s) & Theme(x)(s) & √(e) & Cause(s)(e)]. The first conjunct again expresses the result state brought about by prefixation. The variable P is replaced by the appropriate preposition present in the derivation and the variable √ is replaced by the appropriate root (cf. Basilico 2008, using √ as a variable). The dot stands for concatenation, which is a binary operation producing one word (string) at LF when both sides of the dot are specified. This means that the first conjunct cannot be interpreted until the root is specified. Then, the idiosyncratic meaning of the whole string is used. The second conjunct in the meaning of the prefix says that x is the theme of the result state, which captures the transitivizing effect of prefixation. The third conjunct and the fourth conjunct have the same meaning as in compositional prefixes. As far as superlexical prefixes are concerned, they bear the meaning of compositional prefixes because they can have only a compositional meaning. As discussed in Biskup (2012) with respect to the cumulative na- ‘on’, the only difference between superlexical and compositional lexical prefixes is that superlexical prefixes bear a feature that licenses the presence of the corresponding head in the structure (see also discussion of the cumulative na- in section 4.1). This means that the superlexical meaning is dissociated from the prefix (preposition) itself. Returning to the derivation in (20), the meaning of PrefixP applies to the root ris-, which results in the meaning that the bird is in the state of being on the wall and this state is caused by the event of drawing. Then, the root phrase combines with the verbal head v, which adds the agentive component to the meaning of the root phrase. Functional application between v’ and DP replaces the individual variable y with the entity Oleg, with the result that Oleg is the agent of the drawing event. In the next step, the meaning of vP combines with the perfective meaning of the aspectual head because the tense feature of the aspectual head is valued as bounded by the tense feature on the incorporated preposition, as discussed in the previous section. As far as temporal entities are concerned, we use the standard three times in our ontology: the event time, the reference time and the speech time, following Reichenbach (1947), Klein (1994, 1995), Demirdache and Uribe-Etxebarria (1997, 2000), Giorgi and Pianesi (1997), among others. The aspectual head relates the event time to the reference time and the tense head relates the reference time to the speech time. The perfective aspectual head is usually treated as an operator that localizes the event time within the reference time; see, for instance, Klein (1994), Paslawska and von Stechow (2003). The reference time is represented by the variable t and the event time is obtained by means of the temporal trace function τ, which maps the event e to its “run time”. The meaning of the aspectual head is slightly modified here; we add the existentially bound state variable and the trace function mapping the state s to its time. The time of the event e abuts the time of the state s: τ(e)I3τ(s). The meaning of the aspectual head combines with vP and by means of functional application we receive the meaning of the aspectual phrase. This combines with the tense head, which bears the standard past tense semantics. It existentially binds the time variable t and relates it to the speech time t* by means of the ‘before’ relation > (8a)

Type II is slightly more complex, insofar as it involves absolute overriding forms. German (and other Germanic languages or Basque) illustrates this type.

87. Denumeral categories

1521

(9) Type II. a. Simple ordinals: Ord PHON (X) = X-te b. Ord PHON (ein) = erste, Ord PHON (drei) = dritte c. Complex ordinals: Ord PHON ([X1,… Xn]) = [X1,… Ord PHON (Xn)] d. Overrides: (9c) >> (9a); (9b) >> (9a) The next type is more complicated for what regards suppletion, since it involves both conjunct and absolute forms. Hungarian illustrates this type. (10) Type III. a. Simple ordinals: Ord PHON (X) = X-ik aa′. OrdPHON-CJT(X) = Ord PHON (X) b. Ord PHON (egy) = elsö, Ord PHON (ket) = masodik bb′. OrdPHON-CJT(egy) = egyedik, OrdPHON-CJT(ket) = kettedik c. Complex ordinals: Ord PHON ([X1,… Xn]) = [X1,… OrdPHON-CJT(Xn)] d. Overrides: (10c) >> (10a); (10b) >> (10a); (10bb′) >> (10aa′) The next types concern extended marking. Type IV deals with a case of total extended marking, that of Finnish (cf. section 2.4). Cases of partial extended marking are subcases of type IV. If we leave aside the conditions governing the nature of the exponent and the absolute suppletives, for cardinal numerals with hundreds or higher addends, Czech has to stipulate either Ord PHON ([X1,… Xn−1 + Xn]) = [Ord PHON (X1),… Ord PHON (Xn )] or [X1,…,Ord PHON (Xn−1), Ord PHON (Xn )]. Only the latter condition is allowed in Polish. (11) Type IV. (partial account) a. Simple ordinals: Ord PHON (X) = X-s b. Ord PHON (yksi) = ensimmäinen, Ord PHON (kaksi) = toinen c. Complex ordinals: Ord PHON ([X1,…, Xn]) = [Ord PHON (X1),…, Ord PHON (Xn)] d. Overrides: (11c) >> (11a); (11b) >> (11a) The last type is that of Celtic languages, where the structure of the complex numeral determines the place of the exponent. In structure [X1 MRK Xn], where MRK notes a set of elements that the exponence rule is sensitive to (preposition war ‘on’ and conjunction ha ‘and’), the exponent is suffixed to X1 cf. (4a). In other complex structures, it is suffixed on the last element, e.g., an daou.ugent-ved 2.20-ORD ‘40 th’ (Stump 2010). French ordinals present a mix of type I and III, since deuxième ‘2 nd’ is both absolute and conjunct and is not overriden by the absolute suppletive second.

2.4. Idiosyncratic cases The mechanisms conceived of up to now are too simplistic to account for stem allomorphy in certain languages. In Finnish, this phenomenon is pervasive and also occurs in inflection, e.g., kieli ‘language’ GEN = kiele-n, PART = kiel-tä. On the model of nouns, cardinal numerals have four stems, which may be distinct, e.g., ST0 (basic form) kuusi ‘6’, ST1 (inflectional stem) kuute-en ‘6-POSS’, ST2 (inflectional stem) kuude-n ‘6-GEN’,

1522

VII. Semantics and pragmatics in word-formation II: Special cases

ST3 (partitive stem) kuut-ta ‘6-PART’, or not, e.g., neljä ‘4’ (Karlsson 1987: 115). When they are declined, cardinals take the stem that the case requires, e.g., kuude-ssa-toista maa-ssa ‘in sixteen countries’ cf. kuusitoista ‘16’. Ordinal numerals also have four stems the form of which results from regular alternations affecting the ending: /Xs/, /Xnte/, /Xnne/, /Xt/. So for ‘3 rd’ we have kolmas (ST0), kolmante (ST1), kolmanne (ST2) and kolmat (ST3), for ‘6 th’ kuudes, kuudente, kuudenne, kuudet, etc. When declined, the ordinal selects the appropriate stem, e.g., joulukuu-n kahdente-na kymmenente-nä kuudente-na päivä-nä december-GEN 26 th-SUPESS day-SUPESS ‘on the 26 th of December’. Note that suppletive forms (11b) may occur as conjunct but only as the rightmost element of complex numerals, e.g., ‘202 nd’ (kahdes | *toinen).sadas. (kahdes | toinen). To cope with the fact that, e.g., kuudestoista is the ordinal corresponding to kuusitoista ‘16’, rule (11) has to be changed and completed taking into account conditions (12), which basically specify that the default stem for complex ordinals is the cardinal’s Stem 2. Hence, Ord PHON (kuusitoista) toista = Ord PHON (kuude) 4 toista = kuudestoista. (12) a. Simple ordinals: Ord PHON (X) = Ord PHON (XST2) aa′. Ord PHON (XST2) = XST2-s c. Complex ordinals: Ord PHON ([X1,… Xn]) = [Ord PHON (X1),…, Ord PHON (Xn−1), Xn] if Xn = toista, else = [Ord PHON (X1),…, Ord PHON (Xn)] Many Slavic languages also show stem alternation. Cardinal numbers decline on the model of adjectives for the first four Numbers and of nouns for the others, following a well-known typological cline. With higher numerals, a lot of idiosyncrasies crop up. Ordinals of tens are formed on the cardinal’s stem, e.g., Polish pięćdziesiąt ‘50’ → pięćdziesiąty ‘50 th’ (but dwudziesty ‘20 th’ ← dwadzieścia ‘20’), Czech padesát → padesátý ‘50 th’, Serbian pedèsēt → pedèsētī ‘50 th’, Russian pjat’desjat’ → pjatidesjatyj ‘50 th’. In Russian, but not in Czech nor Serbian, a similar phenomenon takes place for hundreds and thousands: the stem of the unit looks like a genitive form (so traditional grammars say), e.g., Russian tri.sta ‘3.102’ → trëx.sot-yj 3\GEN-102\GEN-AZR ‘300 th’. In Polish it occurs when sto ‘100’ is involved, e.g., Polish sto tysiący ‘100.103’ → stu.tysięcz\ny 100\GEN-103\AZR ‘100,000 th’ but cztery tysiące ‘4.103’ → czter-otysięcz\ny 4-RFX-103\AZR ‘4,000 th’. In any case, forms such as Russian trëx, četyrëx, etc. are instances of inherent inflection. It is then less problematic to say that they are morphomes. Postulating allomorphic stems for Greek ordinals belonging to the series of tens and hundreds seems to be the best way to handle the variations we observe. For the tens, the alternation can be formulated as (i) /Xnta/~/Xkos/, e.g., triánta ‘30’ → triakos\tós ‘30 rd’ but we have to resort to suppletion for saránta ‘40’ / tessarakos\tós ‘40 th’. For hundreds, the alternation is completely regular (ii) /Xsia/~/Xsios/, e.g., tetrakósia ‘400’ → tretrakosios\tós ‘400 th’. The last alternation regards thousands: the form of their first element is that of the multiplicative numeral (section 4.2) when it ranges from 4 to 9, e.g., pent-ákis instead of ordinal pémptos for pénte ‘5’ (section 2.2). We need to stipulate that the multiplicative form is used instead of the ordinal in this context. In languages with extended exponence, it happens that the ordinal exponent does not appear on all elements in long ordinals. In addition to Czech, in Finnish long ordinals may occur with the exponent on the last element only, e.g., kolmetuhatta sata kolmekymmentänejänne-n ‘3,134 th-GEN’ (Karlsson 1987: 119), but such forms are considered un-

87. Denumeral categories

1523

grammatical by many speakers. This practice is the rule in Spanish and Portuguese, at least for long ordinals. A form such as Portuguese décimo milésimo seiscentésimo quinquagésimo quarto ‘10,654 th’ is extremely formal and because people know only a handful of simple ordinals (all inherited from Latin), they shift to the cardinal, e.g., Spanish el veinticinco aniversario ‘the 25 th birthday’ instead of el vigésimo quinto aniversario. The ordinal exponent may be affixed to non-numeral bases, mainly (interrogative) words whose meaning has to do with ranking, e.g., Hungarian hánya-dik ‘how-ORD’, French combien(t)-ième ‘how-ORD’, quel-ième ‘which-ORD’, n-ième ‘nth’ C’est la combiennième inondation dans la région ? ‘Which number flood is this in the region?’ (Fradin and Saulnier 2009).

3. Fractional numerals Fractional numerals are two-part expressions, insofar as fractional numbers involve a numerator and a denominator. While the numerator is always a cardinal numeral, the denominator, which is the part that interests us here, is constructed either on a cardinal or an ordinal. The first option is chosen in Hungarian (X → X-Ad cf. Table 87.1), e.g., négy öt-öd ‘4/5’, Basque (X → X-en), e.g., zazpi-r-en bat ‘1/7’ ← zazpi ‘7’, German (X → X-tel ), e.g., ein Zehn-tel ‘1/10’ ← Zehn ‘10’, Czech (X → X-ina), e.g., jedna sedm-ina ‘1/7’ ← sedm ‘7’, Welsh (X → X-fed ), e.g., tri wythfed ‘3/8’ ← wyth ‘8’. The second option is the norm in most languages as shown by the numerals expressing ‘1/7’ and ‘5/10’ in the following sample: Greek én-a évδom-o one.F seventh-NEU, pénte dékat-a five tenth-PL; Polish jedna siódm-a one.F seventh-F, pięć dziesiątych five tenth.F.GEN.PL; Lituanian vienà septint-óji one.F seven-ORD.F, peñkios dešim ˜ t-osios five ten-ORD.F.PL; Swedish en sjunde-del one seven-ORD, fem tiondel-ar five ten-ORD.PL; Dutch een zeven-de one seven-ORD, vijf tien-de five ten-ORD; French un sept-ième one seven-ORD, cinq dix-ièmes five ten-ORD.PL; Romanian o septime ‘one seventh’, cinci zecim-i ‘five tenth-PL’; Albanian një e shtat-a ‘one seventh.F ’, pesë të dhjet-at ‘five tenth-PL’. When the first ordinal numerals are suppletive forms the latter are frequently used to build fractional numerals, e.g., Italian un quint-o one fifth.M ‘1/5’, tre quart-i three fourth-PL ‘3/4’. But in many languages, the first fractional numerals have specific suppletive forms for ‘1/2’, e.g., Hungarian fél, English, Dutch half, Swedish halv, German Halb, Albanian gjysmë, Portuguese meio, metad, French demi, Italian metà, Turkish yarım, Russian, Serbian polovina, Maltese nofs, Welsh hanner, Basque erdi; for ‘1/3’ Russian tret’, Welsh traean, Basque heren, Portuguese terço, Italian terzo, French tiers, Maltese terz; for 1/4 Russian četvert’, Basque laurden, French quart, Italian quarto, Maltese kwart, Albanian çerek. Higher denominators are usually regular numerals, except Basque bortzen ‘1/5’. The Slavic languages also have a special numeral for ‘1 1/2’, e.g., Russian poltora, Polish półtora, etc. The phrases expressing fractional numerals vary from language to language. The following patterns have been observed: (a) CARD FRAC with no variation at all, e.g., German ein Zwanzigs-tel ‘1/20’, dreizehn Zwangzigs-tel ‘13/20’. (b) CARD, FRAC:FLX where the two elements agree since FRAC is a plain noun governed by CARD, e.g., Czech jedn-a sedm-ina ‘1/7’, třináct dvacet-in ‘13/20’. (c) CARD ORD:FLX, the cardinal governs

1524

VII. Semantics and pragmatics in word-formation II: Special cases

the ordinal (a noun or an adjective) which must be inflected, e.g., Portuguese um terço ‘1/3’, cinco décimo-s ‘5/10’; French un dixième ‘1/10’, six dixième-s ‘6/10’. (d) The fractional numeral is an NP explicitly headed by the noun part, or its equivalents, which regularly agrees with the ordinal DET (CARD) ORD:FLX part, e.g., Spanish la quinta parte ‘1/5’, las siete doceava-s parte-s ‘7/12’. (e) The fractional numeral is an NP whose head (= part) has been elided but whose constituents nevertheless agree according to the rules in force with numerals CARD:FLX ORD:FLX (part), e.g., Russian odn-a pjat\aja (čast’) one-F.SG five\AZR.F.SG (part[F.NOM.SG]) ‘1/5’, pjat’ šest\yx (čast-ej) five six\AZR.GEN.PL (part-F.GEN.PL) ‘5/6’. (f ) CARD-FLX CARD, the relation between the numerals expressing the numerator and the denominator is marked by a case, e.g., Turkish üç-te iki threeINESS two ‘2/3’. Structure CARD bölü CARD, e.g., iki bölü beş 2 on 5 ‘2/5’ is also available.

4. Non-strict numeral denumerals 4.1. Derivation of collectives Whereas cardinal numerals denote combinations of individual entities, collective numerals denote combinations of groups of entities (Ojeda 1997). Hence constrasts such as Islandic tveir sokkar ‘two socks’ (cardinal) vs. tvennir sokkar ‘two pairs of socks’ (collective). In addition to counting groups of individuals, collectives may be used to count individuals belonging to the same group (or kind), e.g., Serbian sedmoro dècē ‘7.COLL children’. The semantic variation observed with collectives is tied to the way the group of entities is constituted. Besides Icelandic, e.g., einn/enir, tveir/tvennir, ƿrír/ƿrennir, fjórir/fernir, collective numerals are found in the Balto-Slavic languages. The derivational nature of collective numerals can be ascertained only if they form long enough series and present a recurrent pattern of affixation. This is the case in Polish and in Serbian, e.g., Serbian dvoje, troje, četvoro, petoro, šestoro, sedmoro, osmoro, devetoro, desetoro, pedesetoro …; Polish dwoje, troje, czworo, pięcioro, sześcioro, siedmioro, ośmioro, dziewięcioro, jedenaścioro, dwanaścioro … dziesięcioro, dwadzieścioro, trzydzieścioro, etc. In Russian and Czech, they seem limited to the first ten numbers. It is generally assumed that collective numerals used alone imply the notion of ‘group’, e.g., Serbian devedesetoro ‘group of 90 people’. It is not so when they modify a noun, since their use is almost always grammatically constrained: they must be chosen when the noun is a pluralia tantum, e.g., Russian troe sanej ‘3:COLL sledge:GEN.PL’ vs. *tri sanej, ‘3:CARD ~’ or when it is a neuter noun denoting humans, e.g., Polish Widzę pięcioro dzieci vs. *Widzę pięć dzieci ‘I see (five.COLL | *five.CARD) children’. On the basis of such cases, Saloni argues that collective and cardinal numerals constitute one and the same paradigm in Polish since the distribution of the former is stricly conditioned by the (sub-)gender of the noun they apply to (Saloni 2010). This view is supported by the fact that collective numerals in Polish, do occur with complex numbers, in contradistinction to Russian, e.g., Polish pięćdziesięci-oro czw-oro dzieci ‘54 children’ (with exponent on both addends). In Serbian, however, collective numerals are restricted neither to human, nor to plurale tantum, e.g., Serbian pet-oro pasa ‘5 dogs’. Lithuanian has two series of collective numerals. Series (i) vieneri, dveji, treji, ketveri, penkeri, šešeri, septyneri, aštuoneri, devyn-

87. Denumeral categories

1525

eri (from 1 to 9) is used with pluralia tantum and pairs, while series (ii) dve˜jetas, tre˜jetas, ke˜tvertas, peñketas, še˜šetas, septynetas, aštuonetas, devynetas is used for groups. The above mentioned numerals in Polish and Serbian convey the additional meaning that the NPs they occur in refer to sexually mixed groups of persons. In Bulgarian, collectives ending in -mina have the same effect or exclude a female only referent. Serbian also has a series of collective (feminine) nouns, regularly formed by suffixing -ica onto the collective numeral of the first series, e.g., Serbian dvoj-ica ← dvoje ‘2.COLL’, četrnaestòr-ica ← četŕnaestoro ‘14.COLL’, which is used only when the collective number refers to male individuals. These nouns yield an agreement mismatch: their determiner is marked feminine singular while the number of their predicate is singular or plural, e.g., Serbian sv-a su petor-ica (doš-la | doš-li) (Meillet and Vaillant 1952: 128) all-F.SG be:PRS.3pl 5.COLL-NZR.F.SG (come-F.SG | come-M.PL) ‘all five (men) came’.

4.2. Derivation of multiplicatives In most languages, the semantic domain of multiplicatives is temporal and their meaning can be roughly represented as (i) MultT (cardinal_numeral) = ‘ times’. They are generally derived from (a special stem of ) cardinal numerals and their distribution is that of adverbials. Vestigial in English, e.g., twice, thrice, multiplicatives exist in other Germanic languages, e.g., German (X → X-mal ) ein-mal ‘once’, zehn-mal ‘10 times’, Dutch (X → X-maal or X → X-voud ) drie-maal / drie-voud ‘thrice’, vier-maal / viervoud ‘4 times’. We also find them in the following languages: Greek (X → X-ákis), e.g., pent-ákis ‘5 times’ (slightly old-fashioned), Lituanian (X → X-kart), e.g., ke˜turis-kart ‘4 times’, Hungarian (X → X-szOr), e.g., ötven-szer ‘50 times’, Welsh (X → X-waith), e.g., deng-waith ‘10 times’, Akhvakh (X → Xče), e.g., k’e-če ‘2 times’, Czech (X → X-krát), e.g., tři-krát ‘3 times’, Polish (X COLL → X COLL-krotnie, †X-kroć), e.g., trzykrotnie ‘3 times’ (cf. also wielo-krotnie ‘many times’, często-kroć ‘frequently’). The latter forms are semantically related to derived adjectives (X COLL → X COLL-krotny), e.g., Polish pięcio-krotny ‘repeated 5 times’. From five onward, these adjectives are formed on the short stem of collective numerals. Some languages have spatial multiplicatives, the meaning of which can be expressed as (ii) MultS (cardinal_numeral) = ‘has similar parts’. Basque illustrates this case (X → X-koitz), e.g., hiru-koitz as in arrazoinamendu hiru-koitz-a reasoning 3-MULT-DEF ‘a reasoning in three parts’. Polish has two processes of this type, one derives adjectives by circumfixation onto the stem of collective numerals (X COLL → po-X COLL-ny), e.g., Polish po-czwór-ny ‘with 4 parts’ potrójna porcja ‘triple portion’, the other by suffixation onto the same stem (X COLL → X COLL-aki), e.g., dwoj-aki ‘twofold’, czwor-aki ‘fourfold’, etc.

4.3. Derivation of distributives Derived distributives exist in Basque and Akhvakh. Their semantics corresponds to (i) Distr(cardinal_numeral) = ‘ apiece’, e.g., Basque hiru-na ogi three-DISTR bread ‘three (loaves of ) bread apiece’ (Hualde and Ortiz de Urbina 2003: 128). In Basque the rule of exponence suffixes /na/ to the cardinal’s stem (X → X-na), e.g., hiru-na ← hiru

1526

VII. Semantics and pragmatics in word-formation II: Special cases

‘3’ and these forms are determiners. In Akhvakh, it involves the reduplication of the cardinal’s stem, e.g., Akhvakh išt wištuda ‘(give) five apiece (to each)’ ← ištu-da ‘5’.

5. Non-numeral denumerals 5.1. Approximative denumerals Approximatives are attested in Romance and some Slavic languages. As expected, they are mainly based on cardinals corresponding to tens. Their meaning can be uniformly represented as (i) Appr(cardinal_numeral) = ‘ ± n’, where n varies in function of the value of ‘number’, e.g., French une quinzaine de voitures ‘about 15 cars’ i.e. ‘15±2’ vs. une centaine de voitures ‘about 100 cars (±10)’. In Czech (X → X-ka), they are used to refer only to humans, e.g., dvacít\ka ← dvacet ‘20’, but not in Serbian (X → X-ak), e.g., deset-ak ← deset ‘10’, dvadeset-ak ← dvadeset ‘20’, stotin\ak ← sto ‘100’ petnaest-ak kolača ‘about 15 cakes’. In French (X → X-aine1), the series is longer, e.g., huit-aine ← huit ‘8’, douz-aine ← douze ‘12’, cinquant-aine ← cinquante ‘50’, nonant-aine ← nonante ‘90’ (in Belgium and Switzerland), etc. In non-standard French, approximatives can be formed on many more cardinals, e.g., dix-sept-aine ← dix-sept ‘17’, quatre-vingt-aine ← quatre-vingt ‘80’, cinq cent-aine ← cinq cent ‘500’, etc. (Saulnier 2010). Such wealth of forms is observed neither in Ibero-romance, nor in Italian, e.g., Spanish quinc-ena ← quince ‘15’, veint-ena ← veinte ‘20’, cent\ena | cent\en-ar ← ciento ‘100’; Italian dec-ina ← dieci ‘10’, quindic-ina ← quindici ‘15’, vent-ina ← venti ‘20’, cent-in-aio ← cento ‘100’.

5.2. Exhibitive denumerals Exhibitives are derived nouns denoting an entity somehow explicitly correlated with a particular Number. In Slavic languages, names of playing cards are formed on (the stem of the collective) numeral corresponding to the value of the card, e.g., Bulgarian dvojka karo ‘2 of diamonds’, osmor-ka kupa ‘8 of hearts’, Polish trój-ka pik ‘3 of spades’. The same forms are used for the marks given to pupils at school, e.g., Bulgarian učiteljat mu pisa dvoj-ka po matematika ‘the teacher gave him two in mathematics’, Polish dostałem czwór-kę ‘I have got a 4’; also German (Austria) Eins-er, Zwei-er, Drei-er, Vierer, Fünf-er ‘1’, ‘2’, etc. and Czech dvoj-ka, šest-ka ‘2’, ‘6’, etc. The latter also name trams, e.g., Czech jednička, třináctka ‘tram 1, 13’. In Russian exhibitives are used for cards and for grades up to 10. In Serbian, pupils of the nth class can be denoted by an exhibitive noun formed on the corresponding ordinal numeral, e.g., Serbian drug-ak, treć-ak ‘pupil of the 2 nd, 3 rd class’. Finnish developed special forms for school grades, e.g., ykkönen ‘1’, kakkonen ‘2’, nelonen ‘4’, viitonen ‘5’, kuutonen ‘6’, seiska ‘7’, kasi ‘8’, ysi ‘9’, kymppi ‘10’. These forms also denote entities for which the corresponding number is a distinctive feature, e.g., sata kuutonen ‘bus 106’, kymppi ‘10 euros bill’.

87. Denumeral categories

1527

5.3. Appellative denumerals A crucial property of some entities is their being composed of n parts, repeated n times, etc. Appellative denumerals precisely provide the entity a name based on the numeral denoting number n. Groups of people, animals engaged in a common activity are frequently denoted by appellatives: teams, e.g., Russian troj-ka ‘(sledge) drawn by 3 horses’; music groups, e.g., Hungarian vonós négy-es ‘string quartet’, fúvós öt-ös ‘wind quintet’, etc. both derived from cardinal numerals (X → X-As), Italian duetto ‘duet’, quint-etto ‘quintet’, etc. (← ORD-DIM). Activities or artefacts involving n parts, measures, etc. also fall in the realm of appellatives, e.g., Albanian nëntëshe ‘series of 9 prayers’, French siz-ain ‘poem of 6 verses’, vingt-deux-ain ‘cloth with 22 hundreds of thread’, neuv-aine2 ‘pious excercice made during 9 days’, Italian terz-ina ‘tercet’, quart-ina ‘poem of 4 verses’, Serbian sedmer-ac ‘free-throw shot from 7 meters (handball)’. In French suffixing -aire onto the cardinal’s learned stem is a way to form adjectives expressing the age, e.g., quaranten\aire fourty\AZR ‘40 year old’, which coexists with the older form based on loan translations from Latin, e.g., quadragénaire ‘40 year old’ < Latin quadragenarius. By metonymy, these forms also denote the person with the corresponding age (Fradin and Saulnier 2009: 216−220).

6. Numeral-based compounds In all language families but Romance, compounding with numerals is a common way to form adjectives indicating that the noun they modify possesses what the base noun’s referent denotes in n examples. The general pattern looks like (i) NUM-N-AZR, e.g., German fünf-tür-ig five-door-AZR ‘with 5 doors’. The most frequent domains denoted by the noun are: age or time, e.g., Hungarian négy-év-es four-year-AZR ‘4 years old’, Polish dwu-let-ni, Russian dvux-let-nij, Dutch twee-jar-ig two-year-AZR ‘2 year old’, Finnish kolme-vuot-ias three-year-AZR ‘3 year old’, Dutch drie-daag-s ‘3 day long’; part of a functional whole, e.g., Hungarian négy-lab-ú four-leg-AZR ‘quadruped’, három-árbóc-os hajó three-mast-AZR ‘three-masted boat’, Polish cztero-list\ny four-leaf\AZR ‘fourleaved’, Finnish kymmen-ikkuna-inen ten-window-AZR ‘with 10 windows’, Greek examel-is six-member-AZR ‘with 6 members’; others: Russian dvu-jazyč\nyj twolanguage\AZR ‘bilingual’, dvoe-muž-ie 2.COLL-husband-NZR ‘biandry’, kuusi-lapsi-inen six-child-AZR ‘with 6 children’. Derived adjectives also exist, e.g., French bis\annu-el 2\year-AZR ‘lasting 2 years’. Some of these compounds have been lexicalized as nouns, e.g., Greek exá-psalm-os six-psalm-AZR ‘hymn of 6 psalms’, Serbian tro\međa three\border ‘3 border (point)’, Spanish quince-añ-er-a 15-year-NZR-F ‘15-year-old girl’.

Acknowledgements I gratefully thank my informants, Astrid Alexander-Bakkerus, Peter Arkadiev, Michel Aurnague, Gabriela Bîlbîie, Denis Creissels, Hans-Olaf Enger, Mats Forsgren, Zsuzsana Gécseg, Teresa Giermak-Zielińska, Aslı Göksel, Anna Mańkowska, Alexandru Mardale, Dragomir Milošević, Kaarina Pitkänen-Heikkilä, Tomorr Plangarica, Svetlana Sokolova, Dejan Stosic, Jana Strnadová, Sophie Vassilaki, Madeleine Voga.

1528

VII. Semantics and pragmatics in word-formation II: Special cases

7. References Corbett, Greville 2010 Canonical derivational morphology. Word Structure 3: 141−155. Fradin, Bernard and Sophie Saulnier 2009 Les cardinaux et la morphologie constructionnelle du français. In: Bernard Fradin, Françoise Kerleroux and Marc Plénat (eds.), Aperçus de morphologie du français, 199−230. Saint-Denis: Presses Universitaires de Vincennes. Göksel, Aslı and Celia Kerslake 2005 Turkish. A Comprehensive Grammar. London/New York: Routledge. Hualde, José Ignacio and Jon Ortiz de Urbina 2003 A Grammar of Basque. Berlin/New York: Mouton de Gruyter. Huddleston, Rodney and Geoffrey K. Pullum (eds.) 2002 The Grammar of the English Language. Cambridge: Cambridge University Press. Karlsson, Fred 1987 Finnish Grammar. 2 nd ed. Helsinki: Söderström. Meillet, Antoine and André Vaillant 1952 Grammaire de la langue serbo-croate. 2 nd ed. Paris: Champion. Ojeda, Almerindo E. 1997 A semantics for the counting numerals of Latin. Journal of Semantics 14(2): 143−171. Saloni, Zygmunt 2010 So-called collective numerals in Polish (in comparison with Russian). Studies in Polish Linguistics 5: 51−64. Saulnier, Sophie 2010 Les nombres. Lexique et grammaire. Rennes: Presses Universitaires de Rennes. Stump, Gregory T. 2010 The derivation of compound ordinal numerals: Implications for morphological theory. Word Structure 3(2): 205−233. Trépos, Pierre. 1994 Grammaire bretonne. 3 rd ed. Brest: Nevez and Breiz.

Bernard Fradin, Paris (France)

88. The semantics and pragmatics of Romance evaluative suffixes 1. 2. 3. 4. 5. 6. 7. 8. 9.

Introduction Pragmatic analysis ante litteram European structuralism The shortcomings of structural linguistic analysis Morphopragmatics Semantics and pragmatics Denotative variation and lexicalization Oral familiar ingroup- and outgroup communication References

88. The semantics and pragmatics of Romance evaluative suffixes

1529

Abstract The article provides an historical outline of the theoretical work on evaluative suffixes in Romance. The discussion spans over the first efforts made within stylistics (Leo Spitzer), early pragmatics (Amado Alonso), European structuralism (Eugenio Coseriu) and recent pragmatic and morphopragmatic approaches. It will be shown that the respective theories account for crucial shortcomings.

1. Introduction Evaluative suffixes are essentially used to vary in a subjective way and for specific pragmatic purposes the concept expressed by a word (e.g., Sp. abogado ‘lawyer’ → abogadillo ‘strange, bad and possibly ridiculous lawyer’). Subjectivity may be faded out when the new word is meant to denote a specific object, as in Sp. tornillo ‘screw’ ← torno (13th century) ‘turning instrument’ or It. telefonino ‘mobile phone’ ← telefono ‘telephone’. In this case, the diminutive suffix loses its subjective pragmatic force, and even the conceptual restriction of ‘smallness’ may be reduced in favor of naming of a special type (variety) of object, although smallness is generally respected in terms of prototypicality, that is, Sp. tornillo will usually and prototypically refer to a small screw. The traditional terms diminutive and augmentative focus on a semantic concept, but ignore important pragmatic aspects. In this sense, the generic term evaluative suffix opens the perspective for subjective and pragmatic aspects. Moreover, scientific research was traditionally concentrated on diminutives and, subsequently, their conceptual counterpart, the augmentatives, consigning other evaluative suffixes to the sidelines. Hence, the term evaluative suffix itself sheds new light on a group of morphemes that was basically considered as the class of diminutives plus satellites. In view of this, Charles Bally’s (1965: 248−252) consideration of the entire group of evaluative suffixes (Fr. suffixes appréciatifs) marked a theoretical progress in history. The Swiss linguist defined them as suffixes “that express emotions or evaluations triggered by the concept of the stem” [my translation]. This matches with his morphological analysis: evaluative suffixes modify (“determine”) the stem, which functions as head of the word-formation. In line with this, Romance evaluative suffixes usually adopt the gender of the noun that underlies the stem, but gender variation is not excluded (Dressler and Merlini Barbaresi 1994: 94−96; e.g., Port. uma mulher f. ‘a woman’ → um mulherão m. ‘impressive woman’). Bally subcategorized evaluative suffixes into diminutives and augmentatives, on the one hand, and amelioratives and pejoratives, on the other. The first group associates positive or negative affection with a dimensional concept, whereas the second group apparently relates to affective evaluation only, but dimension and intensity play a certain role in the second group as well (e.g., Fr. faiblard ‘too weak’ ← faible ‘weak’). Most authors opt for a ternary subcategorization into diminutives, augmentatives, and pejoratives (e.g., Real Academia Española 2009; Lázaro Mora 1999). Mihatsch (2010: 114−118) underlines the semantic and pragmatic effect of “approximation”. If the successors of the Latin elative -issimus are included, as I claim, a fourth group of intensifiers appears, whereas amelioration seems to be a secondary effect of diminutives, augmentatives and intensifiers. Amelioration as a primary effect falls essentially into the domain of prefixation (e.g., Fr. super-, hyper-, méga-, extra-).

1530

VII. Semantics and pragmatics in word-formation II: Special cases

Evaluative suffixes display an important morphological and functional variety. It is difficult, therefore, to give a brief account of the relevant morphemes. Lüdtke (2011: 453−486) provides a recent diachronic and synchronic synthesis of evaluative suffixes in Latin and Romance (see also the classic study by Hasselrot 1957). Depending on the context, all productive diminutives and augmentatives may convey negative (pejorative) or positive (ameliorative) evaluation, with suffix-specific preferences. In Romance, and especially in Brazil, augmentative suffixes are not restricted to pejorative functions, as Schneider (1991: 239) assumes from a universal point of view. Hence, evaluation is the central axis of the group from which positive or negative evaluations emerge as connotative variations according to established preferences, context and interpretation (cf. Merlini Barbaresi 2004: 283, for It. -ino). We can even say that the axis has one endpoint where negative evaluation passes from connotation to denotation. This is the case of pejoratives. Curiously, the positive denotative equivalent, that is, amelioration, does not exist in the domain of suffixation. Intensity appears to be a second axis, since upgrading and downgrading are possible variants, and the forms that developed from Lat. -issimus may be considered a denotative endpoint for high intensity. Finally, we might view diminution and augmentation as variations situated on the axis of dimension, with denotational diminution and augmentation as the two opposed endpoints. This axis includes the metaphorical interpretations ‘diminution → mitigation’ and ‘augmentation → intensification/emphasis’, which bring them in touch with the second axis. If we exclude lexicalized words, we may say that along all three axes, all variants, including the denotational ones, involve a personal, subjective interpretation or belong to a language marked by familiarity that allows for subjectivity. In fact, dimension, degree and intensity are not expressed as descriptive (objective) but as evaluative (subjective) features. Hence, subjective evaluation belongs to the class-meaning of these suffixes. The subjective implicatures of all evaluative suffixes have direct consequences for the commitment of speaker and hearer. While the speaker may emphatically convey a strong personal commitment, the hearer’s commitment is left up to him, since it is personal and subjective in the case of evaluative suffixes. The other side of the coin is the lack of commitment to truth or, in semantic terms, to description, that is, the subjective effort to represent things with a maximum amount of objectivity. Even if not intended by the speaker, the hearer may infer a rather loose relation to truth and objective description. Therefore, the effect of approximation may affect both the semantic description and the personal attitude of speaker and hearer. This is not the case with evaluative adjectives, for example in Sp. Juan es tonto ‘Juan is stupid’, where the speaker not only expresses a personal point of view but also affirms it in terms of truth. This affirmation challenges the hearer, while evaluative suffixes create mutual tolerance. No matter if one uses pequeñito (← pequeño ‘small’ + diminutive) or pequeñazo (‘small’ + augmentative), the judgment is subjective, be it placed on the conceptual, the affective or the interactive level of commitment. In the case of evaluative suffixes, communication is not committed and not committing. Hence, mitigation and approximation cannot be exclusively reduced to metaphorical extensions of the concept of smallness, as held by Jurafsky (1996), but have to be related to the categorial function of subjective evaluation. Again, the fact that most studies only focus on diminutives accounts for methodologically biased results. The only thing we can say is that diminutives are particularly suitable for mitigation, as minoration and lack of commitment go hand in hand, thus reinforcing the mitigating effect. Dressler and Merlini Barbaresi (1994: 144) and Waltereit (2006: 112−118) stress

88. The semantics and pragmatics of Romance evaluative suffixes

1531

that diminutives are prototypically used in “non-serious” communication. While mitigation may be more prototypical for diminutives, the lack of gravity is certainly valid for all evaluative suffixes. Textlinguistic data reflect the subjective intention of evaluative suffixes. The more objective a text is intended to be, the less (productive) evaluative suffixes are used (Würstle 1992: 233). In legal texts, for example, no productive evaluative suffixes should be found.

2. Pragmatic analysis ante litteram 2.1. Idealism and early pragmatics: Amado Alonso The Spanish-Argentinian linguist Amado Alonso published two articles on diminutives in Spanish. The first (1930), which was a first version of the second (1961 [1935]), is almost never quoted, whereas the second turned out to be the landmark study on the topic for Spanish and even Portuguese until today. Interestingly, this first theoretical essay on evaluative suffixes in Romance adopted a radical pragmatic point of view. In contrast to the general tendency of the 20th century, where pragmatics followed structuralism, Alonso (1961 [1935]) claimed very early that diminutives have no conceptual fundament but only context dependent pragmatic functions driven by emotion. He thereby challenged the traditional assumption that affectivity develops metaphorically from conceptual smallness. Alonso did not use the term pragmatics, nor did he refer to Karl Bühler, but his analysis was clearly inspired by Bühler’s (1982: 24−33) communicative semiotic triangle (“Organonmodell”), which is today considered a precursor of the communication models used in pragmatics. Alonso was a disciple of Menéndez Pidal, the spiritus rector of historical linguistics in Spain, but as far as the relevant theory is concerned, he was strongly influenced by the idealistic approach of Karl Vossler and the psycholinguistic theory of Karl Bühler (see Neumann-Holzschuh 2009). Both theories appear to be fruitful for the analysis of evaluative suffixes, but conflict on a crucial point, as we shall see below. As for the primacy of emotional functions, Alonso was inspired by Wrede’s analysis of German diminutives (1908: 127−144). Wrede claimed that all Germanic diminutives derive from hypocoristic names without original diminutive function. It should be noted as well that Alonso, despite overtly claiming to analyze only the function of diminutives, integrated examples of all evaluative suffixes into his work. Hence, his theory should be valid for the entire group. According to Bühler, a linguistic sign represents a real world object or state of affairs (symbolic function), reflects the state of mind of the speaker (symptomatic function), and appeals to the hearer (signaling (appealing) function). Alonso stressed the appealing function, alluding to the “active” force of diminutives, a function we might now name illocutionary force. Moreover, diminutives are supposed to highlight the referent, providing exceptional communicative relevance to the symbolic dimension as well. Finally, the evaluative suffix is symptomatic insofar as it mirrors the emotional attitude of the speaker. In sum, evaluative suffixes provide a subjective-emotive perspective on the referent that aims at exerting a perlocutionary effect on the interlocutor. Consequently, Alonso’s analysis combines the pragmatic strategies directed to the interlocutor with the idealistic standpoint that views the object as a subjective creation

1532

VII. Semantics and pragmatics in word-formation II: Special cases

of the speaker rather than a representation of an independently given object. More than representing a referent, a word like Sp. abogadillo indeed creates a subjectively colored thing meant. This is the point where the idealistic approach conflicts with semiotic models that consider meaning as a mere symbolic representation of a referent. Dressler and Merlini Barbaresi (1994: 50, 87) recognize the pioneering work of Alonso for the analysis of pragmatic strategies, but they overlook his not less important idealistic approach. Notwithstanding the justified critiques directed to linguistic idealism in other domains, it is essentially adequate for the study of evaluative suffixes. Though the symbolic function presupposes a sign-specific fusion of objectively represented and idealistically projected semantic features, the second prevails with evaluative suffixes (cf. recently Fretel 2010, for a not overtly declared idealistic analysis). Obviously, nouns like idiot express subjective evaluations as well. In the case of evaluative suffixes, however, the thing meant is suggested by language and context rather than concretely expressed. This is the point where Alonso’s idealistic view meets pragmatics. The fact that the interlocutors are obliged to retrieve the vaguely suggested subjective thing meant, explains the manifold iridescent implicatures of evaluative suffixes, and especially the difficulties of interpretation that immediately create an “active” suspense between speaker and hearer. Using abogadillo, the speaker playfully invites the hearer to find out what he could mean, creating a situation of intimate complicity (see section 8). Excluding polemically the conceptual function of evaluative suffixes, Alonso completely eliminated the symbolic function itself, at least insofar as the representative function is concerned. In Dressler and Merlini Barbaresi’s (1994) one-sided reception of Alonso’s analysis, evaluative suffixes are considered linguistic signs that are functionally loaded by the communicative tension that links speaker and hearer in a given situation. For Alonso, evaluative suffixes are more than coined pragmatic strategies: they are linguistic signs that convey emotions, albeit not concepts. Alonso’s classification of the pragmatic effects and situations was almost tentative (cf. Walsh’s 1944 similar classification of American Spanish examples; for Portuguese see Skorge 1956/58): − − − − − −

Politeness: ¡Entre usted despacito! ‘Come in. Take your time’ (← despacio ‘slowly’) Intimacy: Ya estamos los dos solitos ‘Now we are finally alone’ (← solo ‘alone’) Tenderness: Juanito (hypocoristic name) ‘dear Juan’ (← Juan) Mockery: abogadillo ‘lousy lawyer’ (← abogado ‘lawyer’) Begging: ¡Una monedita, por favor! ‘Some money, please!’ (← moneda ‘coin’) Modesty (cf. mitigation): bastantito ‘a good deal of something’ (← bastante ‘quite a lot’, un favorcito ‘a (small) favor’ (← favor ‘favor’) − Humour: see Spitzer’s example in section 2.2 − Ludic variation: chiquito, chiquitiquito, chiquitiquillo, etc. (← chico ‘small’)

2.2. Stylistics and early textlinguistics: Leo Spitzer The playful idealistic projections of evaluative suffixes on referents are richly explored in literature. Consequently, Alonso’s pragmatic starting point joins the prior tradition of stylistic analyses in Romance philology, namely the positions of the Austrian Leo Spitzer. In fact, Spitzer (1921) motivated Alonso (1930), to which Spitzer (1933) in turn

88. The semantics and pragmatics of Romance evaluative suffixes

1533

replied. Spitzer underlined the ludic character of evaluative suffixes and their capacity to reflect the speaker’s mood. In the following Spanish canto popular, a mother prays to Saint Christopher, asking him to give her a good son-in-law, and, having got one, gives a sad mood variant of the same poem (Spitzer 1921: 202): San Cristobalito Manitas, patitas, Carita de rosa, Dame un nobio pa mi niña, que la tengo mosa.

Dear Saint Christopher Nice hands, nice feet Sweet, rosy face Give me a son-in-law for my daughter, ‘cause she’s unmarried.

San Cristobalón Manazas, patazas, Cara de cuerno, Como tienes la cara me distes el yerno.

Damned Saint Christopher Rough hands, rough feet Horned face How could you do this to me and give me this son-in-law.

In the first verse, the diminutive -ito matches the hopeful praying of the mother, while in the second, the augmentatives -ón and -azo convey her disappointment. To put it in Spitzer’s terms: tonality changes from major to minor. In the example, the evaluative suffixes not only modify the word they belong to, as one would expect according to Bally’s definition, but a whole sentence or text. Spitzer therefore suggests the terms Satzdiminutive (sentential diminutives) and Satzpejorative (sentential pejoratives) that are located in one word but “color” a whole sentence (1918: 108−110, 1921: 201−202, and 1933). Analogously to Alonso, Spitzer explicitly focuses on diminutives but considers all other evaluative as well. More specifically, Spitzer (1918, 1921, 1933) and Alonso (1961 [1935]) claimed that sentential coloring is a general feature of evaluative suffixes that combine with adverbs, which almost always involves diminutives. In Spanish, adverbial diminutives are usually perceived as an American Spanish peculiarity, e.g., Si acasito muero ‘If I should die for some reason’ (← acaso ‘by chance’), probably because the American varieties have conserved the original rural traditions better than European Spanish. Spitzer’s examples show that adverbial diminutives are also commonplace in Portuguese. Modern linguistics extends the sentential function to that of “pragmatic markers” in the interactive situation of speech (e.g., Günthner and Mutz 2004; Mutz 2000). Merlini Barbaresi (2004: 280) even claims that this is a general feature of all productive usages. The inclusion of sentential evaluative suffixes determined by text and pragmatic evaluative suffixes determined by situation invalidates Bally’s definition that limits the functioning to two components of the word: the stem and the suffix. In his review of Alonso (1930), Spitzer (1933) points out that Alonso’s position differs from his own insofar as the Spanish-Argentinian linguist does not limit the function of the evaluative suffixes to the speaker’s mood, but stresses their active role with respect to the hearer. Spitzer (1933) essentially coincides with Alonso, although he argues that the symptomatic and signaling functions are two sides of the same coin. According to Spitzer, it is difficult to decide whether linguistically expressed emotion reflects the speaker’s mood or is used as a strategy that acts on the hearer. Consequently, evaluative suffixes color discourse as a whole. The different positions of these authors

1534

VII. Semantics and pragmatics in word-formation II: Special cases

can be attributed to the fact that Alonso’s argumentation is based on (a vision of ) real communication, whereas Spitzer’s stylistic studies focus on literary communication where a fictitious real life communication is mediated by both the narrator’s external presentation and the reader’s external perception. Hence, Bühler’s appealing function cannot be negated if we consider Spitzer’s example as real acts of communication. However, in literature the appeal is not directed to the reader but to a character of the text. Yet, from a stylistic point of view, the reader’s outside perception is decisive. Hence, they are felt as mirrors of mood rather than “active” devices.

3. European structuralism Structural linguistic analysis displayed on the one hand an increase of theoretical coherence compared to Alonso and Spitzer. On the other hand, it had to pay the price for its theoretical limitations, especially those that would be pointed out by modern pragmatics. Not surprisingly, Alonso’s pragmatic standpoint provoked a lively theoretical discussion amongst structural linguists following the tradition of Ferdinand de Saussure (1983). Again, the argumentation was mainly based on Spanish, and secondarily on Portuguese. Max Leopold Wagner (1952) was the first to oppose Alonso’s view that diminution cannot be ruled out as a function of diminutives, as diminutives occur without affective connotations as conceptual diminishers. He recognizes, however, that the affective function can stand alone as well. In the same vein, Monge (1965) claimed that Alonso was correct insofar as the level of parole (speech, utterance) is concerned. At the level of langue (language system), however, the Spanish linguist claimed that diminutives have both a subjective and a diminishing function, and, alluding to Bally, that they share the first feature with all evaluative suffixes. Monge presented this standpoint in 1962 at the 10th International Congress of Romance Linguistics and Philology. During the discussion of Monge’s paper, the upcoming leading authority of structural linguistics in Romance, Eugenio Coseriu, held that the “objective” diminishing function should be considered the basic meaning of diminutives at the level of system, and the affective component an occasional feature at the level of utterance. He argued that different values like affection, irony, aversion, and disdain are necessarily determined by context, not by the language system, since they can be expressed by the same diminutive (Monge 1965: 147). Later, the Romanian linguist specified that the affective features are triggered in those cases where objective conceptual diminution is not possible (1988: 189−192). Coseriu’s position was criticized by Hummel (1994), still from a structural linguistic standpoint. Affective evaluation, Hummel argued, is often combined with objective diminution. Moreover, all linguists agree that diminutives like It. -ino, Pg. -inho, and Sp. -ito prototypically combine both features, ‘small’ and ‘nice’. Hence, there is clear empirical evidence against the assumption that the affective “meaning” is triggered only in those cases where objective diminution is impossible. Hummel (1994) suggested an empirical approach. According to the principles of Saussure, linguistic function and meaning should appear as an invariable feature in representative linguistic data. Ettinger (1980) tested Coseriu’s hypothesis on the basis of a large corpus of literary data from French, Portuguese, Spanish, and Romanian (for Romanian he used an inverse dictionary). In his conclusion, he stated that linguistic research had not been able to describe diminu-

88. The semantics and pragmatics of Romance evaluative suffixes

1535

tives and augmentatives at the level of language system (1980: 198). Hummel (1994) argued against Coseriu and Ettinger that two features might be considered invariable: productive evaluative suffixes always highlight a word and signal to the hearer that a subjective evaluation has to be chosen according to the co-text and the pragmatic conditions of the utterance. These features are not specific for diminutives but categorial features that hold for all evaluative suffixes. Consequently, instead of considering affectivity to be an occasional feature of diminutives at the level of parole, as assumed by Coseriu and Monge, Hummel attributed it not only to the system meaning of diminutives, like Monge, but to the category of evaluative suffixes, that is, to a level superior to the subgroup of diminutives. As for diminutives, Hummel argued that the contextual effects can be explained by the interplay of general categorial features, the subgroup-specific concept of diminution (including metaphor, irony, etc.), contextual and situational factors. Coseriu’s argumentation was based on the evidence that contradictory values like affection, irony, aversion, and disdain could only be attributed to context. This evidence is clearly misleading, since it fails if more abstract features like subjective evaluation and highlighting, which are shared by affection, irony, aversion, and disdain, are taken into account. The highlighting function is assumed by Spitzer for Romance (“Individualisierung”), by Alonso for Spanish (“el destacar la representación del objeto”), and as the most important feature of French diminutives by Weber (1963, “surparticularisation”). Alonso was inspired in this respect by Wrede’s (1908: 135) analysis of German suffixes (“verschärfte Individualisierungen”).

4. The shortcomings of structural linguistic analysis In a certain sense, the shortcomings of the structural linguistic approach to evaluative suffixes provide more insights into the phenomenon than its results.

4.1. The exclusion of the referential function of linguistic signs The structuralist analysis yields systemic linguistic relations and excludes reference to the extra-linguistic reality. Thus, the referential and pragmatic motivations of using a linguistic sign are not focused on by this line of research, even if the theoretical concept of parole would include them. Now, referential and pragmatic features are crucial for the understanding of evaluative suffixes. In the Spanish NP casa pequeña ‘small house’, the morphologically free diminishing adjective pequeño does not modify the word casa, but its referent, that is, the house that is referred to in a given utterance. In contrast to this, in casita the morphologically bound diminutive suffix -ito modifies first the stem (cf. Coseriu himself in Monge 1965: 147; Schneider 1991: 234). This explains why words that are modified with an evaluative suffix tend towards lexicalization, while NPs like casa pequeña do not. The only point in common is that casa is the head for both casa pequeña and casita. With evaluative suffixes, the type of modification that first concerns the stem and then the possible referents supports the idealistic standpoint of Alonso. This does not apply to casa pequeña, which provides a more descriptive representation of the object. The same holds for the opposition of the relative superlative to

1536

VII. Semantics and pragmatics in word-formation II: Special cases

the elative (e.g., It. il piú alto ‘the highest’ vs. altissimo ‘extremely high’). Now, if the structuralist analysis ignores the referential function, the notion of modification becomes unspecific, since it is not clear what is modified, the stem or the referent. The specific function of evaluative suffixes is better reflected by terms like alteration (Dressler and Merlini Barbaresi 1994, by analogy to the usual term It. alterazione) or variation (Gauger 1971: 124−125, 136). Variation seems to be more adequate, since it insists on the active role of the speaker, and it is traditionally open for two types of results: productive variants and lexicalized varieties, whereas alteration has passive and even negative connotations.

4.2. The limitation to diminutives The second shortcoming stems from the fact that the scope of interest was reduced to the study of diminutives, and, occasionally, augmentatives. This reductionist approach certainly has a longer tradition, but, in the case of structural linguistics, we also have to relate it to the exclusive interest for semantics at the expense of pragmatics. From the four groups mentioned by Bally (see section 1), only diminutives and augmentatives can easily be considered from a semantic conceptual point of view, whereas amelioratives and pejoratives entail rather vague concepts, if any at all. They are more oriented towards pragmatic conditions than to semantic concepts. In this context, it does not seem to be coincidental that the structuralist approach opted for a narrow view on diminutives and augmentatives, since these are exactly the evaluative suffixes that are closest to a conceptual definition. In fact, structural linguistics aimed at excluding the emotional features (Monge 1965: 138). The exclusion of emotional (subjective) and referential features reduces the analysis to what might be identified as “meaning” or “concept”. As we have seen, this attempt fails. Yet, this does not necessarily mean that semantics has no role to play. The very problem concerned the limitation to rational, anti-affective concepts like those of tallness or smallness. Features such as ‘of little value’, ‘meager’ or ‘measly’ are not less semantic. If emotion could not be expressed by concepts, words like love and hate would not exist. Be this as it may, the analysis of single evaluative suffixes has to be integrated into a categorial view that accounts for general features like ‘highlighting’, ‘subjective evaluation’ and ‘variation’. The early approach of Bally, Hummel’s balance of structuralist analysis, and the morphopragmatic approach of Dressler and Merlini Barbaresi (see section 6) coincide on this point, not to speak of their implicit treatment as a group by Spitzer and Alonso. In the same vein, the new grammar of the Real Academia Española (2009: 627−662) dedicates a chapter to evaluative suffixes (“derivación apreciativa”). In his chapter on semantics, Bloomfield (1963: 146) suggested the almost forgotten term class-meaning as a complement for the analysis of word meaning. This is clearly the case for evaluative suffixes in Romance. Highlighting, subjective evaluation, and variation are constant semantic features of evaluative suffixes. From a methodological point of view, the theoretical focus that structural linguistics placed on the “functional oppositions” of signs seems to be co-responsible for the fact that the class-meaning was ignored in favor of the differentiating features.

88. The semantics and pragmatics of Romance evaluative suffixes

1537

4.3. Word-orientation The third shortcoming deals with a rather word-oriented analysis that reduces the role of context and situation to triggers of semantic information given by the word. In the same vein, Bally’s definition of evaluative suffixes only refers to the stem of a word (see section 1). By contrast, Spitzer clearly pointed out that evaluative suffixes may color utterances and discourse as a whole. In addition to that, Alonso argued that evaluative suffixes are used to negotiate the speaker’s relation with the interlocutors. However, as we have seen in section 4.1, there is evidence for the fact that evaluative suffixes are more word-oriented than adnominal adjectives. They are also able to create a range of ideas about the object that subsequently guide the specific pragmatic features of communication. Consequently, the solution is neither to reject word-orientation nor to refute pragmatic specification, but to admit the interplay of a (rather suggestive) idealistic input given by the word, a series of possible semantic-pragmatic patterns for interpretation, and the concrete features of co-text and situation.

5. Morphopragmatics The shortcomings of the structuralist linguistic analysis provide indirect evidence for pragmatically based approaches. Indeed, Merlini Barbaresi’s (2004: 279) observation that the best guide to the meaning of evaluative suffixes in Italian is context in its broadest sense starts at the point where structuralist linguistics falls short. Dressler and Merlini Barbaresi (1994: 52−54, 84) suggested an ambitious “morphopragmatic” theory of Romance evaluative suffixes that relates morphology with pragmatics. They define morphopragmatics as “morphologized pragmatics”, that is, “a certain type of grammaticalized pragmatics”. Thus, they are less interested in the grammaticalization process than in its result. Further, priority is given to rule-guided results over lexical idiosyncrasy. Moreover, the semantics of evaluative suffixes is minimized in favor of predictable pragmatic strategies. Finally, despite the focus on diminutives and suffixes, they are considered a part of the general group of evaluative suffixes, including the elative It. -issimo. Dressler and Merlini Barbaresi’s analysis proceeds step by step from denotation and connotation to morphopragmatics. At the level of denotation, the stem conditions possible effects for connotative or pragmatic effects. To give an example, the intensifying function of It. -ino is only possible if the basic concept of the stem is that of smallness: piccolo ‘small’ → piccolino ‘very small’ or ‘a bit small’, in contrast to altino ‘a bit tall’ ← alto ‘tall’, where the basic concept of tallness excludes the intensification ‘very tall’ (1994: 118). As for Spanish, this may reflect a general tendency as well, but fresquito always means ‘very fresh’ (← fresco ‘fresh’), blanquito may mean ‘very white’ (← blanco ‘white’), and grandecito (← grande ‘big, tall’) may be ironically intensifying, including denotative effects: grandecito may refer to a small child, signaling that s/he is quite tall for his/her age (cf. Real Academia Española 2009: 653). The same applies to the corresponding adjectives in Portuguese: fresquinho/fresquito, branquinho, grandezinho (cf. Sten 1944: 72 and Skorge 1956/58: 80−81, 261−263, 285−293). Hence, Portuguese and Spanish differ from Italian insofar as intensification occurs frequently, even in those cases where minoration and intensification do not behave like vectors that add conceptual features included in the stem to those of the suffix.

1538

VII. Semantics and pragmatics in word-formation II: Special cases

By contrast, Wagner (1952: 465) argued that the intention of the Portuguese (fresquinho) and Spanish (calentito ‘very hot’ ← caliente ‘hot’) diminutives is not intensification but the highlighting of a property. Given the importance of inference, it is hard to separate highlighting from intensification, especially because speakers perceive the intensifying effects and may reinforce them with intensifying prosody. However, Wagner’s observation allows us to explain the intensifying effect with the class-feature of highlighting, which in turn would explain the larger extension of intensification in Spanish and Portuguese. Highlighting effects seem to be possible with It. altino as well, at least if [i] is reinforced by prosody. In any case, it is certainly true that the general highlighting function of evaluative suffixes meets intensification, especially when notional diminutive or augmentative features reinforce the effect, that is, when ‘smaller/ bigger’ parallel intensification, as in Sp. pequeñito ‘very small’ (← pequeño ‘small’) and montón ‘very much’ (← monte ‘mountain’). Importantly, this analysis supports the assumption that the combination of available semantic features of both, the evaluative suffixe and the stem, produces specific effects. Hence, semantics explains basic effects produced by evaluative suffixes. Sp. pequeñito ‘very small’ clearly offers other conditions for pragmatic strategies than grandecito ‘a little bit tall’. According to Lüdtke (2011: 480−481), the intensification of Sp. -ito only occurs with adjectives that denote a positively connoted quality. Contrariwise, we might suppose that negatively connoted concepts are intensified with augmentatives, like in Sp. cabrón ‘bastard’ (← cabro ‘male goat’). However, intensifying effects are possible in Sp. ¡Qué feíto! ‘how ugly!’ (← feo ‘ugly’), but it is correct that the context should be positively connoted, e.g., in baby-talk. Consequently, the denotation and the connotation conveyed by the stem are relevant for the semantic and pragmatic effects, but probably not fully determining, as intonation ([i] would receive high pitch in feíto) and situation are relevant as well.

6. Semantics and pragmatics The fact that each evaluative suffix guides specific pragmatic strategies calls into question Dressler and Merlini Barbaresi’s exclusive focus on pragmatics. The theoretical problem is that we do not know exactly where semantics begins and pragmatics ends, or vice-versa. Fundamentally, meaning is always morphologized pragmatics. Even a word like apple concentrates and coins pragmatic experience in a morpheme. Meaning is, essentially, a heuristic device to explain reference. Now, reference is a pragmatic category. Children acquire meaning through referential experience. So meaning in language acquisition is coined by pragmatic experience. Processes of semantic abstraction, extension, extrapolation, metaphor or metonymy may follow and create semantic realities of their own, but pragmatic input cannot be negated, and the same would apply for evaluative suffixes. The concept of meaning tries to explain reference in terms of coined semantic features that allow for reference. Evaluative suffixes have to be considered as meaningful linguistic signs, since they not only accept and convey pragmatic strategies determined by cotext and situation, but allow for, control and guide these strategies. If this were not so, no difference could be observed between It. -ino and -one. Dressler and Merlini Barbare-

88. The semantics and pragmatics of Romance evaluative suffixes

1539

si agree with this, since they assume morphologized pragmatics. But what is the difference between morphologized pragmatics and meaning? There must be a crucial overlap of both. However, the present article encounters a similar aporia, since, on the one hand, I have argued that a pragmatic approach is needed because the semantic approach of structural linguistics has failed. Now I argue that the definition of morphologized pragmatics is not far from that of semantic meaning. Hence, it would be premature to declare that semantics or pragmatics is the dominant force. But how can this quandary be resolved? The problem and its solution depend crucially on the kinds of semantics and pragmatics that are used. The structuralist semantic approach failed because it was restricted to descriptive conceptual semantics based on features like ‘small’ or ‘big’, thus excluding affective features. On the other hand, Dressler and Merlini Barbaresi (1994: 132) object to the hypothesis of “emotive or affective invariant connotation” on the grounds that this semantic feature would be “clearly pragmatically based”. This is accurate, but is it not exactly this kind of semantic feature that would be needed for evaluative suffixes? A similar case is deixis, where semantic features of signs like now, there, he, etc. are bound to be directed from the speaker’s hic et nunc to the surrounding situation. If meaning and pragmatic function were contradictory, deictic signs would not be possible. In a certain sense, evaluative suffixes may be viewed as deictic devices for subjective evaluation based on the speaker’s hic et nunc. Hence, linguistic analysis has to account for both, the sign-based semantic features of evaluative suffixes and their utterance-specific interplay with stem, context, discourse strategy and situation. Features like ‘subjective evaluation’ and ‘highlighting’ clearly belong to the class-meaning of evaluative suffixes (cf. Rainer 1993: 198−199). They convincingly explain the alleged “imprecision of meaning” of evaluative suffixes (Dressler and Merlini Barbaresi 1994: 130−132 and passim). On the one hand, subjective evaluation with evaluative suffixes may be perceived as less precise than the objectified description with adjectives. On the other hand, in using evaluative suffixes speakers are voluntarily imprecise insofar as the exact determination of the pragmatic strategy is left to the hearer. Hummel (1994) therefore qualifies them as interpretive suffixes. Hence the lack of precision belongs to the intension of evaluative suffixes, that is, semantics. The point is that evaluative suffixes subjectively evaluate, conveying an idea of the referent, in contrast to the corresponding adjectives that describe more objectively, that is, try to represent a referent. This explains why Sp. un añito ‘a short or insignificant year’ (← año ‘year’), un litrito ‘a liter perceived as something harmless’ (← litro ‘liter’), un quilito (← quilo ‘kilo’) or It. milioncino (← milione ‘billion’) are commonplace, whereas *un pequeño año or *un pequeño litro are not. If (semantic) class-meaning and the (pragmatic) principle of co-text-, situation- and experience-guided determination of the communicative strategies associated with evaluative suffixes come out clearly from the analysis of the bibliography, there still remains the thorny issue of the distinctive features that allow for the specific properties of each evaluative suffix. In the case of diminutives and augmentatives, the semantic features ‘subjective diminution’ and ‘subjective augmentation’ account for their basic properties, including mitigation, intensification and emphasis. The differences between competing series of diminutives like Sp. -ito, -illo, -ico, -uco, -ino, -iño, etc. can be explained as affinity for additional features like dialect, negative/positive evaluation, animate/inanimate referents, etc. They appear as variationist contrasts rooted in the combined effect

1540

VII. Semantics and pragmatics in word-formation II: Special cases

of class-meaning with subjective diminution. The same holds for augmentatives. The latter has been demonstrated by Spitzer (1921) for -one (< Lat. -o, -onis) in Romance. But what about pejorative evaluative suffixes? The problem disappears if we do not consider pejorative or ameliorative effects as only connotations. Although negative evaluation is connotative or contextual in Sp. abogadito, as the suffix itself is not pejorative, this is not the case for pejorative suffixes like Sp. -ete. Pejorative suffixes have a pejorative meaning, that is, what ‘subjective diminution’ is for diminutives translates as ‘negative evaluation’ for pejoratives. In other cases, the concept of debility may be associated with pejorative connotations (e.g., Fr. -ard ). Possibly, a more polysemic approach would be needed to account for parallel, internally analogous functional series based on the same suffix, such as the downgrading of It. -ino (cf. Rainer 1993: 584−585). However, it would be impossible to separate polysemic differentiation from polyfunctional pragmatic strategies. Possibly, Dressler and Merlini Barbaresi’s focus on coined pragmatic strategies would have more explanatory force if it were not applied to morphologized pragmatic strategies in general, which implicitly refer to the complete semantic and pragmatic potential of one suffix (= the morphologizing morpheme), but restricted to the internal polysemic-pragmatic differentiation of a suffix and the corresponding creation of analogous series of words. Historically speaking, early pragmatic and stylistic approaches like those of Alonso and Spitzer often provided better results than recent analyses in structuralist semantics and pragmatics. The latter two may be considered superior to the former insofar as they are based on explicit theories. On the other hand, the clear-cut theoretical separation of semantics from pragmatics explains the shortcomings of these approaches. Hence, the theoretical effort has not only been dedicated to the formulation of a semantic or a pragmatic theory, but also to separate the two. However, linguistic theory cannot ignore the interfaces between semantics and pragmatics. Consequently, the study of the interfaces between seems to be a major desideratum for modern investigation on evaluative suffixes.

7. Denotative variation and lexicalization Denotative variation is one of the patterns systematically realized by evaluative suffixes. In this case, the word denotes a specific class or subclass of objects, e.g., Sp. tornillo, carrito ‘shopping cart, trolley, (baby) stroller’ (← carro ‘cart’), It. telefonino, libretto ‘libretto’ (← libro), Pg. kitchinete (French loan word), Fr. cigarette ‘cigarette’ (← cigare ‘cigar’), Fr. jeton ‘token’ (← jeter ‘to calculate’), etc. The objects usually are small, at least when compared to other subclasses (e.g., Fr. cigarette vs. cigare), but effects of subjective evaluation are not intended. Speakers would have to add a second suffix in order to add evaluative features (e.g., Sp. tornillito). Dressler and Merlini Barbaresi (1994) and other authors exclude this pattern from their analysis, claiming that these words are lexicalized, and thus not productive. However, there must be an underlying productive basis. In the case of evaluative suffixes, referential effects are directly linked to the categorial functions of variation and highlighting, which are used to select a special referent and to stress its subjective communicative relevance. The highlighting of a special type of object is productive in ojillos (← ojos ‘eyes’), which in standard

88. The semantics and pragmatics of Romance evaluative suffixes

1541

Spanish refers to specific eyes whose peculiarity is determined by context. In cases like tornillo, we have to assume that the lexicalization of a peculiarity was intended at the very origin. Words like tornillo are often intentionally created as names of classes of objects, just as German Handy or It. telefonino ‘mobile phone’. In this sense, naming has to be considered a productive variant of a denotative pattern that specifies a variety of objects. Now, productive reference is a basic aspect of pragmatics, since it takes place in communicative situations. Hence, Dressler and Merlini Barbaresi (1994) inappropriately exclude this pattern. What could be excluded from pragmatics is the development from productive reference to general usage. This is in fact the very domain of lexicalization, since the name one productively gives to an object does not necessarily enter into common vocabulary. Furthermore, frequency itself does not drive lexicalization, since frequent diminutives like Sp. hijto ‘dear son’ (← hijo ‘son’) are common but not semantically specified, that is, they correspond to what would be expressed by productive use. Consequently, lexicalization necessarily refers to the common use of a semantic or denotative variety that develops from the basic concept expressed by the stem. This could not happen if semantic or referential specification were not productively intended during the lexicalization process. Curiously, structural linguistics excluded or neglected reference for theoretical reasons, whereas pragmatics seems to prefer all contextual aspects of language (situation, discourse, interaction of speaker and hearer, register, politeness, etc.) to reference, possibly because reference and naming are too traditional. The consequence is that the crucial linguistic function of reference is located in between linguistic approaches and tends to be thrown to the oubliettes or taken for granted. Most authors assume that the lexicalization properties of Sp. -illo are rooted in the fact that it is older than -ito. However, just as with frequency, age does not obligatorily imply lexicalization, since the productive process is certainly not younger and would not lead to specifications if they were not productively intended at some time in concrete speech. In the case of Spanish, the argument fails because -ito has the same etymology as Cat. -et, Fr. -ette, It. -etto/-itto, and Port. -ito. In studies dedicated to Spanish, -ito is normally not considered a Romance suffix (cf. González Ollé 2007; Skorge 1956/58: 55), probably due to a confusion involuntarily caused by Menéndez Pidal (2007: 69) who declared it as “not Latin”. As a matter of fact, -ittu was used in Latin for hypocoristic names (e.g., Julitta, Bonitta (possibly > Sp. bonito)), as pointed out by Menéndez Pidal himself. Hence, what he intended to say is that authors like Meyer-Lübke (1972: § 505) considered the Latin suffix as “un-Latin”, that is, not looking like a Latin word and probably loaned to Latin from an unknown source. Later, the suffix was transmitted from Latin into Romance. Therefore, neither Menéndez Pidal nor Meyer-Lübke intended to say that Spanish did not inherit the suffix from Latin.

8. Oral familiar ingroup- and outgroup communication Prototypically, evaluative suffixes belong to oral colloquial communication between persons who are familiar to each other. Hence, their interpretation presupposes an intimate knowledge of a familiar world shared by the speakers. The affinity of evaluative suffixes with informal oral conversation comes out clearly in novels, where they prevail in direct speech (Lukas 1992: 155−156). As a consequence, evaluative suffixes are generally

1542

VII. Semantics and pragmatics in word-formation II: Special cases

associated with words that belong to a familiar domain. From a broader perspective, the social-psychological distinction of ingroup and outgroup behavior seems to be crucial (cf. Brown and Levinson 1987: 107−112). The Spanish suffix -ito tends to produce hypocoristic effects when it is directed to a member of a group, but pejorative effects when it is directed to members of a negatively perceived outgroup (e.g., Ha venido con sus abogaditos ‘He has come with his lousy lawyers’). In this case, -ito has the same pejorative function as -illo and does not necessarily refer to small persons. The concept of objective smallness is not relevant in this context, but is transposed metaphorically to a minorative appreciation. This analysis provides objective evidence for the emphasis Alonso placed on the “active” character of evaluative suffixes. Dressler and Merlini Barbaresi (1994: 233) recognize the ingroup effect but neglect the outgroup effect. In familiar ingroup communication, -ito is often lexicalized with hypocoristic functions, especially with names of persons, pets, etc. Juanito may be a name for a person that everybody uses in a given family. Similarly, pets will receive evaluative suffixes, but unknown animals will not, especially when they do not have a usual name (Skorge 1956/58: 251−254; Ettinger 1980: 146, 194; Waltereit 2006: 115−117). By contrast, evaluative suffixes rarely combine with abstract words (Lázaro Mora 1976), with Italian behaving less restrictively (e.g., weekendino ← weekend; cf. Merlini Barbaresi 2004: 267, 269, 282). Again, the suffixes behave specifically, that is, they are signs that produce effects of contrast in a given domain. In standard European Spanish, -ito tends to express positive feelings toward small objects (e.g., ojitos ‘small (and nice) eyes’), whereas -illo appeals to create a rather strange representation of an object (e.g., ojillos ‘strange eyes’) (cf. Rainer 1993: 540−542). However, this process also depends on the emotions associated with the type of object involved. To give an example, the probability of finding positive emotions associated with abogadito is lower than with ojito, but abogadillo and ojillo should almost always have a pejorative function (for Italian, see Merlini Barbaresi 2004: 284). According to Rainer (1993: 582), abogadito shares the pejorative effect with masculine professional designations (cf., however, abogadita). Importantly, these communicative effects cannot be fully explained by the properties of the word stem and the suffix. When ojito refers to an object that is regarded unemotionally, for instance a painting in a museum, it may simply express ‘small eyes’. In the same vein, pedrecilla ‘small stone’ (← piedra ‘stone’) may refer to a small or a small and strange stone. On the contrary, a nickname like El abogadito may be a sign of respect in a given community (cf. Placencia 2010). Hypocoristic features are automatically relevant, when ojitos refers to one’s children. The interpretation depends thus on the emotional attitude towards the object one has in mind. This means that the ingroup situation and the emotions felt or not felt towards the referent are decisive for the interpretation. The fact that hypocoristic evaluative suffixes are characteristic for the language adults direct to children can be explained as an extension of the familiar domain. However, the hypocoristic interpretation would change if the word referred to an annoying child.

9. References Alonso, Amado 1930 Para la lingüística de nuestro diminutivo. Humanidades 21: 35−41. Alonso, Amado 1961 [1935] Noción, emoción, acción y fantasía en los diminutivos. In: Amado Alonso (ed.), Estudios lingüísticos. Temas españoles, 195−229. 2nd ed. Madrid: Gredos.

88. The semantics and pragmatics of Romance evaluative suffixes

1543

Bally, Charles 1965 Linguistique générale et linguistique française. 4 th ed. Berne: Francke. Bloomfield, Leonard 1963 Language. 2nd ed. New York: Holt, Rinehart and Winston. Brown, Penelope and Stephen Levinson 1987 Politeness. Cambridge: Cambridge University Press. Bühler, Karl 1982 Sprachtheorie. 3rd ed. Stuttgart/New York: Fischer. Coseriu, Eugenio 1988 Einführung in die Allgemeine Sprachwissenschaft. Tübingen: Francke. Dressler, Wolfgang and Lavinia Merlini Barbaresi 1994 Morphopragmatics. Diminutives and intensifiers in Italian, German, and other languages. Berlin/New York: Mouton de Gruyter. Ettinger, Stefan 1980 Form und Funktion in der Wortbildung. Die Diminutiv- und Augmentativmodifikation im Lateinischen, Deutschen und Romanischen (Portugiesisch, Spanisch, Italienisch und Rumänisch). 2nd ed. Tübingen: Narr. Fretel, Hélène 2010 Le suffixe diminutif: Un marqueur d’appropriation du signifiant. In: Gabrielle Le TallecLloret (ed.), Vues et contrevues, 413−423. Limoges: Lambert-Lucas. Gauger, Hans-Martin 1971 Durchsichtige Wörter. Heidelberg: Winter. González Ollé, Fernando 2007 Origen de -ito, con una revisión histórica de otros sufijos diminutivos románicos. In: Emili Cásanova i Herrero and Xavier Terrado i Pablo (eds.), Studia in honorem Joan Coromines, 157−177. Lleida: Pagès. Günthner, Susanne and Katrin Mutz 2004 Grammaticalization vs. pragmaticalization? The development of pragmatic markers in German and Italian. In: Walter Bisang, Nikolaus Himmelmann and Björn Wiemer (eds.), What makes grammaticalization?, 77−107. Berlin/New York: Mouton de Gruyter. Hasselrot, Bengt 1957 Études sur la formation diminutive dans les langues romanes. Uppsala/Wiesbaden: Lundequistska Bokhandelen/Harrassowitz. Hummel, Martin 1994 Diminutive als Apreziativa: Zur Theorie der Diminutive im Spanischen. Romanistisches Jahrbuch 45: 243−261. Jurafsky, Daniel 1996 Universal tendencies in the semantics of the diminutive. Language 72(3): 533−578. Lázaro Mora, Fernando 1976 Compatibilidad entre lexemas nominales y sufijos diminutivos. Thesaurus 31: 41−57. Lázaro Mora, Fernando 1999 La derivación apreciativa. In: Ignacio Bosque and Violeta Demonte (eds.), Gramática descriptiva de la lengua española. Vol. 3, 4645−4682. Madrid: Espasa Calpe. Lüdtke, Jens 2011 La formación de palabras en las lenguas románicas. Su semántica en diacronía y sincronía. México, D.F.: El Colegio de México. Lukas, Ulrike 1992 Die Funktion der Diminutive in zeitgenössischen Romanen andalusischer und kastilischer Autoren. Frankfurt/M.: Lang. Menéndez Pidal, Ramón 2007 Historia de la lengua española. Vol. 1. 2 nd ed. Madrid: Fundación Ramón Menéndez Pidal.

1544

VII. Semantics and pragmatics in word-formation II: Special cases

Merlini Barbaresi, Lavinia 2004 Suffissazione. In: Maria Grossmann and Franz Rainer (eds.), La formazione delle parole in italiano, 189−492. Tübingen: Niemeyer. Meyer-Lübke, Wilhelm 1972 Grammatik der romanischen Sprachen. Vol. 2: Romanische Formenlehre. 2nd ed. Hildesheim/New York: Olms. Mihatsch, Wiltrud 2010 “Wird man von hustensaft wie so ne art bekifft?” Approximationsmarker in romanischen Sprachen. Frankfurt/M.: Klostermann. Monge, Félix 1965 Los diminutivos en español. In: Georges Straka (ed.), Actes du X e Congrès International de Linguistique et Philologie Romanes (Strasbourg 1962). Vol. 1, 137−147. Paris: Klincksieck. Mutz, Katrin 2000 Die italienischen Modifikationssuffixe. Synchronie und Diachronie. Frankfurt/M.: Lang. Neumann-Holzschuh, Ingrid 2009 Alonso, Amado. In: Harro Stammerjohann (ed.), Lexicon grammaticorum. Vol. 1, 33− 34. 2nd ed. Tübingen: Niemeyer. Placencia, María Elena 2010 “¿Qué dice flaco?” Algunos aspectos de la práctica social de apodar en Quito. In: Martin Hummel, Bettina Kluge and María Eugenia Vázquez Laslop (eds.), Formas y fórmulas de tratamiento en el mundo hispánico, 965−992. México, D.F.: El Colegio de México/ Karl-Franzens-Universität Graz. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Real Academia Española and Asociación de Academias de la Lengua Española 2009 Nueva gramática de la lengua española. Vol. 1. Madrid: Espasa. Saussure, Ferdinand de 1983 Cours de linguistique générale. Édition critique préparée par Tullio de Mauro. Paris: Payot. Schneider, Klaus P. 1991 Affektive Lexik: Kognitive, semantische und morphologische Aspekte. In: Eberhard Klein, Françoise Pouradier Duteil and Karl Heinz Wagner (eds.), Betriebslinguistik und Linguistikbetrieb. Vol. 1, 233−241. Tübingen: Niemeyer. Skorge, Silvia 1956/58 Os sufixos diminutivos em português. Boletim de Filologia 16: 50−90, 222−305; 17: 20−53. Spitzer, Leo 1918 Aufsätze zur romanischen Syntax und Stilistik. Halle/S.: Niemeyer. Spitzer, Leo 1921 Das Suffix -one im Romanischen. In: Ernst Gamillscheg and Leo Spitzer (eds.), Beiträge zur romanischen Wortbildungslehre, 183−205. Genève: Olschki. Spitzer, Leo 1933 Amado Alonso, “Para la lingüística de nuestro diminutivo” [review]. Literaturblatt für germanische und romanische Philologie 9−10: 320−323. Sten, Holger 1944 Les particularités de la langue portugaise. Kobenhavn: Munksgaard. Wagner, Max Leopold 1952 Das ‘Diminutiv’ im Portugiesischen. Orbis 1(2): 460−476. Walsh, Donald 1944 Spanish diminutives. Hispania 27: 11−20.

89. Morphopragmatics in Slavic

1545

Waltereit, Richard 2006 Abtönung. Zur Pragmatik und historischen Semantik von Modalpartikeln und ihren funktionalen Äquivalenten in romanischen Sprachen. Tübingen: Niemeyer. Weber, Marcel 1963 Contributions à l’étude du diminutif en français moderne. Ph.D. dissertation, University of Zürich. Wrede, Ferdinand 1908 Die Diminutiva im Deutschen. Marburg: Elwert. Würstle, Regine 1992 Überangebot und Defizit in der Wortbildung. Eine kontrastive Studie zur Diminutivbildung im Deutschen, Französischen und Englischen. Frankfurt/M.: Lang.

Martin Hummel, Graz (Austria)

89. Morphopragmatics in Slavic 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Introduction The function of word-formation formatives The classical approach: stylistics Pragmatics vs. semantics Diminutives and similar phenomena Augmentatives and similar phenomena Language of love Irony and banter Word-formation in the context of other rhetorical devices References

Abstract Apart from a few individual articles that have appeared since the end of the 1990s, the term “morphopragmatics” has not yet become established in works on Slavic wordformation. One reason for this is undoubtedly the long tradition that functional stylistics has enjoyed in Slavic studies which anticipated a series of pragmatic questions. This article proposes a possible, albeit quite broad, pragmalinguistic approach to Slavic word-formation phenomena both on the level of the language system (dealing with, e.g., pragma-stylistically marked affixes expressing emotion and judgement) as well as on the discourse level (considering esp. family talk, child speech, ironic or playful attitude of the speaker, etc.).

1. Introduction What pragmatics and semantics have in common is that they apply to both the lexical and grammatical system of a language. Due to its liminal character, word-formation is

1546

VII. Semantics and pragmatics in word-formation II: Special cases

concerned with issues of semantics and pragmatics in the lexicon (since derivatives are a part of the lexicon) and in morphology. The latter applies to affixation and other wordformation processes whose role is greater than just the simple naming of phenomena. Morphopragmatics, which is a relatively new discipline developed in response to a new approach towards language and incorporates language users’ intentions and communication strategies as well as context-dependent social conventions of speaking, will be the main focus of this article. The question is to what degree these strategies make use of fixed and codified language elements, and to what degree they are pragmatically dependent on the changing context and the situation (cf. discussion concerning this matter in section 4). Examples from Slavic languages, which are characterized by a plethora of expressive morphological means, will be used as illustration in this article. The following overview does not exhaust the entire spectrum of morphopragmatics. An analysis of textual metaphors, such as the transposition of derivatives denoting real things to abstract domains, is not discussed here. The potential for associations (including humorous ones) within the frame of these concepts would demand the exploration of a cognitive approach.

2. The function of word-formation formatives It is important to begin the review of word-formation phenomena which may be included as parts of language pragmatics with a brief survey of opinions about what happens during the derivation act. Kuryłowicz (1963) introduced a division between syntactic derivation (shift of syntactic function) and lexical derivation (shift of meaning). Within the terminology of Slavic word-formation the former is known as pure transposition. The derivative differs from its base only in its morphosyntactic characteristics. According to Dokulil (1962) this is the case with abstract nouns derived from verbs (Czech padnout/ padat ‘to fall (pf.)/to fall (ipf.)’ → padnutí/padání ‘a fall/falling’) and from adjectives (Czech tekutý ‘liquid’ → tekutost ‘liquidity’). A transposition may also occur in case of adjectives derived from adverbs, cf. Czech rychle ‘fast (Adv)’ → rychlý ‘fast (A)’. The author considers these shifts to be cases of hypostasis in which a concept gains certain independence but whose status as an onomasiological category is only conditional (Dokulil 1979: 65). The assumption that a transposition does not entail semantic consequences has however been challenged by cognitive semantics. Moreover, abstract nouns are proper to nominal style, hence they are not stylistically irrelevant, even if it is assumed that the base word and the derivative share the same deep structure. Apart from syntactic transposition understood in this manner there are other productive derivation techniques the aim of which is not to name, but to give a language more economy and brevity. This may be illustrated by the univerbation (in Slavic wordformation defined as combination of ellipsis and suffixation) of multi-word structures, cf. Polish colloquial skarbówka ‘tax office’ and the official name izba skarbowa lit. ‘tax chamber’, Czech generálka (← generální zkouška ‘dress (lit. general) rehearsal’); by the use of abbreviations, cf. Russian SPID ‘AIDS’, zavlab (← zavedujuščij laboratoriej ‘chief of laboratory’); as well as by the use of the so-called złożeniowce (midway structures between compounds proper and (syllable) acronyms) with a reduced first element, cf. Polish examples modeled on Russian: specsłużby (← służby specjalne ‘special for-

89. Morphopragmatics in Slavic

1547

ces’). (Zemskaja 1992: 9 described this word-formation function as compressive.) It should also be noted that the compression of a form is related to the character of a communication and its products are present only in certain types of texts. Univerbation is characteristic of colloquial language, abbreviations are found mostly in written language, while złożeniowce function in technical jargons. Kuryłowicz’s “lexical derivation” is presented, following Dokulil, as an instance of a complete change of meaning (cf. Polish ryba ‘fish’ → rybak ‘fisherman’), or as a modification of meaning (cf. Polish ryba → rybka ‘small fish’), though their complete separation is still debatable. Modification entails the addition of a feature, but one which renders the derivative still part of the same onomasiological conceptual category: rybka ‘a small fish’ is, after all, still a fish. Mutation, however, transfers the derivative into a new conceptual class. However, this classification has proven to be too narrow. In order to maintain it, new mixed classes have been introduced to account for the combined transpositional and modificational derivatives, as well as for the combined mutational and expressive derivatives in Polish word-formation (Grzegorczykowa, Laskowski and Wróbel 1998: 377). The former type includes an additional element of axiological judgement, e.g., Polish bieganina (← biegać ‘to run’) ‘running + chaotic, aimless’. The latter may be represented by Polish agent nouns like pracuś (← pracować ‘to work’) − a mutation of the meaning ‘somebody who works a lot’ which, depending on the speaker’s intention, incorporates an ironic or humorous element. Sokolová (2007: 186) introduced into the description of Slovak an intermediate type combining modification and mutation in which she includes negated derivatives, such as neúspech ‘failure; lit. non-success’. Hence it is clear that the neat division into transposition, mutation and modification is undermined by subjective elements of meaning expressing axiological judgments or emotions. These elements do not fit the referential semantics that deals with objective, verifiable states of affairs. This problem could only be addressed with the appearance of pragmalinguistics and its orientation to the speaker’s point of view and intention, i.e. on the “here and now” of the actual context of a given act of speech. Within the field of derivational theory, the introduction of a new approach has lead to a shift in orientation from the description of a system to an examination of texts and non-standardized vocabulary. An analysis of neologism databases, including wordformation neologisms (Martincová 1998, 2004; Liptáková 2000; Jadacka 2001; Uluchanow and Belentschikow 2007; and many others) has shown the variety of its uses, especially the strong need of expression. A noteworthy example is provided by a gradual shift from agent nouns to nouns where human beings are no longer profiled as active participants in an event, but as passive bearers of a relationship towards an object. This in turn entails a shift from a verbal to a nominal derivation base, and reflects a certain axiological position, cf. Polish styropianowiec (from styropian ‘styrofoam’) ‘a person involved in the (anti-Communist) opposition, the symbol of which (among many others) is styrofoam used as bedding during the occupational strikes at Gdańsk shipyard in 1981’. An expansion of the colloquial register in all Slavic languages has also lead to an observable tendency to replace neutral words with their more expressive synonyms, cf. Russian vodila instead of voditel ‘driver’, or Czech lokomotivák instead of lokomotivář ‘engine driver’ (Neščimenko 2004: 87).

1548

VII. Semantics and pragmatics in word-formation II: Special cases

3. The classical approach: stylistics In the past, the subjective and emotional markings of derivatives, which were difficult to describe at the level of the word-formation system, were considered stylistic differences (cf. similar remarks concerning earlier studies of different languages in article 62 on the pragmatics of word-formation, in article 88 on the semantics and pragmatics of Romance evaluative suffixes, and in article 128 on word-formation and literature). It is worth mentioning that linguistic stylistics has a long tradition in Russia, as well as in Poland and the Czech Republic, and is, to some extent, a forerunner of pragmalinguistics in these countries (cf. Ohnheiser 2003: 198). An examination of lexis according to a given functional style reveals a fundamental division into the spoken (colloquial) and the written (literary) register. Especially the latter has been rigorously codified and regulated. At present there is a perceptible liberalization of language norms in all Slavic languages, increased interference between the two registers, and especially an expansion of colloquial forms. Among the many results of this situation are the increasing diversification of linguistic expression and the presence of numerous doublets in language which differ in their degree of expressivity, especially in contexts where there is no straightforward rule determining correctness. An example of such variation may be the feminine names that occur in colloquial language, and yet are treated with reserve by normative linguistics, cf. Russian feminine personal nouns with the -ša suffix, such as inspektorša ‘woman inspector’, professorša ‘woman professor’, or in Polish with the -ka suffix in words formed from a foreign base: psycholożka ‘woman psychologist’, etyczka ‘woman ethicist’, and filozofka ‘woman philosopher’. Their neutral variants come in the form of analytical expressions, such as pani psycholog lit. ‘lady-psychologist’. Hence a choice between one of the forms may depend to an extent on the style of the message, and definitely on its level of formality. From the point of view of word-formation, expressivity (emotionality) may be expressed both via the base or the formative. Furdík (2004: 123) predicts the following combinations: a) neutral base and marked formative, cf. Slovak Nemč-ur, Polish Niemi-aszek ‘German (n., derogatory)’ from nemec/Niemiec ‘German (n.)’. (Stylistic marking is also present in foreign formatives. For example, in Bosnian, Croatian and Serbian, borrowings from Turkish are stylistically marked, cf. the semantically equal, yet stylistically different suffixes -luk and -nost (Slav.), e.g., bezobrazluk ‘insolence’ vs. bezobraznost, and others, cf. Raecke 2003: 244), b) marked base and neutral formative, cf. place nouns/collective nouns: Slovak babinec, Polish bab-iniec ‘group of women’ (where baba is an expressive variant of the word žena/kobieta ‘woman’), Slovak vagabund-stvo, Polish włóczęg-ostwo ‘vagrancy’ (vagabund/włóczęga ‘vagabond’ is a pejorative term), c) in compounds it is possible to combine two bases, one neutral and one marked, cf. Slovak darmo-žráč, Polish darmozjad ‘freeloader; lit. free eater’ (the Slovak base žrať ‘to eat’ refers to animals). Other criteria of (in)compatibility have been applied in the evaluation of (new) morpheme combinations. For example, hybrid forms and infelicitous metaphoric word-formations, cf. Slovak compounds sexbomba ‘sexbomb’, sexraketa ‘sexrocket’, čisložonglérstwo lit. ‘number juggling’ (cf. English creative accounting) have met with disap-

89. Morphopragmatics in Slavic

1549

proval. In this respect the current norm is much more liberal, and what follows is an interest in substandard vocabulary, group slangs and individualisms. Synonymy is also part of stylistics, and its analysis includes the function of synonymous word-formation and its role in a text. In a narrow sense these are stylistic variants, such as the Polish name for ‘porcini mushroom’ where prawdziwek (← prawdziwy ‘true’) is the colloquial, and borowik the formal name (bór is poetic for las ‘forest’). In a broader understanding, these are neutral derivatives and referentially equivalent descriptive-evaluative structures, cf. Polish solidarnościowiec ‘supporter of the Solidarity movement’ vs. solidaruch ‘supporter of the Solidarity movement + aversion’. Serial expressive derivatives are usually semantically negative (derogatory), cf. the numerous Polish terms for ‘stupid person’: głupi-ec, głup-ek, głup-ol, głupol-ek, głup-tas, głuptasek, głup, przygłup(ek). Apart from the shared invariable modifying function these formatives differ substantially in their expressive load ranging from disrespect, contempt, and pity, to tenderness or even attempts at reducing the negativity of the content (litotes, e.g., głuptas and głuptasek). This however raises questions concerning the place of emotion in the information structure of a statement. Is it a subjective, context-dependent communicative reaction to a given stimulus, or is emotionality part of a given predicate, conventionally ascribed to it − part of its fixed meaning? In the traditional Russian terminology “expressivity” has a mixed character, as it is considered a set of semantic and stylistic features proper to a given linguistic unit (cf. Ohnheiser 2007: 349). Meanwhile Zemskaja (1992: 10−12) distinguishes the expressive function of a formative (expression of a subjective judgment) from its purely stylistic function. This second function consists in choosing linguistic means appropriate for the context of the speech act and the genre. Within the scope of stylistic word-formation the author distinguishes two cases. The first applies to derivatives that differ from the bases only in their stylistic character, e.g., marked as colloquial forms with the -ka suffix used in unofficial circumstances, such as seledka (← sel’d’ ‘herring’), tetradka (← tetrad’ ‘notebook’). In terms of meaning, these forms are equivalent to their unsuffixed counterparts, yet the latter would sound artificial and too literary for spoken communication. The second type creates pairs of derivatives that differ from one another in their stylistic and pragmatic character, such as in the case of the neutral pochoronit’ ‘to bury a deceased person’ and the official zachoronit’ with the za- prefix, analogically zaslušať ‘to hear out’ and proslušat’ ‘to listen to’, začitať ‘to read out’ and pročitat’ ‘to read’. In Polish this mark of formality may be present in some verbs with the prefix u-: ukochać ‘to start loving’, ubogacić ‘to enrich’, instead of the neutral pokochać, wzbogacić, especially prominent in the language of priests delivering sermons.

4. Pragmatics vs. semantics The rise and rapid development of pragmalinguistics, a shift of focus from the system towards the speech act and discourse have enabled a new perspective to be assumed when dealing with the issues discussed above. Cognitive linguistics, which by default blurs all boundaries, has abolished the distinction between semantics and pragmatics. Yet even cognitive linguists agree that meaning is always context-dependent, as its inter-

1550

VII. Semantics and pragmatics in word-formation II: Special cases

pretation is always conducted in a socio-physical environment and takes into account the knowledge and experience of the interlocutors. Thus it seems that it should be possible to isolate the constant and necessary (systemic) linguistic means and the variable and optional (contextual) linguistic means. This task requires an examination of both the code, i.e. the store of invariant word-formation means, and the use of these means, i.e. their function in discourse. It thus comes as no surprise that researchers investigating this issue direct their attention towards the sociopragmatics of various “languages” and sublanguages. Experiments bring forward evidence for some of the claims, for example those concerning the frequency of affective expressions in the language of women. (Dąbrowska 2005: 149 asked a group of men and women to describe Polish Christmas cards. The descriptions provided by women indeed contained numerous diminutives, such as bałwanek ‘snowman’, gwiazdka ‘star, also a familiar name for Christmas’, obrazek ‘picture’, szaliczki ‘scarfs’, and sweterki ‘sweaters’. However, other experiments conducted by her did not confirm women’s predilection for emphasis.) Morphopragmatics includes, apart from typical stylistic markers, emotion and judgment exponents considered as that part of the information structure of the statement that corresponds to its illocution or is a subjective element of the proposition the negation of which does not lead to a contradiction (cf. Nagórko 1997: 264). These means are fully conventional − the speaker has word-formation models, affixes and ready derivatives at his or her disposal. The pragmatics of axiological judgment grants some leeway: a marked linguistic unit usually has a neutral equivalent, cf. the purely descriptive Russian pisatel’, Polish pisarz ‘writer’ and the emotional and evaluative Russian pisaka, Polish pismak or pisarzyna ‘hack writer’. The pragmatic content spans a scale between two poles: negative and positive. The context may cause the reversal of the vector from positive to negative, e.g., Polish lal-unia is a tender name for a doll (lala) in children’s speech, but a patronizing name for a woman when uttered by a macho man, or alternatively a term of endearment a mother uses to call her baby girl. Many word-formation formatives have a complex descriptive and evaluative character and their productivity increases constantly. This is connected to the already mentioned need for expression fuelled by the new media including the Internet. It is well known that emotions have a persuasive force. A good example of this are the numerous abstract nouns formed on the basis of the names of politicians and other prominent people used by their critics in journalistic texts and on online forums. This is a pan-Slavic tendency, cf. Czech klausiánství/klausizmus ‘politics inspired by Václav Klaus’, havlomanie from Havel; Polish wałęsizm from Wałęsa, kwaśniewszczyzna from Kwaśniewski; Russian brežnevščina from Brežnev; Ukrainian kučmunizm from Kučma (which might be regarded as a blend of Kučm-a and komm-unizm). Many among these neologisms have a temporary character and lose their motivation with the withdrawal of that person from the political scene. A characteristic feature shared by the Slavic languages is the ease in creating diminutive and augmentative forms. They are created not only from nominal parts of speech (nouns and adjectives), but also from verbs and adverbs, cf. Kashubian chrüstac ‘to bite’ and diminutives, used in child or caretaker language chrüszczekac, chrüszczeczkac, Russian nemnogo ‘a bit’ → nemnožko, nemnožečko ‘a little bit’ and Polish trochę → troszkę, troszeczkę ‘id.’. Diminutives and augmentatives create a polar scale spanning from the “small” to the “big” pole. Their prototypical meanings have been expanded by means of semantic operations to pragmatic emotional and evaluative meanings − as an

89. Morphopragmatics in Slavic

1551

additional description of the derivative or as its sole function. Formatives that combine with a broad range of bases offer means of expressing an entire gamut of attitudes, cf. the pancategorial suffix -uchn- in Polish, which can attach to any main part of speech, such as an adjective (mały → mal-uchn-y ‘very small’ and ‘evoking tenderness’), noun (buzia → buzi-uchn-a ‘face (with tenderness)’, władza → władz-uchn-a ‘authority (ironic); militia officer (humorous, mocking)’), or the dialectal verb płak-uchn-ać ‘to cry a bit’ in child language. These categories merit a thorough examination (cf. sections 6 and 7). Additionally, on the basis of familial language this article will examine morphological means of expressing empathy and tenderness (cf. section 7). It will also analyze folk attitudes towards language, irony and wordplay using word-formation (section 8), as well as the rhetorical function of derivatives in texts, their uses as hyperboles, litotes, euphemisms and ad hoc metaphors (section 9). These stylistic devices infringe on Grice’s maxim of manner and are also a source of conventionalized implicatures enabling people to speak indirectly, “off the record”. Traditional rhetorical tropes are analyzed within the scope of pragmalinguistics as strategies of politeness minimizing threats to the addressee’s “face” (cf. Brown and Levinson 1987: 211 ff.).

5. Diminutives and similar phenomena Diminutives constitute a modifying category in which the additional predication consists of a quantitative evaluation of a given object as “smaller than the norm” in its class. The information about smallness often entails an emotional component, though pure diminutives are also possible, e.g., use of names of baby body parts in a textbook for midwives. Diminutives usually carry a positive tinge probably related to the fact that things and beings smaller than expected evoke sympathy or conversely do not constitute a threat. The geography of the use of diminutive forms is interesting (cf. Raecke 2004: 170), since it clearly correlates with ethnic cultures. Languages of the so-called “warm” cultures, such as Italian and Slavic languages contain many such forms. They are characteristic of folk dialects and children’s speech − exactly the contexts in which spontaneity and directness are not yet encumbered with intellectual control. Slavic languages have many diminutive suffixes at their disposal, such as the panSlavic -ek, -ik for masculine, -k(a) for feminine, and -k(o) for neuter gender. In some languages the old suffixes -ic(a)/-ic(e) have kept their productivity, cf. Croatian traka ‘tape’ → trakica ‘ribbon’, and the purely expressive majčica, mamica ‘mommy’ (← majka, mama), Russian krepostica ‘small fortress’ (← krepost’), and the poetic Rusyn vodicja ‘water-DIM’ (← voda) (see Chomjak 1997: 80) and -c(a)/c(ja)/-c(e), cf. Russian slezca ‘teardrop’ (← sleza), Rusyn golovcja ‘head-DIM’ (← golova), Croatian pilence ‘chick-DIM’ (← pile). In the Western-Slavic group these latter suffixes have become obsolete, or assumed a new meaning, cf. Czech hlavice, Polish głowica (from hlava/ głowa ‘head’) as part of a machine. The quantitative evaluation of the small size of something against an implied norm carries a risk of a change in the image of the prototype, thus a reversal of motivation is not surprising. An item named by a derivative often becomes a point of reference for the norm and its name loses the meaning of “smallness”. Genetically basic nouns become marked when they begin to be understood as names for big objects, bigger than usual,

1552

VII. Semantics and pragmatics in word-formation II: Special cases

cf. Polish słój ‘big jar’ and słoik ‘jar’, kielich ‘chalice’ and kieliszek ‘vodka/wine glass’, where synchronically the former bases (słój and kielich) have become augmentatives. These shifts indicate that the speaker, while performing a parametric evaluation, refers to the entire implicit scale between diminutive and augmentative. The need to name the “small” pole explains the creation of an entire series of complex suffixes usually with a strong expressive element, hence less popular in common use. This can be seen, for example, in Polish -eczek (kieliszeczek) ‘little wine glass’, -iczek (słoiczek) ‘little jar’, -aszek (kijaszek) ‘little stick’ (← kij); Czech -ičk-, -ečk-, -enk-, -ušk-; Russian -išk-, -ešk-, -ul’k-. They may also form diminutives of the second degree, cf. Czech ryba → ryb-ka ‘small fish’ → ryb-ič-k-a ‘very small fish’, Ukrainian lis ‘forest’ → lis-ok ‘small forest’ → lis-oč-ok ‘very small forest’. The diminutive category has a semantic-pragmatic character. The same word-formation means may be used in concepts only metaphorically or metonymically related to smallness, in which the feature of smallness is less prominent. Among the related categories are a) similatives, as in Polish szyjka butelki ‘bottle neck, similar to an anatomical neck (szyja)’, Croatian ušica od igle ‘eye of a needle; lit. (little) ear of a needle’ (← uho), Russian kolence ‘a dance figure’ (← koleno ‘knee’), Czech ručička ‘clock hand’ (← ruka/ručka ‘hand’); b) singulatives, as in Polish wacik ‘wad of cotton wool, cotton pad’ (← wata), Czech travička ‘blade of grass’ (← tráva), Slovak mrkvička ‘carrot, a single carrot root’ (← mrkva); c) taxonomic names meaning ‘a type of ’, cf. Polish stół ‘table’ and stolik (kawiarniany) ‘coffeehouse table’, szczotka do włosów ‘hair brush’ vs. szczoteczka do zębów ‘toothbrush’; d) names of toys and other items belonging to a child’s world, such as Polish tygrysek ‘tiger’ (We would say: Dziecko bawi się pluszowym tygryskiem ‘The child is playing with a plush tiger-DIM’ rather than *Dziecko bawi się małym tygrysem ‘The child is playing with a small tiger’); e) names of animal offsprings and small animals, such as Polish słoniątko ‘baby elephant’ (← słoń ‘elephant’), Slovak mačiatko ‘kitten’ (← mačka ‘cat’); f ) nouns describing an approximate duration of something, usually decreasing its length, such as Polish kwadransik ‘quarter of an hour’ (← kwadrans), godzinka ‘one hour’ (← godzina), słóweczko (← słówko ← słowo) ‘a word, meaning a short conversation’. Apart from description, i.e. the attribution of traits to the referent of a name, the semantic structure of the derivative may also contain pragmatic information that introduces the addressee into the emotional and evaluative world of the speaker, or that regulates certain conventional behaviors − often with the intent of manipulating the addressee. Oftentimes in these cases pragmatics dominates the objective content. The word-formation systems in Slavic languages contain suffixes with a strong expressive force and stylistic variation, as in literary Russian sud’b-ina ‘fate’ (← sud’ba), folk pal’t-uška ‘cheap coat’ (← pal’to), colloquial del’-ce ‘issue, matter’ (← delo). They may all be called quasi-diminutives, since semantically they do not express smallness, though formally they resemble diminutives. They are very frequent in some sociolects and communication contexts. Among these the most noteworthy is the language of serving staff (waiters, salespersons,

89. Morphopragmatics in Slavic

1553

hairdressers, ticket inspectors) used when speaking to customers and medical staff when talking to patients. This mannerism may be attributed to a certain professional etiquette that aims to minimize the speaker’s benefit and show empathy. This is the comitative function, cf. Milewska-Stawiany (2007: 153). Polish examples include: kotlecik (← kotlet) ‘cutlet’, wódeczka (← wódka) ‘shot of vodka’, herbatka z cytrynką (← herbata z cytryną) ‘tea with lemon’, rachuneczek (← rachunek) ‘check’ (in waiter’s speech), schabik (← schab) ‘pork chop’, jabłuszka (← jabłka) ‘apples’ (salesperson), paszporcik (← paszport) ‘passport’, dokumenciki (← dokumenty) ‘documents’ (police officer). Russian examples include: ponedel’niček (← ponedel’nik) ‘Monday’, časik (← čas) ‘hour’, bočok (← bok) ‘side of the body’ (masseuse to a patient), zubik (← zub) ‘tooth’ (dentist to patient), kotletka s pjureškoj (← kotlet s pjure) ‘cutlet with mashed potatoes’, pirožočki s kapustkoj (← pirožki s kapustoj) ‘dumplings with cabbage stuffing’ (chef to patron). As may be illustrated by early 20th century literary dialogues this is not a new phenomenon (cf. Nagórko 1997: 265), though it seems to be on the rise with the gradual spread of familiar style in public speech acts. A natural environment for quasi-diminutives is undoubtedly the familial language, which will be discussed in section 7. They are also a common feature of authorial irony, if they infringe the restrictions on suffix collocation, cf. Witkacy’s neologisms created on the basis of Polish abstract nouns and individual names: dowcipek (← dowcip) ‘joke’, myślątka (← myśli ) ‘thoughts’, ziemiczka (← Ziemia) ‘Earth’ (for more information see section 8).

6. Augmentatives and similar phenomena The comparatively high number of augmentatives is related to the cognitive attitude toward negative phenomena and to the known semantic asymmetry. Augmentatives exhibit a typically “masculine” emotional model with its roughness and anti-sentimental attitude. This is readily picked up by contemporary youth, hence augmentatives are often stylistically marked as part of a group slang or belong to lower register of the language. Traditional pan-Slavic suffixes -isk- and -išč- may express both an unusual size of an item and a negative attitude of the speaker. If the base word does not refer to an item’s physical feature, the derivative fulfills only its pragmatic function, cf. Upper Sorbian: dub → dubisko ‘oak’, žona → žonisko ‘woman’, psyk → psyčišćo ‘dog’; Polish chłop → chłopisko ‘bloke (about a man)’; Slovak baba → babizňa ‘hag’ (about a woman); Russian anekdot → anekdotišče ‘anecdote’, anketa → anketišča ‘questionnaire’, bardak → bardačišče ‘mess’. An even greater depreciation effect is achieved by switching the gender of the derivative to neuter and adding a modifying adjective, cf. Upper Sorbian stare, špatne žonisko ‘old, evil hag’, Polish to wielkie chłopisko ‘this huge bloke’. Group slangs have seen a rise in popularity of clipped structures, cf. Croatian Amer (← Amerikanac) ‘American’; Russian buk (← bukinističeskij magazin) ‘second hand bookshop’, šiz (← šizofrenik) ‘schizophrenic’, chor (← chorošo) ‘OK’; Czech kama (← kamarád ) ‘buddy/mate’, haš (← hašiš) ‘hashish’, póba (← pobožnost) ‘(religious) service’ (in church slang). In Polish, clippings are relatively rare, cf. in youth language spoko (← spokojnie) ‘chill’, nara (← na razie) ‘bye/so long!’. Clipping is usually accompanied by additional suffixes, such as -ol in psychol (← psychiczny) ‘psycho, some-

1554

VII. Semantics and pragmatics in word-formation II: Special cases

body who is mentally ill’, kibol (← kibic) ‘football hooligan, football fan’, or phonologic alternations in which some linguists notice a sound symbolism, cf. Polish micha (← miska) ‘bowl’, artycha (← artysta) ‘artist’, ciacho (← ciastko) ‘cake’. Analogical Russian structures include igrucha (← igruška) ‘toy’, kartocha (← kartoška) ‘potato’. Zemskaja (1992: 68) interprets them however as suffixal derivatives, distinguishing the -uch(a) and -och(a) suffixes. Some clear examples of clipped and suffixed forms in Russian are alk-aš, alk-ašok (← alkogolik) ‘alcoholic’ and analogously in Serbian and Croatian alk-ić (← alkoholičar). The phonological sequence /Vx/ (where V stands for vowel, x for velar fricative) is in Polish characteristic for both the expressive suffixes such as -ich(a) − Cyganicha ‘Gypsy’, -ach(a) − równiacha ‘cool dude’, -och- − tłuścioch ‘fatso’, and -uch- − staruch ‘old fart’, dzieciuch ‘(spoiled) brat’ and for the suffix-free products of alternation, in which a series of fricatives /s, sz, ś, z/ is replaced by /x/, cf. deska → decha ‘board’, gruszka → grucha ‘pear’, pięść → piącha ‘fist’, wiązka → wiącha ‘bundle’, and so forth. Additionally the process may involve the deletion of a part of the base (usually these are not morphemes, but elements such as /k/, /tk/, /ć/). Szymanek (1996) considered the motivation for negative expression seemingly coded in the velar fricative ch /x/ to belong to the very physiology of emotions. It is indeed striking that this consonant is present in expressive interjections related to sighing or heaving, as in ach!, och!, ech! ‘oh well!’, sequences evoking rude laughter: cha-cha, che-che, chi-chi and other sounds prototypically connected to unpleasant things, such as buch! ‘boom!’ and ciach! ‘swoosh!’. (Strictly speaking other velar consonants seem to symbolize negative attitudes as well, cf. the suffixes -aka-, -aga- and analogous alternations such as Russian beznadežnost’ ‘hopelesness’ → beznadega.) The above derivatives differ from their bases only in their strong deteriorative value. To be precise, this is the way they are received by users of the so called cultured language. From the point of view of an individual speaker, however, their use might be a sign of belonging to a group that uses a particular slang. Only a small part of this vocabulary finds its way to lexicographic analysis which could give it a normative status.

7. Language of love Positive sentiments of love and tenderness are manifest in familial language. Like other variants of colloquial language they are poorly represented in language corpora since the latter mostly register written language. The term familial style may be applied as a general concept, referring to its different manifestations, such as baby talk or caretaker talk. In the same way that augmentatives use a masculine emotional model as a reference point, people’s attitude towards young children and animals functions as a prototype in familial language. It translates also to lovers’ speech, which is rich in so-called affectionate names, i.e. emotionally marked forms of address (cf. Bańko and Zygmunt 2010). The two authors compiled a dictionary of Polish affectionate names, containing extensive lexical material collected via online questionnaires, which merits a comparative analysis. Obviously, not all affectionate terms are products of word-formation. We are interested in hypocorisms, i.e. tender nicknames and diminutives used to address a person in a conversation (or whenever referring to a third party). According to the above mentioned

89. Morphopragmatics in Slavic

1555

questionnaire, these hypocorisms may have bases that stem from various lexical categories: animals − misiaczku (all examples of affectionate names are quoted in their vocative form) ‘teddybear’, koteczku ‘kitty’, sweets − cukiereczku ‘candy’, ptysiu ‘choux bun’, small children − bobasku ‘tot’, aniołku ‘angel’ (in a secular sense), as well as objects that do not have diminutive names for semantic reasons − słoneczko ‘sun’ − or something small by nature, e.g., bąbelku ‘bubble’. No contradiction is involved if such a term has a semantically negative base and expresses tenderness à rebours − brzydalu ‘ugly one’. Similarly, no contradiction is involved if the expression has no literal or structural meaning to start with, in a sort of art-for-art’s-sake coinage, as in the expressive-sounding dziabągu!. Familial language should be distinguished from familiar style which is associated with a negatively received attempt to become familiar with somebody. It should be mentioned that this is the way some people respond to quasi-diminutives used in the above described comitative function, as well as to hypocorisms of first names used in official situations, which is becoming more common especially in TV talk-shows. It is worth noting that the Slavic languages have kept the double system of addressing a person: familiar (Latin tu) and honorific (Latin vos). In addition, some Slavic languages maintain the vocative case that is used in forms of address, especially in familiar context, cf. Slovak švagre ‘brother-in-law-VOC’, priateľu ‘friend-VOC’. In Polish the genetic vocative may function in the subject position, instead of a nominative as its familiarly marked variant, cf. teściu from teść ‘father-in-law’ (Teściu przyszedł ‘The father-in-law came’), and Lechu from Lech (Lechu powiedział ‘Lech said’). These forms may be analyzed as belonging to a pragmatically marked inflexion paradigm, or the -u- forms may be treated as special cases of word-formation. Affective word-formation shows some mechanisms known from child speech. One of the simplest means is reduplication of the form, cf. universally understood names for people in the child’s world: mama ‘mommy’, papa ‘daddy’ and niania ‘nanny’. This word-formation technique is also used to create interjections. Onomatopoeic reduplications in this class of words are quite common, cf. Czech buc, búc ‘plonk’, kap, káp ‘drop’, Polish ple-ple ‘yadda-yadda-yadda’, gul-gul ‘gulp’ (loud swallowing). Suffix extensions with a strong pragmatic effect are also possible, cf. bzz → bzy → bzyk → bzyku ‘buzz’, Russian agú ‘goo-goo (as in baby cooing)’ → agúlen’ki/agúlečki/agúnen’ki/ agúnečki/agúšen’ki/agúški. Reduplications may also be interpreted as linkers or infixes, such as in Czech adjectives/adverbs mal(ý) → mal-ink(ý) → mal-il-ink(ý) ‘small’, and Polish mał(y) → mal-utk(i) → mal-ut-eń-k(i) → mal-ut-eni-eń-k(i) ‘small’. Such forms proliferate in child and folk poetry, as well as in emotional, intimate spoken language. A phonological sign of child speech is in Polish a sequence of palatalized sounds: /ś/, /ź/, /ć/, /dź/, /ń/, since palatalization is a step in ontogenesis that precedes the pronunciation of coronal sounds. This can serve as explanation for emotional hypocorisms such as rąsia ‘hand/arm’, nózia ‘leg’, dziadzio ‘grandpa’, żabcia ‘froggy’, nonio ‘baby’s nose’, which may be derived from first degree diminutives rączka (← ręka ‘hand/arm’), nóżka (← noga ‘leg’), dziadek ‘grandpa’ (← dziad ), żabka ‘froggy’ (← żaba), and nosek (← nos ‘nose’). Child language suspends some of the limitations applying to modificational derivation, which here extends to verbs, cf. Polish spać and spatki/spatuchy ‘to sleep’ (Idziemy spatki ‘We’re going to bed’ − spoken to a child) and examples from Kashubian quoted earlier (cf. section 4).

1556

VII. Semantics and pragmatics in word-formation II: Special cases

There is a thin line between empathy and pity, cf. the Polish suffix -in(a), as in rączyna ‘small, thin hand/arm’ (← ręka), kobiecina ‘poor, pitiable woman’ (← kobieta), from which a negative evaluation of the referent may follow, as in Polish dziennikarzyna ‘poor, bad journalist’ (← dziennikarz). The interpretation of the formant depends greatly on the inner context, i.e. the type of the base. It may be expected that the further its semantics is from the world of a child, the more likely it is to convey a negative attitude, cf. wódka → wódzia ‘vodka’. This incompatibility also explains ironic interpretations, cf. Polish wariat → wariatuńcio ‘madman’ (see section 8). Conversely, an amelioration of the base is also possible in familial language, cf. judgmental descriptions in Polish brudas ‘slob’ (← brud(ny) ‘dirt(y)’), kłamczuch ‘liar’ (← kłamać ‘to lie’) and their derivatives brudasek, kłamczuszek (about a child scornfully, yet with a touch of sympathy). It should be noted that word-formation concerning people’s first names is rich in hypocorisms, as in Slovak Ján → Janíčko, Maria → Marienka; or Russian Aleksandruška/Aleksanja/Aleksanečka/Aleksaša formed from Aleksandr. The number of forms, all of which express a particular attitude towards the addressee, gives interpersonal contacts their emotional character. Some of them are used only with children, or shared between a husband and wife as terms of endearment. It should be noted however that among the forms there exist some that bear a rougher marking, such as Polish Agnieszka → Agniecha, and Katarzyna → Kaśka, allowing the addressees to feel more mature. (They are generally used when speaking to young girls.)

8. Irony and banter Irony is considered a traditional rhetorical device, especially in its literary sense (cf. romantic or Socratic irony). We also use it in daily conversations. Leech considers irony as one of the rules of interpersonal rhetoric − one that allows the interlocutors to maintain a semblance of politeness: “The IP [i.e. irony principle − A. N.] is a ‘second-orderprinciple’ which enables a speaker to be impolite while seeming to be polite” (Leech 1983: 142). It is then an intended deception. Irony is also used for comic and humorous effects. Humor, without which human existence would be insufferable, is a broad term that contains a variety of phenomena that are not always easily separated, such as laughter, mockery, sarcasm, ridicule, or corresponding linguistic genres such as a joke, anecdote, wordplay, pun, and so forth. Word-formation has its share in it as well. Humor, like irony, is culture-dependent, cf. Chłopicki (2005: 139−141). Below are examples illustrating the use of word-formation structures for comic, mocking or humorous effects. Traditionally it is believed that the source of irony lies in the contrast between the literal and intentional meaning. An excellent illustration of this comes from one of the Polish group slangs: roztropa, structurally a ‘sagacious person’ (← roztropny ‘prudent’), means its opposite, viz. ‘clumsy/inattentive person’ (Bogusławski and Wawrzyńczyk 1993: 315). An ironic effect is achieved with the use of the following word-formation means: a) Stylistic or functional contrast between the derivational stem and the formative, e.g., a colloquial stem with trivial meaning and formant marked as literary or foreign, cf.

89. Morphopragmatics in Slavic

1557

wiochmen ‘primitive man’ from wiocha ‘derogative for village’ + English men associated with words such as dżentelmen ‘gentleman’ and biznesmen ‘businesman’; b) Allusion to a hidden semantic connotation in the base, such as the noun rasa ‘race’ and trait czarny ‘black’: A journalist, who had made racist jokes responded to the question Jest pan rasistą? ‘Are you a racist?’ − Tak, jestem rasistą. Nie dopuszczam do siebie czarnych myśli. ‘Yes, I am a racist. I do not allow myself to think in dark [lit. black] colors.’; c) Humorous interpretation of structural meaning; d) Homophony, play on the graphic form revealing the structural potential of a word. Below are some more examples. The first type may be illustrated by faux names of mock-scientific fields with the suffix -logia and faux intellectual currents with the suffix -izm that were created to reveal these “disciplines” as intellectually phony and pseudoscientific, cf. Polish łopatologia ‘*shovellogy’ (from the colloquial łopata ‘shovel’ and the phrase kłaść łopatą do głowy ‘to put things into somebody’s head with a shovel, i.e. to explain something to somebody who is not very smart’). An image of the superficiality of Communism is revealed by Russian agent nouns with the foreign suffix -ant and Russian verb stems: ob’’javljant (← ob’’javljat’ ‘to announce’) ‘somebody who introduces something/somebody publically − *introducer’, vručant (← vručat’ ‘to hand over’) ‘somebody who gives out awards during festivals and galas − *awarder’, polučant (← polučat’ ‘to receive’) ‘somebody who receives these awards − awardee’; cf. the following quote: “Process ceremonii svelsja k rečam „ob”javljantov”, „vručantov”, „polučantov” premii, kotorye podolgu vspominali golodnoe detstvo i govorili ob umiranii teatra.” [The ceremony boiled down to the introducers’, awarders’ and awardees’ speeches, who talked at length about their difficult childhood and the death of theatre.] (Ermakova 2005: 139−140). The humorous effect comes from the fact that the suffix -ant adds to the semantics of base verbs the meaning of habituality in agent nouns, while in this case the activities should be of one-of-a-kind nature. Especially common in this derogatory function are quasi-diminutives that are located close to irony and lack of respect, cf. Russian ėpopejka lit. ‘little epopee’ (← ėpopeja), argumentik lit. ‘little argument, little piece of evidence’ (← argument) in scientific discussion, Polish złodziejaszek (about a petty thief ) (← złodziej), and − analogous to familial papcio ‘pap’ − skarbuńcio (← skarb ‘treasure’). Quasi-diminutives are also used in advertisement, cf. Czech České lázně jsou dnes rodinné střibřičko lit. ‘Today Czech spas are the family silver-QUASIDIM’ (from rodinné stříbro ‘family heirlooms’). The judgmental or playful attitude of the speaker may be expressed in humorous reinterpretations of derivatives or via a reference to folk etymology: in Polish humorous slang komunista is not a communist, but a child receiving his or her first Holy Communion (Polish komunia), Czech mšice ‘a woman, who takes part in every mass [Cz. mše]’ and mšice ‘plant louse’. A sure comic effect is also achieved via homonymy ‘thing’ − ‘person’, cf. Russian jaičnica ‘scrambled eggs’ and colloquial for ‘woman who sells eggs’, analogously Polish herbatnik ‘(tea) biscuit’ and ‘fan of tea’. Folk humor feeds on allusions to the sexual sphere, cf. the Russian compound členovoz ‘government limousine’ analogical to parovoz ‘steamboat’, which uses the homonymy of the base člen ‘member [of the politbureau or USSR government]’ and ‘penis’, Polish smutas ‘officer of security services’ alluding to the words smutny ‘sad’ and kutas, vulgar for ‘penis’ (Bogusławski and Wawrzyńczyk 1993: 338). This latter word is an

1558

VII. Semantics and pragmatics in word-formation II: Special cases

example of contamination/blending that is also used for wordplay, cf. humorous Russian gajdaronomika from the words Gaidar [Russian economist and politician in the beginning of the 1990s] + ėkonomika ‘economy’. In written language a humorous effect may be achieved by alluding to spurious derivational homonymy, such as in the Czech commercial boschské − božské (‘from Bosch’ or ‘from God’ − both sch and ž pronounced as [ʃ]) where the Bosch company was compared to God). Similarly the name of a Warsaw sushi bar Sush-arnia formed with sushi + place noun suffix -arnia humorously alludes to suszarnia ‘drying room’. This is a way unexpected allusions are explored to attract attention of the audience, which is after all the purpose of politics and advertisement.

9. Word-formation in the context of other rhetorical devices As noted by Ermakova (2005: 106), “irony prefers dealing with a high degree of a trait, judgment, or characteristics: it is present in an exaggerated false praise, in playing down a particularly negative trait; it does not respond well to moderation”. Precisely these two opposite strategies − exaggeration and restraint − lie at the base of the related rhetoric devices of hyperbole, litotes, and euphemism. Word-formation hyperbole is achieved via compounds featuring intensifying elements such as super-, hiper- ‘hyper’, mega-, and turbo-, which are meant to bring the addressee to a euphoric state, and are oftentimes used in advertising, cf. Slovak hypermoderný ‘hyper + modern’ or Croatian megapopularan ‘mega + popular’. Comparative and superlative forms of semantically non-gradable adjectives may also serve this function, such as Czech fantastičtější ‘more fantastic; lit. fantastic-COMP’, hedvábnější ‘silkier’ in texts of advertisements. The hyperbolic effect is also achieved via reduplication as means of forming adverbs, especially in folk and child literature, cf. Russian bystro-bystro ‘fastfast’ and Slovak široko-ďaleko lit. ‘wide-far’. The speaker’s restraint is, in turn, expressed via litotes. Its word-formation means include adjectival structures with a negative prefix ne-/nie-, cf. Russian/Polish neglupyj/ niegłupi used instead of the antonym proper mudryj/mądry ‘smart’. To say about somebody who is really smart that he is “not stupid” is definitely an understatement. An examination of word-formation antonymy from a pragmatic perspective is, unfortunately, still only fragmentary (cf. Ermakova 2005: 142 ff.). Yet at first glance a certain regularity may be observed: the function of litotes is performed by a negated element on the negative pole of the antonymic opposition, cf. Polish niebiedny ‘not poor’ (about a wealthy person), nielekki ‘not light’ (about something that is heavy), niezły ‘not bad’. By contrast, the use of a derivative with a negative prefix from the positive antonym creates the effect of euphemism, cf. On jest niebogaty ‘He is not rich’ (mildly about a poor person) and niemądry (mildly about a stupid person).

10. References Bańko, Mirosław 2008 Współczesny polski onomatopeikon. Ikoniczność w języku. Warszawa: WN PWN.

89. Morphopragmatics in Slavic

1559

Bańko, Mirosław and Agnieszka Zygmunt 2010 Czułe słówka. Słownik afektonimów. Warszawa: WN PWN. Barz, Irmhild 1988 Nomination durch Wortbildung. Leipzig: Verlag Enzyklopädie. Bogusławski, Andrzej and Jan Wawrzyńczyk 1993 Polszczyzna, jaką znamy. Nowa sonda słownikowa. Warszawa: Katedra Lingwistyki Formalnej, Uniwersytet Warszawski. Brown, Penelope and Stephen C. Levinson 1987 Politeness. Some universals in language usage. Cambridge: Cambridge University Press. Chłopicki, Władysław 2005 Ironia jako zjawisko kulturowe. In: Marta Dąbrowska (ed.), Język trzeciego tysiąclecia III. Język polski i języki obce − kontakty, kultura, dydaktyka, 135−141. Kraków: tertium. Chomjak, Miroslava 1997 Suffiksy sub”ektivnoj ocenki v sovremennych rusinskich literaturnych tekstach. In: Michał Blicharski and Henryk Fontański (eds.), Zagadnienia słowotwórstwa i składni w opisie współczesnych języków słowiańskich. Vol. 1, 78−87. Katowice: Uniwersytet Śląski. Dąbrowska, Marta 2005 Język płci − prawda czy mit? In: Marta Dąbrowska (ed.), Język trzeciego tysiąclecia III. Język polski i języki obce − kontakty, kultura, dydaktyka, 145−157. Kraków: tertium. Dokulil, Miloš 1962 Tvoření slov v češtině. Vol. 1: Teorie odvozování slov. Praha: Nakladatelství Československé akademie věd. Dokulil, Miloš 1979 Teoria derywacji. Wrocław: Ossolineum. Ermakova, Ol’ga P. 2005 Ironija i ee rol’ v žizni jazyka. Kaluga: Izdatel’stvo KGPU. Furdík, Juraj 2004 Slovenská slovotvorba. Prešov: Náuka. Grzegorczykowa, Renata, Roman Laskowski and Henryk Wróbel (eds.) 1998 Gramatyka współczesnego języka polskiego. Morfologia. Warszawa: WN PWN. Jadacka, Hanna 2001 System słowotwórczy współczesnej polszczyzny (1945−2000). Warszawa: WN PWN. Kiklewicz, Aleksander 2002 Język polski obojga narodów? Wpływ języka polskiego na język białoruskich mediów. In: Władysław Chłopicki (ed.), Język trzeciego tysiąclecia II. Polszczyzna a języki obce. Przekład i dydaktyka, 321−329. Kraków: tertium. Kiklewicz, Aleksander 2007 Zrozumieć język. Łask: Oficyna Wydawnicza Leksem. Kuryłowicz, Jerzy 1960 Dérivation lexicale et dérivation syntaxique. Contribution à la théorie des parties du discours. In: Jerzy Kuryłowicz, Esquisses linguistiques, 16−26. Wrocław/Kraków: Polska Akademia Nauk. Leech, Geoffrey 1983 Principles of Pragmatics. London: Longman. Liptáková, Ľudmila 2000 Okazionalizmy v hovorenej slovenčine. Prešov: Náuka. Martincová, Olga (ed.) 1998 Nová slova v češtině. Slovník neologizmů. Praha: Academia. Martincová, Olga (ed.) 2004 Nová slova v češtině. Slovník neologizmů. 2. Praha: Academia.

1560

VII. Semantics and pragmatics in word-formation II: Special cases

Milewska-Stawiany, Małgorzata 2007 Funkcje derywatów deminutywnych i augmentatywnych w piśmiennictwie górnołużyckim. In: Viara Maldjieva and Zofia Rudnik-Karwatowa (eds.), Słowotwórstwo i tekst, 149−157. Warszawa: Slawistyczny Ośrodek Wydawniczy. Nagórko, Alicja 1997 Zur (west)slavischen Morphologie aus pragmatischer Perspektive. Zeitschrift für Slawistik 42(3): 263−273. Neščimenko, Galina P. 2004 O nekotorych tendencijach v razvitii sovremennogo slavjanskogo slovoobrazovanija. In: Aljaksandr Lukašanec and Zinaida Charytončyk (eds.), Prablemy tėoryi i gistoryi slavjanskaga slovaŭtvarėnnja, 85−97. Minsk: VTAA Prava i ėkanomika. Ohnheiser, Ingeborg 2003 Pragmatiko-stilističeskaja differenciacija slovoobrazovatel’nych sredstv i tendencija ee stiranija. In: Ingeborg Ohnheiser (ed.), Komparacja systemów a funkcjonowania współczesnych języków słowiańskich. 1. Słowotwórstwo/Nominacja, 216−248. Opole: Uniwersytet Opolski. Ohnheiser, Ingeborg 2007 Nominacija i emocija. In: Alicja Nagórko (ed.), Sprachliche Kategorien und die slawische Wortbildung, 347−368. Hildesheim: Olms. Raecke, Jochen 2003 O situativnoj ili stilističkoj podeli južnoslovenskih jezika i o stilističkom razvoju divergencije i konvergencije u oblasti tvorbe reči. In: Ingeborg Ohnheiser (ed.), Komparacja systemów a funkcjonowania współczesnych języków słowiańskich. 1. Słowotwórstwo/ Nominacja, 234−248. Opole: Uniwersytet Opolski. Raecke, Jochen 2004 Bedeutung(en) und Funktion(en) von Deminutiva in der Sprache und in Texten. In: Marion Krause and Christian Sappock (eds.), Slavistische Linguistik 2002, 169−199. München: Sagner. Sokolová Miloslava 2007 Vzťah slovnodruhových a onomaziologických kategoriálnych významov. In: Alicja Nagórko (ed.), Sprachliche Kategorien und die slawische Wortbildung, 179−195. Hildesheim: Olms. Srpová, Hana 2008 Knížka o reklamě. Ostrava: Ostravská univerzita, Filozofická fakulta. Szymanek, Bogdan 1996 The morphology of phonological strings: Polish /Vx/. In: Henryk Kardela and Bogdan Szymanek (eds.), A Festschrift for Edmund Gussmann, 293−308. Lublin: University Press of the Catholic University of Lublin. Švedova, Nataľja Ju. (ed.) 1980 Russkaja grammatika. Vol. 1. Moskva: Nauka. Uluchanow, Igor and Renate Belentschikow (eds.) 2007 Russisch-deutsches Wörterbuch der neuen Wörter. Moskva: Azbukovnik. Zemskaja, Elena A. 1992 Slovoobrazovanie kak dejatel’nost’. Moskva: Nauka.

Alicja Nagórko, Berlin (Germany)