Diachronic Studies on Information Structure: Language Acquisition and Change 9783110227475, 9783110227468

Citation preview

Diachronic Studies on Information Structure

Language, Context, and Cognition Edited by Anita Steube Volume 10

De Gruyter

Diachronic Studies on Information Structure Language Acquisition and Change

Edited by Gisella Ferraresi and Rosemarie Lühr

De Gruyter

ISBN 978-3-11-022746-8 e-ISBN 978-3-11-022747-5 Library of Congress Cataloging-in-Publication Data Diachronic studies on information structure : language acquisition and change / edited by Gisella Ferraresi and Rosemarie Lühr. p. cm. -- (Language, context, and cognition; 10) Includes index. ISBN 978-3-11-022746-8 (alk. paper) 1. Grammar, Comparative and general--Topic and comment. 2. Language acquisition. 3. Language and languages--Variation. 4. Linguistic change. 5. Language change. I. Ferraresi, Gisella. II. Lühr, Rosemarie, 1946P298.D53 2010 401'.93--dc22 2010027102 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. © 2010 Walter de Gruyter GmbH & Co. KG, Berlin/New York Printing: Hubert & Co. GmbH & Co. KG, Göttingen ∞ Printed on acid-free paper Printed in Germany www.degruyter.com

Contents Gisella Ferraresi & Rosemarie Lühr The Role of Information Structure in Language Change: Introductory Remarks ............................................................................. 1 Rolando Féliz Armendáriz Information Structure, Constituent Order, and Case in Warihío ........... 15 Carlotta Viti The Information Structure of OVS in Vedic.......................................... 37 Christoph Gabriel & Esther Rinke Information Packaging and the Rise of Clitic Doubling in the History of Spanish ....................................................................... 63 Marit Westergaard Cue-based Acquisition and Information Structure Drift in Diachronic Language Development .................................................. 87 Dejan Matić Discourse and Syntax in Linguistic Change: Decline of Postverbal Topical Subjects in Serbo-Croat......................... 117 Kristine Gunn Eide Prosody, Information Structure and Word Order Changes in Portuguese ....................................................................................... 143 Melanie Wratil The Development of V-to-C Movement in the West Germanic and Romance Languages................................... 161 Svetlana Petrova & Roland Hinterhölzl Evidence for Two Types of Focus Positions in Old High German ....... 189 Index .................................................................................................. 219 Contributors ........................................................................................ 223

The Role of Information Structure in Language Change: Introductory Remarks 1

Gisella Ferraresi (Frankfurt /M.) & Rosemarie Lühr (Jena) The renewed interest in the interaction between ‘information structure’ and grammar is witnessed by many recent publications which try to disentangle the correlation between syntax, prosody and information structure in modern languages. The aim of this volume is to encourage the discussion of the role of information structure in language change and that of operationable methods which can be applied on corpus languages when working on information structure. Correlations between word order and intonational patterns which, through the question test, enable us to detect the informativity of a sentence (cf. Sgall et al. 1973), for example, are not applicable to corpus languages. Together with discourse semantics, prosody and syntax are the important ingredients for the production and interpretation of information packaging. Different languages, however, make different use of these components. German, for example, operates in the first place at the prosodic level 2 , although it can also make use of marked syntactic constructions like hanging topics or left dislocation (Frey 2005). In Catalan, on the other hand, syntactic constituents are moved in order to constitute the corresponding prosodic constituents (Vallduví & Vilkuna 1998). Other languages again adopt morphological marking. 3 In historical linguistics, the prosodic correlate of syntactic constructions cannot be tested directly. However, it is still possible to reach some relevant generalizations about the realization of pragmatic categories like Topic, Focus and their interaction with word order in written texts. This is facilitated also thanks to the intensive research in the last years of the mapping of pragmatic categories in syntax – an interest which has as1

2 3

Some of the papers contained in this volume have been read at the workshop ‘The role of information structure on language change’ during the annual DGfS Conference held at the University of Siegen (Germany) in 2007. For a discussion on the prosodic correlates of Topic and Focus in German see Féry (1993); Uhmann (1991) See also. u.a. Jacobs (1984); Uhmann (1991).

2

Gisella Ferraresi & Rosemarie Lühr

sumed an important role in formal syntactic theories as well. Chomsky (1995), for example, recognizes that the interface between syntax and discourse is realized in the CP. Likewise coached in generative grammar, the so-called cartographic model (Rizzi 1997 a.o.) holds Topic and Focus to be realized in precise syntactic positions in a split CP. In Rizzi (1997) the CP is therefore further distinguished in (1)

Force TOP FOC FinP

From a diachronic point of view, many structural changes which involve the left periphery can in this way be put down to changes in the mapping of pragmatic information onto syntax. For the ancient Indo-European languages, Kiparsky (1995) assumes a structural differentiation between Topic and Focus. His proposal is based on Hale's assumption that topic and wh-elements move to different positions in these languages (Kiparsky 1995: 253): (2)

S'' (=CP) 3

TOPIC

S'

XPi

3

FOCUS XPii

S 3

ti

tii

In ancient Indo-European languages verb fronting is a strategy of emphasis, whereas clause typing is achieved by particles and sentence prosody. Therefore verb-second phenomena are the rule neither in old Germanic nor in old Romance languages 4 . Gothic, for example, shows a rich array of particles – albeit more limited than other ancient Indo-European languages like Ancient Greek –, many of which induce fronting of the finite verb depending on the clause type on the one hand and on the informational status of the moved element on the other (cf. Ferraresi 2005). Verb movement is found regularly only in imperative clauses. In Old High German and Old English, some of the corresponding particles are still attested (Axel 2007; Fuß & Trips in print; Kemenade & Los 2009). However, verb movement in Old High German declarative sentences is more and more generalized, for example in verb-first structures (cf. Hinterhölzl, Petrova & Solf 2005), especially in so-called thetic sentences. There, it has the function of introducing a new discourse referent while changing the discourse situation, nevertheless maintaining the narration line. Verb4

See a.o. Ferraresi & Goldbach (2002) for Old French, Axel (2007) for Old High German.

Introductory Remarks

3

second word order, on the other hand, has the function of introducing a new referent, which predominantly appears in the first position. In declarative main clauses of Modern German, this word order has been generalized and has become the unmarked one after different discourse particles have got lost. In this sense, the question arises if there really is a transition from discourse-oriented to syntax-oriented languages and, if so, if this transition always takes place in that direction. According to Givón (1979: 98), the word order in the “pragmatic mode” is mostly governed by pragmatic principles, while the word order in the “syntactic mode” displays “case functions”. The degree of accessibility or activation of a discourse referent of a certain phrase, e.g. the correlation with the grammatical function (obliqueness hierarchy: subject > object1 > object2 > ... > adjuncts), actuality, definiteness (cf. Givon 1983; Gundel, Hedberg & Zacharski 1993) is, however, influenced by many different factors. Semantic salience is in any case relevant for the analysis of information structure in ancient Indo-European languages, as it allows for a differentiation between discourse-oriented and syntax-oriented languages (Givón 1979). Personal pronouns, for example, often appear in the topic position in Old Indic, even though ancient Indo-European languages are subject-pro-drop and partially object-pro-drop languages (Luraghi 2003; Lühr 2005; Keydana 2009). Functional tests, especially those operating on the level of text development, in this way allow hypotheses on the informational status of a linguistic unit since they can be supported by semantic determination. The investigation of corpus languages can also cast light on synchronic issues in modern languages. There is some significance, for example, in the fact that Ancient Indic marks contrastive focus specifically (Lühr 2009), supporting the differentiating analysis between New Information and Contrastive Focus against the unified semantic interpretation of focus (Rooth 1992; Krifka 1993; 2007), according to which focus always evokes a set of alternatives. This differentiation is necessary as the intonations of these two foci differ from each another also in modern languages like in German (Kiss 1998; Steube 2002; Späth 2007). In the data material of corpus languages it is thus possible to define which type of focus is realized 5 . According to Umbach (2004), it is unclear how the alternatives are contextually restricted in case of a purely semantically determined focus. Consequently, it has to be assumed that every sentence adds information to the preceding one; this is why continuation sentences are subdivided into established information and new information. The established information is usually expressed via definite nominal or prepositional groups (with definite articles in German and Greek), possessive pronouns or per5

For a good overview of the different categories cf. Féry & Krifka (2008).

4

Gisella Ferraresi & Rosemarie Lühr

sonal pronouns or via anaphorical, often deictic expressions. Among these, the Aboutness Topic is typically a referential DP. Still, the topic may be contrast-focussed and thus be a Contrastive Topic. As compared to the topic, the focus is related to a specific domain, the focus domain. As long as no contrast foci are present, the focus domain includes new information. Under certain circumstances, the background needs to be added; it usually consists of the established information outside the focus domain.To determine the topic, the approach in Frascarelli & Hinterhölzl (2007) is promising, as Aboutness Topics, Contrastive Topics and Familiar Topics are dealt with separately, which are all operationalizable pragmatical categories that can be found in non-Indoeuropean languages as well (Ermisch 2007). When questions are answered in the text itself in corpus languages, a solicited New Information Focus is involved. To determine the unsolicited New Information Focus (new discourse referents as well as new relations between given discourse referents) in narrative texts, a question can be formed that is as universal as possible and that is based on established material (Petrova & Solf 2009 on the quaestio-thesis by Klein & von Stutterheim (1987). Still, when studying ancient Indo-European languages, this procedure can only be a possible test in addition to semantic analyses, e.g. in connection to Dik's (1989: 268) strategies for the introduction of a new discourse referent: 1. meta-linguistic information: ‘I’m going to tell you a story about X’ 2. as an object or second argument in sentences with transitive verbs: ‘In …, we saw X’ 3. as a subject of “an existential or presentation construction” of the type: ‘Once upon a time, there lived/was …’; such constructions are typical of sentences initializing a text 4. with verbs of movement indicating an “appearing on the scene”: ‘Suddenly, right before our eyes, X appeared …’ (Petrova & Solf 2009). To be able to comment, in a second step, on the part of the linguistic system of an ancient Indo-European language which concerns the information structure, the analysis units of information structure first have to be synchronically determined and to be assigned to the language material. When one has undertaken several synchronic sections, diachronic processes become evident and explicable as well. Moreover, comparing different languages makes language-specific and general traits of information structure in ancient Indo-European languages become apparent. As mentioned above, however, the study of ancient Indo-European languages can support the analysis of modern ones.

Introductory Remarks

5

These are only some of the problems which one is exposed to when working on pragmatic categories in corpus languages, and which the contributions of this volume partly address. The papers analyse different phenomena of language change connected to information structure in many different old languages and language families, applying some of the techniques described above for corpus languages. The languages analyzed in the papers belong to the Uto-Aztecan (Armendáriz), Vedic (Viti), Slavic (Matić), Romance (Gabriel & Rinke, Eide) and Germanic language families (Wratil, Westergaard, Petrova & Hinterhölzl). The phenomena considered range from word order to pronouns and verb movement. The paper ‘Information structure, constituent order, and case marking in Warihío’ by Rolando Féliz Armendáriz discusses word order in Warihío, an Uto-Aztecan language spoken in Mexico. The methods applied for this study are both text analysis and direct elicitation. Armendáriz shows that syntactic accounts are not sufficient for understanding the principles underlying the various patterns of word order in Warihío. He therefore proposes a pragmatic explanation based on Lambrecht (1994). Accordingly, focus determines the fronting of constituents, while topical material usually follows. The analysis is mainly conducted on a synchronic level, as Warihío does not possess a long written tradition that permits a reconstruction of previous linguistic stages. However, Armendáriz indicates that some diachronic evolution is observable in the use of a suffix, which is shifting from a syntactic to a pragmatic function of signalling definiteness. Carlotta Viti's paper ‘The information structure of the OVS order in Vedic’ discusses OVS word order in Vedic, which is marginally attested in the Rig-Veda, the earliest Vedic text, and completely unattested in the subsequent Indian linguistic varieties. Owing to this fact, OVS has been considered as being an exception to the basic SOV word order, related to the poetic register of the Rig-Veda. However, data point out that OVS is consistently used to convey a certain type of information structure, where the referent of the object is more topical (i.e. human, specific, definite, persistent, etc.) than the referent of the subject. Accordingly, the synchronic rarity and the diachronic disappearance of the OVS order are due to the fact that its information structure has fewer chances to appear in discourse. In their article ‘Information packaging and the rise of clitic-doubling in the history of Spanish’, Christoph Gabriel & Esther Rinke investigate the diachronic development of clitic-doubling (CD) in Peninsular Spanish. They show how this construction differs from structures with rightdislocated objects (clitic right dislocation; CLRD). Besides prosodic and syntactic differences, both structures differ in their informational status. In CLRD constructions, a right-dislocated object is separated from the core

6

Gisella Ferraresi & Rosemarie Lühr

sentence by a prosodic boundary being located in a position outside vP, whereas in CD constructions a nominal or strong pronominal object forms, together with a co-referential clitic pronoun, one prosodic domain – the doubled object remains in its vP-internal base position. Here, object clitics function as agreement markers. As for information structure, the main difference between CLRD and CD constructions is that in a CLRD structure the object constitutes a dislocated topic, whereas in CD constructions it belongs to the focus domain. In modern Spanish, elements displaying a high degree of ‘topicality’, such as pronouns, definite noun phrases, experiencer arguments and proper names, are more likely doubled than those situated lower in the topicality hierarchy. This seems to confirm that CD constructions follow the topic hierarchy given by Givón (1976). As a whole, contexts with obligatory CD are quite rare and the appearance of the clitic is due to optionality as it is in modern Peninsular Spanish. In the early periods, in the 12th and 13th centuries, resumptive clitics appear in texts that reflect spoken language. However, such examples are far from being systematic or obligatory. Resumptive clitics become more and more frequent, especially during the 16th century. Gabriel & Rinke agree with Fontana (1993) that CD should be analyzed as an agreement phenomenon. They provide evidence from texts from the 15th up to the 18th century for the assumption that there was a diachronic tendency from more topical to less topical elements to co-occur with a resumptive clitic, whereby a topicalization strategy (clitic right dislocation) was reanalyzed into a doubling structure. The reason behind this development is that right dislocation is more marked and less economic than CD. The paper ‘Cue-based acquisition and information structure drift in diachronic language development’ by Marit Westergaard evaluates information structure from another point of view. Within a model of language acquisition and change which recognizes the importance of cues and socalled ‘micro-cues’ in the primary linguistic data (PLD) that children are exposed to, she shows that information structure may be a factor for wordorder change. Following Lightfoot (1999, 2006) she assumes that there is a UG requirement for cues to be obligatory in children's I(nternalized)language grammars. Westergaard analyses word-order variation that is dependent on linguistically relevant sub-categories as pertaining to the Ilanguage. The data from English and Scandinavian languages present mixed grammars which allow two different subject or object positions. One of the positions is preferred for discourse-given elements (typically pronouns) and the other for informationally new or focused elements (typically heavier elements such as full DPs or clauses). The paper shows that word order variation is relatively common in present-day grammars,

Introductory Remarks

7

whereby the choice of a word-order is at least partly dependent on information structure. In accordance with Behaghel, Westergaard demonstrates that discourse-given elements, typically pronouns, appear in a high position, whereas informationally new or focused elements stay in a lower position. Historical data, too, attest that word order variation based on mixed VO/OV is caused by information structure. With respect to child language, Westergaard demonstrates on the basis of Norwegian and Russian data that children have an early sensitivity to patterns of information structure, producing both word orders in a more or less target-consistent way in e.g. Norwegian wh-question and subject-shift constructions. In the second part of her paper, Westergaard investigates the realization of subjects and objects in a sample of Norwegian conversational speech and in a sample of Norwegian and English child-directed speech. The results show a clear preference of putting discourse-given information in subject position and discourse-new or focused information in object position. Thus, for mixed word orders where the choice depends on information structure, word orders linked to informationally given (typically pronominal) subjects should be naturally frequent in the E-language. Word orders linked to new or focused (full DP) subjects should be correspondingly infrequent. For mixed grammars involving two object positions, the situation should be the other way round. The mixed word order system should be easily learnable, and from an acquisition perspective such grammars should also be relatively stable diachronically. However, in this scenario the input varies, since not all children are exposed to exactly the same linguistic data. The low frequency of a construction may fall below the threshold for acquisition in some children, who will develop a grammar with only one subject or object position. The decline in the frequency of word order would be the crucial factor for change. Certain developments can be reversed if external factors and/or internal language development cause the PLD to change as in some present-day Norwegian dialects, where the interference from the standard language seems to cause a return from non-V2 to V2 word order. In his contribution ‘Discourse and syntax: the decline of postverbal topics in Serbo-Croat’, Dejan Matić deals with a word order change that cannot be attributed to syntactic processes and is therefore impossible to phrase in terms of competing grammatical choices. Postverbal topical subjects (PTS), i.e. subjects with topical interpretation placed to the right of the verb, have been attested throughout the history of Serbo-Croat (SC) and seem to be as frequent and of the same range of functions until the 19th century. These constructions are a) inversion, a verb-second-like construction in which subjects appear postverbally after a fronted whword, relative pronoun, quote, and a fronted focus, b) vS, a construction

8

Gisella Ferraresi & Rosemarie Lühr

with a focused postverbal subject, c) VsX, a construction with a topical postverbal subject. They differ in intonation contour, position of sentence adverbials, position of the subject relative to the copula/auxiliary and function as continuous topics. During the 20th century, the frequency and variability of PTSs rapidily declined, even though speakers are still ready to construe all types of clauses with PTSs like the ones attested in the sources from the 19th century. The reasons for this language change are a number of interrelated sociolinguistic factors, initiated by pragmatic changes. In the 20th century, sixty to seventy per cent of all instances of PTSs are more or less petrified verb-subject collocations, which are regularly used in a particular discourse function. Matić suggests to introduce a process of automatisation into the repertoire of diachronic changes, a kind of binding of syntactic structure to certain lexemes. The only change between the 19 and the 20th centuries that seems to have led to something resembling a categorical difference is the change in the function of VsX clauses with PTSs, which are no longer markers of resultative and consecutive events in narratives. The syntax remains unchanged, only discourseorganization principles change. Kristine Eide, in her paper ‘Prosody, information structure and word order change in Portuguese’, considers the word-order changes which have taken place in the period from Classical Portuguese to Modern European Portuguese, which – according to Eide – have their reason in a prosodically driven change of discourse patterns. This change caused a fixation of the subject in the preverbal position, which then developed from a topic position to a more neutral one. Classical Portuguese is a language with topic-verb-XP structure, in which inversion takes place in case the first position is not occupied by the subject. Eide observes a decrease in the postverbal position of subjects of unaccusative verbs which represent old information. Topics in Classical Portuguese always move to the beginning of the sentence, whereas the rest of the background follows the finite verb. This creates many inversion structures, in case the topic is not a subject. In Modern European Portuguese only new information can occur postverbally. Melani Wratil's paper ‘The development of imperative V-to-C movement in the West Germanic and Romance Languages’ contributes to the discussion about the movement of the imperative into the C-position. In the Indo-European languages the imperative as a directive speech act constitutes a genuine directive operator represented by a zero-expression. With respect to head movement, imperative and indicative behave in the same way. In VSO languages they occupy the regular clause-initial position, in SVO languages they are positioned immediately after the subject, and in SOV languages they appear in sentence-final position. But in the

Introductory Remarks

9

West Germanic SOV languages, as for example in German, as well as in the West Germanic SVO languages, as for example in Yiddish, imperative verbs normally precede their complements and adjuncts. Here, the exceptional structure of the West Germanic and Romance imperatives is a result of the emergence of V2 phenomena. Root clauses are reanalyzed as V2 structures as soon as the finite auxiliary or lexical verb follows an initial focalized or topicalized constituent in the second position, whereby the fronted focus phrases are considered as the scope-marking specifier of a new functional projection CP. The V2 properties of the modern West Germanic and Romance languages are strongly connected with the rise of V1 clauses. V1 structures are the structural outcome of a focalization process. Head movement with imperative verbs had been solidified by analogy with the finite verbs of V1 interrogative and declarative clause before V2 structures became basic structures. Wratil shows why especially the V1 structures of imperative clauses and therefore the imperative verb movement to C has been able to survive in the West Germanic and Romance languages up to this day. In modern Standard English, in all nonnegated and non-emphasized imperative clauses V-to-C movement applies at LF, whereas in negative and emphasized affirmative imperative clauses do moves to C before Spell out. In their paper ‘Evidence for two types of focus positions in Old High German’, Svetlana Petrova & Roland Hinterhölzl – using data from the Old High German Tatian translation – propose a basic VO order and an informational account in explaining word-order variation in subordinate clauses in Old High German. This view is in contrast with two other positions found in the theoretical discussion on basic word order in (old) Germanic languages. The first one is based on a basic OV order with different possible extraposition movement operations, the other one assumes two underlying base orders, which are due to different grammars (the socalled ‘Double Base Hypothesis’). Petrova & Hinterhölz's argumentation is supported by a battery of syntactic tests which show a contradictory picture for both hypotheses. These tests include the analysis of materials following the finite verb in subordinate clauses, such as DPs with different functions, PPs, predicative adjectives or nouns, Verb Raising and Verb Projection Raising. Root patterns (Verb-first and Verb-second) in subordinate clauses do not confirm the hypothesis of root patterns, either. The information-structural approach proposed in this paper considers the distribution of background material in comparison to asserted information and narrow focus patterns. Background materials tend to occupy the Wackernagel position, which in subordinate clauses is the position immediately following the subordinating conjunction. In most of the cases, pronouns in the corpus data are positioned in the Wackernagel position

10

Gisella Ferraresi & Rosemarie Lühr

also against the Latin original. A similar observation is made for accessible or discourse-anaphoric full lexical phrases: they are moved from a postverbal position to the position after C°. In postverbal position of subordinate clauses, arguments of the verb and non-finite parts of complex verbal predicates are considered. Arguments in postverbal positions represent new information, non-finite parts of complex verbal predicates constitute the asserted part of the proposition; in particular it is shown in the paper that contrastive focus regularly appears left-adjacent to the finite verb. Petrova & Hinterhölzl propose a structure where the verb moves to the head of a Focus phrase at the left edge of the middle field. Whereas contrastive elements move to SpecFoc, new objects remain in the scope of the focus head. Last but not least, we would like to thank all our authors, who cooperated with us with great patience. We also thank the series editors, above all Anita Steube. Many thanks also to the person in charge of linguistics, Heiko Hartmann, as well as to Andreas Brandmair at de Gruyter's, who assisted us in all steps of the publication process.

References Axel, K. (2007): Studies on Old High German Syntax: Left Sentence Periphery, Verb Placement and Verb Second. Amsterdam and Philadelphia: John Benjamins. Bosch, P., Rozario, T. & Zhao, Y. (2003): Demonstrative Pronouns and Personal Pronouns. German der vs. er. In: Proceedings of the EACL 2003 Budapest. Bybee, J. (2007): Frequency of use and the organization of language. Cambridge: Cambridge University Press. Chomsky, N. (1995): The Minimalist Program. Cambridge, Mass.: MIT Press. Denison, J. D. (1996): The Greek particles. Hackett. Dik S. C. (1989/1997): The Theory of Functional Grammar. Part I: The Structure of the Clause. Berlin & New York. Ermisch, S. (2007): Issues in the Left Periphery. A Typological Approach to Topic and Focus Constructions. Frankfurt a.M. and Bern: Peter Lang. Ferraresi, G. (2005): Word Order and Phrase Structure in Gothic. Leuven: Peeters. Ferraresi, G. & M. Goldbach (2002): V2 Syntax and Topicalization in Old French. Linguistische Berichte 189/2002, 2-25. Féry, C. (1993): German Intonational Patterns. Tübingen: Niemeyer. Féry, C. & M. Krifka (2008): Information Structure. Notional distinctions, ways of expression. In: P. van Sterkenburg (ed.), Unity and diversity of languages. Amsterdam and Philadelphia: John Benjamins, 123-136

Introductory Remarks

11

Frey, W. (2005): Pragmatic properties of certain German and English left peripheral constructions. Linguistics 43, 89-129. Fontana, J. (1993): Phrase Structure and the Syntax of Clitics in the History of Spanish. Ph.D. Diss. Pennsylvania: University of Pennsylvania. Frascarelli M. & Hinterhölzl, R. (2007): Types of Topics in German and Italian. In: S. Winkler and K. Schwabe (eds.), On Information structure, Meaning and Form, 87-116. Amsterdam and Philadelphia: John Benjamins. Fuß, E. & C. Trips (in print): The syntax and semantics of the temporal anaphor 'then' in Old and Middle English. In: T. McFaddan and F. Schäfer (eds.), Proceedings of CGSW 21+22. Amsterdam and Philadelphia: John Benjamins Givón, T. (1976): Topic, pronoun, and grammatical agreement. In: C. N. Li (ed.), Subject and Topic, 149-188. New York: Academic Press. Givón, T. (1979): From discourse to syntax: grammar as a processing strategy", in: T. Givon (ed.), Syntax and semantics, Vol. 12, Discourse and syntax, 81-112. New York: Academic Press. Givón, T. (1988): The pragmatics of word order: predictability, importance and attention, 147-187. In: M. Hammond, et al. (2007), Studies in syntactic typology 18, 5-41. Amsterdam and Philadelphia: John Benjamins. Givón, T. (ed.) (1983): Topic continuity in discourse. A quantitative cross-language study. Amsterdam and Philadelphia: John Benjamins. Gundel, J.K., Hedberg, N. & Zacharski, R. (1993): Cognitive status and the form of referring expressions in discourse. Language 69, 274-307 Hinterhölzl, R., S. Petrova & M. Solf (2005): Diskurspragmatische Faktoren für Topikalität und Verbstellung in der ahd. Tatianübersetzung (9. Jh.). In: S. Ishihara, M. Schmitz & A. Schwarz (eds.), Approaches and Findings in Oral, Written and Gestural Language. Working Papers of the SFB 632, Interdisciplinary Studies on Information Structure 3. Berlin, 143-182. Jacobs, J. (1984):Funktionale Satzperspektive und Illokutionssemantik. Linguistische Berichte 91, 25-58. Kemenade, A. van & B. Los (2009): Discourse adverbs and clausal syntax in Old and Middle English. In: A. van Kemenade and B. Los (eds.), The Handbook of the History of English, 224-248. London: Blackwell. Keydana, Götz (2009): Latente Objecte und altindische Diskursgrammatik, 125142. In: E. Rieken, & P. Widmer (eds.), Pragmatische Kategorien. Form, Funktion und Diachronie. Akten der Arbeitstagung der Indogermanischen Gesellschaft vom 24. bis 26. September 2007 in Marburg. Wiesbaden: Dr. Reichert Verlag. Kiparsky, P. (1995): The Indo-European Origins of Germanic Syntax, 140-169. In: A. Battye. & I. Roberts (eds.), Clause Structure and Language Change. Oxford: Oxford University Press.

12

Gisella Ferraresi & Rosemarie Lühr

Kiss, K.É. (1998): Identificational focus versus information focus. Language 74,2, 245-273 Klein, W. & Stutterheim, Ch. von (1987): Quaestio und referentielle Bewegung in Erzählungen. Linguistische Berichte 109, 163-183 Kroon, C. (1995): Discourse particles in Latin: A study of NAM, ENIM, AUTEM, VERO and AT. Amsterdam:.J. C. Gieben. Lambrecht, K. (1994): Information Structure and Sentence Form. Topic, focus, and mental representations of discourse referents. Cambridge: Cambridge University Press. Lightfoot, D. (1999): The development of language: Acquisition, change, and evolution. Oxford: Blackwell. Lightfoot, D. (2006): How New Languages Emerge. Cambridge University Press Lühr, R. (2005): Der Einfluß der klassischen Sprachen auf die germanische Grammatik, 341-362. In: G. Meiser and O. Hackstein (eds.), Sprachkontakt und Sprachwandel. Akten der XI. Fachtagung der Indogermanischen Gesellschaft, 17.-23. September 2000, Halle an der Saale. Wiesbaden. Lühr, R. (2006): Informationsstrukturelle Merkmale in der Morphologie des altindischen Personalpronomens. Rutgers Optimality Archive. Lühr (2009): Verbakzent und Informationsstruktur. In: R. Lühr and S. Zeilfelder (eds.), Struktur und Semantik der Verbalphrase. Jena (im Druck). Luraghi, S. (2003): Definite referential null objects in Ancient Greek. Indogermanische Forschungen 108, 167-194. Petrova, S. & Solf, M. (2008): Rhetorical relations and verb placement in the early Germanic Languages. A cross-linguistic study, 329-351. In: C. FabriciusHansen and W. Ramm (eds), Subordination’ versus ,Coordination’ in Sentence and Text. A cross-linguistic perspective. Amsterdam and Philadelphia: John Benjamins. Petrova, S. & Solf, M. (2009): On the methods of information-structural analysis in historical texts. A case study on Old High German, 121-160. In: R. Hinterhölzl and S. Petrova (eds.), New Approaches to Word Order Variation and Word OrderChange. Berlin and New York: de Gryuter. Prince E.F. (1981): Toward a Taxonomy of Given – New Information. In: P. Cole (ed.), Radical Pragmatics, 223-255. New York: Academic Press. Sgall, P., E. Hajičová & E. Benesǂvá (1973): Topic, Focus and Generative Semantics. Kronberg Späth, A. (2007): On the Semantic Foundations of the Contrastive Focus. Within a Lexicalist Approach, 245-276. In: A. Späth (ed.), Interfaces and Interface condition. Berlin and New York: de Gryuter.

Introductory Remarks

13

Steube, A. (2000): Ein kognitionswissenschaftlich basiertes Modell für Informationsstrukturierung In: J. Bayer, Ch. Römer (eds.), Von der Philologie zur Grammatiktheorie, 213-238. Tübingen: Niemeyer Steube, A. (2002): Correction by Contrastive Focus. Theoretical Linguistics 27,23, 215-249. Uhmann, S. (1991): Fokusphonologie. Eine Analyse deutscher Intonationskonturen im Rahmen der nicht-linearen Phonologie. Tübingen: Niemeyer. Umbach, C. (2004): On the Notion of Contrast in Information Stucture and Discourse Structure. Journal of Semantics 21, 155-175 Vallduví Enric & Maria Vilkuna (1998): On rheme and kontrast. In: P. W. Culicover & N. McNally (eds.), The limits of syntax. Syntax and Semantics vol. 29, 79-108. London: Academic Press.

Information Structure, Constituent Order, and Case in Warihío Rolando Féliz Armendáriz (Universidad de Sonora)

1 Introduction Warihío is a member of the Sonoran branch of the Uto-Aztecan family, spoken in Sonora and Chihuahua (Mexico) by ca. 1200 speakers (according to the census provided by the National Institute of Indigenous Languages of Mexico). Different syntactic orders have been suggested for Warihío or, more generally, for Uto-Aztecan. Langacker (1977) proposed an SOV order for Proto-Uto-Aztecan. Miller (1984) considers Warihío as being in the process of changing from an SOV to an SVO language, while Barreras (2000) hypothesized a change from SOV to VSO. Such disagreement implies that syntactic principles do not suffice to capture the principles underlying Warihío word order synchronically or diachronically. Accordingly, I describe Warihío as a language with a pragmatically flexible order with some strongly grammaticalized features of an SOV language such as verbal suffixes, postpositions, and a final copula. I will start showing some basic facts of Warihío grammar.

2 Marking participants in Warihío Warihío does not mark case for core arguments, as we can see in (1), where the subject waní ‘John’, the direct object owítiame ‘little girl’, and the indirect object maría ‘Mary’ all appear in their bare form. (1)

waní itočé-re owítiame maría tapaná John send. pfv=1sg.s little.girl Mary yesterday ‘John sent the little girl to Mary yesterday.’ 1 This paper appears posthumously. All the last revisions have been done by Carlotta Viti, whom the editors thank for her help.

16

Rolando Féliz Armendáriz

Warihío only has two sets of personal pronouns that may help to distinguish participants: one set characterizes Subject and Agent, while another set is used for Patient, Theme, Recipient, Possessor, Reflexive, as well as for Subjects of subordinate clauses. Cf. Table 1.

1sg 2sg 3sg 1pl 2pl 3pl

Subject pronouns Free and emphatic Neé muú apoé/puú temé emé a'póe

Non-subject pronouns Bound and non-emphatic =ne =mu -Ø -teme -eme -Ø

no'ó amó Ø, ahpó tamó amó Ø, ahpó

Table 1: Pronominal system in Warihío.

I present some examples of such pronouns in (2a)-(2g). As can be seen, the emphatic S/A pronouns may occur sentence initially (2a). Instead, the non-emphatic S/A pronouns occur as verbal clitics or suffixes (2b). The stressed forms of non non-subject pronouns may occur either sentence initially (2c)-(2d) or in internal position, before the verb (2e)-(2g): (2)

1

a. neé u'má-ru. 1sg.s run-pfv.ev. ‘I ran’ b. wewé-ru=ne waní hit-pfv.ev=1sg.s John ‘I hit John.’ c. no´ó no'nó no'ó wewé-ru 1sg.ns father 1sg.ns hit-pfv.ev ‘My father hit me.’ d. no'ó nené-na=ne ehpého-či 1sg.ns see-prs=1sg.s mirror-loc ‘I see myself in the mirror.’ e. waní no'ó wewé-ru John 1sg.ns hit-pfv.ev ‘John hit me.’

Abbreviations: a agent, adj adjective, adv adverb, appl applicative, caus causative, clf classifier, com comitative, d.d distal demonstrative, emph emphatic, ev evidential, fut future, g genitive, id identificational, inch inchoative, ins instrumental, int intensifier, imper imperative, ipfv imperfective, loc locative, n noun, neg negative, nmlz nominalizer, ns nonsubject, mov movement, o object, obl oblique, opro object pronoun, p patient, pass passive, p.d proximal demonstrative, pfv perfective, S subject, pl plural, prs present, ptcp participializer, purp purpose, quant quantifier, quot quotative, r recipient, rel relational, rel. c. relative clause, sg singular, sub subordinator, t theme, v verb, vi intransitive verb, vt transitive verb.

Information Structure, Constituent Order, and Case in Warihío

f. waní no'ó itočá-re muní John 1sg.ns send-pfv beans ‘John sent me some beans.’ g. pedró no'ó tuyé-ru wewe-míčio Peter 1sg.ns tell-pfv.ev hit-purp Peter told me to hit John.’

17

waní John

3 Constituent order Greenberg (1966:76) proposes that, when a language exhibits several orders, only one of those orders is dominant. This basic order, from which other features of the same language are inferred, is usually established according to three main criteria: 1) least morphological marking; 2) statistical textual frequency, cf. Hawkins 1983; 3) occurrence in most possible contexts, and especially in the unmarked clause type, i.e. main, declarative, affirmative clause. Of course, such criteria are not always easy to apply, so that Mithun (1992:15) questions whether all languages really have a basic order. Givón (1988:243) similarly points out that the idea of a ‘basic word order’ tends to ignore the existence of flexible order-languages, assigning them arbitrarily to one of the rigid type of languages. Payne (1992:1) suggests that the first typological division should be two-fold: (i) languages in which order correlates mainly with grammatical relations or other syntactic factors, and (ii) languages whose principal clause word order correlates mainly with pragmatic factors. Warihío clearly belongs to the second type of languages posited by Payne, since it challenges all the usual criteria for a basic word order. Morphologically, we do not find any extra marking in different constituent orders. As far as frequency is concerned, nearly all logically possible orders appear with some regularity in main clauses, as shown in Tables 1 and 2. As predicted by Du Bois (1987), clauses with one or none participant overtly expressed (Table 2) are more frequent than transitive clauses with two nominal participants (Table 3).

18

Rolando Féliz Armendáriz

Life stories: 389 clauses N% Vi = 188 = 48% Vt = 28 = 7% VS= 59 = 15% SV= 53 = 14% VO= 19 = 5% OV= 27 = 7%

Folk tales: 250 clauses N% Vi = 83 = 33% Vt = 6 = 3% VS = 59 = 24% SV = 31 = 13% VO = 29 = 12% OV = 28 = 11%

Table 2: Percentage of clauses with one or none participant

Life stories: 389 clauses SVO = 6 SOV = 3 OVS = 2 VSO = 1 VOS = 1

Folk tales: 250 clauses SVO = 9 SOV = 1 OVS = 1 VSO = 3

Total: 4% out of 639 clauses SVO = 15 SOV = 4 OVS = 3 VSO = 4 VOS = 1

Table 3: Percentage of clauses with two nominal participants.

Contextual unmarkedness is also difficult to associate with a certain word order in Warihío. Consider the main, declarative, affirmative clause in (3), which represents an instance of a ditransitive construction: (3)

waní ihkó-ke-ru piípi sipičá tapaná obregón hustína John give-appl-pfv.ev one dress yesterday Obregón Agustina ‘John gave Agustina a dress yesterday in Obregón.’

The AVTR sentence in (3) may occur in all possible alternative orders of agent, verb, theme and recipient, of course with semantic/pragmatic differences. Some of these possible orders are shown in (4): (4)

waní piípi sipičá ihkókeru hustína tapaná obregón ihkókeru waní piípi sipičá hustína tapaná obregón ihkókeru piipi sipičá waní hustína tapaná obregón obregón waní hustína ihkókeru piípi sipičá tapaná hustína ihkókeru piípi sipičá tapaná obregón waní piípi sipičá hustína ihkókeru tapaná obregón waní piípi sipičá hustína waní ihkókeru tapaná obregón

ATVR VATR VTAR ARVT RVTA TRVA TRAV

Alternative orders are allowed in noun phrases too: (5)

a. kaaká weméra N Adj ‘new sandals’ b. waní kawáira G N ‘John’s horse’

tepérume paúmpari ‘long years’ nawésarira riosí ‘God’s word’

Adj N NG

Information Structure, Constituent Order, and Case in Warihío

19

Relative clauses can occur postnominally (6a) or prenominally (6b): (6)

a. N Rel. C. tihoé no'ó kompáe me'yá-ka-me man1 sg.ns buddy kill.SG-ptcp-nmlz ‘...the man that killed my buddy.’ b. Rel. C N ahpó kompáe mahi-ri-á sa'pá 3sg.ns buddy grill-pfv-nmlz meat ‘...the meat that his buddy grilled.’

Moreover, the head of a relative clause can be separated from the rest by a complete clause: (7)

N O V-V Rel. C tihoé no'ó meri-núra-ni kuú pehaná kahti-áme man 1sg.ns kill.sg-order-prs tree behind be.seated.sg-nmlz ‘They are asking me to kill the man that is seated behind the tree.’

Even the pronoun in P/R function, which shows the most fixed position in Warihío and regularly appears preverbally (8a), may occur postverbally (8b): (8)

a. waní no'ó wewe-rú John 1sg.ns hit-pfv.ev ‘John hit me.’ b. waní wewe-rú no'ó John hit-pfv.ev 1sg.ns ‘John hit me.’

Opro V

V Opro

The following comparative construction features do not correlate with any constituent order of a fixed type: (9)

adj standard mark no'ó puhkú miísi epečé werumá ki=amó puhkú čitiá 1sg.ns clf cat more big neg=2sg.ns clf like ‘My cat is bigger than your cat.’

The dependent verb can occur before the main verb, but the rest of the dependent clause occurs after the main verb: (10) a. owítiame yau-ká-meka erá-re pedró-ma woman dance-ptcp-like imagine-pfv Peter-com ‘The woman imagined herself dancing with Peter.’ b. Agustína eči-m-ó o'óra-ri suunú ahpó pete-čí Agustina plant-fut-sub try-ipfv corn 3sg.ns house-loc ‘Agustina tried to plant corn in her yard.’ c. Lupita nehi-m-ó o'orá-ni ku-ó ahpó pete-čí Lupita sell-fut-sub plan-prs wood-emph 3sg.ns house-loc ‘Lupita plans to sell wood in her house.’

20

Rolando Féliz Armendáriz

In the face of this syntactic freedom, a promising account seems to be offered by a pragmatic approach to word order. In the following section I talk about the information structure in Warihío in the terms of Lambrecht (1994), relating this variation in constituent order mainly to the pragmatic relation of Focus and to a lesser degree with that of Topic.

4 Information Structure According to Lambrecht (1994: 5), information structure is “that component of sentence grammar in which propositions as conceptual representations of states of affairs are paired with lexico-grammatical structures in accordance with the mental states of interlocutors who use and interpret these structures as units of information in given discourse contexts”. He establishes two pragmatic relations in the information structure: the Focus and the Topic. Lambrecht (1994: 213) conceptualizes FOCUS as “The semantic component of a pragmatically structured proposition whereby the assertion differs from the presupposition.” This conceptualization covers similar functions as Mithun’s newsworhthiness principle (1992). In general terms, he takes the pragmatic relation of TOPIC as “the thing which the proposition expressed by the sentence is ABOUT (Lambrecht, 1994:118). As Dik pointed out (1997: 327), cross-linguistically the Focus function may manifest itself through one or more of the following focalizing devices: prosodic prominence, special constituent order, special Focus markers, special Focus constructions.

5 Methods The present study is based on my analysis of a continuous segment of text, ‘Don Alejandro Ruelas Zazueta: life story’, in which I first identified the distribution of the core participants (§6), and later verified my findings on native speakers of Warihío by direct elicitation (§7). The excerpt of the text contains 139 sentences with more than 150 clauses. Even though the data come from narrative, the analysis is made up at the sentence level. The analysis is directed to two main points: (i) the order of the major clause constituents, that is, participants and predicates, and (ii) how these participants are introduced for the first time, reintroduced, emphasized, etc. The participants in this text can be coded as lexical (N), free pronoun, bound pronoun, or not coded: Ø. The functions coded by these participants are Agent (A), Subject of intransitive (S), Pa-

Information Structure, Constituent Order, and Case in Warihío

21

tient (P), Recipient (R), and Oblique (OBL). Three observations about the excerpt are important: (i) Certain elements that neither denote participant nor represent verbs tend to appear toward the beginning of the sentence, e.g. there, everywhere, after that, nothing, that’s why, again, toward here, until, etc. These elements have the function of ‘orientational material’ (Mithun, 1992). (ii) The life story of Don Alejandro since the very beginning is mainly about living in one place for a job. After the work is done, the whole family used to go looking for work in other towns. Therefore, events as ‘living’, ‘going’, ‘coming’, and ‘working’ -all of them coded by intransitive verbs- were very important in Don Alejandro’s life. The clauses containing these verbs have the VS order with the first person (‘I’ or ‘we’) for S. (iii) Pronouns in P function, ‘me’ and ‘us’, show the most fixed position in Warihío grammar, immediately preverbal.

6 Text analysis 6.1 Exordium Let us consider the beginning of the narrative: (11)

kusí-tere nawá-ka=ne čía ye'yé-a woods-middle born-ptcp=1sg.s say mother-emph ‘I was born in the middle of the woods, my mother told me.’

The first constituent mentioned in (11) is the locative ‘in the middle of the woods’, which on the one hand has the typical distribution of a local setting, and on the other represents a piece of unexpected information. The wood was an unusual place for a birth, since at the time of the speaker Warihío women used to give birth at home. After this beginning, the speaker continues developing the locative: (12)

kuú werí-ata wa'á tree standing.up-quot there ‘Where there was a tree.’

naáti thing

The first constituent in (12), the noun kuú ‘tree’, represents the most salient piece of information in the clause, which has a presentational function. Given the type of narrative -life story- the main participant will be in the first person, and is often coded by a bound pronoun, as in (13). (13)

weikaóba u'pa-re-tú=ni-a wa'á then bring-pfv-pass=1sg.s-emph here ‘Then I was brought along the river bank.’

aki-čí-o river-loc-emph

22

Rolando Féliz Armendáriz

Despite their initial position, elements such as weikaóba, weikó, eikó, translated as ‘then’, do not provide ‘important’ information, but rather are elements of discursive coherence. Therefore, the real first constituent in (13) is the verb that constitutes a passive clause. We have two possibilities for a passive clause in Warihío in terms of the affected participant position. The contrast is clearly evident with pronoun participants: (i) as verbal enclitic, or (ii) as a preverbal free pronoun. The example in (13) is of the type (i). 6.2 Tracking of participants When the same participant occurs repeatedly in the text, it is coded lexically in its first occurrence and anaphorically (by means of pronouns or zero) in its subsequent mentions. In the following examples we may observe such sequences with the noun of the ‘mother’. Although both (14) and (15) express the ‘mother’ by an oblique nominal with a comitative function which is placed in postverbal position, the referent is coded as a full-fledged noun phrase in (14) and as a pronominal anaphoric in (15). (14)

(15)

ohóe-re=pu ye' yé-ma eikáo live-pfv=d.d mother-com then ‘I lived with my mother then.’ wa'á oi-ré=ne pu'-ká aháma eikaóba there walk-pfv=1sg.s d.d-id with then ‘I was with her there then.’

We observe in (14) that first singular person is coded by Ø, something not very common for 1st and 2nd person in Warihío narrative. There comes an intermezzo (16), where the 1st person subject (this time in the plural) is again zero coded. (16)

tehpéi wa'á ohóe long.time there live ‘We lived there for a long time.’

At this point the ‘mother’ reappears. In its first occurrence, reported in (17), it is a lexical noun phrase in postverbal position and in subject function. In the subsequent clause the same referent is still subject (and is also agent here), but is Ø coded (18). (17)

neipá wahká moči-ré=pu last everywhere be.seated-pfv=d.d ye'ye-á=pu weikaóba mother-emph=d.d then ‘After that my mother lived everywhere.’

Information Structure, Constituent Order, and Case in Warihío

(18)

23

tamó weikaóba upá-re wa'á-tepa 1pl.ns then bring-pfv here –up ‘Then she brought us here.’

The pronoun tamó ‘us’ in P function in example (18) is preverbal; as anticipated in §5, pronouns in non-S/A function tend to have the most fixed position for a constituent in Warihío, immediately preverbal. 6.3 Contrastive contexts Agents may occur in initial position. In (19), for example, the first singular free pronoun is also emphasized with the suffix -a: (19)

ní-a kí=nané-ri-a 1sg.s-emph neg=know-pfv-emph ‘I did not know (that place).’

Similarly, a new referent such as the ‘father’ (no'nó) opens the clause in (20). This seems to contradict the ‘avoid new lexical A’ discourse rule given by Du Bois (1987): (20)

no'nó nane-ré=pu no'nó eikó wa'á ohóe-re=pu father know-pfv=d.d father then there live-pfv-d.d ‘(But) my father knew it (because) my father used to live there.’

Two explanations, however, may exist for the initial position of the new agent no'nó. First, the name of the ‘father’ represents an instance of inalienable possession and can be easily anchored to the name of the speaker (ní-a ‘I’) in the previous clause. Second, the two clauses have a contrastive relation: while the speaker did not know the place, the father did. Accordingly, contrastive agents occupy the initial position of the sentence. The discourse continues with the sentence reported in (21), which opens with the usual orientative deictic wa'á, alluding to ‘Campo Nuevo’, and which contains a clause coordination. (21)

wa'á yasa-ré no'nó poní-ra there be.seated-pfv father brother-rel weiká=pu yetó-re=pu-a tekihpaná-mia then=d.d invite-pfv=d.d-emph work-fut ‘My father’s brother was there and invited me to work.’

In the first clause, which presents a VS order, the S participant (‘my father’s brother’) is lexical and it is mentioned for the first time. In this case as well, the referent may be inferred by the name of the ‘father’, already introduced. Another example of contrast between two participants involved in the same event is given in the sequence (22)-(23). In (22) the speaker’s broth-

24

Rolando Féliz Armendáriz

ers did attend school, while in (23), the speaker didn’t; both participants are sentence initial and morphologically emphasized. (22)

(23)

no'ó po-poni-á=pu 1sg.ns pl-brother-emph=d.d ihkwéra-či moči-ré=pu-a=pu school-loc be.seated.pl-pfv=d.d-emph=d.d ‘My brothers were (going) to school.’ ni-á ki ki=asa-ré=ni-a 1sg.s-emph neg neg=arrive-pfv=1sg.s-emph ihkwéra=ni-a=pu school=1sg.s-emph=d.d ‘I didn’t attend school.’

6.4 Unexpected information In an anecdote inside the story ‘mother’ and ‘father’ are reintroduced lexically and sentence initially (24): (24)

kuándo ye'yé weikaóba ko'kó-ri-a when mother then sick-pfv-emph eikó no'nó=pu taha-ré=pu-a=pu then father=d.d burn-pfv=d.d-emph=d.d ‘When my mom was sick my dad burned himself.’

Right afterwards the speaker develops what happened to his father in this way (25): (25)

naásipa naásipa isáwi-či in.the.middle in.the.middle hot.coal-loc ‘(He was) in the middle of the hot coals.’ poí-ri-a eikó pu'-ká-e no'nó lay-pfv-emph then d.d-id-ins father ‘My dad, he was lying on them.’

The initial locative in (25), ‘in the middle of the hot coals’, carries an unexpected piece of information, and its salience is underlined by the iteration naásipa … naásipa. After having brought a doctor to heal his father, the speaker makes this comment: (26)

weikáoba ki=iyoé-ka then neg=heal-ptcp pu'-ká rootóre pu'-ká no'ó no'nó d.d-id doctor nd.d-id 1sg.ns father ‘(He didn´t heal him) the doctor didn´t heal my dad.’

In (26), a rare text case of a clause with two core participants coded by noun phrases, the speaker talks about the unexpected event ‘not having healed (someone)’ with a clause like this: neg=V A P . Of course, calling

Information Structure, Constituent Order, and Case in Warihío

25

the doctor presupposes the expectation that the doctor may heal the dad. This expectation is, however, unsatisfied, and the negative marker, which typically has a focal function in languages, is placed in the initial position of the clause. 6.5 New information After the speaker left the doctor in his house, he was on his way back home when something happened to him. He refers to this passage in this way (27)-(31): (27)

(28)

(29)

(30)

(31)

no'=aká ini-siá=ne 1sg.ns=come.back be-go=1sg.ns eikó simi-yái simi-yái then go-ipfv go-ipfv ‘I was going back, walking and walking.’ weikaóba makoyáwi tewaniá wa'á wanámi wa'á then Macoyagüi called there by.there there ‘(I passed by) there in Macoyagüi.’ wa'á weri-á piípi kuú temóri werumá there stand.up-emph one trunk temóri big pálo blánko tewaniá stick white called ‘There was a huge white-stick trunk.’ kuú temóri=pu-a wa'á eikó trunk temóri=d.d-emph there then ‘It was a huge trunk.’ wa'á nará čiwá katiáme weikáo weikaóba there cry goat like then then ‘Then something like a goat was crying there.’

The new piece of information makoyáwi, the town where this anecdote happened, is fronted with respect to the verb in the Warihío clause in (28), unlike in its English translation. Similarly, the clause in (31) opens with the new sound (nará ‘cry’). Then the speaker went there trying to identify the sound, and found a goat just after delivery. (32)

(33)

simi-ká=ne eikó nené-mia pu'-ká wa'á nené-mia go-ptcp=1sg.s then see-fut d.d-id there see-fut ‘I went to see it.’ wahká poi=pú čiwá tanayáme čia=né eikó over.there lay=d.d goat just.given.birth say=1sg.s then ‘A goat that has just given birth is there, I said (thought) then.

26

Rolando Féliz Armendáriz

6.6 Important information Sometimes the initial constituent of a clause represents an important piece of information in the text. This can be see in cases of lexical iteration such as teme … teme in (34), where the speaker recalls a period of hunger. (34)

teme-á=pu corowá-ni 1pl.s-emph=d.d be.hungry-prs teme-á=pu pukaépa itá-ni sunú 1pl.s-emph=d.d that’s why ask-prs corn ‘We were hungry; that’s why we asked for corn.’

In this piece of the story, the ‘food’ becomes very relevant, and is therefore often found at the beginning of the clause (35). (35)

ko'áme tarí-mia wa'á asi-ré=pu food buy-fut there arrive-pfv=d.d ‘My mom went there to buy food.’

ye'yé-a eikábo mother-emph then

7 Direct elicitation Elicitation seems to confirm the association between focus and initial position that has been observed above in the analysis of the narrative. Several types of focus have been elicited, such as completive focus (§7.1), selective focus (§7.2), and replacing focus (§7.3). 7.1 Completive focus Completive focus implies a question-answer exchange, so that the answer ‘completes’ the proposition conveyed by the previous question. The focused information is not necessarily new in the sense that it is not known or not previously mentioned, or that it cannot be inferred from the context. What is new is the relation with respect to the rest of the clause and to the addressee’s assumptions, as illustrated in §4. 7.1.1 Participant of intransitive inactive event For example, the question in (36a) ‘Who slept here last night?’ has its natural answer in (36b), where the subject constituent is sentence initial. (36)

Focus domain: pronoun-participant of an inactive intransitive event a. ataná koči-ré tukaó iwá who sleep-pfv night here ‘Who slept here last night?’

Information Structure, Constituent Order, and Case in Warihío

b. neé koci-rú 1sg.s sleep-pfv.ev ‘I slept here.’

27

iwá here

Although the initial position for the interrogative marker in questions is also typical of the Indo-European languages, not so typical of these languages is the initial position of the focus in answer. The question in (36a) cannot be answered by clauses where the focused subject is displaced from the beginning of the sentence into a clause internal clitic (36c) or stressed position (36d). At most, the following verb-initial clauses represent answers to the question ‘What did you do here?’ (36) c. koci-rú=ne iwá sleep-pfv.ev=1sg.s here ‘I slept here.’ d. koci-rú neé iwá sleep-pfv.ev sg.s here ‘I slept here.’

7.1.2 Participant of an intransitive active event The focused element in Warihío question-answer pairs can be any constituent in any semantic relation. In (36) we have identified it in the participant of an intransitive inactive event such as ‘sleep’. In (37) we can see that the focus is the participant of an intransitive active event such as ‘run’. The question ‘Who ran in the woods yesterday?’ in (37a) can be answered by a clause in which če'é ‘José’ is focused and initial. (37)

Focus domain:noun-participant of an active intransitive event a. ataná u'má-re kusí-tere tapaná who run-pfv woods-through yesterday ‘Who ran in the woods yesterday?’ b. če'é u'má-rekusí-tere José run-pfv woods-through ‘José ran in the woods.’

The clause in (37b) is pragmatically equivalent to a cleft-sentence in English, such as ‘It is José who ran in the woods’. Alternative orders such as (37c) and (37d) do not answer the question in (37a), but rather other questions such as ‘What did John do in the wood?’ or ‘What happened?’ (37) c. u'má-re če'é kusí-tere run-pfv José woods-through ‘José ran in the woods.’ d. kusí-tere u'má-re če'é woods-through run-pfv José ‘José ran in the woods.’

28

Rolando Féliz Armendáriz

7.1.3 Agent of a transitive event In the following pair (38a)-(38b) the requested constituent is the agent of a transitive event: (38)

Focus domain: noun-agent of a transitive event a. ataná čikó-re kawái who steal-pfv horse ‘Who stole the horse?’ b. waní čikó-re kawái John steal-pfv horse ‘John stole the horse.’

Alternative orders such as (38c) and (38d) do not answer the question in (38a): (38) c. čikó-re kawái waní steal-pfv horse John ‘John stole the horse.’ d. kawái waní čikó-re horse John steal-pfv ‘John stole the horse.’

7.1.4 Patient of a transitive event Somebody could say that initial position is reserved for S/A, giving to the language an SVO order, but the next two examples show otherwise. In (39b), the PATIENT kawái ‘horse’ is in initial position: (39)

Focus domain: noun-patient of a transitive event a. ihtaná tara-ré waní what buy-pfv John ‘What did John buy?’ b. kawái tara-ré waní horse buy-pfv John ‘John bought a horse.’

An adequate translation of the answer in (39b) would be the cleft sentence ‘It is a horse that John bought’. As expected, alternative orders do not answer the question in (39a), but rather the question ‘What did John do?’ (39c) or ‘Who bought a horse?’ (39d): (39) c. tara-ré kawái waní buy-pfv horse John ‘John bought a horse.’ d. waní kawái tara-ré John horse buy-pfv ‘John bought a horse.’

Information Structure, Constituent Order, and Case in Warihío

29

7.1.5 Recipient of a ditransitive event In (40) the focused, initial constituent is the RECIPIENT: (40)

Focus domain: noun-recipient of a ditransitive event a. ataná rewe-rú=mu ihpíčira who lend-pfv.ev=2sg.s broom ‘Whom did you lend the broom to?’ b. maría rewe-rú=ne ihpíčira Mary lend-pfv.ev=1sg.s broom ‘I lent the broom to mary.’

Alternative orders do not answer the question in (40a): (40) c. ihpíčira maría rewe-rú=ne broom Mary lend-pfv.ev=1sg.s ‘I lent the broom to Mary.’ d. rewe-rú=ne ihpíčira maría lend-pfv.ev=1sg.s broom Mary ‘I lent the broom to Mary.’

7.1.6 Predicate The verbal constituent may also represent the focused piece of information. Cf. (41): (41)

Focus domain: V-Predicate a. ihtaná yowa-ré pedró kawái aháma what do-pfv Peter horse com ‘What did Peter do to the horse?’ b. neha-ré pedró kawái aháma sell-pfv Peter horse com ‘Peter sold the horse.’

Alternative orders do not answer the question in (41a): (41) c. kawái neha-ré waní aháma horse sell-pfv John com ‘John sold the horse.’

7.1.7 Adverbials This special order works not only for major constituents but for adverbials also, as I will show with different examples for relations of place (42)-(43) and of time (44). In (42) the question ‘Where did John go Saturday?’ is answered by the clause ‘He went to Alamos’: the local adverbial ‘Alamos’, which here plays the role of the Goal, is focused and sentence initial:

30

Rolando Féliz Armendáriz

(42) a. akaná oi-ré waní sawára-či where walk-pfv John Saturday-loc ‘Where did John go Saturday?’ b. alamó-či oi-ré=pu-a Alamos- loc walk-pfv=d.d-emph ‘He went to alamos.’

Focus domain: NPGoal

The same occurs in (43), where the adverbial o'oránačí ‘on the stove’ plays a locative stative function: (43)

Focus domain: NP-loc-Place a. akaná wera-ré sikorí maría where set.sg-pfv pot Mary ‘Where did Mary put the pot?’ b. o'orána-či wera-ré sikorí stove-loc set.sg-pfv pot ‘She put the pot over the stove.’

Alternative orders (43c)-(43d) do not answer the question in (43a): (43) c. wera-ré sikorí o'orána-či set.sg-pfv pot stove-loc ‘She put the pot over the stove.’ d. sikorí wera-ré o'orána-či pot set.sg-pfv stove-loc ‘She put the pot over the stove.’

In (44) we have a temporal focused constituent, such as tapaná ‘yesterday’: (44)

Focus domain: ADV -Time a. ačintió simi-ré pedró sawanánto when go-pfv Peter San Bernardo ‘When did Peter leave for San Bernardo?’ b. tapaná simi-ré=pu=a sawanánto yesterday go-pfv=d.d=emph San Bernardo ‘He left for San Bernardo yesterday.’

Alternative orders do not answer the question in (44a): (44) c. simi-ré=pu=a sawanánto tapaná go-pfv=d.d=emph San Bernardo yesterday ‘He left for San Bernardo yesterday.’ d. sawanánto simi-ré=pu=a tapaná San.Bernardo go-pfv=d.d=emph yesterday ‘He left for San Bernardo yesterday.’

7.1.8 Adjectives Similar considerations hold for relations of manner, which may be expressed by adverbs as well as by adjectives. Given that in Warihío most adjectives are formally nouns and quantifiers may function as participants,

Information Structure, Constituent Order, and Case in Warihío

31

these same elements can be in focus relation independently of the noun they modify. If we observe the response in (45b) to the question in (45a), the adjective tahtáme is displaced from the modified noun re'ečú: (45)

Focus domain: ADJ a. ačitiamé nahki=mu re'ečú how want=2sg.s milk ‘How do you want the milk?’ b. tahtáme nahkí=ne re'ečú hot want=1sg.s milk ‘I want the milk hot.’

Otherwise the adjective may occur after the noun, as in (45c), but this time the adjective is not focused and the clause does not answer the question in (45a): (45) c. nahkí=ne re'ečú tahtáme want=1sg.s milk hot ‘I want the hot milk.’

It is the same with the colors ohčóname and sióname in (46) and (47), respectively: (46)

Focus domain: ADJ a. číntiame pu'-ká pu'seré-na=mu kawái which d.d-def like-prs=2sg.s horse ‘Which horse do you like?’ b. ohčóname pu'seré-na=ne kawái black like-pres=1sg.s horse ‘I like the black horse.’

The example in (46c) does not answer the question in (46a): (46) c. pu'serénane kawái ohčóname ‘I like the black horse.’ (47) Focus domain: ADJ a. číntiame pu'seré-na sipičá which like-prs dress ‘Which dress does Petra like?’ b. si'óname pu'seré-na sipičá green like-prs dress ‘Petra likes the green dress.’

peterá Petra peterá Petra

The example in (47c) does not answer the question in (47a): (47) c. peterá pu'seréna sipičá si'óname ‘Petra likes dress green.’

32

Rolando Féliz Armendáriz

7.1.9 Relative clauses Even the elements constituting a relative clause may be sensitive to this initial position (in Warihío the nucleus and the relative clause are normally contiguous and the latter appears to the right). We observe in (48) first the nucleus of the relative clause, then the main clause, and at the end the relative clause: (48)

N O V-V Rel. C tihoé no'ó meri-núra-ni kuú pehaná kahti-áme man 1sg.ns kill.sg-order-prs tree behind be.seated.sg- nmlz ‘They are asking me to kill the man that is seated behind the tree.’

7.2 Selective focus Another type of focus relation is one that chooses between two or more options. In this case the element in focus relation has been previously mentioned and it is known. What is new, again, is the relation established in the sentence. The element selected from the options in (49a) is going to be in focus relation in (49b) and occurs in initial position: (49) a. ihtánata tara-ré waní piípi what buy-pfv John one ‘Did John buy a horse or a donkey?’ b. piípi kawái tararé ‘He bought a horse.’

kawái piípi horse one

u'urú donkey

7.3 Replacing focus The replacing focus type invalidates any information already given by placing the non-expected information in initial position, as is shown in (50b): (50) a. waní simi-ré sawanánto John go-pfv San Bernardo ‘John left for San Bernardo.’ b. kaí imi-ré sawananto neg go-pfv San Bernardo obregoni kipa-ré Obregón leave-pfv ‘No, he didn´t leave for San Bernardo, he left for obregón.’

Information Structure, Constituent Order, and Case in Warihío

33

The construction in (50c) does not represent an adequate replacing focus for (50a): (50) c. kaí simi-ré sawananto neg go-pfv San Bernardo kipa-ré obregoni leave-pfv Obregón ‘No, he didn´t leave for San Bernardo, he left for Obregón.’

8 Topic The Topic, in the sense of ‘aboutness’ (Lambrecht, 1994), may be also introduced in initial position when it has a low degree of accessibility. In the subsequent clauses, though, the Topic appears in post-verbal position. All this can be seen in the following excerpt from ‘The Coyote and the buddy Fox’: (51)

simi-ká-ta puarí go-ptcr-quot too ‘...he went too, wahká kawí-tere werí-ri-ata far away hill-between be standing-pfv-quot far away he was standing between the hills, kawí komí-ka hill hold-ptcp propping up a hill, wa'á tewa-ri-áta=pu 'wí-a there find-pfv-quot=d.d coyote-emph the coyote found him (the fox) there.’ kompáe či-atá keóči-a buddy say-quot fox-emph ‘Hey buddy!’ the fox said: wiči-ná ihí kawí fall down-prs p.d hill ‘that hill is falling down,’ amó-či u'mí-ma=pu 2sg.ns-loc fall down-fut=d.d it is going to fall down on you.’ či-atá keóči-a say-quot fox-emph ‘The fox said.’ na'pá-ka komi-ri-áta pu'-ká kawí both-between hold-pfv-quot d.d-id hill ‘Both of them held the hill up.’

34

Rolando Féliz Armendáriz

weikáo kawé čapi-mitia then well hold-imper ‘Then, ‘Hold it up well!’ tamó-či ru'mí-ma-pu kawí kompáe=pu 1pl.ns-loc fall down-fut-d.d hill buddy=d.d ‘The hill is going to fall down on us, buddy.’ či-atá keóči-a say-quot fox-emph ‘The fox said.’

As can be seen, in the first part of the text the noun kawí ‘hill’ is preverbal, while in the subsequent clauses it is found after the verb.

9 Case marking Although Warihío has no case marking for core nominals, as shown in §2, it is worth mentioning a cognate object case marking –ka, which occurs in the determinants in Yaqui and Pima Bajo, and which can be suffixed to a demonstrative in Warihío. This suffix is taken as an object case marker by Barreras (1990) and Miller (1996). My claim is that such object case marking seems to be in process of change from a syntactic to a pragmatic function. In discourse, when a demonstrative accompanies a noun previously mentioned or identified-inferred, it takes the suffix –ka, it does not matter if the noun is in S, A or P function or even in oblique (locative and instrumental) function (Félix Armendáriz, 2006). We can see examples of this suffix in (52) and (53): (52)

(53)

eikó asi-ré=pu pu'-ká pete-čí then arrive=d.d d.d-id house-loc ‘...then he arrived at the house weikáoba ki=iyoé-ka then neg=heal-ptcp pu'-ká rootóre pu'-ká no'ó no'nó d.d-id doctor d.d-id 1sg.ns father ‘he didn´t heal him, the doctor didn´t heal my dad. ’ kuu-é wepa-ká stick-ins hit-ptcp pu'-ká-e pu'-ká-e napawi-ré pu'-káa aroso d.d-id-ins d.d-id-ins gather-pfv d.d-id rice ‘...they hit it with a stick and got the rice.’

Information Structure, Constituent Order, and Case in Warihío

35

10 Conclusions Warihío is an Uto-Aztecan language characterized by verbal suffixes, postpositions, and a final copula. These are strongly grammaticalized features characteristic of an SOV language, although in Warihío all alternations of orders are in principle possible in texts or even via direct elicitation. In our study it was not possible to definitively establish a basic word order in S, V, and O terms. That is to say, Warihío word order doesn’t help us establish S and O as grammatical relations, and may be considered as being pragmatically determined. In particular, it seems that different pragmatic functions such as Focus and Topic show a preferential order. Initial position is reserved for the FOCUS relation, independently of the type of syntactic constituent (subject, object, etc.) or the semantic relation (agent, patient, etc.) in question. Instead, the topic is often post-verbal. The fact that pragmatic relations influence the relative order of constituents in the clause may compensate the impossibility of using morphology to signal syntactic relations. At the same time, we have observed that even in the domain of morphology a certain suffix is shifting from marking grammatical relations to signalling pragmatic functions. Thus, pragmatic functions seem to be active also in the change and diachrony of this language.

References Barreras Aguilar, I. (1990): Esbozo Gramatical del Guarijío de Mesa Colorada. Ms. Universidad de Sonora. Barreras Aguilar, I. (2000): Orden de Palabras en el Guarijío de Sonora. In: Casad, E. and T. Willet (eds.), Uto-Aztecan, Structural, Temporal, and Geographic Perspectives,125-138. Hermosillo: Editorial UniSon. Bybee, J. (2002): Main clauses are innovative, subordinate clauses are conservative: consequences of the nature of constructions. In: Bybee, J. and M. Noonan (eds.), Complex Sentences in Grammar and Discourse,1-17. Amsterdam and Philadelphia: John Benjamins. Dik, S. (1980): Studies in Functional Grammar. London and New York: Academic Press. Dik, S. (1997): The Theory of Functional Grammar. Berlin and New York: Mouton de Gruyter. Du Bois, J. (1987): The discourse basis of ergativity. Language 64: 805-855. Félix-Armendáriz, R. (2006): A Grammar of River Warihío. Ph.D. Dissertation: Rice University.

36

Rolando Féliz Armendáriz

Givón, T. (1988): The pragmatics of word order: Predictability, Importance and Attention. In: Hammond, M., E. Moravcisk and J. Wirth (eds.), Language Typology and Language Universals, 243-248. Amsterdam and Philadelphia: John Benjamins. Givón, T. (2001): Syntax. Vol. II. Amsterdam and Philadelphia: John Benjamins. Greenberg, J. (1966) Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements. In: Greenberg, J. (ed.), Universals of Language, 73-113. Cambridge: Massachusetts Institute of Technology. Hawkins, J. (1983): Word order Universals. New York: Academic Press. Lambrecht, K. (1994): Information Structure and Sentence Form. Cambridge: Cambridge University Press. Langacker, R. (ed.) (1977): Studies in Uto-Aztecan Grammar, Vol. 1, An Overview of Uto-Aztecan grammar. Dallas: Summer Institute of Linguistics and University of Texas at Arlington. Miller, W. (1984): Subordinate Verbs in Guarijío: Evidence of a Shift from SOV to SVO in Uto-Aztecan Language. Mid-America Linguistics Conference Papers. Boulder: University of Colorado. Mithun, M. (1992): Is basic word order universal? In: Doris L. Payne (ed.) Pragmatics of word order flexibility, 15-61. Amsterdam and Philadelphia: John Benjamins. Mithun, M. (1991): The role of Motivation in the Emergence of Grammatical Categories: the grammaticization of subjects. In: Closs Traugott, E. and B. Heine (eds.), Approaches to Grammaticalization, focus on types of grammatical markers, vol.II,161-184. Amsterdam and Philadelphia: John Benjamins. Payne, D. (1992): Introduction. In: Payne, D. (ed.), Pragmatics of word order flexibility,1-14. Amsterdam and Philadelphia: John Benjamins.

The Information Structure of OVS in Vedic Carlotta Viti (Zurich)

1 Introduction This paper discusses the OVS word order in Vedic, one of the earliest recorded Indo-European (IE) languages, dating back to the second half of the second millennium B.C. Vedic is traditionally described as an SOV language (Delbrück 1888: 15-16), and represents a crucial piece of evidence for the reconstruction of SOV as the basic word order of ProtoIndo-European (PIE) (Lehmann 1974; Watkins 1993: 85). In addition to SOV, however, all other logically possible arrangements of subject, verb, and object (SVO, VSO, OSV, OVS, and VOS) are also attested as marked variants in Vedic, and presumably were also used in PIE. Among these orders, OVS is especially rare synchronically, and subject to decay diachronically from Vedic to Classical Sanskrit. Neither Indian grammarians nor western Indologists have completely clarified the principles underlying word order flexibility in Vedic. They occasionally ascribe syntactic fronting to emphasis. For example, Macdonell observes that the subject regularly begins the sentence in Vedic, but it may be preceded “by any other member of the sentence intended to be strongly emphasized”. (1916: 284) The argument for emphasis, however, is circular: a constituent is placed in initial position as long as it is considered emphatic, and at the same time it is considered emphatic as long as it is placed in initial position. This is because emphasis, salience, and topicality are abstract concepts that cannot be directly detected in a text and that can be judged subjectively, unless they are related to an objectively observable parameter. Some heuristic measures of topicality may be identified in specificity, humanness, definiteness, etc. of a participant and of the noun phrase de

I express my gratitude to the scholars who read this paper and provided me with very helpful comments: Rolando Félix Armendáriz, Eystein Dahl, Gisella Ferraresi, Leonid Kulikov, Rosemarie Lühr, Dejan Matić, Anita Steube and the reviewers of the series Language, Context and Cognition.

38

Carlotta Viti

noting it (cf. Myhill 1992). A singular noun phrase referring to a specific individual is usually more topical than a plural noun phrase having a generic referent. Pronouns are usually more topical than nouns, and nouns of humans are usually more topical than nouns of inanimates, according to Silverstein’s (1976) Animacy Hierarchy reported in (1). Cf. also Timberlake (1977); Hopper and Thompson (1980); Mallinson and Blake (1981: 158ff); Comrie (1981: 178ff); Lazard (1984); Croft (1990: 111117), etc.: (1)

First / second person pronouns > third person pronouns > proper names > common nouns of human beings > common nouns of animate, nonhuman beings > common nouns of inanimate beings > mass nouns

Of course, topicality is an utterance-level phenomenon, and therefore cannot be completely reduced to certain semantic features of the arguments. One should look at the context of each clause to establish which nominal is the topic and which is not. Beside this syntagmatic definition, however, it is undeniable that topicality also has some paradigmatic aspects. If a view of topicality as “aboutness” is adopted (i.e. if the topic is considered as being what the sentence is about), it is clear that a discourse is more often about human, animate, and specific referents than about inanimate or generic referents. Accordingly, the nominals located on the left or high part of the Animacy Hierarchy in (1) have more chances to be topics than nominals located on the right or low part of the Animacy Hierarchy. Moreover, a topic is not only definite with respect to what precedes, but is also persistent with respect to what follows (cf. Givón 1988). The repeated mention of a noun phrase entails that the denoted participant has an important role in the text. The relative topicality of the clause arguments may be also considered as being a useful criterion to assess the level of transitivity of the clause. Despite the different views of transitivity that have been propounded, all scholars agree that in the prototypical transitive clause properties such as humanness, individuation, agentivity, and volitionality are more present in S than in O. Dowty (1991) argues that the Agent Proto-Role, which largely correlates with the subject, exists independently of the event named by the verb, and causes a change of state in another participant. Instead, the Patient Proto-Role does not exist independently of the event, is causally affected by another participant, and undergoes a change of state. As far as “the defining properties of transitivity are discourse-determined” (Hopper and Thompson 1980), the degree of transitivity of a sentence may be relevant to identify the information structure carried by a particular construction. These studies of animacy, transitivity and topicality are here used in the analysis of Vedic OVS clauses. It appears that the noun phrases that

The Information Structure of OVS in Vedic

39

are semantically higher on the Animacy Hierarchy or contextually more topical (in the sense that the sentence is about them) tend to be fronted. Accordingly, OVS seems to be used when the object is more topical than the subject and the clause has a low level of transitivity. In the following sections, we will illustrate the corpus and the method used to obtain a text count of the OVS order (§2). We will analyze the morpho-syntactic, semantic, and pragmatic features of OVS (§3) with respect to different word orders (§4). A comparison will be made with genetically and areally unrelated languages where OVS is the basic order (§5). Finally, we will suggest a pragmatic explanation of the fact that certain word order patterns such as OVS diachronically decay, while others remain in the path of syntactic change from PIE to the daughter languages (§6).

2 Materials and method Our observations on the OVS word order are based on the entire RigVeda, the earliest Vedic text, even though the statistics here presented are only based on the data from book I to book V. The Rig-Veda is a collection of hymns, mainly of eulogistic character, addressed to various deities such as Indra, Agni, etc. One might object that the Rig-Veda is a poetic text, and that word order in poetry is often distorted. Although in general the basic word order of a language is more faithfully manifested in a prose text, in the specific case of Vedic it is difficult to choose between an earlier poetic text such as the Rig-Veda and a later prose text such as the BrāhmanҜas. The BrāhmanҜas contain descriptions and comments on the religious ritual; they are written in a poetic prose and in a high linguistic register, and are therefore also susceptible to display word order deviation with respect to the spoken language. Moreover, syntactic categories such as subject, object, and predicate may be represented by lexemes with very different prosodic structure, so that neither style nor meter is completely responsible for the major constituent order in the Rig-Veda. OVS is a verb-medial order, like SVO. It has been stated that, when verb-medial clauses appear in Old Indian, the post-verbal constituent represents a sort of afterthought with respect to a clause that is already syntactically complete (Gonda 1959). This does not undermine the prevalence of the verb-final order, since the postverbal subject in OVS would represent an appositional adjunct of an OV structure. It is therefore necessary to consider the cases where S itself has the function of an argument and does not represent an extra-position that is coreferential with a preverbal constituent. In this, the metric organization of the text may be of some help, since the basic syntactic structure of the clause is usually com-

40

Carlotta Viti

prised inside the borders of the stanza or of the pāda, the metric unit of Vedic poetry (Gonda 1958). Both an absolute and a modified version of OVS may be identified. In the absolute version, S and O are unambiguously separated by the verb, as in (2), where S and O are represented by plain nominals, and in (3), where they are coded by full noun phrases, accompanied by various modifiers. Even in this case, all modifiers of the subject appear to the right of the verb, and all modifiers of the object appear to the left, so that S and O are syntactically continuous: (2)

(3)

dívamҕ jinvanty agnáyahҜ sky(M).ACC.SG animate.PR.IND.3PL fire(M).NOM.PL “Fires animate the sky.” (1.164.51) asmé vatsám pári ߈ántamҕ ná us.LOC calf(M) ACC.SG around being.ACC.M.SG NEG vindann ichánto víśve find. IPF.INJ.3PL searching.NOM.M.PL all.NOM.M.PL amŕલtā ámūrāhҜ immortal(M).NOM.M.PL wise.NOM.M.PL “All wise immortals did not find the calf that is around us, although they were searching (it).” (1.72.2)

In the modified version of the OVS order, S and/or O are represented by discontinuous constituents, whose members appear on either side of the verb. This is what Schäufele (1991: 182) calls “partial extra-position”. In this case, we have included only those passages where the verb is preceded by the head of the object noun phrase and is followed by the head of the subject noun phrase, as in (4). (4)

índramҕ víśvā avīvrલdhan Indra.ACC all.NOM.F.PL increase.AOR.IND.3PL samudrávyacasamҕ gírahҕ expansive.as.the.sea.ACC.M.SG song(F).NOM.PL “All songs have increased Indra, expansive as the sea.” (1.11.1)

Here the noun phrase with the function of direct object includes both the pre-verbal proper name índram and the post-verbal adjective samudrávyacasam lit. “having the extension of the sea”. Although they appear on opposite sides of the verb, índram is the head of the noun phrase, and therefore the passage can be considered as having a preposed object. Similarly, the noun phrase with the function of the subject includes both the adjective víśvāh̠ “all” and the common noun gírah̠ “songs”. Since the latter is the head of this noun phrase, the subject can be computed as postposed to the verb. This interpretation is supported by similar contexts where songs that increase (i.e. make powerful) gods are represented by simple noun phrases in the OVS order, as we will see below in (7).

The Information Structure of OVS in Vedic

41

The category of modifiers does not comprise demonstratives (5), which in Vedic may have the function of third person pronouns (Whitney 1889: §495; Macdonell 1910: §392), and therefore function as authentic arguments. Accordingly, a verb that is preceded by an accusative demonstrative pronoun and is followed by a co-referential noun is considered as governing a preposed direct object, and is included in the group of OVS. The passage in (6), where the pre-verbal demonstrative pronoun tám is cataphorically resumed by a post-verbal proper name (índram), has been analyzed in the same way as the clause in (5). The placement in two different metric units highlights the loose linkage between the argument demonstrative and the proper name, which functions as an apposition: (5)

(6)

tám íd gachanti juhvàs this.ACC.M.SG PTC go.PR.IND.3PL spoon(F).NOM.PL “To him (lit. ‘to this’) the spoons go.” (1.145.3) támҕ vrલtrahátye ánu this.ACC.M.SG Vrલtra.fight(N).LOC.SG beside tasthur ūtáyahҜ / stay.PF.3PL help(F).NOM.PL śúsҖmā índram avātā¸ strong.NOM.F.PL Indra.ACC invincible.NOM.F.PL áhrutapsavahҜ fresh.NOM.F.PL “Strong, fresh, invincible helps stay beside him (lit. ‘beside this’), Indra, in the fight with Vrલtra.” (1.52.4)

By means of such criteria, we counted 119 cases of the OVS order in books I-V of the Rig-Veda.

3 Results 3.1 Formal correlates of OVS The subject and the object of OVS clauses are quite consistent in their morphological features, in that the subject is a plural constituent and the object is a singular constituent in 60 clauses, which correspond to 50% of cases 1 . Cf. (2)-(6). This is the opposite of what we expect in a typical tran1

The cases with SG object and PL subject are the following: 1.5.8, 1.11.1, 1.22.20, 1.32.2, 1.32.8, 1.32.10, 1.45.6, 1.50.1, 1.52.4, 1.68.9, 1.72.2, 1.73.7, 1.85.2, 1.89.2, 1.89.5, 1.90.6, 1.96.3, 1.100.13, 1.122.14, 1.125.7, 1.132.5, 1.136.7, 1.141.2, 1.144.5, 1.145.3, 1.152.5, 1.159.4, 1.164.7, 1.164.43, 1.164.50, 1.164.51, 2.4.2, 2.18.3, 2.33.11, 2.35.3, 2.35.4, 3.2.5, 3.10.6, 3.26.5, 3.28.4, 3.34.8, 3.36.7, 4.2.12, 4.2.14, 4.7.4, 4.25.5, 4.31.9, 4.33.9, 4.42.1 (twice), 4.50.9, 5.8.3, 5.14.2, 5.30.13, 5.32.11, 5.48.1, 5.53.10, 5.56.2, 5.63.5, 5.82.1.

42

Carlotta Viti

sitive clause, where the subject is supposed to be presented as more specific than the object (Hopper and Thompson 1980). Such clauses, where S is singular and O is plural, represent the lowest percentage (7%) in the group of OVS clauses, as can be seen in Table 1.

RV I S: SG O: SG S: PL O: PL S: SG O: PL S: PL O: SG

N (%) 29 (24%) 13 (11%) 8 (7%) 60 (50%)

O/S: DU

9 (8%)

TOTAL

119 (100%)

Table 1: Grammatical number of S and O in the OVS order

Of course, the notion of “singular vs. plural” reference does not completely overlap with the notion of “specific vs. generic” reference: plural NPs may be specific, and singular NPs may be generic in reference. However, a certain correlation between singularity and specificity on the one hand, and between plurality and generality on the other may exist, in the sense that – ceteris paribus – a definite singular NP is presented as more specific than a definite plural NP, and an indefinite singular NP is presented as more specific than an indefinite plural NP. From this point of view, grammatical number may be considered as being a heuristic measure of topicality similarly to other morpho-syntactic features such as gender or case. The nominative, for example, is usually treated as correlated to agentive and topical roles in nominative-accusative languages, although strictly speaking not all nominatives and not all subjects are agents or topics. 3.2 Semantic correlates of OVS The higher referentiality of the object with respect to the subject emerges if we take into account the lexical-semantic features of these noun phrases in OVS clauses. In 41 cases the object occupies a higher position than the subject on the Animacy Hierarchy reported in (1) 2 . The prototypical situa2

The passages where O ranks higher than S on the Animacy Hierarchy are the following:

The Information Structure of OVS in Vedic

43

tion is that of an O represented by a personal or demonstrative pronoun and of an S represented by a common noun with an inanimate referent, as in (7), which represents the first instance of the OVS order in the RigVeda: (7)

tvāғmҕ vardhantu no gírahҜ you.ACC increase.IPV.3PL our song(F).NOM.PL “May our songs increase you.” (1.5.8)

From a syntactic or prosodic point of view, stressed pronouns such as tvā̗m “you” display the same free distribution as full-fledged nouns, and may open the clause or the verse 3 . The preposed object of OVS may also be the proper name of a human being or anthropomorphic god, which similarly presents a high degree of individuation, as in (4) and in (8). Here Sūrya, the sun, is described as a conscious god, rather than as a natural force, as can be observed from the epithet Jātavedas lit. “the one who knows (vedas) all that is born (jātá-)”. (8)

úd u tyámҕ jātávedasamҕ devámҕ up PTC that Jātavedas.ACC god(M).ACC.SG vahanti ketávahપ bring.PR.IND.3PL ray(M).NOM.PL “The rays bring up the god Jātavedas.” (1.50.1)

The fact that the patient is denoted by a personal pronoun or by a proper name referring to the addressee of the hymn, while the non-human referent of the agent is presented as generic, is a signal of the low transitivity of the clause. The presence of a negative operator, as in (3) and (9), or of an imperative illocutionary force, as in (7) and (10), may also decrease transitivity. It is well known that imperative clauses and negative clauses are less transitive than declarative clauses and clauses with a positive polarity (Hopper and Thompson 1980):

3

1.5.8, 1.11.1, 1.32.8, 1.32.10, 1.45.6, 1.50.1, 1.52.4, 1.96.3, 1.100.13, 1.117.19, 1.125.7, 1.136.7, 1.144.5, 1.145.3, 1.164.45, 1.167.4, 2.4.2, 2.18.3, 2.23.6, 2.26.4, 2.33.11, 2.35.3, 2.35.4, 3.2.5, 3.10.6, 3.20.1, 3.31.10, 3.34.8, 3.56.1, 4.25.5, 4.31.10, 4.47.2, 4.50.9, 4.53.1, 5.8.3, 5.14.2, 5.30.13, 5.32.11, 5.33.5, 5.53.10, 5.56.2. The “object before subject”-position is even more typical for clitic pronouns (cf. 1.105.8: sám mā tapanty abhítah̠ sapátnīr iva párśavah̠ “Ribs oppress me on every side like rival wives”). Clitic pronouns have been excluded from our counting, since their occurrence in the second slot of the clause (usually after a preverb) is syntactically determined by Wackernagel’s Law, rather than by pragmatics. It may be argued, however, that this syntactic tendency represents the grammaticalization of the same functional principle identifiable in the preferred fronted position of stressed pronouns.

44

Carlotta Viti

(9)

(10)

ná te vájram ánv aśnoti NEG your deadly.weapon(M).ACC PRE reach.PR.IND.3SG káś caná anybody.NOM.M.SG “Nobody reaches your deadly weapon.” (2.16.3) áprલnҕantam abhí sámҕ yantu śókāhҕ stingy(M).ACC.SG on with go.IPV.3PL pain(M).NOM.PL “May pains meet onto the stingy.” (1.125.7)

The scarce transitivity of the situation described in (10) emerges from the professional translations, where the human patient is promoted to the subject position and the inanimate agent is relegated to the status of a locative role: “Let not the liberal sink to sin and sorrow” (Griffith 1889: 87); “Die Spender sollen nicht in Sünde und Schuld fallen” (Geldner 1951: 174) 4 . One third of OVS clauses (47 occurrences, corresponding to 39%) is taken by declarative clauses with a present indicative, which contrast with the only 3 occurrences (2.5%) of the aorist indicative 5 . In Vedic, the present is used for imperfective and atelic situations, while perfectivity and telicity determine the use of the aorist (Gonda 1962). Since perfective aspect and telic actionality are exhibited by prototypical transitive clauses, the rarity of OVS clauses with an aorist verb is a further signal of low transitivity. This also comes out if one considers the remaining OVS clauses which contain an imperfect or a perfect. Although imperfect, perfect, and aorist are indifferently used as past tenses in Old Indian, they still display dissimilar aspectual and actional values in Early Vedic, where the perfect and the imperfect denote stative or durative situations, like the present and 4

5

Western translations rephrase the Vedic expression also as far as negation is concerned. In Vedic we have “may pains meet the stingy”, where negation has narrow scope on the object noun: áprલn̙at- “stingy” is literally a participle of the root p‫ސ‬ઓ “fill” with privative a- meaning “not filling”. Instead, Griffith (1889) and Geldner (1951) extend the scope of negation to the whole clause: “may the liberal not sink” etc. Accordingly, the passage in (10) presents affixal negation (cf. Bhatia 1995: 3ff), in addition to the imperative, as a transitivity decreasing factor. Sentential negation may be also observed in 3.28.4, 3.32.11, 3.56.1, 4.25.5, 4.31.9, etc. The imperative mood proper occurs in 1.122.14, 1.183.2, 2.23.6, 2.33.11, 3.10.6, 3.37.2, 4.31.10, 5.50.3, etc.; an imperative illocutionary force may be also conveyed by optative and subjective moods, which in Geldner (1951) are often translated as möchten-Sätze and sollen-Sätze, cf. 1.120.2, 1.136.7, 4.2.11, 4.16.1, 5.21.4, 5.30.6, 5.33.5, 5.48.1, etc. The only occurrences of the aorist indicative are in 1.11.1, 4.53.3, and 5.81.2. The present indicative occurs in 1.22.20, 1.32.8, 1.32.10, 1.45.6, 1.48.8, 1.50.1, 1.89.3, 1.89.5, 1.90.6, 1.100.13, 1.105.15, 1.123.6, 1.144.5, 1.145.3, 1.152.2, 1.159.4, 1.164.7, 1.164.51, 2.16.3, 2.25.1, 2.26.4, 2.35.4, 3.3.2, 3.20.1, 3.26.5, 3.28.4, 3.34.8, 3.36.7, 3.56.1, 4.7.11, 4.25.5, 4.42.1, 4.42.2, 4.47.2, 4.50.9, 5.3.10, 5.8.3, 5.14.2., 5.39.5 (twice), 5.53.6, 5.53.10, 5.63.2, 5.63.5 (twice), 5.81.2, 5.82.1.

The Information Structure of OVS in Vedic

45

unlike the aorist (cf. Viti 2007: 104ff; a comprehensive discussion of tenses in the Rig-Veda can be found in a recent study of Dahl 2008). Atelic actionality is expressed by the imperfect or perfect of īl̜ “pray” (1.96.3), tap “heat” (3.31.10), duh “milk” (3.31.11), bhrલ “carry” (1.60.1, 4.7.4), ks̚ar “flow” (1.84.4), jusҖ “enjoy” (1.152.5, 1.165.2, 4.33.9), vid “know” (1.164.45, 4.26.5, 5.40.6), sad “sit” (1.65.2, 1.89.2), pū “purify” (3.2.9), yam “stretch” (4.2.14), sthā “stay” (1.52.4, 2.35.3), hve “invoke” (1.117.19), etc. In the remaining occurrences, the imperfect and the perfect mainly have a middle voice, which expresses a less transitive situation than the active voice, implying that the process takes place in the personal sphere of interests of the subject (cf. Kemmer 1993; Kaufmann 2004)6 . It may be significant that the OVS order is particularly frequent with predicates of motion, such as i ‘go’ (1.32.8, 1.125.7, 1.145.3, 2.35.4, 4.47.2, 5.53.6, 5.53.10), gam ‘id.’ (1.32.2, 5.56.2), car ‘move’ (1.32.10, 5.63.2, 5.63.5), etc., as we have seen in (5) and in (10). Strictly speaking, accusative nominals selected by verbs of going are not bona fide direct objects, or at least they lack several features of direct objects. For example, they cannot be promoted to subjects of passive in Vedic, as Kulikov (2001) properly observes. From this point of view, they should not be even included in our corpus of OVS clauses. However, we included predicates of motion because their accusatives of goal do not differ from more typical accusative objects as far as word order is concerned. Owing to the relative frequency of these objects, their exclusion would make us miss a further piece of evidence indicating the low transitivity of OVS. Low transitivity is especially evident in the predicate sac ‘follow’ (11), where the object referent is conceived as the one who directs the motion of the subject referent: (11)

támҕ sacante sanáyas this.ACC.M.SG follow.PR.IND.3PL booty(F).NOM.PL támҕ dhánāni this.ACC.M.SG treasure(N).NOM.PL “Booties, treasures follow him.” (1.100.13)

3.3 Pragmatic correlates of OVS In 59 cases, the higher salience of the object as compared to the subject is not immediately evident from lexical semantics, since O shows an equal or minor degree of animacy than S on the Hierarchy in (1), and may have a 6

The middle voice with the imperfect and the perfect is found in 1.73.7, 1.85.2, 1.164.5, 1.68.9, 1.141.2, 1.164.43, 1.164.50, 3.1.3, 3.2.5, 3.31.11, 3.48.3, 4.26.5, 4.33.9, 5.32.11, 5.44.9, 5.81.1, etc.

46

Carlotta Viti

non-human referent. Nevertheless, the consideration of the entire context clarifies that O is more salient than S under a precise set of circumstances. 7 First, the non-human object noun may be a metaphor of a human being or god, who is currently the topic of the discourse. In the clause illustrated in (3) “The immortals did not find the calf”, the object noun vatsám “calf” is a typical image of the god Agni (cf. the passages 1.95.1 and 1.96.5), which is the addressee of the hymn, and therefore is more topical than the gods (amŕલtāh̠). The latter are presented as generic entities that did not do something, rather than as specific agents. 8 Second, the non-human referent of the object may be contrasted with another referent: contrast is inherently related to salience, and triggers fronting cross-linguistically (Myhill 1992: 208ff.). The example in (2) is extracted from a contrastive situation, which is rendered by means of an adversative clause in scholarly translations: “While the rain clouds vivify the earth, the flames of the sacrifice vivify the sky.” (Brown 1968: 218; emphasis added) Earth and sky represent an established couple in many languages, and especially in Vedic imagery, where they often form copulative compounds such as dyā̗va-prલthivī or instances of natural coordination such as dyaúś ca prલthivīv ca, which imply a semantic component of contrast (Viti 2006). Third, an object phrase may be fronted when it (or part of it) falls into the scope of focus, expressed by a concessive particle such as even, only, etc. (cf. 5.32.3 tyásya cin maható nír mrપgásya vádhar jaghāna távi߅ībhir índrahપ ‘Indra with his forces hit the weapon even of that mighty wild creature’) or by the marker of a wh-question. In (12), for example, the object noun bráhmān̠i “sacred words” is modified by the interrogative pronoun kásya, and therefore is moved at the beginning of the sentence (cf. also 5.48.1):

7

8

The passages where O > S can only be assessed on the basis of the context are the following: 1.22.20, 1.60.1, 1.72.2, 1.89.2, 1.89.3, 1.89.5, 1.90.6, 1.120.2, 1.122.14, 1.132.5, 1.141.2, 1.152.2, 1.152.5, 1.159.4, 1.164.5, 1.164.43, 1.164.50, 1.164.51, 1.165.2, 1.183.2, 2.16.3, 2.25.1 (five times), 3.2.9, 3.3.2, 3.3.11, 3.26.5, 3.28.4, 3.32.11, 3.35.5, 3.36.7, 3.37.2, 3.48.3, 4.2.11, 4.2.12, 4.2.14, 4.7.4, 4.7.11, 4.31.9, 4.33.9, 4.42.1, 4.42.2, 5.3.10, 5.21.4, 5.30.6, 5.32.3, 5.39.5 (twice), 5.40.6, 5.48.1, 5.50.3, 5.63.2, 5.63.5 (twice), 5.81.2, 5.82.1. The idea that O rather than S is currently the topic of discussion is often suggested by the cumulative use of modifiers, complements or appositions for O, as in 1.60.1: “Mātariśvan brought Bh‫ގ‬gu, as a gift, the glorious conveyer of oblations, banner of sacrifice, zealous messenger that immediately reaches his goal, child of two births, praised like richness”, etc. Here the object váhnim “draught animal, horse that conveys oblations” (typically said of Agni) together with its attributes builds a long noun phrase consisting of 33 out of the 44 syllables of the stanza, and denotes the addressee of the hymn, which is further mentioned in the subsequent verses. Instead, the subject noun phrase is only represented by the noun mātariśvan, which decays in the following (cf. also 1.22.20, 1.89.3, and 1.89.5).

The Information Structure of OVS in Vedic

(12)

47

kásya bráhmānҜi jujusҜurҜ whose sacred.words(N).ACC.PL enjoy.PF.3PL yúvānah young(M).NOM.PL “Whose sacred words did the young ones enjoy?” (1.165.2)

The fact that focused constituents are fronted is not at odds with the fronting of topical objects in Vedic. These ideas would be difficult to conciliate if one adopts a view of topic that is mutually exclusive with that of focus: in Lambrecht (1994), for example, a topic is the piece of information that the speaker judges to be identifiable by the hearer on the basis of the previous context, while the focus is the piece of information on which the hearer has presumably not drawn his/her attention. However, as anticipated in §1, we consider the topic as being “what the discourse is about” (cf. Dik 1995), and this can be either definite or indefinite information. Vedic does not possess specific presentative constructions such as English there is to introduce new referents in the discourse: when a referent that is “important” (i.e. a referent that represents a persisting piece of information) is introduced, it is fronted in the same way as an important referent that conveys old information. Syntactic fronting does not occur if the new referent immediately decays. Fourth, object fronting may be determined by numerals, as in (13). Cf. also 1.141.2 and 1.164.5. It may be argued that numerals increase the specificity and individuation of a noun phrase, in that they explicit the exact size of the referent: (13)

tisró yahvásya samídhahҜ three young.GEN.M.SG piece.of.firewood(F).ACC.PL párijmano’ ৴gnér apunann wandering.GEN.M.SG Agni.GEN purify.IPF.IND.3PL uśíjo ámrપtyavaь Uśijs.NOM.PL immortal.NOM.M.PL “The immortal Uśijs purified three pieces of firewood of the young, wan der ing Agni.” (3.2.9)

Fifth, the iteration of the object noun reveals its topic-worthiness. In (14) both S and O are represented by common nouns of inanimate items: S refers to the “rivers” (síndhavah̠), and O refers to a “sweet substance, mead, honey” (mádhu). In the three verses of the stanza, however, the subject changes, and only the noun-adjective mádhu persists:

48 (14)

Carlotta Viti

mádhu vāғtā sweets(N).ACC.SG wind(M).NOM.SG rલtāyaté mádhu order.keeper(M).DAT.SG sweets(N).ACC.SG ksҕaranti síndhavaь pour.PR.IND.3PL river(F).NOM.PL mā¸dhvīr nah Ҝ santv sweet.NOM.F.PL us.DAT be.IPV.3PL ósҕadhīhҜ plant(F).NOM.PL “The wind (brings) sweets to the man who keeps the order; the rivers pour sweets; may the plants be sweet for us.” (1.90.6)

The object noun mádhu is also iterated in stanzas 7 and 8: “Sweet (mádhu) may be the night and the dawns, full of sweets (mádhumat) the terrestrial atmosphere; sweet (mádhu) may be Father Heaven to us. Full of sweets (mádhumān) may be the tree for us, full of sweets (mádhumān) the sun; sweet (mā¸dhvīh̠) may be our milk-cows.” The fact that a mass noun such as mádhu may be topicalized shows the importance of taking into account discourse pragmatics to investigate the principles underlying word order. In (15), the name of an inanimate object (yajñá- “sacrifice”) appears contextually even more topical than the name of an animate subject (devā̗h̠ “gods”), as indicated by the figura etymologica: (15)

yajñéna yajñám sacrifice(M).INSTR.SG sacrifice(M).ACC.SG ayajanta devā¸ь sacrifice.IPF.IND.3PL.MID god(M).NOM.PL “With the sacrifice the gods sacrificed the sacrifice.” (1.164.50)

The same passage recurs in RV 10.90.16. Scholars acknowledge that this hymn ushers in a new conceptualization of the sacrifice, which is no longer seen as a means to worship the gods, but rather as an independent deity who deserves to receive its own worship (Sani 2000: 248-49). When the object is involved in a figura etymologica, it is syntactically fronted even without a precise lexical repetition. Cf. 1.159.4 “The refulgent poets weave a web for ever new in the sky, in the sea” (návya‫؃‬-navya‫ ؃‬tántum ā̗ tanvate diví samudré antáь kaváyaь sudītáyaь); 1.183.2 “may this song attend your glory with glory (vápur vapu߅yā̗ sacatām iyá‫ ؃‬gī̗ь); 2.25.1 “He with his seed spread forth beyond another’s seed, whomever BrahmanҜaspati takes for his friend” (jāténa jātám áti sá prá sars‫ގ‬te yá‫؃‬-ya‫ ؃‬yúja‫ ؃‬k‫ذގ‬uté bráhma‫ذ‬as pátiь); 4.2.11 “May the wise discern discernment and nondiscernment of men, like straight and crooked backs of horses” (cíttim ácitti‫ ؃‬cinavad ví vidvā̗n p‫߮߅ގ‬héva vītā̗ v‫ގ‬jinā̗ ca mártān); 4.7.11 “When, thirsty, (consuming) the food with its thirsty (flame), he grows, the young

The Information Structure of OVS in Vedic

49

Agni makes the thirsty (Wind) his messenger” (t‫߅ގ‬ú yád ánnā t‫߅ގ‬ú‫ذ‬ā vavák߅a t‫߅ގ‬ú‫ ؃‬dūtá‫ ؃‬k‫ذގ‬ute yahvó agníь), etc. These factors are often cumulative. In (16), for example, the object noun phrase, denoting “threads” (tántūn), is fronted with respect to the human subject. Two pragmatic principles may underlie this fronting: the object is specified by a numeral (saptá “seven”) and forms a figura etymologica with the verb, which is equally derived from the root tan “to tend”: (16)

vatsé basҜkáyé 'dhi saptá calf(M).LOC.SG yearling.LOC.M.SG above seven tántūn ví tatnire thread(M).ACC.PL PRE weave.PF.MID.3PL kaváya ótavāғ u sage(M).NOM.PL weave.IF PTC “Above the yearling calf the sages have woven the seven threads to form a web” (1.164.5).

On the whole, a situation where the object ranks higher than the subject in salience because of its lexical-semantic or discourse-pragmatic features appears in 100 out of 119 passages, which correspond to 84% of the analyzed cases of OVS, as can be seen in Table 2: RV I

N (%)

O>S

100 (84%)

S>O

19 (16%)

TOTAL

119 (100%)

Table 2: Information structure of OVS in Rig-Veda I-V

4 Different word orders in the Rig-Veda The peculiar character of the OVS order especially emerges insofar as it is compared with the semantic and pragmatic features typically associated with alternative word order patterns. Here we report the first example occurring in the Rig-Veda for each type of major constituent order where subject, object, and verb are explicitly represented, in patterns such as SOV (17), SVO (18), VSO (19), OSV (20), and VOS (21). The first OVS clause has been reported already in (7):

50 (17)

(18)

(19)

(20)

(21)

Carlotta Viti

SOV sá devāғn éhá vaksҜati he.NOM god(M).ACC.PL PRE.here carry.AOR.SB.3SG “May he carry here the gods.” (1.1.2) SVO tvámҕ valásya gómato you.NOM Vala.GEN rich.in.cows.GEN.M.SG 'pāvar adrivo bílam open.AOR.IND.2SG lord.of.thunder.VOC cave(N).ACC.SG “Lord of the thunder, you opened the cave of Vala rich in cows.” (1.11.5) VSO gámad vāғjebhir āғ come.AOR.SB.3SG strength(M).INSTR.PL here sá nahҕ he.NOM us. ACC “May he come to us with his strength.” (1.5.3) OSV stómā avīvrપdhan tvāғmફ੡ you.ACC praise(M).NOM.PL increase.AOR.IND.3PL tvāғm ukthāғ śatakrato you.ACC hymn(N).NOM.PL having.hundredfold.insight: VOC.M.SG “Praises increased you, hymns (increased) you, who have hundredfold in sight.” (1.5.8) VOS vāғya ukthébhir Vāyu.VOC hymn(N).INSTR.PL jarante tvāғm glorify.PR.IND.3PL.MID you.ACC áchā jaritāғrahҜ PTC singer(M).NOM.PL “O Vāyu, the singers glorify you with their hymns.” (1.2.2)

All orders where the subject precedes the object, such as SOV (17), SVO (18), and VSO (19), present the aorist stem, which – as we have seen in §3.2 – typically expresses perfectivity and telicity, and therefore characterizes highly transitive clauses. Instead, the orders where the object comes first typically display either a present tense, as in the VOS clause in (21), or an imperative mood, as in the OVS clause in (7) (for the aorist tense in OSV, see below). We mentioned that the durative actionality of the present and the irrealis modality of the imperative indicate a low degree of transitivity. This is confirmed if we take into account the ranking of S and O on the Animacy Hierarchy in (1). In SOV and SVO, the subject usually ranks higher than the object on the Animacy Hierarchy. In the SOV order reported in (17), S is a third

The Information Structure of OVS in Vedic

51

person pronoun (sáh̠ “he”) and O is a common noun (devā̗n “gods”). Similarly, in the SVO order in (18), O is a common noun (bílam “cave”) and S is a second person pronoun (tvám “you”). It appears that SOV differs from SVO in the relative topicality (in the sense of relative ranking on the Animacy Hierarchy) of the object with respect to the subject. In SVO, a high difference in topicality exists between the subject and the object, and the object is commonly not topical at all, that is, it is located on the very low part of the Hierarchy. In (18), O is represented by the common noun of an inanimate item such as bílam “cave”. In SVO, V behaves as a borderline between the preposed topical information and the postposed non-topical information. The idea that non-topical material is relegated in the post-verbal position has also been assessed in the major constituent order of Ancient Greek (Dik 1995: 111ff; Matić 2003). In the Rig-Veda, the information structure conveyed by SVO can be considered the mirror-image of the information structure of OVS, the other verb-medial pattern, analyzed in §3. We have seen that a striking difference in the pragmatic status of subject and object exists in OVS, where the subject is the non-topical constituent. In the example of OVS in (7), O is a second person pronoun (tvā̗m “you”), while S has an inanimate referent, such as gírah̠ “songs”. Unlike in SVO, in SOV the object is assigned a certain degree of topicality, that is, it usually refers to human or animate items, which are represented higher than inanimate items on the Hierarchy. In (17), the animate referents of the object, i.e. the gods, persist as topical participants in this hymn: cf. stanza 4 “he goes among the gods” (sá íd devés̚u gachati); 5 “he, the god, may come here with the gods” (devó devébhir ā̗ gamat). A different situation can be identified in OSV, the other verb-final construction. On the one hand, in this order O outranks S in topicality, similarly to what happens to other object-initial orders. The OSV clause “Praises increased you” in (20), for example, is coordinated with a similar OVS clause such as “May our songs increase you” reported in (7). On the other hand, transitivity is relatively higher in OSV than in OVS. In our examples, (7) shows an imperative, while (20) has the aorist, which per se is more typical of subject-initial orders than of object-initial orders. In an analysis of the OSV word order in the Rig-Veda (Viti 2009), we observed that in the majority of cases the subject has a human referent like the object in this order, as can be seen in the second OSV clause that is attested in the Rig-Veda: “Indra the singers … glorified” (índram íd gāthíno ... anū߅ata). Thus, OSV is similar to other object-initial orders such as OVS in presenting the object as more topical than the subject, but at the same time the difference in topicality between S and O is not as dramatic in OSV as it is in OVS.

52

Carlotta Viti

The two verb-initial patterns, VSO and VOS, can be also conceived as mirror-images of each other. In VOS (21), the object pronoun tvā̗m “you” is clearly more topical than the subject common noun jaritā̗rah̠ “singers”. VOS shares this pragmatic situation with other orders such as OSV and OVS, where the object is placed before the subject. Instead, the consideration of the context clarifies that S outranks O in topicality in the VSO word order in (19). Here we have a third person subject (sáh̠ “he”) and a first person object (nah̠ “us”), so that the latter seems to rank higher from a purely semantic point of view, that is, considering the ranking of the Animacy Hierarchy (1). However, it is enough to take a look at the whole sentence, reported in (22), to see that the third person subject occurs in an anaphoric chain, and is more persisting and topical than the first person object. Pragmatic studies acknowledge that, in case of a clash between lexical semantics and discourse pragmatics, the latter prevails in orienting the word order of a noun phrase. “If a language should be discovered where some order facts appear to be describable in semantic terms, but in a given context a certain pragmatic principle would predict an alternative order, the pragmatic principle will most surely win out. The overall governing principles would thus be pragmatic, and order would only appear to be semantically based in the norm because of the large majority of cases where semantic and pragmatic principles converge on the same structure.” (Payne 1992: 3-4): (22)

sá ghā no yoga āғ he.NOM PTC our performance(M).LOC.SG PRE bhuvat sá rāyé be.AOR.INJ.3SG he.NOM wealth(M).DAT.SG sá púramҕdhyām he.NOM abundance(F).LOC.SG gámad vāғjebhir come.AOR.SB.3SG strength(M).INSTR.PL āғ sá nahҜ here he.NOM us.ACC “May he stand by us in our performance, he in abundance for our wealth, may he come to us with his strength”. (1.5.3)

This information structure puts VSO on the same plan of the other orders where the subject is placed in front of the object, such as SOV and SVO, in which the subject is also more topical than the object. However, verbinitial orders are substantially different from other patterns, in that here the verb is really the most salient piece of information of the clause. A predicate may be considered a salient piece of information when attention is given to the way the denoted action takes place, by means of modifiers

The Information Structure of OVS in Vedic

53

or complements that add some details of that action. By means of such adverbs the composer may insist on a verbal concept 9 . In the VSO order in (19), both S and O are pronouns, which represent highly accessible topics. The imperative illocutionary force of the subjunctive mood gámad “may he come” lays emphasis on the predicate, and the instrumental complement vā̗jebhih̠ “with strength” lingers on the description of the manner of coming. Similarly, in the VOS order in (21), the object is a pronoun, and the noun with the function of the subject, jaritā̗rah̠ “singers”, does not add anything new to the information of the preceding verb jarante “they sing”. The fact that S is built on the same root jrલ “sing” as the verb indicates that S is completely predictable. Here the verbal action is highlighted by the addition of the instrumental complement ukthébhih̠ “with hymns”, which portrays the manner of singing. The instrumental case has the function of a manner adverb qualifying verbal action in Old Indian. In the rubric entitled “instrumental, the howcase”, Speyer (1886: §63) observes: “It is always used when it is wanted to express the circumstances, instruments, means, ways, properties accompanying the action and qualifying it. In other terms, the instrumental has the duty of telling the how of the action or state, expressed by the verb or verbal noun, it depends on.” The fact that adverbials draw attention to verbal action may explain the cross-linguistic finding, noted by Bubenik (1991: 19), that adverbials are typical of the VSO order. This may also explain why two similar constructions, quoted by Bubenik (1991: 20) and here reported as examples (23) and (24), may present two different word orders: (23)

(24)

āғ gachanti īm ávasā PRE come.PR.IND.3PL him.ACC aid(N).INSTR.SG “They come to him with aid.” (1.85.11) pári sҖīmҕ nayanti PRE him.ACC lead.PR.IND.3PL “They lead him around.” (1.95.2)

The verb-initial clause in (23) is accompanied by an instrumental complement (ávasā “with aid”) that specifies the circumstances of the action of coming, similarly to (19) and (21) above. The idea that verb-initial pat9

More specific conditions of the verb-initial order may vary from language to language. In Vedic and in Homeric Greek this order seems to be used when the sentence has a “thetic” function, that is, when the subject belongs to the focus, rather than to the topic, of the clause (Viti 2008). In other languages, initial verbs may have a text-connecting function, as Reis (2000) and Lühr (2007a) show for German and Old Frisian, respectively. In Old Frisian, conditional clauses with initial verbs are found in the middle of a discourse, and are distinguished from conditional clauses with generic relative pronouns, which occur at the beginning of a text section. This phenomenon may be considered a development peculiar to the Germanic languages.

54

Carlotta Viti

terns are related to the salience of the predicate is supported by Bubenik’s (1991: 22) data, according to which Vedic favors VO in jussive clauses. In the same vein, Klein (1991: 130) states that verb-initial clauses in Vedic are often found with performative predicates, such as “I call”, “we proclaim”, etc., where the utterance of a first-person subject is identical to the performance of the action denoted by the verb. Command and first person can be considered independent signals of verbal salience. The salience of the verb in verb-initial orders may be related to the fact that verbs are accented in Vedic only when they are placed in the first position of a metric or syntactic unit (Macdonell 1916: 466-468). Prosodic prominence is a formal manifestation of importance. Table 3 summarizes the pragmatic situations associated with the different word orders in Vedic. The cases where the predicate is the main focus of attention are signalled by the symbol + V; the cases where this does not occur have the symbol – V. The marks > and >> represent the relative ranking of arguments, in agreement with Givón (2001: 93), who uses these symbols to illustrate the different topicality of the agent and of the patient in active, inverse, passive, and anti-passive constructions: 10 ORDER

ARGUMENTS

PREDICATE

SOV

S>O

-V

SVO

S>>O

-V

OSV

O>S

-V

OVS

O>>S

-V

VSO

S>O

+V

VOS

O>S

+V

Table 3: Information structure of the major constituent orders

Table 3 shows that the relative ranking of S and O is less dramatic when they are adjacent, i.e. in SOV, OSV, VSO, and VOS. Instead, when they 10 Particularly, >> signals that the arguments not only have a different prominence in the context, but often belong also to different lexical-semantic classes or to different positions of the Animacy Hierarchy in (1). Since OVS often contains subjects and objects that differ both semantically and pragmatically (cf. §3), the single arrow (>) in Table 2 has been more properly presented here as double arrow (>>) in comparison with the information structure of the other word orders.

The Information Structure of OVS in Vedic

55

are separated by the verb, in SVO and OVS, they differ much more in salience. Since in Vedic the syntactic function of an argument is indicated by its grammatical case, word order (particularly verb-medial word order) is clearly not used to disambiguate subject and object. To the extent that arguments that are semantically and pragmatically similar are placed closer to each other than dissimilar arguments, word order is iconically arranged in this language.

5 Typological parallels The OVS order is rarer than its symmetrical pendant, the SVO order. In the first book of the Rig-Veda, for example, we counted 45 occurrences of OVS and 132 occurrences of SVO. This tallies with the distribution of OVS cross-linguistically: a basic OVS order is so rare that it was assumed not to exist for a long time, until it was identified in Hixkaryana (25) and in other languages of the Amazon basin by Derbyshire and colleagues: (25)

Hixkaryana (Derbyshire 1985: 31; cf. 97ff.) kuraha yonyhoryeno biryekomo bow he made: it boy “The boy made a bow.”

More generally, typological research found that all orders where the object precedes the subject (OSV, OVS, and VOS) are recessive with respect to the orders where the subject precedes the object (SOV, SVO, and VSO; cf. Greenberg 1966: 77). In the sample of Mallinson and Blake (1981: 148), 85 out of 100 languages abide by the principle of subject precedence, and only 4 languages contravene it (the remaining languages have a free word order, where no basic pattern can be identified from the sources). In a larger sample (Dryer 2005: 330-333) consisting of 1228 languages, subject precedence is represented by 1017 languages (83%), while object precedence appears in 39 languages (3%) (the remaining 172 languages lack a dominant word order). Comrie motivates the principle of subject precedence with the universal correlation between subject and agent, which typically is presented as more topical (i.e. as anaphorically more definite and cataphorically more persisting) than the object in transitive clauses. “Explanations for the predominance of word orders where the subject precedes the object seem more likely to have a psychological basis, in terms of the salience of the agent in the agent-action-patient situation, and the high correlation between semantic agent and syntactic subject.” (1981: 20) The semantic motivation of word order is obscured in many languages, where the overlap between agent and subject determines the exten-

56

Carlotta Viti

sion of the word order used in prototypical transitivity to different situations, independently of the degree of agentivity of the subject. This semantic basis, however, can be observed both in Vedic and in Hixkaryana. In Vedic, where word order is not fixed, we have seen that the object precedes the subject in the unusual case of an object referring to a more topical participant than the subject. Hixkaryana, where the establishment of OVS is a relatively recent phenomenon, has an ergative alignment (thus, PVA would be a more appropriate description of this basic word order). It has been observed that ergative constructions often derive from passive constructions with the agent expressed (Mallinson and Blake 1981: 109ff.) 11 . In the passive, the patient is the real topic of discourse. The parallel with the languages of the Americas, where a pragmatic pattern topic – comment or rheme – theme has been often identified (cf. Tomlin and Rhodes 1979; Mithun 1987; Payne 1987, etc.), suggests not to consider merely definiteness as being the driving factor of syntactic fronting in Vedic, as proposed for example by the Prague School. In the Functional Sentence Perspective of the Prague School, theme and rheme correspond to old information and new information, and topicality is identified with giveness (Firbas 1966). This view is based on a number of modern European languages, where the first constituent of a sentence is usually more definite or lower in communicative dynamism than the subsequent constituents. Instead, when we observe that the most topical piece of information comes first in Vedic, we think of the nominal referring to the participant that ranks higher in humanness, specificity, agentivity, and persistence in the discourse. The Rig-Veda contains a quite limited range of issues, inherent to prayer, sacrifice, priests, etc., so that both the subject and the object of a transitive clause often convey given information, and identifiability is not sufficient to predict word order (cf. Lühr 2004). Moreover, even a noun phrase conveying new information may be fronted, as long as it is the most salient argument of the clause in Vedic. For example, the passage “The rays bring up the god Jātavedas” in (8) represents the incipit of a hymn, so that the pre-verbal object jātávedasam necessarily conveys new information, while the post-verbal subject ketávah̠ “rays” are easily inferable from the mention of the sun, on the basis of a part-whole relation. In this case, the placement of the word for the whole in front of the

11 In the case of Hixkaryana, the ergative alignment derives from a passive nominalization, where the patient is encoded as a pre-verbal genitive and the agent appears after the verb as the complement of an adposition. For example, the clause “the enemy destroys the city” would be originally expressed as a nominalized expression such as “the city’s destruction by the enemy” (Gildea 1997).

The Information Structure of OVS in Vedic

57

word for the part is consistent with the Animacy Hierarchy, but is inconsistent with the “old before new” principle. Under this perspective, a topical referent is one that has a prominent position not only, or not necessarily, in the previous part of the discourse, but also in the following. Lexical iteration manifests such prominence: in the Rig-Veda, both content-words and function-words that are iterated are typically placed in the initial position of the linear string. This matches the situation commonly found in non-IE language provided with a flexible word order. In languages such as Ute and Papago, the pre-verbal position is devoted to focused, discontinuous, contrastive, and indefinite noun phrases (Myhill 1992: 209). Indefinite noun phrases, however, are only fronted as long as they persist in the subsequent discourse. Payne (1987: 796) finds out that a substantial amount of post-verbal indefinite noun phrases in Papago are not mentioned again in the discourse. Armendáriz (this volume) notices that the focus relation is coded by the sentence initial position in Warihío. The fact that languages such as Hixkaryana (Carib), Ute (UtoAztecan), Papago (id.), Warihío (id.), etc. belong to non-IE domains should not obscure their similarities with the early IE languages as far as word order is concerned. All these languages, where word order is not subject to strictly syntactic rules, mainly have an oral tradition. A diachronic change is likely to have occurred from the stage of the early IE languages, where fronting was due to topicality in the sense of specificity, animacy, and persistence, to the stage of the modern IE languages, where old information usually precedes new information. The change probably started from enclitic pronouns, which are inherently definite expressions and, at the same time, tend to appear in long chains, so that their referents are contextually important. Since enclitic pronouns grammaticalize their position earlier than full-fledged noun phrases (Lambrecht 1994: 201-202), the association between definiteness and preverbal position easily entrenches. This probably determined the reinterpretation of pre-verbal arguments as definite arguments, as captured by the tenets of the Prague School.

6 Conclusions We have discussed word order in Early Vedic, and particularly the appearance of the pattern OVS in the Rig-Veda. Traditionally, Vedic is assigned a basic SOV word order, which is assumed to reflect the consistent SOV word order of PIE. Alternative arrangements such as OVS are wounded up as exceptions due to poetic license. This interpretation, however, does

58

Carlotta Viti

not capture the generalization that OVS occurs in a precise set of pragmatic situations, where an inanimate agent affects a human patient, independently of the morphological or metrical structure of the arguments. Accordingly, OVS clauses display a low degree of transitivity, which is also indicated by the common association with an irrealis modality and with predicates expressing states, rather than actions, or durative events. We do not deny the marked status of OVS in Vedic, which is also relatively rare in our corpus. Rather, we contest its interpretation: in our opinion, OVS should be considered marked because the pragmatic functions it conveys have less chance to appear in discourse. In Vedic, different word orders are associated with different pragmatic situations, according to two main principles. First, the fronted argument is more specific, animate, and topical (in SOV, SVO, and VSO, the subject is more topical than the object; in OSV, OVS, and VOS, the opposite occurs). The relation between syntactic fronting and perceptual salience has cross-linguistic parallels, and is also demonstrated by cognitive studies, according to which material placed in the initial position of a string plays a privileged role in processes of attention and memory (Gernsbacher 1990). Second, subject and object tend to be adjacent when they are semantically and/or pragmatically similar (in SOV, OSV, VSO, and VOS), while they are usually separated by the verb when they are different from each other (in SVO and OVS). Synchronically, the idea that different word orders are not synonymous is compatible with an approach like that posited in Construction Grammar (cf. Goldberg 1995). Although Construction Grammar strictly speaking does not take word order into account, its assumptions may be extended to situations where word order is syntactically free and is used to convey meaning, as in the case of Vedic. Diachronically, the marked pragmatic conditions in which OVS appears in the Rig-Veda make it predictable that this word order would decay, together with the other patterns where the subject is preceded by the object, such as OSV and VOS. The IE languages grammaticalize the basic word orders SOV, SVO, and VSO, where the subject is more topical than the object, as in typical transitive clauses.

The Information Structure of OVS in Vedic

59

Abbreviations A = agent; ACC = accusative; AOR = aorist; DAT = dative; DU = dual; F = feminine; GEN = genitive; GER = gerund; IE = Indo-European; IND = indicative; INJ = injunctive; INSTR = instrumental; IPF = imperfect; IPV = imperative; LOC = locative; M = masculine; MID = middle; N = neuter; NEG = negation; NOM = nominative; O = object; P = patient; PIE = Proto-Indo-European; PF = perfect; PL = plural; PR = present; PRE = preverb; PTC = particle; PVA = patient-verb-agent; SB = subjunctive; QUOT = quotative; S = subject; SG = singular; V = verb; VOC = vocative.

References Bhatia, T. (1995): Negation in South Asian languages. Patiala: Indian Institute of Language Studies. Brown, N. (1968): Agni, Sun, Sacrifice, and Vāc: a sacerdotal ode by Dīrghatamas. Journal of American Oriental Studies 88, 199-218. Bubenik, V. (1991): Nominal and pronominal objects in Sanskrit and Prakrit. In: Hans H Hock (ed.), 19-30. Comrie, B. (1981): Language universals and linguistic typology. Chicago: The University of Chicago Press. Croft, W. (1990): Typology and universals. 2nd edition. Cambridge: Cambridge University Press. Dahl, E. (2008): Time, Tense and Aspect in Early Vedic Grammar. A TimeRelational Approach to the Morphosyntax-Semantics Interface. PhD Thesis: University of Oslo. Delbrück B. (1888): Altindische Syntax (=Syntaktische Forschungen 5). Halle: Verlag der Buchhandlung des Weisenhauses. Derbyshire, D. (1985): Hixkaryana and linguistic typology. Arlington: The University of Texas. Dik, H. (1995): Word order in Ancient Greek. A pragmatic account of word order variation in Herodotus. Amsterdam: Gieben. Dowty, D. (1991): Thematic proto-roles and argument selection. Language 67, 547-619. Dryer, M. (2005): Order of subject, object, and verb. In: M. Haspelmath., M. Dryer, D. Gil and B. Comrie (eds.), The World Atlas of Language Structure, 330-333. Oxford: Oxford University Press.

60

Carlotta Viti

Firbas, J. (1966): On defining the theme in functional sentence analysis. In: Travaux linguistique de Prague, 267-80. Geldner K. F. J. (1951): Der Rig-Veda aus dem Sanskrit ins Deutsche übersetzt und mit einem laufenden Kommentar versehen von Karl Friedrich, I-III; IV Index by Nobel J. (1957). Cambridge (Mass.): Harvard University Press. Gernsbacher, M. A. (1990): Language comprehension as structure building. Hillsdale: Erlbaum. Gildea, S. (1997): Introducing ergative word order via reanalysis. Word order change in the Cariban family. In: J. Bybee, J. Haiman and S. Thompson (eds.), Essays on language function and language type dedicated to T. Givón, 145161. Amsterdam and Philadelphia: John Benjamins. Givón, T. (1988): The pragmatics of word order: predictability, importance and attention. In: M. Hammond, E. Moravcsik, and J. Wirth (eds.), Studies in syntactic typology, 243-84. Amsterdam and Philadelphia: John Benjamins. Givón T. (2001): Syntax. Revised edition, Volume II. Amsterdam and Philadelphia: John Benjamins. Goldberg, A. (1995): Constructions: A Construction Grammar approach to argument structure. Chicago-London: The University of Chicago Press. Gonda, J. (1958): Syntax and verse structure in the Veda. Indian Linguistics 20, 35-43. Gonda, J. (1959): On amplified sentences and similar structures in the Veda. In: Four studies in the language of the Veda. The Hague: Mouton, 7-70. Gonda, J. (1962): The aspectual function of the Rigvedic present and aorist. The Hague: Mouton. Greenberg, J. (1966): Some universals of grammar with particular reference to the order of meaningful elements. In: J. Greenberg (ed.), Universals of language. 2nd edition, 73-113. Cambridge (Mass.): MIT Press. Griffith, R. (1889): The hymns of the Rig-Veda. Delhi: Motilal Banarsidass. Hock, H. (ed.) (1991): Studies in Sanskrit syntax. Delhi: Motilal Banarsidass. Hopper, Paul and S. Thompson (1980): Transitivity in grammar and discourse. Language 56, 251-99. Kaufmann, I. (2004): Medium und Reflexiv. Ein Studium zur Verbsemantik. Tübingen: Niemeyer. Kemmer, S. (1993): The middle voice. Amsterdam and Philadelphia: John Benjamins. Klein, J. (1991): Syntactic and discourse correlates of verb-initial sentences in the Rig-Veda. In: H. H. Hock (ed.), 123-143. Kulikov, L. (2001): The Vedic -ya- presents. PhD dissertation: Leiden University.

The Information Structure of OVS in Vedic

61

Lambrecht, K. (1994): Information structure and sentence form. Topic, focus and the mental representations of discourse referents. Cambridge: Cambridge University Press. Lazard, G. (1984): Actance variations and categories of the objects. In: F. Plank (ed.), Objects. Toward a theory of grammatical relations, 269-292. London: Academic Press. Lehmann, W. P. (1974): Proto-Indo-European syntax. Austin-London: University of Texas Press. Lühr, R. (2004): Thematische Rollen und Kasus: Zu Agens und Patiens beim Passiv im Altindischen. In: A. Steube (ed.), Grammatik und Kontext: Zur Interaktion von Syntax und Semantik und Prosodie bei der Informationsstrukturierung. Linguistische Arbeitsberichte 81, 99-126. Lühr, R. (2007a): Bedingungssätze in altfriesischen Rechtstexten. In: R. Bremmer, S. Laker and O. Vries (eds.), Amsterdamer Beiträge zur älteren Germanistik. Band 64. Amsterdam: Rodopi, 213-238. MacDonell, A. (1910): Vedic grammar. Straßburg: Trübner. MacDonell, A. (1916): A Vedic grammar for students. London: Oxford University Press. Mallinson, G. and B. Blake (1981): Language typology. Amsterdam: New Holland. Matić, D. (2003): Topic, focus, and discourse structure. Ancient Greek word order. Studies in Languages 27, 573-633. Mithun, M. (1987): Is basic word order universal? In: Tomlin, R. (ed.), Coherence and grounding in discourse, 281-328. Amsterdam and Philadelphia: John Benjamins. Myhill, J. (1992): Typological discourse analysis. Oxford: Blackwell. Payne, D. (1987): Information structuring in Papago narrative discourse. Language 63, 783-804. Payne, D. (1992): Introduction. In: Payne, D. (ed.), Pragmatics of word order flexibility. Amsterdam and Philadelphia: John Benjamins. Reis, M. (2000): Anmerkungen zu Verb-erst-Satz-Typen im Deutschen. In: R. Thieroff et al. (eds.), Deutsche Grammatik in Theorie und Praxis, 215-227. Tübingen: Niemeyer. Sani, S. (2000): RલJveda. Le strofe della sapienza. Venezia: Marsilio. Schäufele, S. (1991): Verb-medial clauses in Vedic: some theoretical implications. In: H. H. Hock (ed.), 177-196. Silverstein, M. (1976): Hierarchies of features and ergativity. In: R. Dixon (ed.), Grammatical categories in Australian languages, 112-171. Canberra: Australian Institute of Aboriginal Studies.

62

Carlotta Viti

Speyer, J. S. (1886): Sanskrit Syntax. Leyden; repr. 1973, Delhi: Motilal Banarsidass. Timberlake, A. (1977): Reanalysis and actualization in syntactic change. In: C. Li (ed.), Mechanisms of Syntactic Change, 141-177. Austin and London: University of Texas Press. Tomlin, R. and R. Rhodes (1979): An introduction to information distribution in Ojibwa. Chicago Linguistics Society 15, 307-320. Viti, C. (2006): AND in the early Indo-European languages. Archivio Glottologico Italiano 91, 129-165. Viti, C. (2007): Strategies of subordination in the Rig-Veda. Milano: Angeli. Viti, C. (2008): The verb-initial word order in the early poetry of Vedic and Homeric Greek. In: K. Jones-Bley, M. Huld, A. Della Volpe and M. Robbins Dexters (eds.), Proceedings of the 19th Annual UCLA Indo-European Conference (Los Angeles, November 2nd – 3rd 2007), Selected Papers, 89-111. Viti, C. (2009): A quantitative analysis of the OSV word order in Vedic. In E. Rieken and P. Widmer (eds.), Pragmatische Kategorien. Form, Funktion und Diachronie, 307-322. Wiesbaden: Reichert. Watkins, C. (1993): Proto-Indo-Europeo. In: A. Giacalone Ramat and P. Ramat (eds), Le lingue indoeuropee, 45-90. Bologna: Il Mulino. Whitney, W. (1889): Sanskrit Grammar, 2nd edition, Cambridge (Mass.): Harvard University Press.

Information Packaging and the Rise of Clitic Doubling in the History of Spanish Christoph Gabriel (Hamburg) & Esther Rinke (Frankfurt M.) 1 , 2

1 Introduction This paper is concerned with the diachronic development of clitic doubling (CD) in Peninsular Spanish. With CD we refer to the occurrence of a nominal or strong pronominal object together with a co-referential clitic pronoun in one and the same prosodic and syntactic domain. This construction must be distinguished from structures with right-dislocated objects (clitic right dislocation; CLRD) that display the same surface order of constituents, but differ in both prosodic phrasing and internal syntactic structure. The two constructions also differ with respect to the informationstructural status of the (nominal or pronominal) object: While the object constitutes a dislocated topic in a CLRD structure and thus obligatorily forms part of the presupposition, it is preferably construed as belonging to the focus domain in CD constructions. An interesting correlation arises with regard to the nature of the elements that undergo CD in modern Spanish: Although doubling implies a focus interpretation of the object in most cases, elements displaying a high degree of ‘topicality’, such as pronouns, definite noun phrases, experiencer arguments and proper names (Givón 1976), are more likely to undergo doubling than those situated lower in the topicality hierarchy. We interpret this finding in diachronic 1 2

We would like to express our gratitude to Dejan Matić, Kristine Eide and three anonymous reviewers as well as to the editors for helpful comments. The term ‘packaging’ of information was introduced by Chafe (1976). Vallduví (1993), who uses the term and elaborates upon it further, provides the following explanation: “A sentence, in one of its facets, may be viewed as a structural vehicle used to transfer some piece of knowledge (a proposition) from speaker to hearer. Information packaging is the speaker’s tailoring of this structural vehicle to suit some ‘communicative’ aspect of the transfer of knowledge (propositional content) to the hearer. In other words, when communicating a proposition p a given speaker may encode p in different sentential structures according to his/her beliefs about the hearer’s knowledge and attentional state with respect to p” (Vallduví 1993: 2).

64

Christoph Gabriel & Esther Rinke

terms as evidence in favor of the hypothesis that clitic doubling has diachronically evolved from right dislocation. Our paper is organized as follows: In section 2, we clarify the notions of CD and CLRD before giving a brief overview of clause-internal doubling in contemporary Peninsular Spanish. We will point out the close relationship between the contexts in which doubling occurs and the topicality of the doubled constituent. Section 3 provides the results of our quantitative and qualitative analyses of Spanish texts from the 15th to 18th century and sketches the diachronic development of clitic doubling in Spanish. The earliest ‘true’ CD constructions are found as late as the 16th century, whereby doubling first applies for strong pronominal forms before spreading over to nominal objects. We will sketch possible motivations for the diachronic change and provide further arguments in favor of our hypothesis. Section 4 summarizes our argumentation and offers some concluding remarks.

2 Object doubling in contemporary Peninsular Spanish 2.1 Clitic doubling vs. clitic right dislocation Modern Spanish displays two types of constructions that involve both a nominal or strong pronominal object and a co-referential clitic pronoun, namely dislocation structures on the one hand and constructions with clause-internal clitic doubling on the other. Despite the fact that both constructions can be superficially identical (i.e. displaying the same surface ordering, at least in the case of a right-dislocated object), they differ decisively from one another regarding (a) prosody, (b) syntax, and (c) information structure. Following Fernández Soriano, we define CD as the occurrence of a clitic and a co-referential object within one and the same clause (“en la misma oración”, 1999: 1248), with the object XP occupying its canonical position. In order to distinguish these cases from CLRD, we have to rely on prosody, in the sense that the notion of CD is limited to cases where object and co-referential clitic co-occur within the same prosodic unit. This is exemplified in (1a), where the indirect object a Juan is doubled by the dative clitic le (3sg). In contrast, as shown in (1b), a right-dislocated object is separated from the core sentence by a prosodic boundary and is in most cases characterized by a flat intonational contour (low plateau). Here and in the following, the doubled XP is underlined (irrespective of

The Rise of Clitic Doubling in the History of Spanish

65

its categorial status), the clitic appears in italics, and capitalization indicates the position of the nuclear stress (i.e. sentence accent). 3 (( Pedro le dio una manzana a JUAN)ip )IP | | LL% ‘Peter gave an apple to John.’ )ip)IP b. (( Pedro le dio una manZAna,)ip ( a Juan | | | L- low plateau L-L% ‘Peter gave him an apple, to John.’

(1) a.

CD (iO)

CLRD (iO)

Although the examples in (1) display the same linear order, we assume that they differ from one another with respect to their internal syntactic structure: While the doubled object stays in its vP-internal base position in a CD construction (see 2a, below), it is located in a position outside vP, when it undergoes CLRD as in (2b). In this case, we assume that the vPinternal complement position is filled by pro. The same holds for constructions that display a clitic, but lack an overt XP in the canonical object position (2c). (2)

a. Pedro le dio T [vP dio una manzana dio a JUAN] CD b. Pedro le dio T [vP dio una manZAna dio proi], a Juani CLRD only clitic, no c. Pedro le dio T [vP dio una manZAna dio pro] nominal object realized

Considering (2a), we can say that the dative clitic le agrees in its case 4 and phi-features (3sg) with the nominal indirect object a Juan, which is located in its vP-internal base position. In examples (2b) and (2c), on the other hand, the clitic agrees with the phonetically null pronoun pro, which fills the argument position. However, the two structures differ insofar as pro is referentially linked to the dislocated DP a Juan in (2b), whereas in (2c) the null pronoun is interpreted with respect to a discourse antecedent. Since Kayne’s (1975) and Strozer’s (1976) major works, the debate concerning the syntactic derivation of constructions with pronominal clitics can largely be characterized as a competition between two contrasting analyses, namely movement and base generation approaches. For the 3

4

The notation of prosodic contours follows the conventions of the Autosegmental-Metrical (AM) framework; see Hualde (2003) for a concise introduction with special reference to Spanish. Without any further discussion, we assume an intermediate phrasal (i.e. ‘small’ ip) boundary triggered by the insertion of a low intermediate boundary tone (L-) at the left edge of the right-dislocated material, following recent work by Astruc (2005). See Gabriel (2007: chap. 3.1.3.2) concerning the insertion of a high intermediate boundary tone (H-) at the right edge of a left-dislocated constituent. Note that in leísta dialects, the object clitic is often opaque regarding its case feature, given the fact that direct objects [+anim] correspond with the dative clitic le instead of the appropriate accusative clitic lo, e.g. ¿Conoce a Juan? Sí, le conozco. See Bleam (1999).

66

Christoph Gabriel & Esther Rinke

purposes of this paper, we follow the latter and assume that Spanish pronominal clitics are base-generated as agreement markers on the verb. 5 It should be pointed out, however, that this interpretation of Spanish object clitics as agreement markers does not necessarily entail an affix analysis of the elements concerned (see section 3.4). Concerning the structural position of the dislocated DP, we follow the general line of argumentation proposed by Suñer (2006), who adopts a base generation approach for dislocated constituents. She observes that in Argentinean Spanish, a doubled object in a dislocated position can cooccur with a coreferential epithet as in [dislocated XP A Juan]i, loi vi ayer [epithet a este loco]i ‘John, I saw that crazy guy yesterday’. Assuming that the epithet [a este loco] occupies the vP-internal object position, these constructions suggest that the dislocated constituent a Juan cannot be base-generated in the vP-internal object position as well. Suñer proposes that the topicalized constituent is instead merged directly into the relevant target position. Although Suñer (2006) concentrates on left dislocation structures and does not make any claims concerning the structural analysis of CLRD, we adopt her analysis and assume that CLRD is not derived via movement, but rather results from the base generation of the relevant constituent in its vP-external position. 6 The third distinction between CD and CLRD that must be mentioned here concerns information structure: In a CLRD structure, the doubled object is necessarily interpreted as a dislocated topic (3a), whereas an object which is clause-internally doubled by a pronominal clitic can perfectly be understood as a focus or part of the focus domain (sketched in 3b). Note that the focus interpretation of the doubled object is not compulsory, given the fact that a non-focused indirect object, which is not

5

6

For different versions of the base generation approach towards Romance pronominal clitics see, among others, Suñer (1988), Kaiser (1992), and Barbosa (2000); for a general overview of recent syntactic approaches to pronominal clitics see Manzini (2003). It has to be mentioned that other analyses assume movement of right-dislocated constituents. These analyses can be divided into two groups, which either derive CLRD from clitic left dislocation or explain it in terms of rightward linearization. The first option is taken by Cecchetto (1999), who assumes that a dislocated constituent first moves to a left-peripheral topic position (Rizzi 1997). This movement is followed by remnant movement of the remaining TP to a higher topic position. The second option is adopted by Villalba (2000) and López (2003, 2006, 2009), who identify the landing site of the dislocated element with a peripheral position within the vP domain (see also Belletti 2005). In order to allow for rightmost linearization of the respective constituent, López (2006) interprets Kayne’s (1994) Linear Correspondence Axiom (LCA) as an OT style constraint that occupies a low position in the assumed constraint ranking. Violating the LCA therefore does not entail a fatal violation when a higher ranked constraint is satisfied instead.

The Rise of Clitic Doubling in the History of Spanish

67

separated from the core sentence by a prosodic boundary, undergoes clitic doubling in the same way a focal one does (3c). 7 (3)

a. (‘What did Peter give to John?’) CLRD (iO) Pedro le dio una manZAna, [topic a Juan] b. (‘To whom did Peter give an apple?’ / ‘What did Peter do?’ / ‘What hap pened?’) Pedro le dio una manzana [F a JUAN] CD (iO) Pedro [F le dio una manzana a JUAN] CD (iO) [F Pedro le dio una manzana a JUAN] CD (iO) c. (‘Who gave an apple to John?’) CD (iO) [F PEdro] le dio una manzana a Juan

The assumption that the clitic pronoun and the respective (nominal or pronominal) object XP occur in one and the same clause implies that object clitics function as agreement markers, whereas the nominal object XP satisfies the thematic and case conditions within vP. Against the backdrop of this analysis, it would be expected that clitics occur in every sentence containing a transitive verb. The distribution of clitic doubling, however, is determined by a number of additional factors: In most cases, the appearance of the clitic is subject to considerable optionality, even within one and the same variety. 8 The next section therefore considers the distribution of CD in modern Peninsular Spanish in a more detailed manner. 2.2 The distribution of clitic doubling in Peninsular Spanish Cross-linguistically, clitic doubling is characterized by great variability. In contemporary Peninsular Spanish, the obligatory occurrence of CD is limited to some special contexts; in most cases, it applies optionally. Under certain circumstances, it can even be excluded completely. Strong pronominal objects constitute obligatory contexts for doubling, irrespective of their status as direct or indirect objects. As shown in (4a, b), the absence of the clitic pronoun would turn the relevant sentences into ungrammatical structures. The only exceptions are constituted by the pronoun of polite address usted and impersonal ello. While the former optionally occurs without a doubling clitic (4c), CD is strictly excluded from the latter (4d).

7 8

Bracketing with the subscribed index ‘F’ indicates the width of the focus domain; the whquestions in parentheses prior to each example illustrate the possible pragmatic contexts. It should be pointed out that the variability of CD does not pose any problems for the base generation approach we pursue here, given the fact that the use of the relevant markers is not strictly obligatory but rather depends on different factors, even in languages with affixal object agreement; see Kuchenbrandt (2009).

68 (4)

Christoph Gabriel & Esther Rinke

a. Pedro le dio una manzana a él. / *Pedro dio una manzana a él. ‘Peter gave an apple to HIM.’ b. Pedro lo vio a él. / *Pedro vio a él. ‘Peter saw HIM.’ c. Le agradezco a Usted su atención. / Agradezco a Usted su atención. ‘I thank you for your attention.’ d. *Le dedicaré a ello el siguiente capítulo. / Dedicaré a ello el siguiente capítulo. ‘I will dedicate the next chapter to this issue.’

CD with nominal objects depends on their syntactic function (direct vs. indirect object), their definiteness/indefiniteness, and their thematic role. With direct objects, CD is ungrammatical (5a), the only exception being proper names, where a doubling clitic is marginally acceptable (5b). (5)

a. *Pedro lo vio al profesor. / Pedro vio al profesor. ‘Pedro saw the professor.’ b. ?Pedro lo vio a Juan. ‘Pedro saw Juan.’

With indirect nominal objects, CD applies optionally in principle, but there is a strong tendency to realize the clitic. Some slight differences in acceptability come into play, however, given the fact that non-doubled proper names and definite noun phrases (6a, b) are generally judged to be less acceptable than un-doubled indefinite nominal objects (6c). (6)

a. Pedro le dio una manzana a Juan. / ?*Pedro dio una manzana a Juan. b. Pedro le dio una manzana al profesor. / ?*Pedro dio una manzana al profesor. c. Pedro le dio una manzana a un profesor. / ?Pedro dio una manzana a un profesor. ‘Peter gave an apple to John / the professor / a professor.’

Doubling of indirect objects is also affected by theta-roles. As shown in (7), the clitic compulsorily appears with experiencers, benefactives and inalienable possessors (see Fernández Soriano 1999: 1250). (7)

a. Le encanta el teatro a Juan / al profesor. experiencer *Encanta el teatro a Juan / al profesor. ‘John / The professor likes theatre.’ b. Le hizo los deberes a Juan / al niño. benefactive *Hizo los deberes a Juan / al niño. ‘He did the homework for John / the child.’ c. Le cortó las uñas a Juan / al niño. inalienable possessor *Cortó las uñas a Juan / al niño. ‘He cut John’s fingernails / the child’s fingernails.’

69

The Rise of Clitic Doubling in the History of Spanish

Table 1 summarizes the distribution of clitic doubling in Peninsular Spanish 9 as discussed so far:

direct object indirect object

personal pronoun

proper name

full noun phrase [+def]

full noun phrase [-def]

obl. obl.

(3) 3

* 3

* 3

obl. = CD is obligatory; 3 = CD is optional; (3) = CD is marginally acceptable; * = CD is ungrammatical Table 1: Clitic doubling in Peninsular Spanish

It is striking that the elements which preferably or even obligatorily undergo clause-internal clitic doubling are exactly the ones that are more likely to perform the role of the topic in a given sentence. In other words, these elements display a high degree of topicality according to the relevant hierarchies proposed by Givón (1976). (8) a. b. c. d.

Tendencies for topics (according to Givón 1976: 152, 160) Pronouns > Full noun phrases Agent > Benefactive > Dative > Accusative Definite > indefinite More involved participant > less involved participant

Looking back at the examples given in (4-7), the correlation between topicality and clitic doubling becomes obvious. Pronouns constitute typical topics rather than full noun phrases (8a), and they are more likely to undergo CD 10 . While doubling of strong pronouns is compulsory in all possible contexts, irrespective of their status as direct or indirect objects (see 4), the co-occurrence of a doubling clitic with a full nominal object depends on several factors, among them grammatical function (case), referential properties like definiteness, and the semantic role of the object in question. The strong tendency of doubling to apply to indirect rather than direct objects (see 5, 6) is thus mirrored in the higher degree of topicality 9

As already stated, we concentrate here on Standard Peninsular Spanish. Other varieties differ from the Castilian norm with respect to the use of both personal a and clitic doubling. One of these varieties is porteño Spanish, the prestigious urban dialect of Buenos Aires that is spoken today in large parts of Argentina. In porteño Spanish, inanimate objects can be marked with a (Dumitrescu 1998) and undergo clause-internal clitic doubling. See Parodi (1998) and Suñer (2006) concerning the exhaustive use of doubling clitics in this dialect. 10 Dejan Matić points out that in languages displaying a contrast between strong and clitic pronouns, the latter are seen as preferred topics in the sense of Lambrecht (1994: 172ff), whereas the former are preferred in contrastive contexts. Both types, however, are more topical than nominal DP-objects and thus occupy a higher position in the hierarchy mentioned above.

70

Christoph Gabriel & Esther Rinke

of dative constituents (8b). Furthermore, the obligatory doubling of indirect objects that bear an experiencer, benefactive or inalienable possessor role is covered by the fact that benefactives (8b) and, more generally, highly-involved participants (8d) are more likely to be topics than other entities. Given the fact that CLRD is a possible topicalization strategy both in medieval (Rini 1992, Fontana 1993, Eberenz 2000: 180f) and in modern Spanish, we assume that CD has evolved from CLRD. We will discuss the diachronic development of CD in more detail in the following section.

3 The diachronic development of clitic doubling In the first part of this section, we will give an overview of the literature. In 3.2, we present the results of our own empirical study, which is based on textual sources from the 15th to the 18th century. Finally, in section 3.3, we discuss the results with respect to our hypothesis. 3.1 Survey of the literature 3.1.1 The early periods It has been mentioned in the literature that resumptive clitics already appear in early texts, especially in written sources that reflect spoken language (Marcos Marín 1978, Eberenz 2000, Riiho 1988, Rini 1992), as in the following example of a clitic left dislocation structure from the Poema del mio Cid taken from Riiho (1988: 33). (9)

a. la tierra del Rey Alfonso esta noche the land of-the King Alfons this night la podemos quitar cl.ACC can.1PL leave ‘The land of King Alfons, this night, we can leave it.’

(12/13th c.)

However, such examples are rare in the earliest textual sources. According to Fontana (1993: 261), clitics and strong pronominal objects are usually found in complementary distribution in medieval texts. This is shown in (9b, Poema del mio Cid) and (9c, General Estoria), where the strong pronominal objects occur without a doubling clitic. The same is true for direct and indirect nominal objects, irrespective of their placement to the right or to the left of the verb; see (9d, e), both taken from the Estoria de España.

The Rise of Clitic Doubling in the History of Spanish

(9) b. Oyd ami albarfanez | & todos los caualleros listen to-me albarfanez &all the knights ‘You, Alvar Fañez, and the rest of the knights, listen to me’ c. & ael llamauan otrossi amosis amosis and him called.3PL also ‘And they also called him Amosis’ d. segund que a moysen =lo mandara nuestro sennor. as that to Moses=it ordered our lord ‘As our Lord ordered Moses to do it’ e. & dixo assu marido como era doliente and told.3sg her husband how was.3SG ill ‘and told her husband that she was ill’ (all examples taken from Fontana 1993: 262)

71 (12/13th c.)

(13th c.)

(13th c.)

(13th c.)

Fontana (1993) cites the following examples with resumptive clitics from the Cid (10a, b) and the General Estoria (10c, d). (10) a. Esto les demando | a ifantes this them demanded.3SG to princes de Carrion (12/13th c.) of Carrion ‘He bade the princess of Carrion to do this’ b. Gran Iantar le fazen | al buen canpeador (12/13th c.) big meal him prepare.3PL to-the good fighter ‘They are preparing a big meal for Cid, the good fighter’ c. Et mando=l israhel un dia a Josep que fuesse and bade-him israhel one day to Josep that went ueer como to see how ‘And one day Israhel bade Josep to go and see how …’ d. que assi=l fiziemos ael & assu pueblu that thus-him did.1pl to-him & to his people & asu tierra & to-his land ‘… that we did this to him, his people and his land’ (all examples taken from Fontana 1993: 263, ‘|’ indicates the caesura)

He states “that in at least three of the examples …, there are some indications suggesting that the doubled NP could be in fact a right dislocated element, hence analyzable as an adjunct in an A-bar position and possibly not derived via movement” (1993: 263). The author mentions the caesura in the first two examples taken from the Cid as one indication of right dislocation. The caesura usually indicates an intonation break in the recitation, which is also typical for the prosodic realization of right dislocation structures in modern Spanish (low intermediate phrasal boundary L-). Furthermore, he interprets the placement of the doubled constituent to the right of the adverbial expression un dia ‘one day’ in example (10c) as

72

Christoph Gabriel & Esther Rinke

an indication that the indirect object a Josep ‘to Joseph’ does not occupy its canonical vP-internal position, since the adverbial phrase most likely occupies a position adjoined to vP. Examples (10a-c) would thus correspond to the CLRD structures of modern Spanish, as exemplified in the first section (see 1b, 2b, 3a). The only example that potentially allows for an interpretation as a case of ‘true’, i.e. clause-internal doubling is (10d), where the indirect object clitic -l attaches to the preceding adverb. 11 It should be pointed out, however, that we are dealing with a coordinate structure here. Coordination would be impossible with only a clitic pronoun: *que assil fiziemos & assu pueblu (further examples are given in Rini 1992: 128ff and Eberenz 2000: 196). Although the examples in (10) probably should not be regarded as instances of CD, some studies mention the existence of doubling structures in early texts; Kuchenbrandt (2009), for example, investigates the distribution of clitic object pronouns across three medieval texts from the 12th and the 13th centuries and reports that clitic doubling occurs, though to a much lesser extent than resumptive clitics with dislocated objects. She cites 12 examples of clitic doubling out of 738 occurrences of clitic object pronouns in her texts; 12 in comparison, there are 48 cases which show a resumptive clitic together with a dislocated object. Silva-Corvalán (1984) does not explicitly distinguish between dislocated and doubled objects. She reports three examples of doubled postverbal direct objects out of 319 sentences including a direct object in the 12th century (El Cid) and 9 cases out of 59 postverbal indirect objects in the same text. Doubling thus seems to occur in a restricted number of cases, though it is far from being systematic or even obligatory in the early texts, either with nominal or pronominal objects. Some apparent cases of clitic doubling can be explained as instances of right dislocation, as we have seen with regard to Fontana’s examples in (10). The observation that resumptive clitic pronouns in medieval Spanish almost exclusively appear in dislocation structures is confirmed by Eberenz (2000), who points out that until the end of the middle ages, doubling of postverbal nominal constituents usually indicated the status of the relevant information as given (aforementioned or inferable). According to him, doubling occurs as a discourse strategy, marking the doubled entity as a topic (“un evidente motivo de estrategia discursiva: el concepto al que se 11 Given the fact that the direct object pronoun lo usually does not undergo elision, -l must be interpreted as an indirect object clitic which doubles the strong pronoun el. Note that this implies the assumption of a direct null object pronoun in the structure. 12 Two cases concern a strong pronominal direct object and two cases a strong pronominal indirect object. Four examples include a nominal direct object and four examples a nominal indirect object.

The Rise of Clitic Doubling in the History of Spanish

73

refiere el sustantivo en cuestión ya ha sido evocado en una frase anterior”, 2000: 191). 3.1.2 Further diachronic development Resumptive clitics became more and more frequent in the course of time, especially during the 16th century. Company-Company (2003: 230) reports that in her data, the percentage of doubled nominal indirect objects increases from 6% (13th c.) and 19% (16th) up until 40% (18th) and even 83% (20th). However, no distinction is made between pre or post-verbal objects, and the proportion of (left) dislocation structures included in the data remains unclear. The data analyzed by Fontana (1993) only contain two attestations of doubling in the 13th century (6.6%) and two occurrences in the 15th century (10%). In his 16th century data, however, the number of tokens jumps to an overall total of 7 out of 26 (26%). Fontana assumes that these numbers, although too small to be used reliably to substantiate any claim, may indicate a diachronic tendency towards increased doubling, which, according to him, becomes the norm in the 17th and 18th centuries. 13 Saéz Rivera (2003) compares different editions of French-Spanish phrase books from the 17th and 18th century and detects a considerable increase in CD with strong pronouns during the 18th century. This finding at least partly contradicts Fontana’s (1993) claim that CD became compulsory as early as the 17th and 18th century. With respect to the explanation of the diachronic change, Fontana (1993) points out that Spanish has evolved from a system in which strong pronominal and nominal objects are found in complementary distribution with clitic pronouns into a system which is characterized by the obligatory co-occurrence of the clitic with pronominal strong (indirect and direct) objects and (in most dialects) with the indirect nominal objects of ditransitive verbs. He interprets this change in terms of a diachronic evolution of Old Spanish second position clitics 14 into verbal affixes that enter into an object agreement relationship with their agreement positions. One argument in favor of this explanation is that the increase in doubling in the texts goes hand in hand with an abrupt decrease in interpolation structures at the end of the 15th century.

13 Fontana also notes a tendency to double even direct nominal objects in the 16th century. We will disregard these special cases for the moment. 14 The author characterizes these so-called 2P-clitics as prosodically deficient phrasal constituents that are displaced from their internal argument positions.

74

Christoph Gabriel & Esther Rinke

We agree with Fontana (1993) and many others regarding the assumption that CD should be analyzed as an agreement phenomenon. As stated in the previous section, this analysis implies that clitics are syntactically interpreted as part of the agreement system, irrespective of their exact morphological status. The development of clitic object pronouns must therefore be regarded as a prerequisite for the emergence of CD. According to Fontana (1993), this condition is met at least at the end of the 15th century. The different contexts in which CD occurs in Modern Spanish remains an open question. As shown in the previous section, the relevant contexts in which doubling is theoretically possible vary with respect to their obligatoriness: CD is obligatory with pronominal objects and experiencer arguments of verbs like parecer (‘to seem’), gustar (‘to like, to please’) or encantar (‘to delight’). Additionally, it is strongly preferred with nominal indirect objects of ditransitive verbs. As noted in the previous sections, these elements are typically involved in topicalization – although they are not interpreted as topics in the CD structures – and they mirror different degrees of topicality. Silva-Corvalán interprets the fact that certain factors such as animacy, definiteness or recency of reference in the discourse are relevant in determining the occurrence of a co-referential clitic “as a manifestation of O-V agreement in a process of diffusion which has spread first to NPs high in topicality and appears to be gradually extending to NPs which are lower in topicality, i.e. to postverbal, non-human DOs” (1984: 568). We agree with her regarding her statement that resumptive clitics appear first with proto-typical topics and only later with less topical elements. However, our interpretation of this finding differs from Silva-Corvalán’s insofar as we strictly differentiate between dislocation and doubling structures because of the fundamental prosodic, syntactic, and information structural differences between the two types of constructions, already outlined in section 2.1, above. We assume that the diachronic tendency from more to less topical elements co-occurring with a resumptive clitic may be interpreted in terms of a reanalysis of a topicalization strategy (clitic right dislocation) as a doubling structure. We will come back to these issues in more detail and discuss further arguments after having provided the results of our empirical investigation in the following section. 3.2 Data base Our data base consists of 13 texts from the 15th up until the 18th century; references are given in the bibliography. Electronic versions of the texts were taken from the electronic library of the Fundación Biblioteca Virtual

75

The Rise of Clitic Doubling in the History of Spanish

Miguel de Cervantes, accessible through http://www.cervantesvirtual.com. Since doubling may first occur in texts which reflect spoken language to a certain extent, we selected texts which are as close as possible to spoken language, although we are well aware of the fact that there is never a oneto-one ratio between oral and written sources, even if this is intended. Our data base primarily consists of speeches, letters, and drama in prose (texts in verse were explicitly excluded). For the purpose of our study, we extracted all instances of a) strong pronominal direct and indirect objects in post-verbal position, b) full nominal indirect post-verbal objects and c) experiencer arguments. The data were coded with respect to their co-occurrence with a clitic pronoun and with respect to the features person (pronouns), number, syntactic function, definite/indefinite article, and proper name and [r animate]. 3.3 Quantitative results and generalizations The quantitative analysis reveals that doubling steadily increases in the course of diachronic development. The following table summarizes the quantitative results of our empirical study. century 15th 16th 17th 18th

pronominal objects All 42 100% 14 100% 38 100% 80 100%

+CD 7 17% 11 79% 27 71% 71 89%

-CD 35 83% 3 21% 11 29% 9 11%

indirect full nominal objects all 105 100% 40 100% 62 100% 62 100%

+CD 1 1% 4 10% 13 21% 11 18%

-CD 104 99% 36 90% 49 79% 51 82%

experiencer arguments all 10 100% 7 100% 14 100% 1 100%

+CD 3 30% 6 86% 6 43% 1 100 %

-CD 7 70% 1 14% 8 57% 0 0%

Table 2: Distribution of strong pronominal (direct and indirect) objects, indirect full nominal

objects and experiencer arguments with and without a clitic (15th - 18th centuries)

Diagram 1 illustrates the diachronic development of clitic doubling with pronominal and indirect nominal DP-objects (we ignore experiencer arguments because the total number of tokens is too low).

76

Christoph Gabriel & Esther Rinke

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

Pronominal objects Indirect nominal objects

15thCent.

16thCent.

17thCent.

18thCent.

Diagram 1: Diachronic development of clitic doubling with pronominal and

indirect nominal objects

Concerning our 15th century data, we found only one instance of a resumptive pronoun with a postverbal indirect nominal object (11a). In all other cases, these constituents occur without a co-referential clitic pronoun, as shown in (11b):

(11) a. y no le dé Dios más salud and neg cl.dat may-give God more sanity al alma to-the soul ‘and God should not give more sanity to the soul’ b. Et dio a la muger el anillo and gave.3sg to the woman the ring del olvido of-the oblivion ‘And he gave the woman the ring of oblivion.’

(15th c.)

(15th c.)

Strong pronominal objects and experiencer arguments also occur predominantly in a complementary distribution with clitics, as shown in example (12a, b). There are, however, some cases of doubling. Two relevant examples are given in (12c, d). In (12c), the strong pronoun is an experiencer argument. (12) a. Soy çierto que non fallesçerá a ty be.1SG sure that neg will-lack.3SG to you por lo que dieres a my by what will-give.2SG to me ‘I am sure that you will not miss it because you give it to me.’ b. Quien aplaze a los otros más que a who appeals to the others more than to

(15th c.)

The Rise of Clitic Doubling in the History of Spanish

77

sí oneself (15th c.) ‘Who appeals more to the others than to oneself?’ c. E non le paresçía mucho a él and neg cl.DAT seemed.3SG a lot to him proveer a los clérigos e scholares de su provide to the priests and disciples of his Arçobispado (15th c.) archbishopric ‘And it didn’t seem much to him to supply the priests and the scholars of his archbishopric.’ d. No les está menos bien el perdón neg cl.DAT be.3SG less good the mercy a los poderosos cuando son deservidos, to the powerful when be.3PL not obeyed que a los pequeños la venganza than to the small the vengeance cuando son injuriados (15th c.) when be.3Pl insulted ‘Mercy does not beseem powerful people when they are disobeyed less than vengeance small people when they are insulted.’

From the 16th century on, pronominal objects are systematically doubled. It is also in the 16th century that experiencer arguments are doubled in most cases (13a). There are, however, some instances of experiencer arguments which are not doubled – especially in the archaizing 17th century texts, see (13b). The overall picture remains therefore somewhat unclear due to the low number of tokens. (13) a. que a Vuestra Señoría no le pareciere bien that to your lordship NEG cl.DAT will-seem.3SG well ‘that it may not seem well to your lordship’ b. empresa que a Horacio pareció dificultosa undertaking that to Horacio seemed hard ‘an undertaking that seemed hard to Horacio’

(16th c.)

(17th c.)

In contrast to strong pronominal objects, nominal indirect objects start being doubled as late as the 17th and 18th century. Doubling amounts to a percentage of more or less 20% in these contexts. Two examples from the 18th century, one with a co-referential clitic and one without, are given in (14). (14) a. aquella carta que está sobre la mesa, that letter that is on the table dásela al mozo de la posada give-cl.dat-cl.acc to-the boy of the hostel ‘That letter that is on the table, give it to the boy of the hostel.’

(18th c.)

78

Christoph Gabriel & Esther Rinke

b. yo escribiré al señorAznar para que te dé I will-write to Mister Aznar in order that cl.DAT give.3SG cien doblones de orden mía (18th c.) hundred doubloons of order mine ‘I will write to Mister Aznar so that he gives one hundred doubloons to you by my order.’

In the next section, we will come back to our hypothesis that clitic doubling has evolved from dislocation. 3.4 Discussion In the previous section, we have seen that the co-occurrence of a clitic and a strong pronominal object, a nominal indirect object or an experiencer argument is virtually absent in our 15th century data. From the 16th century on, doubling becomes an option with increasing frequency. We have also seen that the first elements to be doubled are strong pronouns and probably experiencer arguments and that the doubling of indirect nominal objects comes into play only later. There is, however, no indication that indirect nominal objects on the one hand and indirect strong pronominal objects on the other would differ with respect to their internal syntactic structure or regarding their prosodic properties. If the doubling of indirect pronominal objects is clearly the preferred option in the 16th century, how can it be explained that it is very rarely attested with nominal objects? As already mentioned in the first section, the two types of objects differ with respect to their suitability to fulfill certain information structural functions. Recall in this context the topicality hierarchy given in section 2.2, according to which a) pronouns are more likely to be topics than noun phrases and b) more involved participants are generally more topical than less involved participants. The problem with this observation is that it is not clear why elements with a higher degree of topicality should diachronically be the first elements which appear in CD constructions, since it is precisely in these constructions that they are not interpreted as topics. We argue that the different degree of topicality and the possibility of undergoing clause-internal doubling can easily be reconciled when we assume that CD constructions result from a syntactic reanalysis of clitic (right) dislocation structures. If this interpretation is true, which additional facts could support this hypothesis? One observation in favor of our hypothesis is the fact that clitic (left and right) dislocation structures appear diachronically before CD occurs. The 12/13th century examples (10a, b) indicate that in the Poema del Mio Cid, a co-referential clitic pronoun occurs almost exclusively when the nominal object is separated by a caesura, i.e. in contexts comparable to the

The Rise of Clitic Doubling in the History of Spanish

79

prosodic realization of CLRD structures in contemporary Spanish. However, in our 15th century texts, clear indications of CLRD are difficult to find, because topicalization without a resumptive clitic was still an option until the 16th century, occurring predominantly in the 15th century texts. One indication that dislocation structures might be older than CD is provided by the 16th century texts, where we systematically find clitic left dislocation structures as in (15a, b). (15) a. al pobre y que no sabe tanto, to-the poor and that neg know so much envíanle a que tome el hábito sent.3PL.CL.DAT to that take.3SG the habit en San Francisco o en Santo Domingo (16th c.) in San Francisco or in Santo Domingo ‘The poor man who does not know a lot, they sent him to San Francisco or to Santo Domingo (in order to join the congregation).’ le ahorran en esto (16th c.) b. y lo que al Rey and what to-the king CL.DAT spare.3Pl in that ‘and what they spare the king with this’

The reinterpretation of right dislocation as doubling may become plausible when we assume that right dislocation is more marked and less economic than CD (see below for details). It should be pointed out, however, that this statement does not entail the idea that marked structures necessarily disappear in favor of less marked ones. They may rather serve, under appropriate conditions, as a starting point for the emergence of new structures, which constitute the less marked option and are at the same time characterized by a more economic syntactic derivation. We exemplify this reanalysis here with an example taken from Amadís de Gaula (15th century): (16) a. Y así se despidió el rey Períon del hermitaño y tornóse a las tiendas en que su muger y compaña avía dexado […] Estando en su lecho en gran plazer díxole a la Reina lo que los maestros avían declarado de su sueño. ‘And so the King Perión took farewell of the eremite and returned to the tents, where he had left his wife and companion […] Later on, when he rested on his bed with great de light, he told the queen what the wise men had said about his dream.’ (15th c. [1482-1492, ed. 1508], Garci Rodríguez de Montalvo, Amadís de Gaula, quota tion following Eberenz 2000: 192). b. su mugeri … DÍxolei ) [Topic a la reina]) [CP lo que …] | | HH‘His wife … He told her, i.e. to the queen, what the wise men had said ’

80

Christoph Gabriel & Esther Rinke

c. su muger …

díxolei a la REInai )[CP lo que …] | H‘His wife … He told the queen what the wise men had said …’

Given the fact that the context is pretty clear (la reina should easily be interpreted as the king’s wife, su muger), one is likely to interpret the example as a simple sentence with a clause-internally doubled indirect object rather than a segmented structure with a right dislocated object that would function here as some sort of insertion (‘he told her, i.e. the queen’). With respect to the internal syntactic structure, it is arguably more economic to have a simple core sentence with the object in its vP-internal base position than a complex structure that involves the vP-internal generation of a phonetically empty pronoun and a dislocated object that is adjoined in a structurally higher position. Note that this does not depend on the exact analysis of CLRD structures: Irrespective of the object’s status as an element base-generated in its target position (as argued by Suñer 2006) or as a constituent moved to some specifier position (following the different accounts mentioned in footnote 6), the derivation of a CLRD construction involves more syntactic operations and is hence less economic than the derivation of a simplex sentence with a vP-internal object. With respect to prosody, right dislocation involves an additional phrasal boundary; the dislocated material usually undergoes deaccentuation, though not in all cases. The prosodic identification of a right dislocated XP as such thus relies exclusively on the perception of the intermediate phrasal boundary (L-). On the basis of its rather vague character and given the fact that the right edge of an utterance is associated with a (linear) focus position (Lambrecht 1994, Astruc 2005: 60), the linear ordering ‘[vP Cli+V proi], objecti’ can easily undergo re-interpretation as a construction with a vP-internal object, i.e. as [vP Cli+V objecti]. This non-dislocated object can then be construed as part of the focus domain. The latter point is of direct relevance for the question of to what extent information structure may be an important factor in language change. In this context, we argue that right dislocation is also the marked option with respect to information packaging in the sentence, since the right edge of the sentence is usually reserved for focused material, at least in Spanish (e.g. Lambrecht 1994, see also Zubizarreta 1998, Gabriel 2007 with special reference to contemporary Spanish). Another definition of markedness relies on the number of features a given element is endowed with. From this point of view, topicalization can also be regarded as more marked compared to focalization, as topicalized constituents are always specific, whereas focused constituents are unmarked with respect to the feature [r specific].

The Rise of Clitic Doubling in the History of Spanish

81

A last aspect that should be addressed is the morphosyntactic status of object clitics in Spanish. In section 2.1, we already pointed out that we interpret clause-internal clitic doubling as an agreement phenomenon. This assumption, however, does not necessarily entail an interpretation of the Spanish object clitics as verbal affixes on a par with the verbal (subject) agreement suffixes. Strong evidence against an affix analysis of object clitics comes from prosody: ‘True’ affixes, such as the nominal plural marker, are integrated into the domain of the prosodic word, as can be seen in the accent shift that regularly occurs in words with antepenultimate stress (palabras esdrújulas, 17a). In contrast, a clitic attached to the verb does not affect the position of the word accent, yielding so-called palabras sobresdrújulas, which exclusively emerge from verb forms plus enclitic pronouns (17b). 15 Interestingly, there are some tendencies to integrate pronominal clitics into the prosodic word in Argentinean porteño, a Spanish variety which is characterized by an intensified and more regularized use of doubling object clitics (Parodi 1988, Suñer 2006). In example (17c), taken from the Argentinean 19th century short story El casamiento de Laucha 16 , the version with accent shift (preguntaseló), indicating the integration of the clitic into the prosodic word, occurs along with the nonshifted version (preguntáselo): 17 (17) a. régimen ‘regime’ (singular), regímenes ‘regimes’ (plural) b. ¡Pregunta! ¡Pregúntale! ¡Pregúntaselo! ‘Ask!’ / ‘Ask him!’ / ‘Ask it to him/her!’ c. ¡Preguntáselo al cura! ‘Ask it to the priest!’ ¡Andá, preguntaseló al nuevo! ‘Go and ask it to the new one!’

Of course, a further development of object clitics to verbal affixes cannot be excluded. At least for standard Peninsular Spanish, however, their status as affixes does not yet seem to be on par with ‘true’ affixes such as the morphological markers of plural or subject agreement.

15 See Kuchenbrandt (2009) concerning the general properties of the prosodification of object clitics in Peninsular Spanish. 16 The Casamiento de Laucha, published in Buenos Aires in the late 19th century by Roberto Payró (1867-1928), is typical of the so-called literatura gauchesca, a literary genre that is characterized by large dialogue passages reflecting colloquial language use and making exhaustive use of argentinismos. 17 Note that in porteño Spanish, imperative forms have final instead of penultimate stress (i.e. ¡preguntá! instead of Peninsular Spanish ¡pregunta!).

82

Christoph Gabriel & Esther Rinke

4 Concluding remarks The goal of the present paper was to highlight the role of information structure in the diachronic development of clitic doubling (CD) in Peninsular Spanish. We pointed out that CD and clitic right dislocation structures (CLRD) differ with respect to their prosodic and syntactic properties as well as regarding information packaging, i.e. concerning the possibility of the doubled object constituent being interpreted as belonging to the focus domain. It was argued that if an object that co-occurs with a coreferential clitic cannot be interpreted as a dislocated topic, but rather forms part of the focus domain, we are dealing with an instance of clauseinternal doubling and hence with a CD construction in the modern sense. This change in information structural categories then acts as an indicator for syntactic change, which should be interpreted in terms of a reanalysis of the former CLRD structure as a CD construction, the latter being less marked with respect to prosody and information packaging and more economic regarding derivational economy. In order to support our claim, we scrutinized data from the 15th up until the 18th century and showed that the increase in CD is in accordance with the topicality hierarchies proposed by Givón (1976), insofar as the first entities to be doubled, i.e. strong pronouns and experiencer arguments, are exactly the ones that display a high degree of topicality. Finally, we argued that the Spanish object clitics form part of the agreement system. This assumption, however, does not entail an affix interpretation of the elements in question.

References Primary sources Acosta, J. de (1539-1600): Escritos menores. Edición digital a partir de Obras del P. José de Acosta. Madrid: Atlas. 1954, 250-386. Calderón de la Barca, P. (1600-1681): Andrómeda y Perseo. Edición digital a partir de la edición de J.M. Ruano de la Haza. Pamplona: Universidad de Navarra; Kassel: Reichenberger, 1995. Cruz, R. de la (1731-1794): El hospital de la moda. Edición digital a partir de la de Emilio Cotarelo y Mori, Sainetes de Don Ramón de la Cruz en su mayoría inéditos. Madrid: Bailly-Bailliere, 1915-1928 (Nueva Biblioteca de Autores Españoles; 23 y 26). Cruz, R. de la (1731-1794): Los dos libritos. Edición digital basada en la edición de Valencia: Yernos de J. Esteban, 1813.

The Rise of Clitic Doubling in the History of Spanish

83

Fernández de Moratín, L. (1760-1828): El sí de las niñas. Reproducción digital a partir del microfilm de la edición original de Obras dramáticas y líricas de D. Leandro Fernández de Moratín. Tomo II. París: Augusto Bobée, 1825, 201338. Localización: Biblioteca Nacional (España). Sig. R. 2571/73. Fernández de Moratín, L. (1760-1828): La comedia nueva. Reproducción digital a partir del microfilm de la edición original de Obras dramáticas y líricas de D. Leandro Fernández de Moratín. Tomo I. París: Augusto Bobée. 1825, 161249. Localización: Biblioteca Nacional (España). Sig. R. 2571/73. Fernández, L. (1474-1541): Auto de la Pasión. Edición digital a partir de la edición facsímil de la de Salamanca. 1514, realizada por Emilio Cotarelo y Mori (Madrid, Real Academia Española, 1929) y cotejada con la edición crítica de Mª Josefa Canellada (Madrid, Castalia, 1976, 211-237). Martínez de Toledo, A. (1398-1468): Vida de Sanct Isidoro. Edición digital basada en la de Vidas de San Ildefonso y San Isidoro, edición, prólogo y notas de José Madoz y Moleres, (S.I.) Madrid, Espasa-Calpe, 1962. Pulgar, F. del (1430-1493): Letras. Edición digital a partir de la edición de Jesús Rodríguez Bordona, Madrid, Espasa-Calpe, 1958. Rueda, L. de (1505?-1565): Comedia Armelina. Reproducción digital a partir de “Las primeras dos elegantes y graciosas comedias del excelente poeta y representante Lope de Rueda” en Las quatro comedias y dos Coloquios pastoriles, En Valencia, en casa de Ioan Mey, 1567, fols.34 y ss. Localización: Biblioteca Nacional. Sig. R. 12055. Saavedra Fajardo, D. de (1584-1648): Idea de un príncipe político cristiano representado en cien empresas. Edición digital a partir de Empresas políticas, tomos I - II, Madrid, Editora Nacional, 1976. San Pedro, D. de (1437-1498): Cárcel de amor. Edición digital basada en la de Sevilla, Cuatro compañeros alemanes [=Pablo de Colonia.Juan Pegnitzer, Magno Herbst y Tomás Glockner] (3 marzo, 1492). Localización: Biblioteca Nacional (España). Sig. I-2134. Villena, E. de Aragón, Marqués de (1384-1434): Tratado de Astrología. Edición digital a partir de Obras Completas. Tomo I. Madrid: Tuner 1994, 400-556. Localización: Biblioteca General de la Universidad de Alicante. Sig. FL LI134.2/VIL/ENR V. 1.

Scientific Literature Astruc, L. (2005): The intonation of sentence-external elements in Catalan and English. Ph.D. Diss. Cambridge: Univ. of Cambridge.

84

Christoph Gabriel & Esther Rinke

Barbosa, P. (2000): Clitics. A Window into the Null Subject Property. In: J. Costa (ed.), Portuguese Syntax. New Comparative Studies. Oxford: Oxford University Press, 31-93. Belletti, A. (2005): Extended Doubling and the VP Periphery. Probus 17, 1-35. Bleam, T. M. (1999): Leísta Spanish and the Syntax of Clitic Doubling. Ph.D. Diss. Delaware: University of Delaware. Cecchetto, C. (1999): A comparative analysis of left and right dislocation in Romance. Studia Linguistica 53, 40-67. Chafe, W. (1976): Givenness, contrastiveness, definiteness, subjects, topics, and point of view. In: C. Li (ed.), Subject and topic, 25-55. New York: Academic Press. Company-Company, C. (2003): Transitivity and Grammaticalization of Object. The Diachronic Struggle of Direct and Indirect Object in Spanish. In: G. Fiorentino (ed.), Romance Objects. Transitivity in Romance Languages, 217260. Berlin: Mouton de Gruyter. Dumitrescu, D (1998): ‘A’ personal, duplicación clítica y marcadez: Español porteño vs. español madrileño. In: A. M. Ward (ed.), Actas del XII Congreso de la Asociación internacional de hispanistas. 21-26 de agosto de 1995, 140-152. Birmingham. Tomo 1: Medieval y lingüística. Birmingham: Department of Hispanic Studies. Eberenz, R. (2000): El español en el otoño de la edad media. Madrid: Gredos. Fernández Soriano, O. (1999): El pronombre personal. Formas y distribuciones. Pronombres átonos y tónicos. In: I. Bosque & V. Demonte (eds.), Gramática descriptiva de la lengua española, 1209-1273. Tomo 1: Sintaxis básica de las clases de palabras. Madrid: Espasa. Fontana, J. (1993): Phrase Structure and the Syntax of Clitics in the History of Spanish. Ph.D. Diss. Pennsylvania: University of Pennsylvania. Gabriel, C. (2007): Fokus im Spannungsfeld von Phonologie und Syntax. Eine Studie zum Spanischen. Frankfurt a.M.: Vervuert. Givón, T. (1976): Topic, pronoun, and grammatical agreement. In: C. N. Li (ed.), Subject and Topic, 149-188New York: Academic Press. Hualde, J. I. (2003): El modelo métrico y autosegmental. In: P. Prieto (ed.), Teorías de la entonación, 155-184 (Ariel lingüística). Barcelona: Editorial Ariel. Kaiser, G. A. (1992): Die klitischen Personalpronomina im Französischen und Portugiesischen. Eine synchronische und diachronische Analyse. Frankfurt a.M.: Vervuert. Kayne, R. S. (1975): French Syntax. The Transformational Cycle. Cambridge, MA: MIT Press. Kayne, R. S. (1994): The Antisymmetry of Syntax. Cambridge, MA: MIT Press.

The Rise of Clitic Doubling in the History of Spanish

85

Kuchenbrandt, I. (2009): Prosodische Aspekte in der Entwicklung der spanischen und französischen Klitika. Ph.D. dissertation. Hamburg: University of Hamburg. [http://www.sub.uni-hamburg.de/opus/volltexte/2009/4070/]. Lambrecht, K. (1994): Information structure and sentence form. A theory of topic, focus, and the mental representations of discourse referents. Cambridge: Cambridge University Press. López, L. (2003): Steps for a well-adjusted dislocation. Studia Linguistica 57, 193231. López, L. (2006): Ranking the LCA. Ms., Paper presented at the DEAL Workshop ZAS Berlin 2006. López, L. (2009): A derivational syntax for information structure. Oxford: Oxford University Press. Marcos Marín, F. (1978): Estudios sobres los pronombres. Madrid: Gredos. Manzini, M. R. (2003): Syntactic approaches to cliticization. In: L. Cheng and R. Sybesma (eds.), The second glot international state-of-the-article book. The latest in linguistics, 367-387. Berlin: Mouton de Gruyter. Parodi, T. (1998): Aspects of clitic doubling and clitic clusters in Spanish. In: T. Parodi et al. (eds.), Models of Inflection, 85-102. Tübingen: Niemeyer. Riiho, T. (1988): La redundancia pronominal en el iberoromance medieval. Tübingen: Niemeyer. Rini, J. (1992): Motives for Linguistic Change in the Formation of the Spanish Object pronouns. Newark: Juan de la Cuesta. Rizzi, L. (1997): The Fine Structure of the Left Periphery. In: L. M. Haegeman (ed.), Elements of Grammar. Handbook in Generative Syntax, 281-337. Dordrecht: Kluwer. Ronat, M. (1979): Pronoms topiques et pronoms distinctifs. Langue Française 44, 106-128. Saéz Rivera, D. M. (2003): La duplicación de clíticos en la obra de Francisco Sobrino. Res Diachronicae Virtual 3, 332-342. Silva-Corvalán, C. (1984): Semantic and pragmatic factors in syntactic change. In: J. Fisiak (ed.), Historical Syntax, 555-573. Berlin and New York: Mouton de Gruyter. Strozer, J. (1976): Clitics in Spanish. Ph.D. dissertation, Los Angeles: UCLA. Suñer, M. (1988): The role of agreement in clitic-doubled constructions. Natural Language and Linguistic Theory 6, 391-434. Suñer, M. (2006): Left dislocations with and without epithets. Probus 18, 127158. Vallduví, E. (1993): Information packaging: A survey. University of Edinburgh: Centre for Cognitive Science. Human Communication Research Centre.

86

Christoph Gabriel & Esther Rinke

Villalba, X. (2000): The Syntax of Sentence Periphery. Ph.D. Diss. Barcelona: Universitat Autònoma de Barcelona. Zubizarreta, M. L. (1998): Prosody, Focus, and Word Order. Cambridge, MA: MIT Press.

Cue-based Acquisition and Information Structure Drift in Diachronic Language Development Marit Westergaard (Tromsø – CASTL)

1 Introduction Within a model of language acquisition and change that recognizes the importance of cues and what is referred to as ‘micro-cues’ in the primary linguistic data (PLD) that children are exposed to, this paper argues that patterns of information structure may be a factor contributing to word order change. The examples discussed are taken from mixed grammars which allow two different subject or object positions, in both historical and present-day data, mainly from English and the North Germanic languages. In these grammars children are exposed to the cue for a particular word order in only some of the relevant contexts in the input. The paper explores how children deal with this kind of input and how information structure may affect the frequency of a particular word order in the PLD, which in turn may lead to diachronic language change. The following examples provide evidence of such mixed word order grammars from the history of English. In (1) and (2) we see that Old English (OE) allowed both OV and VO, while (3) and (4) illustrate that also V2 and non-V2 word orders were attested during the same time period: (1)

(2)

(3)

(4)

We ne magan eow neadian (OE) we neg can you constrain ‘We cannot constrain you. . . ’ (Pintzuk 2005: 129) Se wolde gelytlian þone lyfigendan hælend he would diminish the living saviour ‘He would diminish the living saviour . . . ’ (Pintzuk 2005: 117) On his dagum sende Gregorius us fulluht. in his days sent Gregorius us baptism ‘In his time, Gregory sent us Christianity.’ Hiora untrymnesse he sceal ðrowian on his heortan. their weakness he shall atone in his heart ‘He shall atone in his heart for their weakness.’ (Haeberli 2002: 88-90)

88

Marit Westergaard

The choice of word order in these mixed systems is often at least partly dependent on information structure, in such a way that one of the subject and object positions is preferred for discourse given elements (typically pronouns) and the other for informationally new or focused elements (typically heavier elements such as full DPs or clauses). It is often claimed that subjects tend to be given information and objects more often new. Investigating samples of natural spoken language from Norwegian and English child-directed speech, this paper shows that this is indeed the case in the input to children. While e.g. approximately 90% of all subjects are pronouns, the situation is reversed for objects, which are expressed by pronouns only about 20-35% of the time. This means that in systems allowing two subject positions, the one preferred for discourse given (often pronominal) subjects should be naturally more frequent in the E-language that constitutes the input to children in the acquisition process. 1 Conversely, the position for informationally given objects should be relatively infrequent. Investigating some child language data, the paper shows that children easily acquire mixed word orders. Thus, it is argued that they are sensitive to syntactic micro-cues and patterns of information stucture from early on. Nevertheless, word orders with an extremely low input frequency are argued to be vulnerable to change, in that children may simply ignore the corresponding micro-cues and develop a grammar with a default word order. This syntactic freezing of certain word orders is the result of what is referred to as Information Structure Drift. While other factors (e.g. dialect contact) may reverse such changes, the paper shows that in the examples at hand, the direction of the historical change indeed corresponds to the prediction of this Information Structure Drift. The paper is organized as follows: In the next section I outline the theoretical background for this study, which is an approach to language acquisition and change that is based on the existence of certain micro-cues in children’s I-language grammars, expressed by certain triggers in the input. These micro-cues may account for the word order variation that is found in many languages and the seemingly gradual change that is sometimes attested in historical data. In section 3 I present some examples of word order variation in present-day grammars, involving constructions that allow two subject or object positions, and argue that this variation is dependent on information structure. In section 4 I provide similar data from historical languages. Section 5 presents child language data, indicating that patterns of information structure are easily acquired. I then turn 1

See Chomsky (1986) or Lightfoot (2006) for a discussion of the distinction between Elanguage and I-language, ‘externalized’ vs. ‘internalized’ language.

Cue-based Acquisition and Information Structure Drift

89

to an investigation of some child-directed speech in section 6, showing that subjects typically express given information while objects normally convey new information. Section 7 discusses the concept of Information Structure Drift, arguing that this may be responsible for frequency shifts in the input data. Section 8 contains a brief summary and conclusion.

2 Theoretical Background The theoretical foundation for this study is a cue-based model of language acquisition and change, inspired by Lightfoot (1999, 2006), but also in important ways different from this approach. According to Lightfoot’s model, children scan the PLD for designated cues, which are abstract pieces of I-language structure. This means that cues are not surface strings, but small syntactic trees that are produced in children’s I-language grammars on exposure to certain triggers in the input. Examples of such cues in Lightfoot’s model are provided in (5) and (6) for an OV and a V2 grammar respectively: (5) (6)

Cue for an OV grammar: Cue for a V2 grammar:

VP

[DP V] [XP V]

CP

The formulation of the cues in (5) and (6) is in line with much traditional work within the generative framework which assumes the existence of major word order parameters, e.g. +/- head final or +/-V2. That is, a grammar should choose only one parameter setting, and exceptions to this are considered to simply be E-language phenomena. One important reason for postulating the existence of major parameters was to explain the ease and speed of language acquisition, under the assumption that there is poverty of stimulus in the input: Children are assumed to set parameters early (see e.g. Wexler 1999), and the relevant word order should then fall into place in various contexts, with very little or perhaps no exposure in certain cases. Likewise, Lightfoot (1999) also argues that there must be a UG requirement that cues are obligatory in children’s I-language grammars. The reason for this is to prevent child languages with mixed word orders, since, due to the lack of negative evidence in acquisition, children are assumed not to be able to recover from such grammars. The cue-based model also argues that cues must be relatively robustly expressed in the PLD for children to acquire them. Structures with a low input frequency, on the other hand, could be ignored by children in the acquisition process and may as a consequence disappear from the I-language grammar of the next generation. This means that word order variation in the input to children must

90

Marit Westergaard

be considered to be part of the E-language only, and frequencies of the “irrelevant” word orders may vary over time. Recent work on language acquisition (e.g. Westergaard 2008a) has shown that young children easily learn constructions that have a very low frequency in the input, even exceptions to major, general rules. That is, children seem to be sensitive to relatively minor categories that distinguish between different word orders at a very early stage. For example, children acquiring English have no problem with word order variation that is dependent on clause types (questions vs. declaratives) or different classes of verbs (auxiliaries vs. lexical verbs) in the acquisition of verb movement. According to Radford (1992) and Roeper (1999, 2007), once English children learn subject-auxiliary inversion in questions, it is not overgeneralized to other clause types or verb types. Nor do children acquiring certain Norwegian dialects experience any problem distinguishing subcategories of wh-elements in the acquisition of word order in wh-questions, long wh-elements requiring V2, while short ones allow either V2 or non-V2 (see e.g. Westergaard 2008a). Furthermore, Rodina (2008) shows that Russian children distinguish between different noun classes from early on with respect to gender assignment in cases where there is a mismatch between morphological and semantic cues. In response to these findings, Westergaard (2008a, 2009a) has developed a model of ‘micro-cues’ in acquisition, which argues that the cues that children search for are much smaller entities than what was previously assumed. That is, children do not only pay attention to major categories such as Nouns and Verbs, but easily distinguish between linguistically relevant classes of categories, such as auxiliaries vs. lexical verbs, various sub-classes of nouns, different clause types, or long vs. short wh-words (representing strong and clitic-like elements respectively, see Cardinaletti and Starke 1999). Examples of these micro-cues are provided in (7) and (8), see also Westergaard (2009a). In a Split-CP model, the topmost head in the CP domain in Rizzi’s (1997) system, the ForceP, is divided into different heads depending on clause type, so that declaratives are Top(ic)Ps, wh-questions are Int(errogative)Ps, etc. Thus, while children learning other Germanic languages produce the cue formulated in (7) in their I-language grammars, a trigger for this is not provided in the PLD that English children are exposed to. Furthermore, the cue for verb movement in wh-questions must be even more specific, as illustrated by (8), which is a formulation of what children learning present-day English are exposed to. That is, (8) specifies that only elements that are found in I, i.e. auxiliaries and be, undergo verb movement to the relevant head: (7) (8)

Cue for V2 in declaratives: Cue for V2 in wh-questions (English):

[XP Top° V...] [(wh) Int° I...]

TopP IntP

Cue-based Acquisition and Information Structure Drift

91

In contrast with Lightfoot’s original cue-based approach, word order variation in the present model that is dependent on linguistically relevant subcategories is considered to be part of the I-language, and children are argued to be sensitive to these minor distinctions early. Nevertheless, it will be assumed that micro-cues that are extremely infrequent in typical childdirected speech may be vulnerable to change. This model of language acquisition has consequences for how one views language change. It is no longer necessary to argue that gradualism in historical data is simply an E-language phenomenon (as in e.g. Lightfoot 2006) and that real language change constitutes ‘catastrophes’, i.e. major I-language differences between two generations of speakers. In this extended version of the cue-based model, gradualism can be seen as the result of many ‘small catastrophes’, reflecting change or loss of several I-language micro-cues in succession. In the next section I consider some word order variation in presentday languages and argue that information structure plays a certain role in the choice of one over the other. Section 4 provides similar data from historical languages.

3 Word Order Variation and Information Structure Although it has been common within generative syntactic theory to assume that grammars may only consist of major, global parameters to be learnable grammars, it is not difficult to find languages which allow a considerable amount of word order variation. English, for example, has V-to-I movement only for a certain class of verbs (auxiliaries have/be plus main verb be), which may thus (like modals) appear in front of an adverbial such as never, illustrated in (9 and (10). English also displays a kind of V2 word order (subject-auxiliary inversion) in questions, but not in declaratives, as shown in (11) and (12). That is, there are minor syntactic restrictions on these word order parameters, V-to-I movement and V2, distinguishing between e.g. verb classes, clause types, or other linguistically relevant sub-categories. (9)

a. b. (10) a. b. (11) (12) a. b.

Peter has never tried this wine. Peter is never late. Peter never drinks this wine. *Peter drinks never this wine. When will Peter drink this wine? On weekends Peter will drink this wine. *On weekends will Peter drink this wine.

92

Marit Westergaard

Furthermore, two different word orders are also possible in particle shift double object constructions, see (13) and (14): (13) a. b. (14) a. b.

She looked up the word. She looked the word/it up. I gave Peter a glass of wine. I gave a glass of wine to Peter.

In grammars displaying such mixed word orders, patterns of information structure tend to govern the choice of the two word orders, according to Bresnan and Nikitina (2003, 2007) and Bresnan et al (2007). They show that in double object constructions, the order IO-DO (DP-DP) is preferred when the indirect object is short and informationally given (often a pronoun) and the direct object is correspondingly long and informationally new, as in (15), while the other word order (DP-PP) is chosen when the informational status of the two objects is the other way around, shown in example (16). In fact, 94% of all datives in a corpus investigated by Bresnan et al (2007) can be predicted by a variety of pragmatic factors: discourse accessibility, length, definiteness, animacy, etc: (15) (16)

He gave DP[her] DP[the wine he had bought yesterday]. He gave DP[the wine] PP[to all the students present].

Furthermore, Bresnan and Nikitina (2003) claim that semantic constraints cannot account for the word order preferences displayed by speakers in actual production. A common explanation for the DP-PP construction is that it must express some directionality, which should account for the ungrammaticality of (17b). However, by providing an authentic example such as (18), Bresnan & Nikitina (2003) show that this construction may be perfectly grammatical given the right pragmatic context: (17) a. The movie gave DP[me] DP[the creeps]. b. *The movie gave DP[the creeps] PP[to me] (18) Stories like these must give DP[the creeps] PP[to people whose idea of heaven is a world without religion...]

As these double object constructions have not been considered to be part of a major parameter, this pragmatic account is generally not considered to be controversial. However, word order variation is also found in other present-day languages in contexts that are typically assumed to be the result of a parameter setting. One example is Norwegian, where many dialects have what at first sight looks like optional V2 in wh-questions. This variation has been argued to reflect a stage in a diachronic development towards non-V2 (see Vangsnes 2005, Westergaard 2005, 2009a). In the Tromsø dialect, for example, both V2 and non-V2 are possible in sentences with monosyllabic wh-elements, such as those in (19), and when

Cue-based Acquisition and Information Structure Drift

93

these are uttered in isolation, speakers are generally unable to identify any difference in meaning between the two: (19)

Kor bor du? /Kor where live you/ where ‘Where do you live?’

du you

bor? live

In Westergaard (2003), a sample of spontaneous adult speech was investigated, and it was shown that the choice of the two word orders was not random. V2 was typically chosen when the subject was a full DP and the verb be, while non-V2 was preferred with pronominal subjects and any other verb than be, as illustrated by the two sentences in (20a, b): (20) a. Kor er restauranten? where is restaurant.DEF ‘Where is the restaurant?’ b. Ka vi skal spise? what we shall eat ‘What will we eat?’

These preferences were analyzed in terms of information structure, in that non-V2 was argued to be used with discourse given subjects (often pronouns), and V2 when the subject was informationally new or focused (often DPs). Thus, word order in wh-questions is not optional in Norwegian dialects, but subject to constraints related to information structure. In Westergaard (in press), the two different word orders in (20) are analyzed syntactically in terms of movement to a projection in the CP domain that is sensitive to information structure, called the TopP (see also Rizzi 1997, 2001 for the original version of the idea of a Split CP). This projection has a feature that attracts informationally given elements, either a light verb (typically be) to its head position or a light subject (typically a pronoun) to its Spec position. Thus, there are two subject positions in whquestions in Norwegian dialects, one below the finite verb (in the IP domain, see below) and one above the finite verb, i.e. SpecTopP. This means that the micro-cue for V2 in wh-questions with monosyllabic wh-elements (analyzed here as heads in Int°) could be formulated as in (21), i.e. with verb movement to the Top° head only in case the subject is new information or focused (marked by a [+FOC] feature here): (21)

Cue for V2 in questions with monosyllabic wh-elements (Tromsø): IntP[ Intº [wh ] TopP[ TopqV XP[+FOC]] ...

Two subject positions are also visible in other constructions in the language where there is variable word order. In V2 constructions with negation, i.e. main clause questions and non-subject-initial declaratives, the subject may appear either preceding or following negation, as illustrated in (22) and (23):

94 (22)

(23)

Marit Westergaard

Kan ikke Peter gjøre dette? /Kan Peter ikke gjøre dette? can not Peter do this ‘Can’t Peter do this?’ Dette kan ikke Peter gjøre. /Dette kan Peter ikke gjøre. this can not Peter do ‘This Peter can’t do.’

These two subject positions have been discussed by several scholars, e.g. Holmberg (1993), Nilsen (1997), Svenonius (2002), Westergaard and Vangsnes (2005), Wiklund, Hrafnbjargarson, Bentzen and Hróarsdóttir (2007), and Bentzen (forthcoming), and it has been recognized that the two are distinguished by certain discourse factors. Westergaard and Vangsnes (2005) argue that informationally given subjects precede, while informationally new or focused subjects follow negation and sentence adverbs. This is supported by the fact that pronominal subjects, which are inherently given information, tend to appear in front of negation, as shown in (24a). In fact, a pronoun is ungrammatical in the position following negation unless it is stressed, as illustrated in (24b): (24) a. Dette kan du ikke gjøre. this can you not do ‘This you can’t do.’ b. *?Dette kan ikke du gjøre. /Dette kan ikke DU gjøre. this can not you do ‘This you can’t do.’ /This YOU can’t do.’

There seems to be no agreement about the exact syntactic position of these elements. While in the analysis of e.g. Wiklund, Hrafnbjargarson, Bentzen and Hróarsdóttir (2007) and Bentzen (forthcoming) both subject positions are in the CP domain, Westergaard and Vangsnes (2005) argue that they are located in the IP domain, which is split into a lower T(ense)P and higher In(ner)TopP. Thus, informationally new or focused subjects stay in the lower position, while informationally given subjects appear in the higher position. This has been referred to as ‘subject shift’ in Westergaard (2008b). Finally, two subject positions are also found in embedded clauses, again before and after certain sentence adverbs or negation, one of these elements directly following the complementizer. This is illustrated in (25) and (26): (25)

(26)

Vi vet at studentene ikke ville drikke denne vinen. we know that student.DEF.PL. not would drink this wine ‘We know that the students wouldn’t drink this wine.’ Vi vet at ikke studentene ville drikke denne vinen. we know that not student.DEF.PL. would drink this wine ‘We know that the students wouldn’t drink this wine.’

Cue-based Acquisition and Information Structure Drift

95

According to Nilsen (2003), these two positions are also distinguished by information structure, the higher one being used for given information and the lower one for new or focused information. Therefore, weak pronouns are virtually impossible in the low position. Westergaard and Vangsnes (2005) argue that these subject positions are the same as the ones involved in sentences such as (22) and (23), i.e. the Spec positions of the two projections in the IP domain, TP and InTopP. There may nevertheless be a certain distinction between the sentence pair in (25) and (26) on the one hand and (22) and (23) on the other: To me (as a native speaker of the language) the low position seems generally odd in embedded contexts, and this is corroborated by findings reported in Garbacz (2004). He has studied the word order of embedded clauses in two Norwegian corpora, the Big Brother corpus of spoken Norwegian (various dialects) and the Oslo corpus of written Norwegian. It is found that the word order Neg-S is extremely rare, as low as 7.3% (14/191) of all that-clauses in the spoken corpus and only 2.0% (239/12,049) in the written corpus. 2 I now turn to constructions where two object positions are visible. One well-known example concerns so-called object shift in the Mainland North Germanic languages. In sentences containing a single transitive verb (no auxiliary) and negation, pronominal objects must shift and appear preceding negation, while full DPs and strong (stressed) pronouns follow. This is illustrated in examples (27) and (28): (27)

(28)

Peter så ikke boka. Peter saw not book.DEF Peter didn’t see the book.’ Peter så den ikke. /Peter så Peter saw it not /Peter saw ‘Peter didn’t see it. /Peter didn’t see that.’

ikke not

DEN. that

With respect to the landing site of the object in object shift constructions, several analyses have been proposed. In earlier accounts (e.g. Holmberg and Platzack 1988), shifted objects are taken to appear in a position adjoined to VP, while more recent approaches assume that the object shifts into the specifier of some higher functional projection above VP, e.g. AgrOP (Déprez 1989, Bobaljik 1995), TP (Bobaljik and Jonas 1996), or some IP-internal TopicP (Josefsson 2001). In any case, it seems clear that the two object positions to some extent involve information structure, in that the higher position is reserved for informationally light objects (i.e. pronouns), while DPs and stressed pronouns appear in the lower position. 2

The figures from the written corpus have been arrived at by adding up the numbers presented in Garbacz’s Tables 2, 3 and 4 (representing press, prose and fictional texts respectively).

96

Marit Westergaard

Finally in this section, I would like to point out an example of word order variation that has been attested historically in many languages, viz. the alternation VO/OV, which was a common feature of e.g. Old/Middle English (see e.g. Pintzuk 2005), Middle High German (e.g. Hinterhölzl 2004) and Middle Norwegian (Sundquist 2002). While this is no longer found in the Germanic languages, it is a typical feature of present-day Russian, a language with much freer word order than both English and the North Germanic languages. As shown in (29) and (30), Russian allows both OV and VO, and the variation is dependent on discourse factors, pronominal objects virtually always appearing in OV constructions (see e.g. Diakonova 2003): 3 (29)

(30)

Ivan potseloval tsarevnu. Ivan kissed princess.ACC ‘Ivan kissed the princess.’ (from Diakonova, 2003: 11) Ivan ee potseloval. Ivan her kissed ‘Ivan kissed her.’

To sum up, in this section we have seen that word order variation is relatively common in present-day grammars, which display several constructions that reveal the existence of two subject or object positions. There also exist various syntactic analyses of these subject and object positions. The exact syntactic location of these elements is not crucial for the argument made in this paper. In the following I will therefore simply assume that the two positions are structurally distinct, in that the rightmost subject or object position is lower in the syntactic tree than the other one, and I will refer to them as the high and low positions. The choice of the two word orders in these cases is at least partly dependent on information structure, in such a way that discourse given elements, typically pronouns, appear in the high position. This, I assume, is a result of syntactic movement, triggered by some feature that attracts informationally light elements (as in the case of V2 vs. non-V2, discussed in Westergaard 2009a). Informationally new or focused elements, on the other hand, stay in the low position. This pattern corresponds to the well-known pragmatic principles of end focus and end weight, which go back at least to Behaghel (1932). These principles are also at the heart of the Functional Sentence Perspective of Prague school linguistics, according to which the Communicative Dynamism of a sentence develops from elements with low to elements with high information value, see e.g. Firbas (1992), or Quirk, Leech and Svartvik (1972) for these principles applied to English. Word order variation of

3

Thanks to Yulia Rodina for providing me with the example in (30).

Cue-based Acquisition and Information Structure Drift

97

this kind seems to be even more frequent in historical languages, and some examples of this are considered in the next section.

4 Diachronic Word Order Variation Several scholars have argued that word order variation typically seen in historical data may be dependent on information structure. One example of this is Hróarsdóttir (2004) on the choice of VO vs. OV in Icelandic, a variation which lasted up to the beginning of the 19th century. She shows that in Old Icelandic, the order OV was chosen if the object was given information (often a pronoun), while VO was used if the object was heavy and/or conveyed new information (often a full DP). Examples are provided in (31)-(32): (31)

(32)

að hann hafi hana drepið (Old Icelandic) that he had her killed ‘that he had killed her’ (from Hróarsdóttir 2004: 141) hvört hann vilji ei kaupa þræla whether he wanted not buy slaves ‘whether he didn’t want to buy slaves’ (from Hróarsdóttir 2004: 145)

Mixed VO/OV also existed for an extended period of time in the history of English, as described in numerous publications, e.g. Pintzuk (1991, 2005), Roberts (1997), etc. Various syntactic factors have been argued to play a role in the choice of word order at the different stages of Old and Middle English (OE/ME), e.g. whether the object was negated or contained some kind of quantification. The development from OV to VO has often been argued to be due to the loss of inflectional case morphology (e.g. Roberts 1997). Nevertheless, throughout the period when both word orders are attested, there are certain preferences that must also have been subject to information structure: Pronominal objects virtually always appear in OV constructions, while heavy DPs are strongly preferred in VO structures. This is illustrated in examples (33)-(34): (33)

(34)

We ne magan eow neadian (OE) we neg can you constrain ‘We cannot constrain you. . . ’ (from Pintzuk, 2005: 129) Se wolde gelytlian þone lyfigendan hælend he would diminish the living saviour ‘He would diminish the living saviour . . . ’ (from Pintzuk 2005: 117)

A similar situation existed in Middle Norwegian (MNw), according to Sundquist (2002). That is, the word orders OV and VO were both possible, and the choice depended to a large extent on the type of object, generally pronominal vs. full DP, but also subcategories of pronouns (per-

98

Marit Westergaard

sonal, reflexive etc.) and DPs (definite/indefinite, quantified etc.) were found to play a role for the word order chosen. During an extended period of the variation, pronouns were clearly preferred in preverbal position, while DPs were much more common in postverbal position. A couple of examples are provided in (35)-(36): (35)

(36)

ok þui ualðe sem han heuir oss fengit (MNw) and this power which he has us given ‘...and with this power which he has given us.’ at hann hafðe suaret þænna sama vithnisburð aðr that he had sworn that same testimony before ‘that he had sworn that same testimony before’ (from Sundquist 2002: 101-2)

Returning to the history of English, we consider some constructions where it is the subject that may appear in two different positions. It is well known that there is variation between V2 and non-V2 word orders in declaratives throughout the OE and ME periods. It has also been recognized by many scholars that V2 was preferred with DP subjects, while non-V2 mainly appeared with pronouns (e.g. van Kemenade 1987, Pintzuk 1991, Kroch and Taylor 1997). This is illustrated in (37)-(38): (37)

(38)

On his dagum sende Gregorius us fulluht. in his days sent Gregorius us baptism ‘In his time, Gregory sent us Christianity.’ Hiora untrymnesse he sceal ðrowian on his heortan. their weakness he shall atone in his heart ‘He shall atone in his heart for their weakness.’ (from Haeberli 2002: 8890)

A common analysis of this pattern argues that pronominal subjects are clitics attached to a position in front of the moved verb, so that even examples such as (38) are considered to display V2 word order. However, there are many problems with the clitic hypothesis, one of them being that there is a considerable number of full DP subjects appearing with non-V2 word order. Thus, both Bech (2001) and Westergaard (2009b) argue that the choice of V2 vs. non-V2 in OE/ME was dependent on information structure: The word order XSV (non-V2) was preferred if the subject was informationally given (often a pronoun) and XVS (V2) if the subject was new information or focused in some way (often a full DP). Note that this is very similar to the subject preferences with these two word orders in whquestions in present-day Norwegian dialects, as discussed in section 3 above (cf. example (20)). Finally, we consider a further example from the history of English, the two subject positions found in embedded clauses, discussed in van Kemenade and Los (2006). In OE, the subject may appear either above or be-

Cue-based Acquisition and Information Structure Drift

99

low certain adverbs, most notably þa/þonne ‘then’, which van Kemenade and Los consider to be focus particles. As in the present-day examples from Norwegian discussed in section 3 (cf. (25)-(26)), the higher subject position is preferred when the subject refers to an informationally given referent (often a pronoun), as illustrated in (39), which incidentally also contains a pronominal object preceding þa/þonne. The lower subject position is typically chosen when the subject is discourse new or focused (normally a full DP), as illustrated in (40): (39)

(40)

He ne mihte swaþeah æfre libban, þeah ðe he hine he not-could nevertheless ever live though that they him þa ut alysde then released ‘Nevertheless, he could not live forever, though they then released him’ Gif him þonne God ryhtlice & stræclice deman wile. if him then God justly and strictly judge will ‘if God will then justly and strictly judge him’ (from van Kemenade and Los 2006: 231)

Also for these historical grammars, there exist various syntactic analyses of the positions that the two subjects or objects are located in. As for the present-day data, I will simply assume that the positions are structurally distinct and that the higher one is typically chosen for informationally given elements and the lower one for new or focused elements. The development of these historical mixed grammars is well known in these cases: English, Icelandic and Norwegian generally lose OV word order and develop into relatively strict VO languages. This means that it is the low object position that survives, the one that is closely linked to heavy or new/focused elements. With respect to mixed V2/non-V2 in the history of English, V2 word order was generally lost in declaratives during the ME period, which means that it is the higher subject position which survives. This is also the case in present-day Norwegian dialects, according to the analysis of the variation found in the synchronic data (see e.g. Westergaard 2009a). Finally, in embedded clauses in the history of English, van Kemenade and Los (2006) show that it is the higher subject position that survives, in that this is the preferred position for both pronouns and full DPs in ME. 4 However, it should be mentioned that the frequency of þa/þonne drops considerably from OE into ME, so that the evidence that there are two subject positions in embedded clauses in the language is virtually nonexistent in the later period. In any case, this development means that the subject position that survives in all three cases is the one that is preferred for informationally given subjects in the mixed grammar. 4

Note that this corresponds to the fact that also in present-day Norwegian, it is the higher position that is preferred, see section 3 above.

100

Marit Westergaard

To sum up this section, we have briefly discussed three historical cases where the low object position survives (loss of OV in English, Norwegian and Icelandic), and three cases where the high subject position survives (loss of V2 in English declaratives and Norwegian wh-questions as well as loss of two subject positions in embedded clauses in English). It may of course be the case that these developments are simply coincidental, since examples of the opposite change also exist in historical data, e.g. the development from a mixed VO/OV system into a pure OV grammar in the history of German (see e.g. Hinterhölzl 2004, 2009). It may also be the case that these developments are due to different external effects, as explored in e.g. Hróarsdóttir (2004) for Icelandic and Sundquist (2002) for Middle Norwegian. However, the changes discussed in this section take place at different times and seem to be unrelated developments. Thus, an obvious question to ask is whether these developments could be attributed to particular word order preferences in children’s acquisition process. In the next section we therefore explore some child language data from present-day grammars which display similar word order variation, more specifically on the acquisition of the present-day word order facts discussed in section 3. It will be shown that, although the child data reveal a certain delay in the acquisition of the high subject and object positions, this is unlikely to be the cause of the observed historical changes. The subsequent section therefore considers whether part of the explanation may be sought in the nature of the input.

5 Child Language Data The acquisition data discussed in this section are mainly from Norwegian, from a corpus of child language collected in Tromsø. 5 The corpus contains spontaneous data from three children between the ages of approximately 1;9 and 3, as shown in Table 1: Name of Child Ina Ann Ole Total

Age 1;8.20-3;3.18 1;8.20-3;0.1 1;9.10-2;11.23

Files Ina.01-27 Ann.01-21 Ole.01-22

Child Utterances 20,071 13,129 13,485 46,685

Table 1: Norwegian corpus of child language, Tromsø dialect.

5

Apart from 10 files that have been collected and transcribed by the author, the corpus has been collected by Merete Anderssen. See Anderssen (2006) for further information about the corpus.

Cue-based Acquisition and Information Structure Drift

101

Recall from section 3 that the dialect allows both V2 and non-V2 in whquestions, with clear preference patterns in the adult grammar for subject and verb types appearing with the two word orders, argued to reflect patterns of information structure, see the sentences in (20) above. An investigation of the child data reveals that the children produce both word orders from early on, with the same preferences for subject and verb types as those found in the adult data, i.e. V2 with be and full DP subjects and non-V2 with pronominal subjects and all other verbs than be. Typical examples from the children’s production are provided in (41)-(42). (41)

(42)

kor er Ann sin dukke hen? (Ann, age 1;11.0) where is Ann POSS doll LOC ‘Where is Ann’s doll?’ ka du skal finne? (Ina, age 2;0.5) what you shall find ‘What do you want to find?’

Like the adults in the corpus, the children also produce the two word orders with other subject and verb types if the information structure of the sentence requires it. In (43), the verb is not be and the subject is a pronoun (han der ‘he there’), but nevertheless, V2 word order is used. This is appropriate since the child in this situation is pointing and thus putting emphasis on the subject. In (44), on the other hand, non-V2 word order is chosen although the subject is a full DP (løva ‘the lion’). In this situation the subject has been mentioned in the immediately preceding discourse and is therefore given information. Thus, the word order chosen is again in accordance with the pattern of information structure, V2 with new or focused subjects and non-V2 with subjects that convey given information. (43)

(44)

ka hete han der? (Ina.07, age 2;1.23) what is-called he there ‘What is he called?’ ka løva like å spise mamma? (Ann 2;6.21) hat lion.DEF. likes to eat mommie ‘Mommie, what does the lion like to eat?’

Based on these findings, it is argued in Westergaard (2003) that the children distinguish between informationally given and new subjects in the two types of wh-questions from early on. This indicates that children have an early sensitivity to patterns of information structure. Given very different frequencies of the two word orders produced by individual speakers in the adult data (Westergaard 2009a), there is no adult standard to compare the children’s production to, and it is therefore impossible to state whether there is a delay with respect to the production of one of the two word orders in the child data.

102

Marit Westergaard

Also in the children’s early production of non-subject-initial declaratives, there is an indication of their early sensitivity to information structure. In this case the adult language requires V2 only, which means that there is no input for a distinction between given and new subjects in this clause type. Nevertheless, the children’s early non-subject-initial declaratives occur predominantly with be and full DP subjects, as in (45), while their occasional non-target-consistent (non-V2) constructions appear with other verbs than be and pronominal subjects or the child’s own name (i.e. typical given information), as illustrated in (46); see Westergaard (in press) for further details. (45)

(46)

her er sekken. (Ann, age 1;10.2) here be.PRES backpack.DEF. ‘Here is the backpack.’ der Ann har et. (Ann, age 2;1.28) there Ann have.PRES. one ‘There Ann has one.’ Target form: Der har Ann et.

Thus, the subject and verb combinations illustrated for wh-questions in (41)-(44) are preferred also in declaratives, at least for a short period of time (up to approximately age 2;4). This also means that the patterns found in declaratives in early child language turn out to be similar to what is found in the historical English data (cf. examples (37)-(38) above). We now turn to a construction where a certain delay is attested in the child data, viz. so-called subject shift. This was illustrated in examples (22)-(24) in section 3 and involves movement of a pronominal (or informationally light DP) subject to a position above negation in questions and non-subject-initial declaratives, i.e. clauses that display V2. In Westergaard (2008b) it is shown that all three children in the corpus at an early stage produce a predominance of examples such as (47), where the pronominal subject has failed to undergo movement. A delay in subject movement has also been attested in German child language, see Clahsen, Penke, and Parodi (1993/94): (47)

har ikkje han fota her? (Ina, age 2;5.25) have.PRES. not he feet here ‘Doesn’t he have feet here?’ Target form: Har han ikkje fota her?

A possible analysis of this word order is that it is the result of negation behaving as a clitic in the early child grammar and moving together with the verb across the subject. However, this analysis is rejected in Anderssen, Bentzen, Rodina and Westergaard (2008), mainly because Norwegian children investigated in an experimental study on subject and object shift were found not to pronounce the negation as a clitic in these construc-

Cue-based Acquisition and Information Structure Drift

103

tions. The word order in (47) is attested only for a brief period of time, as between the ages of 2;6 and 3, the subject-shift constructions falls into place and the children generally produce target-consistent word order, as illustrated in (48): (48)

nei, nå må han ikke no now must he not ‘No, now he must’t touch (it).’

røre. (Ina, age 2;9.18) touch

Likewise, the children show an early preference for the low subject position also in embedded clauses, as illustrated in (49), cf. examples (25)-(26) in section 3. Recall that, although this word order is grammatical in the adult grammar, it is extremely dispreferred, appearing only 7.3% in a corpus of spoken Norwegian. Given that embedded clauses are generally late acquisitions and thus very infrequent in early child language, the present corpus, which contains data only until about the age of three, cannot say anything about when this word order becomes target-like: 6 (49)

…ho si at ikkje det er min kjæreste. (Ina, age 3;3.18) …she says that not it is my sweetheart ‘She says that it isn’t my sweetheart.’ Target form: Ho sir at det ikkje er min kjæreste.

We then turn to constructions that provide evidence of two object positions. Recall that in object shift constructions, a pronominal object moves to a position above negation, cf. examples (27)-(28) in section 3. Again, a certain delay is attested in the child data, all three children displaying a preference for the low position at an early stage, as illustrated in (50): (50)

eg finn [>] . (Ina, age 2;5.25) I find not him ‘I can’t find him.’ Target form: Eg finn han ikkje.

Occasional target-consistent forms are produced, see (51), but the construction is in general extremely infrequent both in child-directed speech and in the child data. At the end of data collection for this corpus (around age 3), the proportion of non-target forms is still higher than the targetconsistent production, indicating that the object shift construction is more delayed than subject shift. This finding is confirmed in the experimental study carried out in Anderssen et al (2008). It should be noted that a similar delay is attested also in other languages, e.g. for object shift in Swedish (Josefsson 1996), and for object shift and object scrambling in Dutch and German (Schaeffer 2000, Barbier 2000):

6

See also Westergaard and Bentzen (2007) for a discussion of children’s word order in embedded clauses.

104

Marit Westergaard

(51)

ho har den ikkje på sæ . (Ina, age 2;11.26) she has it not on REFL. ‘She doesn’t have it on.’

Finally, we will have a look at some Russian child data, where both VO and OV word orders are permitted in the adult grammar, cf. examples (29)-(30) in section 3. Diakonova (2003), who studied the word order produced by one Russian child Varvara from the CHILDES database (MacWhinney 2000), finds that both word orders are produced from early on, and that pronominal objects generally appear in front of the verb, see (52). However, there are occasional examples of VO with pronominal objects at an early stage, again indicating a certain delay of movement and an early preference for the low position. This is illustrated in (53): 7 (52)

(53)

Ja ego polozhila. (Varvara.05, age 2;0.01) I him put.PAST ‘I put him.’ On pochinit eto. (Varvara.05, age 2;0.01) he repair.FUT. this ‘He will repair it.’

Summing up, it has been shown in this section that children have an early sensitivity to patterns of information structure, producing both word orders in a more or less target-consistent way in e.g. Norwegian whquestions (V2 vs. non-V2) and subject shift constructions. The findings from the Russian child data (VO vs. OV) provide further evidence for this. To the extent that children produce non-target-consistent word order at all, they have an early preference for the lower position of two available ones, be it a subject or an object position. This was shown to be the case in subject and object shift constructions, as well as the subject position in embedded clauses. This delay has been argued in Westergaard (2008b) to be due to a general principle of economy of movement in the acquisition process, and not a problem with pragmatic principles of information structure. In the historical data discussed in section 4, the mixed grammars developed into systems where the high subject position and the low object position survived. Assuming that what is late acquired is also vulnerable to change (see e.g. Pinker 1999), this means that the preference in early child language for the low object position could be a factor for the diachronic development from OV to VO word order. Another argument in favor of such an explanation is that the delay with respect to object positions seems to be more persistent than the delay in subject positions (object shift vs. 7

Diakonova does not provide any examples from the child data of this delay. The examples in (52) and (53) have been extracted from the Varavara corpus by Yulia Rodina. According to her, the sentence in (53) is grammatical if the pronoun or the verb is stressed.

Cue-based Acquisition and Information Structure Drift

105

subject shift in the Norwegian data). However, the children also have an early preference for the low subject positions, which means that this acquisitional preference would have nothing to say about the historical development of subjects, where it is typically the high position that survives. A natural next step is therefore to consider children’s E-language input and investigate whether typical frequencies of the two word orders in the PLD that children are exposed to could shed some light on the historical development. This is the topic of the next section.

6 Frequencies in Child-directed Speech Recall from section 2 that the theoretical background for this study is an approach to language acquisition and change which is based on the existence of micro-cues in children’s I-language, expressed by the relevant input. That is, children easily acquire mixed systems and even exceptions to general rules, if the variation is based on natural classes of linguistic (sub-)categories or if exceptions are relatively frequent in the input (see Westergaard 2008a, 2009a). In the previous section, we have also seen that children relatively easily learn mixed grammars where the variation is dependent on information structure. Nevertheless, even in this model, a very low input frequency of a construction which expresses a particular micro-cue will make it vulnerable to change over time, and the micro-cue may eventually be lost from the language. It is therefore important to investigate relative frequencies of the two word orders in the input that children are typically exposed to in the acquisition process. Recall that the high position, be it a subject or an object position, tends to be used for discourse given elements, very often realized by pronouns. As mentioned above, this is in line with well-known pragmatic principles such as end focus and end weight. It is also commonly assumed that in language use, subjects tend to refer to given information, while objects more often express new information. If this is the case, then the higher subject position should be more common than the lower subject position in general language use, and consequently also in the PLD that children are exposed to. Conversely, for the two object positions it is the lower position that should be the more frequent one. In this section I have therefore investigated some spontaneous speech from a corpus of adult conversations, and also some samples of childdirected speech in both Norwegian and English. As the informational status of an element is often unclear (and very time-consuming to investigate), I have divided subject and objects types into general categories, i.e. whether they are realized as pronouns, full DPs, or elements such as

106

Marit Westergaard

it/there/that. Pronouns are considered to be typical cases of given information. They are then compared to the number of clauses with full DP subjects or objects, which more often convey new information (although DPs may obviously also convey given information, see e.g. example (44) above). Table 2 displays the results of such counts done on a sample from a corpus of spoken Norwegian (the NoTa corpus - Oslo dialect), more specifically a conversation between two middle-aged adults (age 39 and 52).

Subjects Objects

Pronouns (pers/refl) 340/0 (57.8%) 11/22 (15.4%)

det (‘it/that’) 204 (34.7%) 37 (17.3%)

DPs/clauses

Misc/Other

Total

35/0 (6.0%) 104/38 (66.4%)

9 (1.5%) 2 (0.9%)

588 (100%) 214 (100%)

Table 2: The realization of subjects and objects in a sample of Norwegian conversational speech

(speakers 119 and 120 in the NoTa corpus).

In Table 2, the element det ‘it/that’ has been counted separately, as it is not always clear in the transcriptions whether it is an expletive or referential pronoun (in which case it would tend to pattern with discourse given elements) or the demonstrative (in which case it would often pattern with new or focused elements). Disregarding this element, we see that subjects are realized by pronouns as much as ten times more often than full DPs or clauses (340 vs. 35). Objects, on the other hand, are realized by DPs/clauses 4-5 times more often than by pronouns (104+38=142 vs. 11+22=33). As a subject, the element det tends to be an expletive or a referential pronoun, while as an object it tends to be a demonstrative. Thus, including det and counting it as given information when it is a subject and new information when it is an object, would give an even more uneven distribution of informationally given and new elements in the two positions. In any case, it seems certain that in conversational Norwegian, subjects are realized very differently from objects, displaying a clear preference for pronouns, which reflects the status of this function as a position for informationally given elements. Objects, on the other hand, are normally realized by full DPs or clauses, indicating that this position normally expresses new or focused information. But conversations between adults may be very different from childdirected speech. So in order to say something about typical input to children in the acquisition process, one also needs to consider samples of adult data spoken to children. Table 3 provides an overview of the realization of subjects and objects in a sample of child-directed speech from the Norwegian acquisition corpus (the production of the investigator in file Ole.15,

107

Cue-based Acquisition and Information Structure Drift

age of child 2;7.20). This sample has been hand-counted in the same way as the sample from the adult-adult conversation.

Subjects Objects

Pronouns (pers/refl) 272/0 (67.5%) 26/3 (16.4%)

det (‘it/that’)

DPs/clauses

Misc/Other

Total

99 (24.6%) 40 (22.6%)

29/0 (7.2%) 74/32 (59.9%)

3 (0.7%) 2 (1.1%)

403 (100%) 177 (100%)

Table 3: Overview of the realization of subjects and objects in a sample of Norwegian child-

directed speech (Investigator in file Ole.15).

Table 3 shows that the patterns for the sample of child-directed speech is very similar to those found in the conversation between adults. Disregarding the element det ‘it/that’, subjects are again realized by pronouns about ten times more often than as full DPs (272 vs. 29), while objects are realized by DPs/clauses 3-4 times more often than as pronouns (106 vs. 29). This shows that it is in fact the case that subjects tend to refer to given information (often realized as pronouns), while objects tend to be new (often realized as full DPs), also in the input that children are typically exposed to. Although this has now been attested for Norwegian, we may still wonder to what extent the tendency for such patterns can be found in other languages. To investigate this question, two samples of English child-directed speech have been hand-counted in the same way as the Norwegian data. Tables 4 and 5 provide an overview of the production of two adults in the Brown corpus from the CHILDES database (Brown 1973, MacWhinney 2000), the mother in file Adam.10 (age of child 2;7.14) and the mother in file Eve.15 (age of child 2;1). For the English data the elements it and there have been counted together, as they normally pattern with informationally light elements, while that has been counted as a separate category. 8

8

As was the case with the element det ‘it/that’ in Norwegian, the informational status of that seems to be different in subject and object position. As an object, that most likely conveys some kind of focused information, as in e.g. you don’t use that either. When that is used as a subject, on the other hand, there is normally another element in the sentence that conveys new information, as in e.g. that’s not shampoo, that’s dressing. For this reason, that has been disregarded in the calculations. As with the element det in Norwegian, including that would lead to an even clearer distinction between subject and object positions with respect to information structure.

108

Subjects Objects

Marit Westergaard

Pronouns (pers/refl) 193/0 (65.4%) 7/1 (6.2%)

it/there

that

16/2 (6.1%) 14/0 (10.8%)

46 (15.6%) 10 (7.7%)

DPs/ clauses 35 (11.9%) 96 (73.8%)

Misc/ Other 3 (1.0%) 2 (1.5%)

Total 295 (100%) 130 (100%)

Table 4: Overview of the realization of subjects and objects in a sample of English child-directed

speech (Mother in file Adam.10).

Subjects Objects

Pronouns (pers/refl) 397/0 (68.4%) 44/0 (16.6%)

it/there

that

66/10 (13.1%) 63 (23.8%)

55 (9.5%) 7 (2.6%)

DPs/ clauses 44 (7.6%) 134 (50.6%)

Misc/ Other 8 (1.4%) 17 (6.4%)

Total 580 (100%) 265 (100%)

Table 5: Overview of the realization of subjects and objects in a sample of English child-directed

speech (Mother in file Eve.15).

Counting all pronouns and the elements it/there together, and disregarding that, we see from Table 4 that subjects are realized by pronouns or it/there about six times more often than as full DPs (193+18=211 vs. 35). In the sample displayed in Table 5 the differences are even more striking, as subjects are realized by pronouns almost 11 times as often as full DPs (397+76=473 vs. 44). Objects, on the other hand, are realized by DPs/clauses four times more often than as pronouns (96 vs. 22) in Table 4, while in Table 5 the difference is not quite so great, only about 20% more (44+63=107 vs. 134). Including the element that in the count would presumably give even clearer results of the difference between the informational status of subjects and objects (see previous footnote). Thus, the patterns found for Norwegian can also be said to hold for English childdirected speech: There is a clear preference for putting discourse given information in subject position and discourse new or focused information in object position. In fact, this is not simply a tendency, but a very robust pattern in both languages. For mixed word orders where the choice is to some extent dependent on information structure, as in the present-day systems discussed in section 3 and the historical data presented in section 4, this means that word orders that are linked to informationally given (typically pronominal) subjects should be naturally frequent in the E-language. Word orders linked to new or focused (full DP) subjects should be correspondingly infrequent. For mixed grammars involving two object positions, the situation should be the other way around: The word order preferred with dis-

Cue-based Acquisition and Information Structure Drift

109

course given objects should be infrequent in the natural spoken language, while the word order typically chosen for informationally new objects should be considerably more frequent. In the next section we consider how such a situation over time may affect the PLD that children are exposed to.

7 Information Structure Drift We saw in section 5 that children easily learn different word orders that are dependent on information structure, e.g. V2 vs. non-V2 in Norwegian or VO vs. OV in Russian. This has also been attested in several other recent studies for other languages, e.g. de Cat (2003) for the topichood of subjects in French, and Gordishevsky and Avrutin (2003) for the acquisition of null subjects in Russian. This means that the mixed word order systems discussed for present-day and historical data in sections 3 and 4 should be easily learnable, and from a pure acquisition perspective such grammars should also be relatively stable diachronically. This is the reason why it is important to also consider the nature of the input. The theoretical model of cue-based acquisition and change sketched in section 2 assumes that different word orders are distinguished by various syntactic micro-cues (Westergaard 2009a, Lightfoot and Westergaard 2007). These micro-cues may specify the clause type that the cue is relevant in (e.g. V2 in questions in English but not in declaratives), minor classes of categories that it may apply to (e.g. verb movement applying to auxiliaries in English but not lexical verbs), or patterns of information structure that are relevant (as in many of the examples discussed in section 3). Many of the historical examples in section 4 are also distinguished by such syntactic micro-cues at the different stages, e.g. certain initial elements such as þa/þonne ‘then’ requiring V2 in declaratives in OE, but not in late ME (Bech 2001, Westergaard 2009b), or negated and quantificational objects requiring OV word order at a certain stage in the history of English when other object types mainly appeared in constructions with VO (Pintzuk and Taylor 2006). As mentioned in section 2, under this view of language acquisition, gradualism in historical data is considered to be the result of ‘many small catastrophes’, reflecting the loss of or a change in one micro-cue at a time. And even though children are sensitive to very minor distinctions in language, a low frequency of the triggers for a particular cue in the PLD, and a correspondingly high frequency of a conflicting trigger, will make a micro-cue vulnerable to change. That is, structures with an extremely low

110

Marit Westergaard

input frequency may be ignored by children in the acquisition process and disappear from the I-language grammar of the next generation. In this connection, the patterns of information structure in typical conversational language or child-directed speech are important. As we saw in the previous section, the predominance of informationally given subjects and informationally new objects should be robust in children’s input. And with respect to constructions that allow mixed word orders dependent on the information value of these elements, the word orders linked to informationally given subjects and new or focused objects should be considerably more frequent than the other possible word orders. In this scenario, a crucial factor is the fact that input varies and all children are not exposed to exactly the same PLD. Some children get a lot of input, some children less. Children also typically get input from many different speakers, both adults and other children. And the already low frequency of a construction in the spoken language generally may cause the cue for it to fall below the threshold for acquisition for some children, who will then develop a grammar with only one subject or object position. Since children also often provide input for each other, even short stages of a non-target system in the grammars of some children may have a more long-term effect for other children. The point here is that, even though the potential for change is always present in the acquisition process, many other factors need to be favorable for a stage in the child grammar to persist and to also spread in the population. Different stages of grammars that are distinguished by information structure have been identified in synchronic Norwegian data, where different frequencies of the variable word order in wh-questions have been identified as three distinct grammars; one truly mixed V2/non-V2 grammar (where the two word orders are clearly distinguished by information structure), a default V2 grammar (where V2 appears with all subject types regardless of information value and non-V2 is reserved for special cases, i.e. only with pronominal subjects), and a default non-V2 grammar (where non-V2 appears with all subject types and it is V2 that only survives in special cases), see Westergaard (2009a). That is, even though the three grammars are distinguished by further syntactic micro-cues (e.g. types of wh-elements or different classes of verbs affected by verb movement), the different patterns of information structure play an important role. It should also be noted that the patterns of information structure are operative in relation to individual micro-cues, so that it is possible for a speaker to have e.g. a default V2 grammar in some clause types and a default nonV2 grammar in others. In the history of English, different stages of similar V2/non-V2 grammars have been identified as separate systems distinguished by information structure (Westergaard 2009b).

Cue-based Acquisition and Information Structure Drift

111

This means that, at any given time, different speakers of a language may have slightly different grammars. That is, there may be speakers who have truly mixed grammars (based on information structure), while other speakers have developed grammars where certain information structure patterns have become syntactically “frozen”. A situation where children are exposed to input like this provides the potential for further change over time. The effect that information structure may have over an extended period of time has been referred to as Information Structure Drift (Westergaard 2005). It is important to note, however, that this is not considered to be an inevitable development once it has been initiated, as in a common definition of ‘drift’ in studies of historical language development. In the present model, where language acquisition and language change are closely interrelated, such drift, which often spans several hundred years, is impossible, simply because children are only exposed to one stage of the development and obviously have no information about previous stages. Given the idea of micro-cues instead of major parameters, all stages of a development from e.g. a V2 to a non-V2 grammar or an OV to a VO grammar are learnable, and thus in principle stable. Consequently a certain development could be reversed, if external factors and/or internal language development cause the PLD to change in a certain way. Such a reversal may in fact be taking place in some present-day Norwegian dialects, where interference from the standard language seems to cause a return from nonV2 to V2 (Westergaard 2005). As mentioned in section 4, in historical data, there are obviously also cases of mixed word order systems that develop in the opposite direction of what has been shown for the examples discussed in this paper, e.g. mixed OV/VO in Middle High German developing into a more or less fixed OV grammar in Modern German (Hinterhölzl 2004, 2009). In such cases, external factors or the specific development of certain syntactic micro-cues must have had a more powerful effect than Information Structure Drift, which in this paper is argued to be one factor (presumably among many) in diachronic language development.

8 Summary and conclusions In this paper I have argued that, in mixed grammars where the syntax allows variable word orders involving two subject or object positions, the two orders are normally not produced randomly, but are dependent on various syntactic micro-cues. Furthermore, patterns of information structure typically distinguish between the two, one of them being preferred for

112

Marit Westergaard

informationally given subjects or objects and the other for informationally new elements. Examples of such mixed word orders that are discussed in the paper are provided both from present-day languages and historical data, mainly from Norwegian and English. The relevant diachronic word order changes involve survival of (high) subject positions that are reserved for informationally given elements and (low) object positions that are typically used for informationally new objects. It has also been shown that mixed word orders are easily learnable by children, indicating an early sensitivity to information structure. In this perspective, mixed word order systems should be stable historically. However, within a model of acquisition that is based on micro-cues, it is argued that, when the expression of a micro-cue drops below a certain frequency in the input, it becomes vulnerable to change. Thus, the main argument of the paper is that an explanation of the diachronic development in the examples discussed may be found in the nature of the input. It is shown that, in the E-language that children are typically exposed to, subjects are predominantly given information (often pronouns) and objects predominantly new (often full DPs or clauses). This is a very robust pattern both in English and in Norwegian, which results in a higher frequency of certain word orders in the PLD, viz. word orders that are linked to informationally given subjects and word orders that are linked to informationally new or focused objects. Over time, this may cause what is referred to as Information Structure Drift, resulting in syntactic freezing of certain word order patterns.

References Anderssen, M. (2006): The Acquisition of Compositional Definiteness in Norwegian. Doctoral Dissertation, University of Tromsø. Anderssen, M., K. Bentzen, Y. Rodina, and M. Westergaard (2008): The Acquisition of Apparent Optionality: Word Order in Subject and Object shift Constructions in Norwegian. Ms., University of Tromsø. Barbier, I. (2000): An Experimental Study of Scrambling and Object Shift in the Acquisition of Dutch. In: S. M. Powers and C. Hamann (eds.), The Acquisition of Scrambling and Cliticization, 41-69. Dordrecht: Kluwer Academic Publishers. Bech, K. (2001): Word Order Patterns in Old and Middle English: A Syntactic and Pragmatic Study. Doctoral dissertation, University of Bergen.

Cue-based Acquisition and Information Structure Drift

113

Behaghel, O. (1932): Deutsche Syntax. Eine Geschichtliche Darstellung, Band IV: Wortstellung. Periodenbau. Heidelberg: Carl Winters Universitätsbuchhandlung. Bentzen, K. (Forthcoming): Subject positions and their interaction with verb movement. Studia Linguistica. Bobaljik, J. D. (1995): Morphosyntax: The syntax of verbal inflection. PhD dissertation, MIT. Bobaljik, J. D. and D. Jonas (1996): Subject positions and the roles of TP. Linguistic Inquiry 27.2, 195-236. Bresnan, J. and T. Nikitina (2003): On the gradience of the dative alternation. Ms., Stanford University. Bresnan, J. and T. Nikitina (2007): The Gradience of the Dative Alternation. In: L. Uyechi and L. H. Wee (eds.), Reality Exploration and Discovery: Pattern Interaction in Language and Life. Stanford: CSLI Publications. Bresnan, J., A. Cueni, T. Nikitina, and H. Baayen (2007): Predicting the Dative Alternation. In: G. Boume, I. Kraemer, and J. Zwarts (eds.), Cognitive Foundations of Interpretation, 69-94. Amsterdam: Royal Netherlands Academy of Science. Brown, R. (1973): A First Language: The Early Stages. Cambridge, MA: Harvard University Press. Cardinaletti, A. and M. Starke (1999): The typology of structural deficiency: On the three grammartical classes. In: H. van Riemsdijk (ed.), Clitics in the Languages of Europe 8, Empirical Approaches to Language Typology, 145-233. Berlin: Mouton de Gruyter. Chomsky, N. (1986): Knowledge of Language: Its Nature, Origin and Use. New York: Praeger Publishers. Clahsen, H., M. Penke and T. Parodi (1993/94): Functional categories in Early Child German. Language Acquisition 3, 395-429. De Cat, S. (2003): Syntactic manifestations of very early pragmatic competence. In: B. Beachley, A. Brown, and F. Conlin (eds.), Proceedings of the 27th Annual Boston University Conference on Language Development, 209-219. Somerville, MA: Cascadilla Press. Déprez, V. (1989): On the typology of syntactic positions and the nature of chains: Move A to the specifier of functional projections. PhD dissertation, MIT. Diakonova, M. (2003): The Acquisition of Word Order in English and Russian. MA thesis, University of Tromsø. Firbas, J. (1992): Functional Sentence Perspective in Written and Spoken Discourse. Cambridge and New York: Cambridge University Press.

114

Marit Westergaard

Garbacz, P. (2004): Bisatsstrukturer som man kan inte bortse från (om bisatsordföljd i de fastlandsskandinaviska språken) [Embedded structures that cannot be disregarded (on embedded clause word order in Mainland Scandinavian)]. Ms., University of Lund. Gordishevsky, G. and S. Avrutin (2004): Optional omissions in an optionally null subject language. In: J. van Kampen and S. Baauw (eds.), Proceedings of GALA 2003, Vol. 1, LOT Occasional series 3, University of Utrecht, 187-198. Available online at http://lotos.library.uu.nl/index.html. Hinterhölzl, R. (2004): Language change vs. grammar change: What diachronic data reveal about the distinction between core grammar and periphery. In: E. Fuss and C. Trips (eds.), Diachronic Clues to Synchronic Grammar, 131-160. Amsterdam and Philadelphia: John Benjamins. Hinterhölzl, R. (2009): The role of information structure in word order variation and word order change. In: R. Hinterhölzl and S. Petrova (eds.), New Approaches to Word Order Variation in Germanic, 45-66. Berlin: Mouton de Gruyter. Holmberg, A. (1993): Two subject positions in Mainland Scandinavian. Working Papers in Scandinavian Syntax 52, 29-41. Holmberg, A. and C. Platzack. (1988): On the role of inflection in Scandinavian syntax. Working Papers in Scandinavian Syntax 42, 23-42. Hróarsdóttir, T. (2004): Cues and expressions. Nordlyd 32.1: Tromsø Working Papers in Language Acquisition, 135-155. Josefsson, G. (1996): The Acquisition of Object Shift in Swedish Child Language. In: C. E. Johnson and J. H. V. Gilbert (eds.), Children’s Language 9, 153-165. Mahwah: Lawrence Erlbaum. Josefsson, G. (2001): The true nature of Holmberg’s Generalization revisited – once again. Working Papers in Scandinavian Syntax 67, 85-102. Kemenade, A. van and B. Los. (2006): Discourse adverbs and clausal syntax in Old and Middle English. In: A. van Kemenade and B. Los (eds.), The Handbook of the History of English, 224-248. Malden, MA and Oxford: Blackwell. Lightfoot, D. (1999): The development of language: Acquisition, change, and evolution. Oxford: Blackwell. Lightfoot, D. (2006): How New Languages Emerge. Cambridge: Cambridge University Press. Lightfoot, D. and M. Westergaard (2007): Language Acquisition and Language Change: Interrelationships. Language and Linguistics Compass 1 (5), 396-415. MacWhinney, B. (2000): The CHILDES project: Tools for analyzing talk. Third Edition. Mahwah, NJ: Lawrence Erlbaum Associates. Nilsen, Ø. (1997): Adverbs and A-shift. Working Papers in Scandinavian Syntax 59, 1-31.

Cue-based Acquisition and Information Structure Drift

115

NoTa – Norsk Talespråkskorpus, Oslodelen [Corpus of Spoken Norwegian, the Oslo part], Tekstlaboratoriet, ILN, University of Oslo http://www.tekstlab.uio.no/nota/oslo/index.html. Pinker, S. (1999): Words and Rules: The Ingredients of Language. London: Weidenfeld and Nicolson. Pintzuk, S. (1991): Phrase Structures in Competition: Variation and Change in Old English Word Order. PhD dissertation, University of Pennsylvania. Pintzuk, S. (2005): Arguments against a universal base: evidence from Old English. English Language and Linguistics 9.1, 115–138. Pintzuk, S. and A. Taylor (2006): The loss of OV order in the history of English. In: A. van Kemenade and B. Los (eds.), The Handbook of the History of English, 249-278. Oxford: Blackwell. Quirk, R., G. Leech and J. Svartvik. (1972): A Grammar of Contemporary English. New York: Seminar Press. Rizzi, L. (1997): The fine structure of the left periphery. In: L. Haegeman (ed.), Elements of Grammar: Handbook of Generative Syntax, 281-337. Dordrecht: Kluwer. Rizzi, L. (2001): On the position “Int(errogative)” in the left periphery of the clause. In: G. Cinque and G. Salvi (eds.), Current Studies in Italian Syntax, 287-296. Amsterdam: Elsevier. Roberts, I. G. (1997): Directionality and word order change in the history of English. In: A. van Kemenade and N. Vincent (eds.), Parameters of Morphosyntactic Change, 396-426. Cambridge: Cambridge University Press. Roeper, T. (2007): What frequency can do and what it can’t. In: I. Gülzow and N. Gagarina (eds.), Frequency effects in language acquisition: Defining the limits of frequency as an explanatory concept, 23-48. Berlin and New York: Mouton de Gruyter. Rodina, Y. (2008): Semantics and Morphology: The Acquisition of Grammatical Gender in Russian. PhD dissertation, University of Tromsø. Schaeffer, J. (2000): The Acquisition of Direct Object Scrambling and Clitic Placement: Syntax and Pragmatics. Amsterdam and Philadelphia: John Benjamins. Sundquist, J. D. (2002): Morphosyntactic Change in the History of the Mainland Scandinavian Languages. PhD dissertation, Indiana University. Svenonius, P. (2002): Subject positions and the placement of adverbials. In: P. Svenonius (ed.), Subjects, Expletives and the EPP, 201-242. New York: Oxford University Press. Vangsnes, Ø. A. (2005): Microparameters for Norwegian wh-grammars. Linguistic Variation Yearbook 5, 187-226. Amsterdam and Philadelphia: John Benjamins.

116

Marit Westergaard

Westergaard, M. R. (2003): Word Order in Wh-Questions in a North Norwegian Dialect: Some Evidence from an Acquisition Study. Nordic Journal of Linguistics 26.1, 81-109. Westergaard, M. R. (2005): Optional word order in wh-questions in two Norwegian dialects: A diachronic analysis of synchronic variation. Nordic Journal of Linguistics 28.2, 269-296. Westergaard, M. (2008a): Acquisition and Change: On the Robustness of the Triggering Experience for Word Order Cues. Lingua 118.12, 1841-1863. Westergaard, M. (2008b). Verb Movement and Subject Placement in the Acquisition of Word Order: Pragmatics or Structural Economy? In: P. GuijarroFuentes, P. Larranaga and J. Clibbens (eds.), First Language Acquisition of Morphology and Syntax: Perspectives across languages and learners, 61-86, Amsterdam and Philadelphia: John Benjamins. Westergaard, M. (2009a): Microvariation as Diachrony: A View from Acquisition. Journal of Comparative Germanic Linguistics 12.1, 49-79. Westergaard, M. (2009b): The Development of Word Order in Old and Middle English: The Role of Information Structure and First Language Acquisition. Diachronica 26.1, 65-102. Westergaard, M. (In press): The Acquisition of Word Order: Micro-cues, Information Structure and Economy. Amsterdam and Philadelphia: John Benjamins. Westergaard, M. R. and Ø. A. Vangsnes (2005): Wh-questions, V2, and the left periphery of three Norwegian dialects. Journal of Comparative Germanic Linguistics 8, 117-158. Westergaard, M. and K. Bentzen. (2007): The (non-) effect of input frequency on the acquisition of word order in Norwegian embedded clauses. In: I. Gülzow and N. Gagarina (eds.), Frequency effects in language acquisition: Defining the limits of frequency as an explanatory concept, 271-306, Berlin and New York: Mouton de Gruyter. Wiklund, A.-L., G. H. Hrafnbjargarson, K. Bentzen and T. Hróarsdóttir (2007): Rethinking Scandinavian verb movement. Journal of Comparative Germanic Linguistics 10.3, 203-233. Wexler, K. (1999): Very early parameter setting and the unique checking constraint: a new explanation of the optional infinitive stage. In: A. Sorace, C. Heycock and R. Shillock (eds.), Language Acquisition: Knowledge Representation and Processing, special issue of Lingua, 23-79. Amsterdam: Elsevier.

Discourse and Syntax in Linguistic Change: Decline of Postverbal Topical Subjects in Serbo-Croat Dejan Matić (Leipzig)

1 Introduction The extensive discussion of the principles of word order change in the last decades has resulted in a number of proposals on how to deal with the mechanisms and the motivation for this kind of diachronic development (see e.g. Lightfoot 1999 for an overview). The proposed solutions range from the allegedly universal principle of harmonic branching (Vennemann 1975) and adaptation to human parsing capacities (Aitchinson 1979, Hawkins 1990, 1994) to switches from one parameter setting to another (e.g. Lightfoot 1979, Kroch 1989). What seems to be common to all the different approaches to word order change is that they take two finite states of grammar as their starting points (the initial state X and the final state Y) and try to account for the stage between X and Y as a transitional state in which speakers can freely decide between at least two different grammatical choices, due either to the existence of more than one grammar in the internal language of the speakers (Kroch 1989, Pintzuk & Taylor 2006), or to the optionality of derivations within one grammar (Wurff 1997). The present paper deals with a word order change that cannot be attributed to syntactic processes and is therefore impossible to phrase in terms of competing grammatical choices. Postverbal topical subjects (PTS), i.e. subjects with topical interpretation placed to the right of the verb, have been attested throughout the history of Serbo-Croat (SC) 1 and

1

I would like to thank the participants and the organisers of the Workshop ‘The Role of Information Structure in Language Change’ at the 29th Annual Meeting of the DGfS for their invaluable input. I am also grateful my Serbo-Croat language consultants, too numerous to be listed here. This paper is dedicated to the memory of Srdjan Rajković (19702006). PTSs are more or less frequent in all Slavonic languages (cf. Bernecker 1900; see also Adamec 1966:66ff. for Russian, Jacennik & Dryer 1992 for Polish), in all languages of the Balkans (cf. Myhill 1986 for Rumanian, Matić 2004 for Albanian and Greek), and in most

118

Dejan Matić

seem to have kept the same frequency and the same range of functions till the 19th century. Then, in the course of the 20th century, the frequency and variability of PTSs rapidly declined, even though speakers are still ready to construe all types of clauses with PTSs attested in the sources from the 19th century. There is no evidence, as we shall see, that the decline of PTSs includes any kind of shift of parametric settings in the syntax. The reasons are rather to be sought in the changes affecting the principles of discourse organisation, changes which, in turn, may perhaps be traced down to a number of interrelated sociolinguistic factors. The reduced use of a syntactic structure thus seems to be initiated by changes in pragmatics. Clearly, it cannot be ruled out that the decline in use with no syntactic consequences – and this is roughly the state in which modern SC is with respect to PTSs – will sooner or later result in a genuine syntactic change. The point is that the idea of a transitional stage between two finite states of grammar, X and Y, in which both X and Y are represented in one way or another, does not seem to be the appropriate explanation for the development of PTSs in SC, and thus a fortiori not a universally applicable explanation of syntactic change. A further question to be addressed in this paper is that of persistent lexical instantiations of a construction, i.e. those that display high frequency of usage while the construction itself is in the process of disappearing. In order to offer an explanation for the phenomenon of persistent collocations, I suggest introducing a process of automatisation, a kind of binding of syntactic structures to certain lexemes, into the repertoire of diachronic changes. The paper is organised as follows: In Section 1, the defining features of postverbal topical subjects are introduced, in order to set this construction apart from other word order patterns with postverbal subjects. Section 2 provides a short description of the material basis and the methodology of the diachronic investigation on which the paper is based. Section 3 is dedicated to the comparative qualitative and quantitative analysis of the lexical, syntactic and discourse-pragmatic features of PTSs in two SC corpora – from the 19th and the 20th/21st centuries. A short overview of the patterns of usage of this construction that are still productive is provided in Section 4. Finally, in Section 5, a tentative explanation of the process in which the productive use of VsX construction has declined is put forward.

ancient Indo-European languages (Dressler 1969; see also Önnersfors 1997 and Hinterhölzl & Petrova 2005 for detailed accounts of PTSs in Germanic).

Decline of Postverbal Topical Subjects in Serbo-Croat

119

2 Postverbal topical subjects 2.1 PTSs in VSX construction In SC, both in the 19th century and in the modern variety, there are at least three structures in which the subject follows the finite verb. For ease of reference, I shall label them as follows (cf. Matić 2004 for a detailed synchronic account): (a) Inversion, a verb-second-like construction in which subjects appear postverbally after a fronted wh-word, relative pronoun, quote, and a fronted focus (cf. Alexiadou & Anagnostopoulou 2001, 2007); (b) vS, a construction with a focused postverbal subject, and (c) VsX, a construction with a topical postverbal subject. The present paper is devoted to the last of these, the VsX construction, in which postverbal subjects have an unequivocally topical interpretation. The label ‘VsX’ is borrowed from Jacennik & Dryer (1992). The presence of ‘X’ refers to the fact that in most cases, an additional element apart from the verb and the subject is present; capitalised ‘V’ and ‘X’, opposed to small ‘s’, are meant to give a rough indication of the intonational pattern characteristic of the construction, with accented verbs (+/- additional elements) and unaccented subjects. Example (1) displays the prototypical characteristics of the construction (the topical subject is italicised, stress is indicated with small capitals): (1)

IMAO sam ja ženu MISIRKU AUX-CL I wife Egyptian had ‘I had an Egyptian wife (= I was married to an Egyptian woman)’ (Andrić, Avlija, 1945)

Before turning to the fate of postverbal topical subjects in VsX, I will try to demonstrate that VsX is indeed a construction in its own right, distinct from other types of structures containing postverbal subjects. In contrast to VsX, which always occurs in matrix clauses without fronted material, Inversion is restricted to clauses in which a leftperipheral sentence slot is occupied by certain predefined types of syntactic objects (wh-words, quotes, etc.). Furthermore, the subject in Inversion may (though it need not) be accented, while the postverbal subject in VsX clauses has to be unaccented. Apart from this, there are some clear interpretational differences between the two constructions, the major one being that the subject in Inversion may, but need not be topical, whereas in VsX its topicality is a precondition for the use of the construction (see Matić 2004: 215-364). In short: Inversion is distinct from the VsX construction, i.e. the subject type used in Inversion is different from PTSs. From the practical point of view, this means that, for the purposes of this paper, the presence of a wh-word, relative pronoun, quote, etc. in the left periphery

120

Dejan Matić

of a clause is a sufficient condition to leave that clause out of consideration. The opposition between VsX and vS is less easy to grasp lexically. The two constructions are clearly distinct as to their interpretations, with VsX allowing only for topical, vS only for focused subjects (narrow focus on the subject or projected wide focus on the sentence). Syntactically and prosodically, there are some different statistical preferences (e.g. presence of an additional element after the subject in VsX vs. avoidance of postsubject material in vS; tendency for VsX clauses to be verb-initial vs. tendency for vS clauses to contain preverbal material, etc., cf. Matić 2004: 163ff., 372ff.), and at least three categorical differences: (a) Intonation contour: In vS clauses, subjects carry the nuclear stress (a stress is indicated with small capitals): (2)

vS (focused subject; all new, ‘thetic reading’) [What happened yesterday?] PETAR Pojavio se appeared REFL Peter ‘PETER appeared’

In VsX clauses with PTSs, the verb regularly gets a characteristic rise-fall intonational contour; if there is some material after the subject, it receives the nuclear stress. The subject itself can never be accented: (3)

VsX (topical subject, broad VP focus) [When did you finally see Peter?] naš Petar oko PET POJAVIO se appeared REFL our Peter around five ‘Our Peter appeared around FIVE’

(b) Position of sentence adverbials: In vS clauses, sentence adverbials may intervene between the verb and the subject, as exemplified in (2): (2’)

vS (focused subject) 3Pojavio se verovatno PETAR appeared REFL probably Peter ‘It was probably PETER that appeared.’

In contrast, the VsX construction does not allow for sentence adverbials to intervene between the verb and the topical subject: (3’)

VsX (topical subject) *? Pojavio se verovatno naš Petar oko pet appeared refl probably our Peter around five intended reading: ‘Our Peter probably appeared around FIVE.’

(c) Position of the subject relative to the copula/auxiliary: With focal subjects (vS construction), the subject may not intervene between the copula and the nominal predicate or the auxiliary and its complement:

Decline of Postverbal Topical Subjects in Serbo-Croat

(4)

(4’)

121

vS (focused subject) 3Nisu dostupni VATROGASCI not-are available firemen ‘FIREMEN are not available’ *Nisu vatrogasci dostupni

Postverbal topical subjects in the VsX construction may, and indeed most often do, intervene between the copula/auxiliary and the nominal predicate/complement: (5)

(5’)

VsX (topical subject) 3NISU dostupni vatrogasci DANAS not-are available firemen today ‘Firemen are not available today’ 3NISU vatrogasci dostupni DANAS

These three features – intonational differences plus syntactic discrepancies with respect to sentence adverbials and nominal predicates – may be analysed in a number of ways. For the present purposes, suffice it to conclude that vS and VsX are not only to be distinguished at the level of interpretation, but they also represent formally distinct constructions, so that their diachronic developments can be safely assumed to have run independently. 1.2 PTSs as continuous topics As will become clear in the section devoted to the diachronic development of the VsX construction (Section 3.5), postverbal topical subjects fulfil a limited set of well-defined discourse functions. I will attempt to show that these functions are all derivable from one basic meaning of the immediately postverbal position in SC – that of denoting continuous topicality within a discontinuous discourse frame. In an incremental model of communication, in which the mutual consensus on the way the world is increases step by step, with every utterance (cf. Stalnaker 1978), the status of topics with respect to this consensual knowledge may be twofold. If the speaker chooses a topic which the hearer is not entitled to expect on the basis of previous communication, the topic is discontinuous (or non-ratified, in the parlance of Lambrecht & Michaelis 1998). If, on the other hand, the speaker assumes that the hearer may expect assertions about a certain topic, it is continuous (or ratified) topics that are used, signalling that the interlocutors have reached the mutual consensus on the further increment of knowledge before the moment of the utterance. These two kinds of choices have some grammatical relevance: discontinuous topics tend to receive secondary stress and are as a rule placed in the left periphery or extraposed, whereas continuous topics

122

Dejan Matić

are typically encoded as unaccented pronouns or, if the language allows for pro-drop, they are not expressed at all (cf. Lambrecht 1994: 172ff.). In SC, rather unspectacularly, discontinuous topical subjects are usually full NPs or accented pronouns and are found in preverbal positions (6); continuous topical subjects are usually pro-dropped (7): (6)

(7)

[Nobody wanted to go to the bar with me] Eva je bila bolesna Eve AUX was ill ‘Eve wasn't well’ [I met Eve yesterday] Bila je bolesna was aux ill ‘She wasn't well’

What is, then, the place of postverbal topical subjects in this system? This is where the notion of discourse frame enters the scene. The term ‘discourse frame’ is used here to denote space and time within one discourse universe, i.e. scene, or one discourse universe among other possible discourse universes, i.e. perspective. I would like to argue that PTSs are used in those cases in which the topic they encode is continuous, but the discourse frame changes: (8)

[Eve had a headache then she got a fever and a bad cough. She could only lie in bed] Nije ona bila teško bolesna ali je NEG.AUX she was hard ill but AUX stalno kukala constantly whined ‘She wasn't seriously ill, but she was whining all the time’

The topic of this constructed passage is continuous, in the sense that the speaker makes a series of assertions about Eve, so that the hearer is entitled to expect further information about her. The discourse frame, however, changes: in the first part (given in square brackets), the speaker assumes the perspective of an observer merely reporting what happened. In the second part (the one in which the PTS ona occurs), the perspective changes: the speaker is not an objective observer anymore, s/he is commenting upon the state of affairs s/he has just described. This combination of topic continuity and discourse frame discontinuity is unambiguously marked by the use of a PTS. Note that (8) can also be expressed with a pro-dropped subject, or, under certain circumstances, even with a preverbal subject. This is because zero topics and preverbal topics are not sensitive to changes of scene/perspective, so that they can be used with both continuous and discontinuous discourse frames; PTSs, on the other hand, are used only with

Decline of Postverbal Topical Subjects in Serbo-Croat

123

the latter, and only in those cases in which the speaker considers it worthwhile to mark discontinuity in the ongoing discourse. In terms of markedness, the behaviour of the three types of topical expressions can be represented as follows: Preverbal Topics Zero Topics Postverbal Topics

Topic Continuity – + +

Discourse Frame Continuity +/– +/– –

Table 1: Meanings of topic expressions in SC

It is important to emphasise that the use of PTSs has always been optional, where ‘optional’ does not mean ‘arbitrary’, but rather ‘regulated by the rules of discourse (as opposed to those of grammar)’. Taking this kind of optionality, i.e. discourse determinedness, as a starting point, I shall try to demonstrate that it was the changes in the conventions of discourse organisation that eventually led to the ousting of PTSs by preverbal and zero topics.

2 Corpus and methods The present study is based on a comparison of two Serbo-Croat corpora, one comprising texts from the 19th, the other from the end of the 20th and the beginning of the 21st centuries. The corpora consist of approximately 10.000 clauses each (19th century SC: 10.012, modern SC: 10.005). An attempt has been made to include the most relevant registers and text types in both corpora, so as to provide for comparability and as broad a coverage of genres as possible. First, two representatives of the journalistic style have been chosen for each corpus (19th century SC: Letopis MS, Danica; modern SC: Vjesnik, Vesti). Second, both corpora contain parts of dramatic texts, which stand for (an imitation of) the spoken language (19th century: Sterija, Rodoljupci, Pokondirena; Nušić, Sumnjivo, Narodni; modern SC: Kovačević, Špijun, Maratonci). Third, instances of predominantly narrative prose writing and of expository prose have been included in both corpora (19th century: Novak, Stipančići, Njegoš, Pisma; modern SC: Pavić, Predeo; Kiš, Grobnica). Finally, as representatives of oral narratives, folk stories collected in the 19th century (Pripovetke) and a 20th century novel whose language is largely based on the techniques of oral storytelling (Andrić, Avlija) have been added. Obviously, introspective and elicited data are not applicable in a diachronic study. However, I have used my own and the judgments of other SC speakers as a kind of additional material in order to gauge to what

124

Dejan Matić

extent certain usages of VsX and PTSs are still acceptable in the modern language.

3 PTSs in the 19th and 20th centuries 3.1 Overall frequency The first step in investigating the development of postverbal topical subjects in SC is to compare their absolute frequencies in the corpora. The relevant data are given in Table 2. Ratio of PTS among overt S Ratio of PTS on the whole

19th century

20th century

9.7% (584) 5.8% (584)

1.4% (79) 0.8% (79)

Table 2: Frequency of VsX clauses with PTSs in the 19th and 20th centuries

The general tendency is obvious: clauses with PTSs are roughly seven times less frequent in modern SC than in the 19th century, both within the category of clauses with overt subjects and in the corpus as a whole. In what follows, I shall present the semantic, pragmatic and syntactic parameters along which this rapid decline in frequency took place. 3.2 Subjects A very conspicuous feature that changed within the time span under consideration is the lexical nature of the subjects. Since PTSs encode continuous topics, it is only natural that the proportion of pronominal subjects in the VsX construction is somewhat higher than average. However, this proportion has dramatically changed between the 19th and the late 20th/early 21st century: Pronominal PST Full NP PST

19th century

20th century

32.0% (187) 68.0% (397)

64.6% (51) 35.4% (28)

Table 3: PTSs in VsX clauses in the 19th and 20th centuries

Whereas pronominal subjects make up approximately one third of all occurrences of postverbal topical subjects in the 19th century, in modern SC they are by far the most frequent PTS type at all: almost two thirds of

125

Decline of Postverbal Topical Subjects in Serbo-Croat

PTSs are pronominal, with only one third being expressed with full NPs. Here are two typical examples: 19th century Neće Ilić biti biran NEG.FUT Ilić be chosen ‘Ilić will not be chosen’ (Nušić, Narodni, 1883) 20th century Neću ja […] da lupam glavu o tome NEG.want I to break head about that ‘I don’t want to wrack my brains with it’ (Andrić, Avlija, 1954)

(9)

(10)

Note that (10) and (11) stand merely for typical PTSs of their respective periods: a clause like (10) is still fully grammatical and possible in modern SC, and examples like (11) are abundantly attested in the 19th century texts. The fact is, however, that the frequency of clauses like (10), with full NP subjects, decreased by approximately fifty per cent in the course of one century. 3.3 Verbs There are also interesting differences in the lexical semantics of verbs with which postverbal topical subjects appear in the VsX Construction, as Table 4 clearly shows. 19th century 20th century

copula 9.9% (58) 40.5% (32)

modal 7.7% (45) 15.2% (12)

cognition 15.1% (88) 17.7% (14)

emotion 13.3% (78) 6.3% (5)

existence 10.5% (61) 10.1% (8)

Ȉ 56.5% (330) 89.9% (71)

Table 4: Semantics of the predicates in VsX clauses in the 19th and 20th centuries

There are five classes of predicates prototypically occurring in VsX clauses with PTSs: the copula (biti), modal verbs (hteti [want/will], moći [can], smeti [may]), and verbs of cognition (znati [know], razumeti [understand]), emotion (uplašiti se [get scared], razljutiti se [get enraged], etc.) and existence (esp. the dynamic ones, such as pojaviti se [appear] or nestati [disappear]). The most conspicuous difference is certainly the one represented in the last column: the five prototypical verb classes cover only little more than a half of the occurrences of PTSs in the 19th century SC, whereas in the modern language they occur in almost 90 per cent of instances. This is mostly due to a strong increase in the incidence of copular predicates (from 10 to 40 per cent) and of modal verbs (from 7 to 15 per cent); the other three classes remain at a similar level (cognition, existence) or even decrease (emotion).

126

Dejan Matić

The interpretation of these facts lies at hand. The difference in the overall frequency of prototypical predicate classes can only be explained by the much wider range of predicates with which PTSs used to be combined in the 19th century. This is an indirect sign of the greater productivity of the structure in which they occur, the VsX construction. PTSs in modern SC appear mostly with the copula, modal verbs and verbs of cognition, which account for more than 70 per cent of all occurrences. All other predicate classes are in use at best sporadically. Two typical examples are given in (11) (non-typical predicates ‘grab’ and ‘bite’) and (12) (typical predicate ‘be’): (11)

(12)

19th century Dugo se boraše Uhvati on zmiju za gušu long REFL fought caught he snake for throat Ujede ga zmija bit him snake ‘They fought for a long tome. He grabbed the snake by the neck. The snake bit him and he ran away…’ (Pripovetke, ca. 1820-1850) 20th century Bilo je to jednog užasno studenog popodneva was AUX that one terribly cold afternoon [after a description of a funeral]‘This happened on a terribly cold after noon’ (Kiš, Grobinica, 1976)

A caveat similar to the one mentioned with respect to pronominal subjects is in order here. There are no grammatical, semantic, or even pragmatic restrictions on the predicates with which PTSs may be used in modern SC. The statistics presented in Table 4 and examples (11) and (12) merely show that speakers of SC have gradually stopped using PTSs with most of the predicates they were actively used with only a century ago: Even though they still can form sentences in which PTSs would be combined with any verb, they seldom do. 3.4 Focus Structure The VsX construction is compatible with two kinds of focus construal: polarity (or verum) focus on the verb and wide focus on the verb and the elements to the right of the subject (see Matić 2004: 182ff. for more detail on these two types of focus construal). An example of wide focus is given in (13), in which the focus scope stretches over the predicate and the direct object to the exclusion of the topical subject (it is asserted about Ante Stipančić that he ordered a particular clothing for himself) 2 . Example (14) 2

Note that the wide focus construal in a VsX clause results in a discontinuous focus domain

Decline of Postverbal Topical Subjects in Serbo-Croat

127

illustrates polarity focus on the verb (among the two possible polarities explicitly given in the question, it is the negative one that is asserted, triggering focus on the negated finite predicate): (13)

(14)

19th century [Preparations of a family for the baptism of their new born son are de scribed] Naručio je bio Ante Stipančić za se odijelo AUX AUX A.S. for REFL clothes ordered senjskih plemića Senj:adk noblemen ‘Ante Stipančić had ordered for himself the uniform of the Senj nobil ity…’ (Novak, Sti pančići, 1899) FOCUS CONSTRUAL [Ante Stipančić]Top [had ordered the uniform]Foc 20th century [A: »Can a…hen, which is worth five cervonecs, be considered equal to a polecat, which … stinks awfully? «] B: »Ne, … NE MOŽE se ravnati kokoška koja vredi no NEG can REFL compare hen which is.wort pet červonaca sa smrdljivim tvorom. « five cervonec with stinking polecat. « ‘»No,…a hen which is worthy five cervonecs cannot be compared to a stinking pole cat.«’(Kis, Grobnica, 1976) FOCUS CONSTRUAL [{a hen can be compared to a polecat}, {polarity = X}]Background [X = ™]Foc

The relative frequency of these two types of focus construal has changed in the last hundred years, as the following table shows: Polarity Focus Construal Wide Focus Construal

19th century 26.1% (152) 73.9% (432)

20th century 70.9% (56) 29.1% (23)

Table 6: Polarity focus and wide focus with PTSs in the 19th and the 20th centuries

What we see is almost a reversal in the frequency of the two focus types: whereas in the 19th century wide focus was the dominant option (some three quarters of all instances), in the 20th century SC it is the polarity focus construal that dominates, also with almost three quarters of all instances. Again, this is not a matter of grammatical or any other restrictions: both the wide and the polarity foci can still be used with PTSs, so that a clause like (14) would be fully acceptable today as it was in the 19th

in which the subject disrupts the flow of focal information (see Lambrecht 1994 and Matić 2003 for more detail on discontinuous focus domains).

128

Dejan Matić

century. However, only the latter option seems to be actually employed with some regularity. 3.5 Discourse Functions As already mentioned, VsX clauses with PTSs perform certain discourse functions in a number of well-defined contexts. The nature of these functions is determined by two factors: (a) by the discourse meaning of continuous topics, which are used in contexts in which the referential frame is stable, while the spatio-temporal stage or the perspective change, and (b) by the focus construal. VsX clauses with polarity focus are employed in two basic groups of functions: first, to express certain types of illocutionary force (confirmation/denial, promise/threat, question/answer/wish, etc. – cf. (14) for the use of PTSs in answers, and (15) for a PTS clause expressing reassurance), and second, as textual cohesion devices (concession, adversativity – cf. example (16) for a VsX clause in a concessive context). (15)

(16)

20th century A: »…što ti nisi najurio tog svog … why you NEG.AUX chase.away that REFL otrova…?« poison…? B: » najurio najurio Njie to tako lako…« chase.away chase.away NEG.COP that so easy…« A: »Ah, šta! Najurio bih ja nju…« ah, what! chase.away AUX I her…« [A conversation about an evil wife] ‘A: »...why didn’t you simply chase away that snake?« ... …B: »Chase away, chase away! It is not that easy....« A: »Non sense! I WOULD have chased her away...!«’ (Andrić, Av lija, 1954) 19th century …može čovek imati kožu meku kao rukavica i …can man have skin soft as glove and mirisati na parfem pa opet da ima revolver smell on parfume and again to have revolver u džepu in pocket ‘… a person may have a skin as soft as a glove and smell of parfume, but still he may have a revolver in his pocket.’ (Nušić, Sumnjivo, 1886)

Wide focus VsX clauses are often used to mark descriptive digressions in narrative chains (cf. ex. (12) for a digressive clause containing the copula), to open or close a paragraph (ex. (17), also with the copula), or to resume an interrupted narrative thread. Furthermore, they appear in clauses that

Decline of Postverbal Topical Subjects in Serbo-Croat

129

encode consecutive actions (ex. (11)) or events representing reactions to other events (cf. ex. (18)). 3 (17)

(18)

20th century (paragraph closing) Bila je to kruna svaga njegova AUX that crown all his was nastojanja da se… svijetu effort that REFL world predoči što je jedan mali narod sposoban stvoriti show what is one small nation capable make [In a funeral speech for the Croatian president Tudjman: ‘Dr. Tudjman returned from Rome tired, but proud of the exhibition which was a con firmation of the Croatian identity, which is more than thousand years old.] It was the culmination of his efforts ... to show the world what a small nation is able to create...’(Vijesnik, 1999) 19th century (reactive event chains) Bio je on poša pred svjatago Nikolaja na AUX AUX he go before holy N. to Stanjeviće […] šilja sam ja za njega momke S. send AUX I after him boys dva puta two times [This is how it all happened.] ‘He had gone to Stanjevići before St. Nicholas’ Day [...] and I sent my boys after him two times.’(Njegoš, Pisma, 1831)

The diachronic development of discourse functions performed by PTSs is presented in Table 7. Note that within the group of VsX clauses with wide focus, a distinction between the clauses with copular predicates and those with all other types of predicates is made, since the diachronic developments of the copular and all other VsX clauses seem to have taken a different course.

3

For reasons of space, not all of these functions can be illustrated with examples and explained in detail. For a detailed account of the situation in modern SC in comparison to Albanian and Modern Greek, readers are referred to Matić (2004: 182-204).

130

Dejan Matić

Polarity Focus: Illocution Polarity Focus: Textual Cohesion Wide Focus: Copular Predicates digression in narration paragraph boundaries Wide Focus: Other Predicates digression in narration paragraph boundaries resumption of narration reactive chains consecutive actions

19th century + +

20th century + –/+

+

+

+

–

Table 7: Discourse functions of PTSs in the 19th and 20th centuries

Polarity focus clauses with PTSs have retained their illocutionary uses to the full extent. Their function to mark textual cohesion via adversative or concessive rhetorical relations has been practically lost: in the modern SC corpus only 3 instances have been found, which stands in sharp contrast to some fifty clauses in the 19th century corpus. The modern language seems to opt for explicit marking via adversative, concessive, etc. particles instead. PTS clauses with wide focus construal have survived as a productive syntactic structure only with copular predicates. If combined with the copula, PTSs are still used in digressions and on paragraph boundaries. The combinations of PTSs and other predicate classes with wide focus construal are not attested in the modern SC corpus. This means that the discourse functions this clause type used to fulfil in the 19th century language are not a living option for the speakers of SC anymore. Thus, at the level of attested examples, only the illocutionary function of polarity focus clauses and two minor functions of the copular wide focus clauses seem to be still productive, at least to a certain extent. However, if native speaker intuitions are taken into consideration, i.e. if we observe not only what speakers actually do, but also what they are able to do, a picture is slightly different. All but two discourse functions of PTSs are still latently possible, i.e. speakers would still be able to use VsX clauses with all kinds of predicates with both polarity and wide focus construals in almost all discourse functions attested in the 19th century. The only exceptions are PTSs used in encoding reactive chains (19) and consecutive actions (12): VsX clauses employed in these functions are felt to be archaic and impossible in modern SC by all the native speakers I consulted.

131

Decline of Postverbal Topical Subjects in Serbo-Croat

3.6 Syntax: Adjacency In Section 1.1 we saw that sentence adverbials between the verb and the subject are ungrammatical. With other types of constituents, this restriction cannot be expressed in terms of grammaticality, but the statistical tendencies are obvious: the finite verb and the subject tend to be adjacent, with all the additional elements of the clause following the subject. This tendency, although present already in the 19th century, seems to have become especially strong in modern SC, as the following table shows: [Finite Verb] [Subject] [X] [Finite Verb] [X] [Subject]

19th century 68.8% (402) 31.2% (182)

20th century 94.9% (75) 5.1% (4)

Table 8: Relative position of V and S in the 19th and 20th centuries

Obviously, the adjacency between the finite verb and the PTS is practically a rule in today’s SC, while it used to be merely a preferred ordering in the 19th century. Examples (19) and (20) are typical in this respect: (19)

(20)

19th century Dakle idem od jutros ja čaršijom so go since morning I town: INSTR ‘So, I’ve been going through the town since morning...’ (Nušić, Narodni, 1883) 20th century Idem ja tako svakog dana Kondino go I so every day Konda:POSS ulicom… street:INST ‘Every day, I go along Konda’s street...’ (Vesti, 2000)

Similar to all other features described so far, this difference between the two varieties of SC is not categorical: If asked to insert an element (other than sentence adverbial) between the verb and the subject, the speakers of modern SC will do so without hesitation, but, as the statistics shows, they do not exploit this possibility in their everyday speech.

4 What is left The statistical data we looked at so far reveal that the variability of the VsX construction, in which PTSs appear, has diminished in the last hundred years, which implies a gradual loss of the productivity of the construction. If we look at the most frequent patterns of usage of PTSs in modern SC, the impression of diminished productivity is confirmed.

132

Dejan Matić

What is left and what accounts for some sixty to seventy per cent of all instances of PTSs in modern SC is a number of more or less petrified verb-subject collocations which regularly receive one focus construal, typical for that particular V-S combination, and are regularly used in one particular discourse function. The most frequent combinations are listed here: (21)

Modal Verb/Verb of Cognition/Emotion + Pronominal Subject (ne) znam ja/znaš ti/zna on ... [(not) know I/know you/knows he ...] ‘I (don’t) know/you (don’t) know /he (doesn’t) know…’ (ne) umem ja / umeš ti / ume on ... [(not) can I/can you/can he ...] ‘I am (not) able/you are (not) able/he is (not) able…’ neću ja / nećeš ti / neće on … [not.want I/not.want you/not.want he ...]’ ‘I don’t want/you don’t want/he doesn’t want…’ (ne) mogu ja /možeš ti / može on …[(not) can I/can you/can he... ] ‘I can(’t)/you can(’t)/he can(’t) …’ (ne) volim ja / voliš ti / voli on [(not) love I/love you/loves he...] ‘I (don’t) like/you (don’t) like/he (doesn’t) like…’

The type of verb-subject combination illustrated in (22) regularly receives polarity focus interpretation and expresses one of the illocutions typically connected with PTSs (confirmation, denial, threat, promise, etc.). It covers some 28 per cent of all instances of PTSs in modern SC. The second frequent type of the VsX construction still very much in use – the copula with a pronominal subject – is confined to wide focus construal and appears exclusively in descriptive digressions in narration and on paragraph boundaries. It accounts for some 35 per cent of all instances of PTSs in modern SC. (22)

Copula + Pronominal Subject bio je to / bila je to / bili su to [was.M/.F/.PL AUX that] It was (him/her/them) …’’ nije to bio / nije to bila / nisu to bili [NEG.AUX that was.M/.F/.PL] ‘It wasn’t (him/her/them)…’ bio sam ja / bio je on / bili su oni [was AUX I/he/they] ‘I was/he was/they were …’

The situation can thus be summarised as follows: Focus Construal Discourse Function Percentage Percentage (Ȉ)

Modal Verb, etc. + Pronoun polarity focus illocution

Copula + Pronoun wide focus digression paragraph boundaries 27,8% (22 tokens) 35,4% (28 tokens) 63,2% (50 tokens) Table 9: Patterns of usage of PTSs in modern SC

Decline of Postverbal Topical Subjects in Serbo-Croat

133

The most interesting row in the table is the last one, which shows that some ten verb-subject combinations account for approximately two thirds of all occurrences of PTSs in modern SC. The productivity of the construction is thus definitively to be considered very low, since most of the instances seem to be taken from the lexicon as ready-made combinations, and are not formed by productive rules of grammar.

5 Interpretation 5.1 Mixed grammars As indicated in the introduction to this paper, word order change has been subject to many interpretations in the last decades, the common denominator of all approaches being the concept of an internally inconsistent transitional state between two finite, internally consistent stages of grammar, the initial stage X and the final stage Y. During this transitional state, the speakers freely use the features of both X and Y, the usual explanation for this apparent inconsistency being the coexistence of both grammars in the internal language of the speakers. For instance, the transition of English from OV to VO order is believed to have included a stage in which the speakers internalised both a left-branching and a right-branching grammar (Pintzuk & Taylor 2006). In other words, word order change is supposed to follow the scheme X > X+Y > Y. Is this teleological view of syntactic change applicable to the SC data presented in the preceding sections? Let us take a closer look at how an explanation in terms of transitional states (henceforth mixed grammars approach, MGA) would work with PTSs in SC. The initial state would be a grammar in which non-focal subjects may follow the finite verb. This would be licensed only under certain discourse conditions, which could be accounted for in terms of discourse-driven syntactic features. Let us call this state VsX Grammar. In the final state, the possibility of non-focal subjects following the verb would be excluded – say, because of the change in strength or loss of the postulated features; trivially, this would be NonVsX Grammar. In this scenario, modern SC is in a transitional state, displaying features of both VsX Grammar and Non-VsX Grammar. The speakers have a choice between the two grammars and tend to employ VsX Grammar with pronominal subjects and copular predicates, as well as with polarity focus, and Non-VsX Grammar in all other cases. The main problem with this kind of approach is the lack of categorical evidence that would support it. Apart from one discourse-related feature (see below), all differences between 19th century and modern SC are grad-

134

Dejan Matić

ual rather than discrete, which is confirmed at two levels. First, all semantic, pragmatic and syntactic features attested in the 19th century texts are still found in modern SC texts, only that their frequency is considerably lower; second, speakers of SC are still able to produce clauses with PTSs with the 19th century features without judging them strange or archaic. This means that PTSs are still a part of the internal language of the present-day speakers, though a part rarely made use of. Now, it could be argued that this is not an argument against MGA: the hidden productivity of PTSs may be attributed to the still-existent VsX Grammar, the scarcity of their attestation to the growing Non-VsX Grammar. This objection suffers from one major methodological flaw: it is not falsifiable (or, for that matter, verifiable). If there are no clues for the existence of different parametric settings in the grammar – and the data presented above provide us with only one syntactic difference, which has a plausible pragmatic explanation (see 5.3) – then to posit two grammatical systems would be a pure stipulation without empirical support. Furthermore, the MGA approach would be circular here: differences in frequency, lexical properties, and discourse functions are explained by a mixed syntax, and they are the only evidence for its existence at the same time. Therefore I conclude that the change observed in the behaviour of PTSs in SC in the last hundred years is not to be handled with any variety of MGA. In fact, I shall argue that it is not a grammatical change at all, i.e. that we are much better off in explaining it if we assume that no difference in syntactic parameters is involved. 5.2 Shifts in Discourse Organisation In this section, I am going to provide a (rather sketchy) account of the changes in discourse organisation which seem to have been decisive in the gradual loss of PTSs. Consider first the only change between the 19th and the 20th centuries that seems to have led to something resembling a categorical difference: VsX clauses with PTSs cannot function as markers of resultative and consecutive events in narratives anymore, the speakers judging them at best as very old-fashioned and, if used in modern discourse, downright weird (Section 3.5). These two discourse functions are responsible for much of the lexical variability and for the frequency of VsX clauses in the 19th century corpus: On a moderately generous interpretation of discourse functions, resultative and consecutive VsX clauses make out some 38% of all instances of PTSs (222: 584). Since they appear in narrative texts, clauses marking resultative and consecutive events may be filled with the full gamut of predicates compatible with narrative contexts; since narrative contexts normally include more than two topical discourse

Decline of Postverbal Topical Subjects in Serbo-Croat

135

referents, so that pronominal reference is often not sufficiently precise, the incidence of full NP subjects in these clauses is much higher than in clauses with other typical PTS functions 4 . Thus, the loss of these two discourse functions implied a significant decline in both the frequency and the lexical variability of clauses with PTSs. Note that no syntactic change is postulated here: the syntax of SC clauses remains the same, what changes is the principles of discourse organisation, and this, in turn, has consequences on the productivity of the construction. The speakers of SC used to mark a series of actions as resulting from one another or as following one another with PTSs (topical discourse referents remain constant, the perspective changes). As indicated in Section 1.2, the use of PTSs was always optional, meaning that the speaker had a choice to mark the change of the discourse frame or to leave it underspecified. In the 19th century, the conventions of creating a narrative chain seem to have favoured the explicit marking; in the 20th century, the only living option is to leave it unmarked. Full NP PTSs in resultative and consecutive contexts are now replaced with preverbal topical subjects, i.e. with the SV order, while the place of VsX clauses with pronominal PTSs is taken by zero subject clauses. The loss of resultative and consecutive functions is not the whole story in the history of the decline of PTSs. Though it does account for much of the loss of productivity of the construction, some minor discourse factors seem to have played a role as well. First, wide focus VsX clauses on paragraph boundaries and in digressions in narration are used only with the copula in modern SC, while 19th century SC made use of all predicates in this function. This seems to be due to another change in discourse organisation: explicit marking of the structure of the paragraph via PTSs is not en vogue anymore. Both the written and the spoken language (as far as can be judged by the language of the theatre plays in my corpus) in the 20th century seem to prefer unmarked paragraph boundaries (usually with zero subjects) and insert digressions without an indication of the change of the discourse frame. My native speaker intuitions are that the use of PTSs in these contexts is felt to be somehow ‘uncool’ and schoolmasterly (though not so quaint and archaic as resultative and consecutive clauses), something you would write for your teacher in a prose composition course but never utter among your friends. Thus, this change in discourse habits contributed further to the decrease in frequency of PTSs. The reason for the persistence of the copula in this function will be the topic of the following section. 4

The overall percentage of full NP subjects in VsX clauses in the 19th century is 68,0% (397), while full NP subjects in resultative and consecutive VsX clauses amount to 88,2% (191).

136

Dejan Matić

Before that, another minor change needs to be briefly mentioned. Concessive and adversative rhetorical relations expressed with polarity focus VsX clauses are still, although rarely, found in texts, but the rhetoric of the written language has shifted to the marking of these relations with conjunctions and/or particles (such as iako, mada ‘although’ or ali ‘but’), usually combined with zero subject clauses. In contrast to the 19th century, concessive etc. clauses with PTSs are not considered very good style in the present-day written language, but seem to be at least marginally acceptable in the colloquial registers. The common feature of all these changes seems to be the tendency to avoid explicit marking of the change of the discourse frame under referential continuity and to leave this feature underspecified. It is unclear what triggered this new preference in discourse organisation. No research whatsoever has been done on the historical pragmatics of Serbo-Croat, so that I can merely offer a couple of hypotheses that may be a part of the explanation. First, the growing influence of Western European languages, which generally leave the change of discourse frame unmarked, may have contributed to the ousting of PTSs. Second, reactive and consecutive VsX clauses are felt to be characteristic of oral narratives and have an unmistakable folklore slant. While this may have been seen as a positive connotation in the time of Romanticism, it was definitely a feature to be avoided in the era of urbanisation and conscious creation of non-rural identity, a social process which gained momentum in the Balkans only in the 20th century. Third, the shift of regional centres: while 19th century SC was mostly based on the central-southern speech of Herzegovina, in which VsX clauses still seem to be a productive discourse device, the linguistic centre later moved further north, to Zagreb and Belgrade. It may be the case that the shift to the north also meant a shift away from the structures which were felt to be especially characteristic of the previous linguistic centre. I am not sure that any of these sociolinguistic factors was decisive, but it is at least plausible to assume that a shift in the pragmatics of a language has its roots in changes of the communicative preferences of its speakers, which may very well be sociologically conditioned. 5.3 Automatisation There is one final issue that needs to be addressed – the issue of the types of PTSs which are still relatively frequent in modern SC. As shown in Section 4, clauses containing them fall into two groups: (1) polarity focus VsX clauses with modal verbs/verbs of cognition/emotion + pronominal subjects marking different illocutions, and (2) wide focus VsX clauses with the copula + pronominal subjects denoting paragraph boundaries or di-

Decline of Postverbal Topical Subjects in Serbo-Croat

137

gressions. Why have precisely these two types remained more or less intact, despite the general tendency in SC to avoid explicit marking of discourse frame discontinuity via PTSs? The answer to this question is rather unspectacular: token frequency. In polarity focus clauses used to express threats, promises, agreement and disagreement, etc., the most frequent lexical items are verbs denoting modality, emotions and propositional attitudes, usually coupled with a 1st and 2nd person pronominal subject, since the aforementioned illocutions normally pertain to ability, obligation, emotions and opinions, and they normally involve the speech act participants. The wide focus clauses appearing in digressions and on paragraph boundaries are almost as a rule descriptive or identificational, which implies the use of the copula, and their topics are usually discourse referents mentioned in the previous clause, so that their subjects tend to be pronominal with a greater than chance frequency. As is often the case in the course of phonological, morphological or syntactic change, this kind of frequency phenomena may have a conserving effect even in a most sweeping language change. In the case of syntactic phenomena like the one we are dealing with, this conserving effect of particularly frequent tokens consists in the method of their storage and retrieval: instead of being accessed as productive constructions via the principle of compositionality, these tokens are both stored and retrieved as quasi-lexical items (they are ‘entrenched’, cf. Bybee 2007: 273). This enables them to survive the modern SC drift toward full underspecification of the changes in the discourse frame and the resulting obsolescence of PTSs. Incidentally, the semi-lexicalisation of the most frequent verb-subject combinations also accounts for the adjacency effects observed in Section 3.6. Even though various lexical items may still intervene between the verb and the PTS, in more than 90% of the cases the verb and the PTS are adjacent, whereas in the 19th century less than 70% of the PTSs directly followed the verb. Recall that the semi-lexicalised verb-subject combinations account for some 63% of all PTSs in the 20th century corpus. With only one exception (ex. (15)), all clauses containing these combinations display verb-subject adjacency. Direct adjacency of this kind is a typical feature of ‘entrenched’ phrases (cf. Bybee 2007: 274), which is only logical in view of their being stored as ready-made phrases, not generated from generalised syntactic schemas. The rise in the ratio of the clauses with adjacent verbs and subjects is thus directly attributable to the rise in the ratio of semi-lexicalised expressions, so that there is no need to resort to incipient parametric changes in order to explain this phenomenon.

138

Dejan Matić

Partial lexicalisation is not the whole story, however. For three out of the eight verb-subject collocations listed in Section 4 (bio je to ‘it was’, nije to bio ‘it wasn’t’, and neću ja ‘I won’t/don’t want to’), I was not able to find one single instance of subject-verb order in the 20th century corpus (even though the order to je bio etc. would be perfectly grammatical). This means that semi-lexicalisation affects not only the collocation itself, but also its component parts. When the speakers are confronted with a lexical choice including a particular verb and a particular subject, they tend to automatically use the verb-subject order. Note that this is a reversal of the normal derivational process out of abstract syntactic schemes. In syntactic derivations a particular structure is paired with a particular information packaging and a particular semantics; these three components together determine the lexical filling and the discourse function of the structure. In the case of the three collocations mentioned above, the process goes the other way around: a particular lexical filling triggers the use of a particular structure. I suggest to call this type of diachronic change automatisation, and to define it as a binding of lexical items to a syntactic structure, such that the use of the lexical items automatically triggers the use of the syntactic structure. Of course, automatisation is only very weak in the case of the Serbo-Croat PTSs, since the number of lexemes affected is rather low and even those few lexemes allow for other kinds of construal, but it does seem to be well under way, at least for this limited set of lexemes.

6 Conclusion The evidence presented seems to suggest that it is more plausible to conceive of the decline of PTSs in SC as a consequence of the change in the principles of discourse organisation than to postulate syntactic, or parametric, or I-language change. It is possible that, if the discourse habits of the speakers of SC remain as they are now for some time, the use of postverbal topical subjects will cease to be a living option not only at the level of discourse, but also at the level of grammar, i.e. what once started as a simple shift in the organisation of discourse will turn into a new parametric setting. If this happens, 20th century SC will turn out to be a kind of transitional stage. However, this would be a transitional stage quite different from the one conceived of as a coexistence of two opposed grammars, since it merely includes a disuse of a grammatically perfectly possible structure for reasons beyond grammar. Of course, this in no way implies that the idea of mixed grammatical parameters is not applicable to all those cases to which it has been applied successfully. The implication is rather that the inventory of transitional

Decline of Postverbal Topical Subjects in Serbo-Croat

139

stages should be widened to include cases in which the initial grammar is still present, but out of use for some reason. And there is one more implication: grammatical loss can have more than just one possible trigger, contrary to the popular claim, neatly exemplified in the following quote: People might start to use some new expression because of the social demands of fashion or because of the influence of speakers from a different community, but people do not cease to say things for that sort or reason. […] Changes involving only the loss and obsolescence of forms need to be explained as a consequence of some change in an abstract, cognitive system. (Lightfoot 2006: 29; cf. also Lightfoot 1999:106)

As I hope to have shown, people may very well stop saying things they used to say for reasons that have to do with social demands of fashion, without deep changes in an abstract, cognitive system. Some of these fashionable changes may, of course, strike roots and lead to deeper grammatical shifts.

References Primary sources Andrić, Avlija (1954): I. Andrić, Prokleta avlija. Sarajevo: Svjetlost, 1981. Danica (1849): Danica ilirska 15, Zagreb, 1849. Kiš, Grobnica (1976): D. Kiš, Grobnica za Borisa Davidoviča. Beograd: BIGZ, 1986. Kovačević, Maratonci (1973): D. Kovačević, Maratonci trče počasni krug. In: Odabrane drame I. Beograd: Stubovi kulture, 2001. Kovačević, Špijun (1982): D. Kovečević, Balkanski špijun. In: Odabrane drame I. Beograd: Stubovi kulture, 2001. Letopis MS (1874): Letopis Matice srpske 116, Novi Sad, 1874. Njegoš, Pisma (1831-2): P. Petrović Njegoš, Izabrana dela: Gorski vijenac, Luča Mikrokozma, Pisma. Beograd: Draganić, 1998. Novak, Stipančići (1899): V. Novak: Posljednji Stipančići. Zagreb: Mosta, 1994. Nušić, Narodni (1883): B. Nušić, Narodni poslanik. Beograd: Rad, 1998. Nušić, Sumnjivo (1886): B. Nušić, Sumnjivo lice. Beograd: SKZ, 2003. Pavić, Predeo (1988): M. Pavić, Predeo slikan čajem. Beograd: BIGZ, 1990. Pripovetke (ca. 1820-1850): V. Djurić (ed.), Antologija narodnih pripovedaka. Novi Sad: Matica Srpska, 1960.

140

Dejan Matić

Sterija, Pokondirena (1838): J. Sterija Popović, Pokondirena tikva. In: Dela. Beograd: Draganić, 2006. Sterija, Rodoljupci (1856): J. Sterija Popović, Rodoljupci. In: Dela. Beograd: Draganić, 2006. Vjesnik (1999): Vjesnik, Zagreb, 14.12.1999. Vesti. (2000): Vesti, Frankfurt a.M., 04.03.2000.

Scientific Literature Adamec, P. (1966): Porjadok slov v sovremennom russkom jazyke. Praha: Academia. Aitchinson, J. (1979): The order of word order change. Transactions of the Philological Society, 42-65. Alexiadou, A. and E. Anagnostopoulou (2001): The subject in situ generalization, and the role of Case in driving computations. Linguistic Inquiry 32, 193-231. Alexiadou, A. and E. Anagnostopoulou (2007): The subject-in-situ hypothesis revisited. In: H.-M. Gärtner and U. Sauerland (eds.), Proceedings of the Workshop on Interfaces + Recursion = Language? Berlin and New York: Mouton de Gruyter, 31-60. Berneker, E. (1900): Die Wortfolge in den slawischen Sprachen. Berlin: Behr. Dressler, W. (1969): Eine textsyntaktische Regel der idg. Wortstellung. Zeitschrift für vergleichende Sprachforschung 83, 1-25. Hawkins, J. (1990): Seeking motives for change in typological variation. In: W. Croft, K. Denning, and S. Kemmer (eds.), Studies in Typology and Diachrony: Papers Presented to J. H. Greenberg on his 75th Birthday. Amsterdam and Philadelphia: John Benjamins, 95-122. Hawkins, J. (1994): A Performance Theory of Order and Constituency. Cambridge: Cambridge University Press. Hinterhölzl, R. and S. Petrova (2005): Rhetorical relations and verb placement in early Germanic languages. Evidence from the Old High German Tatian translation (9th century). In: M. Stede et al. (eds.), Salience in Discourse. Multidisciplinary Approaches to Discourse. Münster: Stichting/Nodus, 71-79. Jacennik, B. and M. Dryer (1992): Verb-subject order in Polish. In: D. L. Payne (ed.), Pragmatics of Word Order Flexibility. Amsterdam and Philadelphia: John Benjamins, 209-241. Kroch, A. S. (1989): Reflexes of grammar in patterns of language change. Language Variation and Change 1, 199-244. Lambrecht, K. (1994): Information Structure and Sentence Form. Cambridge: Cambridge University Press.

Decline of Postverbal Topical Subjects in Serbo-Croat

141

Lambrecht, K. and L. Michaelis (1998): Sentence accent in information questions: Default and projection. Linguistics and Philosophy 21, 477-544. Lightfoot, D. W. (1979): Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Lightfoot, D. W. (1999): The Development of Language: Acquisition, Change, and Evolution. Oxford: Blackwell. Lightfoot, D. W. (2006): Cuing a new grammar. In: A. van Kemenade and B. Los (eds.), The Handbook of the History of English. Oxford: Blackwell, 24-44. Matić, D. (2003): Topic, focus, and discourse structure: Ancient Greek word order. Studies in Language 27, 573-633. Matić, D. (2004): Topics, Presuppositions, and Theticity. An Empirical Study of Verb-Subject Clauses in Albanian, Modern Greek, and Serbo-Croat. Ph.D. Dissertation, University of Cologne. (KUPS publication online: http://kups.ub.uni-koeln.de/frontdoor.php?source_opus=1132) Myhill, J. (1986): The two VS constructions in Rumanian. Linguistics 24, 331350. Önnerfors, O. (1997): Verb-erst-Deklarativsätze. Grammatik und Pragmatik. Stockholm: Almqvist & Wiksell. Pintzuk, S. & A. Taylor (2006): Loss of OV order in the history of English. In: A. van Kemenade and B. Los (eds.), The Handbook of the History of English. Oxford: Blackwell, 249-278. Stalnaker, R. (1978): Assertion. In: P. Cole (ed.), Pragmatics. New York: AP, 315332. Vennemann, T. (1975): An explanation of drift. In: Ch. N. Li (ed.), Word Order and Word Order Change. Austin: University of Texas Press. Wurff, W. van der (1997): Deriving object-verb order in Late Middle English. Journal of Linguistics 33, 485-509.

Prosody, Information Structure and Word Order Changes in Portuguese Kristine Gunn Eide (Oslo)

1 Introduction At the end of the 18th century, Portuguese lost many of its verb second characteristics and went from being what seemed like a verb second language to an SVO language. In this paper it is argued that a syntactic change that took place from Classical Portuguese (16th – 18th cent), henceforth ClP, to Modern European Portuguese (18th -20th cent.), henceforth EP, originated in a change in the discourse pattern that may itself have been prosodically driven. The syntactic change in question is a fixing of the subjects in preverbal position. Preverbal subjects go from being exclusively topics, and old information, to occurring in non-topic contexts as well. Following Galves and Galves (1995) and Galves et al. (2005), I assume that the position of the clitics in ClP was phonologically or prosodically determined and that they would occur after specific prosodic patterns. When the position of the clitics changed around the 2nd half of the 17th century, it was the result of a change in the prosodic pattern. I shall argue that the weakening of sentence-initial prosodic prominence opened up for non-topic subjects to occur preverbally. The observable change in syntax, a reanalysis of TVX (Topic – Verb – Other constituents) to SVO (Subject – Verb – Object) was only possible after the non-topic subjects started to appear before the verb. The reanalysis affected the surface structure of unaccusative verbs in particular because the subjects of these verbs are generated in post verbal position. The change in the syntax is dependent on the previous change in the prosody and information structure.

I wish to thank members of FTL (Forum for theoretical linguistics) and members of the Romanist Workgroup in Oslo, in particular Kristin Hagemann, for comments on earlier versions of this paper. I also wish to thank Christoph Gabriel for his very helpful review as well as my anonymous reviewers.

144

Kristine Gunn Eide

In section 2 of this paper is a short outline of the problems regarding a V2 analysis of ClP. I describe the word order in ClP and show that while a verb second analysis can explain the high frequency of inversion, it is nonetheless problematic because of the large number of verb first and verb third sentences. In section 3 is an outline of the role that information structure plays in determining word order in ClP and EP. In section 4, there is a description of topic realisation in EP, and it is argued that we must take the phonologic realisation of discourse structures into consideration in a historical analysis. In section 5, I present an outline of the position of the clitics in ClP and link this to the sentence prosody. I argue that a change in the phonetic realization of the first constituent in the sentence caused the change to happen. In section 6 I argue in favour of an analysis which combines syntax with prosody and information structure in order to describe the data at hand.

2 The verb second hypothesis ClP brings us right into an ongoing debate of what the verb second phenomenon is and whether there is a possibility that European languages other than the Germanic ones have passed through a stage of V2 before ending up as SVO languages like for instance Hulk and Kemenade (1995) have proposed for English. Traditionally, a verb second language is characterised by the verb moving to C and that the specifier position of CP needs to be filled by something, whether it is a subject, an adverb or an object as in the Norwegian example (2) with the structure as in (2). In this structure, the verb is always in the second position of the clause. When SpecCP is occupied by something other than the subject, then the subject is forced to remain in a lower position, after the verb. This leads to quite a lot of inversion, typical of V2 languages such as German and Scandinavian languages. (1)

Ertestuing et Petter aldri. Mushy peas eats Petter never. ‘Petter never eats mushy peas’

Prosody, Information Structure and Word Order Changes in Portuguese

(2)

145

CP

SpecCP Ertestuing

C’

C et

.

VP

SpecVP Petter

V’

aldri

V’ (Norgård and Åfarli 1990)

Adams (1989) and Vance (1989, 1997) among others have proposed a V2 structure for Old French. Salvi (2004) claims that all Romance languages have passed through such a stage sometime during the Middle Ages, while this has been argued against by for instance Kaiser (2002). It is the large number of subject-verb inversions in clauses with an adverb in the first position that have led to these analyses of the medieval Romance languages. A V2 analysis for Old (Medieval) Portuguese, henceforth OP, has been proposed by Ribeiro (1995) and Torres Morais (1996), but rejected by Fiéis (2003) and Rinke (2007), who argue that the verb remains in IP. As Fiéis (2003) points out, most of the examples of the Adverb – Verb – Subject type are with unaccusative verbs, where the subjects are generated in post-verbal position anyway. It is not necessary to assume a movement of the verb to C to account for the “inverted” word order. In ClP on the other hand, there are frequent inversions with transitive (3), intransitive and predicative (4) verbs as well as with the unaccusative (3) ones: (3)

(4)

1 2

Desta vez ficou o Magor em seus Reinos 1 This time stayed the Magor in his realms ‘This time Magor stayed in his realms’ Chegado a Cabul, tomou o Hamau a rainha Having arrived in Cabul, took the Hamau the queen comsigo 2 with him ‘After he came to Cabul, Hamau took the queen with him’

Diogo de Couto: Décadas. (This and the other examples of ClP are from the digitalised texts in the Tycho Brahe Corpus.) Diogo de Couto: Décadas.

146 (5)

Kristine Gunn Eide

E a ssim era êste bárbaro tão afeiçoado And thus was this barbarian so loving aos Cristãos.. 3 towards the Christians ‘And so this barbarian liked the Christians so much…

Because of this, Portuguese seems closer to a V2 structure in its classical period than in the previous one. We do however encounter the same problems here that are relevant for OP: Quite a number of clauses do not have the verb in second place. There are numerous clauses that begin with a verb (V1) as in (6) and (7) as well as clauses that have the verb in the third or fourth position (V3) as in (8) and (9). As for the V3 clauses, both orders Subject – Adverb – Verb as in (8) and Adverb – Subject – Verb as in (9) are possible. (6)

(7)

(8)

3 4 5 6

Reinou êste Groduxá no Reino de Ormuz trinta anos 4 Reigned this Groduxá in-the Kingdom of Ormuz thirty years ‘This Groduxá reigned in the kingdom of Ormuz for thirty years’ Levava o Governador mais sete caravelas 5 Brought the Governor more seven caravelles ‘The Governor also brought seven caravelles’ As tres naos, despois de venderem The three ships, after sell.INF3.pl aly bem suas fazendas, se foraõ there well their goods, REFL went para Goa com sós os officiaes dellas, 6 to Goa with only the officers of-them ‘After they well had sold their goods there, the three ships sailed off to Goa with only the officers on board’

Diogo de Couto: Décadas. Diogo de Couto: Décadas. Diogo de Couto: Décadas. Fernão Mendes Pinto, Perigrinação. An anonymous reviewer points out that this example can also be interpreted as an example of V2 if the dependent clause “after they had sold their goods” is interpreted as a modifier of “the three ships”. Indeed, this is one of several possible interpretations. Another possible V2 interpretation the one outlined in Section 5, example (24). In this analysis “the three ships” would be a dislocated topic outside the CP and “after they had sold their goods” would be sentence initial. This latter analysis could also work for an example such as the one given below, where the time adverb “on October 19th” can not be interpreted as a modifier of the subject. O Conde de Alba de Liste, e o Marques de Alcanises aos 19 de Octubro The count of Alba de Liste and the Marquis of Alcanises on the 19th of October entrarão pella villa de Ifanes (Manuel de Galhegos, Gazeta) entered through the Vila de Ifanes ‘On October 19th, the Count of Alba de Liste and the Marquis of Alcanises entered the Vila de Ifanes’.

Prosody, Information Structure and Word Order Changes in Portuguese

(9)

147

Primeiramente, nas histórias a que chamam verdadeiras, First of all, in the stories to which they.call true, cada um mente segundo lhe convém, … 7 each one lies as to them it pleases ‘First of all, in the stories that they call true, everyone lies as they please.

An analysis of Classical Portuguese has to account for the large number of V2 clauses with or without inversion with all types of verbs as well as the V1 and V3 clauses. In order to account for these data, the V2 analysis that has been proposed for ClP by Galves and Galves (2005), Galves, Britto et al. (1995) and Galves and Sousa (2005), differs considerably from a V2 analysis of the modern Germanic languages, and somehow whether we choose to call ClP a verb second language or not, depends largely on how we choose to define the V2 phenomenon. I will return to the analysis of these authors in section 5. Apart from Subject – Verb inversion, languages that traditionally have been considered V2 languages, such as German and Scandinavian dialects, typically require SpecCP to have phonetic realisation, and when necessary use expletives to fulfil this requirement. Such overt expletives are not systematically found in the Romance languages, and to maintain a V2 analysis, SpecC’ must in some cases be filled by covert pros or empty operators (Vance 1997). This would account for the V1 clauses, because the verb would structurally occupy the second position. As for the V3 clauses, these will be discussed in section 5 8 .

3 Information structure: given and new information Classical Portuguese had a topic-verb-XP structure (10), where inversion usually occurred when the sentence began with something other than the subject. A similar verb second-like structure can also be found in Old French (Adams 1989) and Old English (Hulk and Kemenade 1995; Bech 2001). Like Modern French and Modern English, Modern Portuguese is now an SVO language (11) without this restriction. (10) (11)

7 8

Topic – Verb – XP (XP) – Subject – Verb – Object

Francisco Rodrigues Lobo, Côrte Na Aldeia e Noites de Inverno Although not a topic in this paper, verb raising of auxiliaries is relevant to a V2 analysis. Typically for V2 languages, the finite verb is raised to C while the non finite verb remains below. This does not happen frequently in ClP, although it does occur sporadically.

148

Kristine Gunn Eide

The traditional explanation of the change from (10) to (11) is that at some point the topic position was reanalysed as subject position. In the case of French and English, nonstressed/clitic subject pronouns started to appear in between the topic and the verb. This expanded to non-clitic subjects (Adams 1988, 1989; Vance 1989: Hulk and Kemenade 1995; Vance 1995, 1997) which in turn led to a reanalysis of this position to subject position: Subjects are usually topics and with the appearance of the preverbal clitic subject pronouns, by far the most subjects would be preverbal. This proportionally large number of preverbal subjects then became the canonical subject position through reanalysis of [S V [IP t t O]] to [IP S V O]. Portuguese could not have followed exactly the same pattern as French and English, because it, at least in its European variety, has always been a null-subject language without unstressed/clitic subject pronouns. In fact, Sousa’s (2003) data on subject positions in Classical Portuguese indicate that in most clauses the word order was not SV(O), either because they were null subject clauses or because the subject was post verbal. We need to find another explanation as to how the subjects became preverbal when there were no unstressed subject pronouns to “lead the way” and increase the number of preverbal subjects. Something else must have been the cause of this development. As mentioned in section 2, Fieis (2003) observed that the post verbal subjects in OP were in most cases subjects of unaccusative verbs. They were simply internal arguments generated in the same position as objects of transitive verbs, similar to what we find in modern EP (Costa 2004). Figure 1 shows the development of the subject position with unaccusative verbs in Classical and Modern Portuguese. There is an increase in post verbal subjects in the 16th century (which is in fact a continuation of the increase seen in the previous centuries as described by Fiéis (2003)). The peak in the second half of the 16th century is followed by a decrease in the following centuries.

Prosody, Information Structure and Word Order Changes in Portuguese

149

100 50 0 1497- 1550- 1600- 1650- 1700- 1750- 1800- 1850- 1900- 19501549 1599 1649 1699 1749 1799 1849 1899 1949 1999

VS

SV

Figure 1: Percentages of VS:SV with unaccusative verbs in main clauses 9 . (Eide 2006: 46)

In Sousa’s (2004) study of Portuguese subject positions 10 , we find a similar development for other types of verbs as well: the preverbal subjects increase and there is a decrease in the post verbal subjects. However, there is a marked change around the end of 17th, beginning of the 18th century, one century before the most marked change in Figure 1. It seems that the change affected the transitive and intransitive verbs before the unaccusative ones. In Figure 2, the development of post verbal subjects of unaccusative verbs is compared to the information value of the subjects 11 and we see that subjects that were old information were far more affected by the change than subjects that were new information 12 . This is true of both the increase in post verbal subjects that we see in the first half of the 15th century as well as of the following decrease. The development of old information subjects of the unaccusative verbs seems to map with the development of the subjects of transitive and intransitive verbs. The subjects that are new information have a slower drift toward the canonical preverbal subject position in EP.

9

The corpus in this study consists of texts from the Tycho Brahe Corpus and from a supplementary corpus from the 19th and 20th century gathered by the author. It consists of historical documents, letters, philosophical writings, itineraries, newspapers, magazines as well as literary texts. 10 Sousa’s study does not distinguish between transitive, intransitive or unaccusative verbs. The data are comparable to those from Eide (2006), in particular because they both use the same text corpus. 11 A complete study of the information structure of non unaccusative verbs has yet to be made. 12 Eide (2006) checked the subjects according to animacy, definiteness, phonologic weight and givenness. While the three former conditions did not seem to have made any significant difference, the change was quite clear when it came to givenness, or information structure, as the figure shows.

150

Kristine Gunn Eide

100 80 60 40 20 0

Old info

New info

Figure 2: Percentage of subjects that are in VS position in main clauses. We see a decrease in VS both for subjects that are old information and new information. For subjects that are old information, there is a decrease from 66% in the second half of the 16th century to 6% in the second half of the 20th century. (Eide 2006: 101)

The present analysis uses a tripartition of the pragmatic status of the arguments: that which is topic, that which is focus and that which is neither, along the lines of Büring (1997). As for topic, it is defined using the traditional definition of “what the sentence is about”. There is a strong correlation between topic and old information on the one hand and between focus and new information on the other hand 13 . The verb will usually be part of background, but it can also belong to either topic or focus. While topic in most cases coincides with old information, arguments that are old information are not necessarily topics. The subject placement in ClP is determined partly by syntactic structure, as indicated in (10) which involves verb fronting and partly by information structure restriction as in (12). At a certain period of time in ClP, only topics/old information could appear in the beginning of the sentence, before the verb 14 . In EP (13), both old and new information occur preverbally. 13 While topic is not necessarily old information, the classification of arguments according to whether they have been previously mentioned in the text or not has proven to be a valuable tool for determining the sentence structure because of this strong correlation between topic and old information. Background is that which is neither topic nor focus, typically, in the corpus I will be referring to, it consists of a previously mentioned subject and the verb (The corpus consists of mostly narrative texts, documents and letters where there is very little dialogue. These texts are presumed to make little use of contrastive stress and other prosodic patterns characteristic of speech, such as left dislocation/right dislocation and marked focus stress where focus is not in its usual position.). In other words, it is old information that does not fall under the classification of “what the sentence is about.” 14 In the corpus studied in Eide (2006), preverbal constituents were almost always old information, that is, they had been previously mentioned in the text. Some exceptions to this are such as the one in the following example:

Prosody, Information Structure and Word Order Changes in Portuguese

(12) (13)

151

Old information – Verb – Old/New information Old Information/New information – verb – New information

But most importantly, the background information, that is, that which is neither topic nor focus, and is supposed to have no prosodic prominence, also occurs preverbally (and sentence initially) in EP. In ClP, whichever was the topic of the sentence would move to the beginning of the clause. The verb would follow the topic and the rest of the background, such as subjects that were not topics, would follow the verb. If we were to take the equivalent of the English sentence, Yesterday, John bought the book, where yesterday is the topic, the book is new information and John bought is the background we would get the following patterns: (14)

(15)

Classical Portuguese: (16th-18th century) Ontem comprou o João o livro. Yesterday bought John the book ‘Yesterday John bought the book.’ Modern Portuguese (19th century onwards) Ontem, o João comprou o livro. Yesterday John bought the book

The subject-verb position of the background information has changed from VS in ClP to SV in EP. Thus, the sentence in (3), repeated below in (16), where the subject is old information and part of the background, is impossible in EP (17) where only new information may occur post verbally (Costa 2004) 15 . (16)

(17)

Desta vez ficou o Magor em seus Reinos 16 This time stayed the Magor in his realms ‘This time Magor stayed in his realms Desta vez o Magor ficou... This time Magor stayed...

Hum ferrador vizinho do Cardeal Paloto desappareceo de Roma (Manuel da Costa: Arte de Furtar) ‘A blacksmith neighbour of Cardinal Paloto disappeared from Rome’ This type of example prevents us from defining the preverbal topic position as containing only old information. In this example the subject is not only new information, it is also indefinite. However, it is the beginning of a story about a certain blacksmith and as such marks a change in the discourse topic. Examples such as this one are quite rare, especially with unaccusative verbs. 15 There are counterexamples to this claim, however they require a marked intonation that is typical of direct speech and is not usually associated with the narrative type of texts in the corpus. 16 Diogo de Couto: Décadas. (This and the other examples of ClP are from the digitalised texts in the Tycho Brahe Corpus.)

152

Kristine Gunn Eide

In ClP, subjects that were not topics were post verbal regardless of whether the verb was preceded by a non-subject topic or not. In other words, non topic subjects of main clauses would not move to SpecCP only to meet a verb second requirement. Typically, postverbal subjects occur with unaccusatives because these verbs often serve to introduce new information, but there is also inversion with transitives. This is the case in example (7), repeated below in (18), where the topic is actually the verb ‘bring’ 17 , the subject is background and sete caravelas is new information. In the EP equivalent (19) the background subject cannot remain in post verbal position becaues it is not new information. (18)

(19)

Levava o Governador mais sete caravelas 18 Brought the Governor more seven caravelles ‘The Governor also brought seven caravelles’ O Governador levava mais sete caravelas The Governor brought more seven caravelles

In EP, which is an SVO language with a relatively free word order, one or more constituents may precede the verb, and the preverbal constituents need not be topics, they can be part of the background as well. The subject, when preverbal, is always placed just before the verb. The subjects may be prosodically neutral or they may be prosodically marked as topics or contrastive topics. While there is a phonetic difference in the pronunciation of a subject topic, there is not always a surface structural difference: In an analysis of EP, both topic and non-topic subjects appear in SpecIP (Duarte 1987) which is the canonical subject position (Costa 2004). The next section deals with the prosodic prominence of topics in EP

4 Prosodic marking of discourse patterns Frota’s (2000), phonetic analysis of EP topics reveals that they are represented by different types of phonological structures, corresponding to different types. In EP, both preverbal in situ topics and dislocated topics have phonological traits that distinguish them from focus and neutral utterances, and these traits occur regardless of which position the topic is in (in situ, left dislocated or right dislocated 19 ). One might say that what characterises them is that they are more carefully pronounced: there is lack of sandhi phenomena, because they constitute separate phonological 17 The context is a listing of what the Governor is taking with him. In that sense, levava is old information and interpretable as topic. 18 Diogo de Couto: Décadas. 19 The present paper is concerned with preverbal in situ topics and left dislocated topics.

Prosody, Information Structure and Word Order Changes in Portuguese

153

phrases, the vowels in the topic are less reduced and there is less sandhi within the topic as well, not just on the border. Frota’s examples are cited in (20), (21) and (22), where the sandhi phenomenon of fricative voicing is marked with an underlined s. (20) is what Frota calls a neutral uttering, and as far as I understand, it corresponds to a broad focus reading and the final –s, usually pronounced [³], is voiced [z] because it is precedes a vowel in the following verb. In (21), where the subject as angolanas is the topic, there is no such voicing. In (22), the topic, aos jornalistas is left dislocated and is characterised by the same lack of sandhi as the topic subject in (21): (20)

(21) (22)

As angolanas ofereceram especiarias aos jornalistas (neutral) The angolans gave spices to the journalists ‘The Angolans gave spices to the journalists’ As angolanas ofereceram especiarias aos jornalistas (topic, in situ) Aos jornalistas, as angolanas ofereceram especiarias (topic, dislocated) Frota (2000 : 252).

The fact that the topic readings need not have fricative voicing is analysed as the topic being a separate intonation phrase. In Frota’s analysis there are other segments that serve to distinguish topic phrases from similar arguments in neutral utterances, for instance the pronunciation of aos is more articulate in the topic than in the focus and neutral utterances. Topics in sentence initial position are also, together with focused utterances, distinct from the neutral with regard to lengthening. Topics may either be followed by an acoustic pause or, when there is no such pause, the last stressed syllable is lengthened. Topic lengthening is longer than focus lengthening. Another important fact is that the speakers that Frota uses do not have the same percentage of “topic readings”, some speakers tend to pronounce topics more markedly than others. There is also variation within the data from each and the same speaker, so that one speaker may for instance have topic vowel lengthening in around 50 % of the topics. When we study past stages of the language we should also consider the possibility of similar variation in the pronunciation of topics. This is in fact, in line with Galves and Sousa’s (2005) proposal that there are two types of topics that occur preverbally: the extra clausal left dislocated ones which are more likely to constitute their own intonation phrase than the ones that are located within the clause. This will be outlined in section 5. While the characteristics of topics were not necessarily the same in ClP as they are in EP, we should take them into consideration in our analysis. Even though we cannot say exactly how the topic was phonetically encoded, we may assume that it was different from the neutral and focus utterances and we should not overlook the possibility that there was variation, not just from speaker to speaker, but also within the prosodic pat-

154

Kristine Gunn Eide

terns of each individual speaker. Also, because of the importance prosodic structures have in the encoding of discourse structures, it is probable that a change in the discourse structures also implies a change in the prosodic structure and the other way around: that a change in the prosodic structure, such as a change in the stress pattern of a sentence, would have effect on the discourse structure. I will argue that the change in the initial sentence prosodic prominence, typical of topics lead to a change in the interaction between topic-comment structure and syntactic structure as outlined below: In ClP, we assume that primary sentence stress was located on the clause final focus, and that the sentence initial position carried some sort of prosodic prominence associated with topic. In V1 clauses this prosodic prominence would fall on the verb. This claim relies on the fact that only topics could occur preverbally, and that in cases where there was no topic, the sentence initial verb became more salient (whether or not we interpret the verb itself as topic). When the initial sentence prosodic prominence was modified, arguments without these traits, such as non topic subjects, could occur preverbally. While topics continue to be sentence-initial they will receive a phonological marking that is different from non-topic subjects (Frota 2000). Thus, the TVX structure with numerous inversions that are found in ClP is replaced by an SVO structure where the subject may or may not be the topic of the sentence. Linked to this change is the change in the placement of the clitic pronouns that in certain contexts go from being almost exclusively preverbal in the 16th century to post verbal in the 19th century. The placement of clitic pronouns as outlined by Galves and Galves (1995) will be discussed in the next section.

5 The position of clitics Following Galves and Galves (1995) and Galves and Sousa (2005), I assume that variation in clitic placement in contexts where variation is possible 20 , is dependent on the structural position of preverbal topics as well as their prosodic realisation. In their analysis of ClP, the clitics are in second position in the clause and can therefore be used as a diagnostic for clause boundary. This is illustrated in examples (23), (24) and (25). In a clause such as (23), the XP topic is necessarily analysed as being within the clause, given that in ClP the Tobler-Mussafia law 21 is valid and no un20 That is, disregarding examples where clitics occur either only preverbally, such as after negations or postverbally, such as in cases of V1. 21 A “law” or rule, stating that the sentence cannot begin with an unstressed word. This rule

Prosody, Information Structure and Word Order Changes in Portuguese

155

stressed element may occur sentence initially. In (24), the first topic XP is outside the clause (where # marks the clause boundary) while the second topic is inside. Similarly, in order to meet the requirement that the clitic is always in second place, the topic XP in (25) must be extra clausal and the rest of the sentence is analysed as V1: (23)

(24) (25)

XP cl verb se acharam [IP os dois homens [CP Na corte At-the court REFL found the two men ‘The two men were (lit. ‘found themselves’) at the Court…’ XP#XP cl verb [CP Na corte [CP os dois homens se acharam... XP # verb-cl [CP Na corte [CP acharam-se [IP os dois homens

This analysis leads to a very large number of extra-clausal topics of a type that is typical for spoken texts rather than the written ones that Galves and Galves’ study is based on 22 . Also, even texts with SVO with a post verbal clitic must be analysed as having the subject in an extra-clausal position. Thus (26), must be analysed as in (27) (26)

(27)

O Magor salvou-se com muito trabalho 23 The Magor saved-REFL with much effort ‘Magor saved himself with a lot of effort.’ O Magor # salvou-se com muito trabalho

The analysis does not specify the placement of the extra clausal topic, only that it is outside the CP. Since it is outside the CP, we would expect it to be its own intonation phrase, and that a new intonation phrase starts with the verb in (25) and (26). This would not be the case in examples (23) and (24). Without the clitics, these sentences have the same linear structure. The only way a learner could perceive of the difference between the extra clausal and the clause internal topic, is if they are pronounced differently. Despite the difficulties with maintaining the V2 analysis, I will assume, along with Galves, that there was variation in the pronunciation of the preverbal constituents, and in the present analysis, focus will be on this variation. To support the claim that the placement of the clitics is affected by stress and/or intonation is the effect that even small words such as the conjunction e can have on the clitic placement. In some cases such as (28), has been used in combination with Wackernagel’s law, which states that the least stressed position in a sentence is the second position, to account for the placement of verbs and clitics. 22 Galves and Galves (1995) and Galves and Sousa (2005) are also based on texts from the Tycho Brahe Corpus. 23 Diogo de Couto, Décadas.

156

Kristine Gunn Eide

a stressed sentence initial e can provoke proclisis, although in most cases it does not, as in (29) where the reflexive pronoun is enclitic to the verb. This indicates that it is the phonetic realization of the sentence initial element (which we know is variable) that may cause proclisis, rather than an obligatory move to a structural position. (28)

(29)

E lhe peço por mercê, que... 24 And him I ask please, that …. ‘And I ask him please, that…’ E assentou-se, que ... 25 And decided-REFL, that... ‘And it was decided that... ’

Most importantly, however, is that the position 26 of these clitics, a position determined by prosody, changed: From being almost exclusively preverbal in some texts to being very much enclitic, or post verbal in later texts. Because of this, we may assume a change, perhaps a subtle one, in the prosodic structure of the sentence. In the following chapter, I show how such a change could affect the discourse structures

6 The change in the discourse pattern As we have seen, the ClP clause would typically begin with a topic, if there was a topic expression, otherwise it would be verb first. This structure involved a large number of inversions and also some cases of verb third. EP in comparison shows a decrease in the number of inversions and verb first clauses and an increase in the number of Verb third. This is what we would expect to happen if we follow a traditional analysis of the change from a V2 structure to an (XP)SVO structure such as it is illustrated in (30) where the preverbal subject, outside IP, possibly in SpecCP or in some other, lower functional projection, such as for instance a topic phrase, is reanalysed as being in the canonical subject position which is SpecIP: (30)

[CP Sj Vi [IP j i O]] > [IP SVO]

It is argued that even in V2 languages, the subject is usually the first argument in the sentence, and this place is then reanalysed as subject position. However, as pointed out in section 2, in a null-subject language like 24 Diogo do Couto, Décadas. 25 Diogo de Couto: Décadas. 26 Sousa (2003) suggests that it was actually the subjects in V1 clauses that changed position and became prevebal and that the clitic remained in the same place: Verb-clitic Subject became Subject Verb-clitic.

Prosody, Information Structure and Word Order Changes in Portuguese

157

Portuguese, most sentences are not likely to have the structure in (30). Nevertheless there is a development from ClP (31) to its modern equivalent (32) (31) (32)

Desta vez ficou o Magor em seus Reinos 27 ‘This time stayed Magor in his Kingdom’ Desta vez, o Magor ficou em seus reinos

If we see a TVX structure, not only as a syntactically structured sentencepattern but rather interacting with the prosodically determined discourse structure of the language, V1 and V3 clauses are easier to explain. We have seen that only topics can precede the verb in ClP. In cases of V1 with postverbal subjects, it is the sentence initial prosodic prominence associated with topics that prevents the non topic subject from rising. This is a valid argument whether we chose to analyse the verb as being in C or in I or in some other functional projection. When the sentence initial prosodic prominence ceases to exist or is weakened, non topic subjects may appear sentence initially 28 . Given the starting point of the evolution towards enclisis, we see that the decline in proclisis started at the end of the 17th century/turn of the 18th century (Galves, Britto et al. 2005), at the same time as the old information subjects became more and more preverbal, while the subjects containing new information changed later and at a much slower pace. If we decide to see the change in clitic placement as a result of a change in the prosody 29 , this indicates that the two changes are interdependent on one another. Galves does not, as far as I can see, specify what process in the prosody causes the clitics to change. Both Frota’s analysis of EP topics and Galves et al.’s (2005) on clitic placement show evidence for variation in the pronunciation of topics. Both analyses show that topics may be separate intonational phrases or be less prominently pronounced. In other words, what we do have is a TVX structure with a topic that has a variable pronunciation, and which is therefore more liable to change. Given the fact that the non-topic subjects start to appear in this position, I suggest that it has to do with a weakening process. Given the difference in pace in which the old and new information subjects change their position, it is clear that the old information subjects move first. It is when they become numerous enough that a reanalysis of the preverbal position from 27 Diogo de Couto: Décadas. 28 This would favour a verb second analysis, otherwise why would the non topic subjects want to move? 29 One process which has not been described in this paper, but which happens more or less at the same time as the change in the clitic placement, is changes in the stress patterns of single words, where the vowels in the non-stressed syllables are weakened and become schwalike. This can be seen as an indication that the language is going through a major stress change.

158

Kristine Gunn Eide

topic to subject can happen. The main indication of the reanalysis is that new information subjects of unaccusative verbs start to appear before the verb. These arguments are internal and thus generated in object position. Because of their information status, they are not likely to appear preverbally unless it is for syntactic reasons.

References Adams, M. (1988): Parametric Change: Empty Subjects in Old French. In: D. Birdsong and J.-P. e. Montreuil (eds.), Advances in Romance Linguistics. Dordrecht: Foris Publications, 1-16. Bech, K. (2001): Word order patterns in Old and Middle English a syntactic and pragmatic study. Bergen: Department of English University of Bergen. Büring, D. (1997): The Meaning of Topic and Focus. The 59th Street Bridge Accent. London and New York: Routledge. Costa, J. (2004): Subject positions and interfaces: the case of European Portuguese. Berlin and New York: Mouton de Gruyter. Duarte, M.I.P. d.S. (1987): A construção de topicalização na gramática do português: Regência, ligação e condições sobre movimento. Dissertação de doutoramento em linguística portuguesa. Faculdade de Letras. (ed.). Lisboa: Universidade de Lisboa. Eide, K. (2006): Word order structures and unaccusative verbs in Classical and Modern Portuguese. The reorganisation of information structure. ILOS: University of Oslo. Fiéis, M.A. (2003): Ordem de palavras, transitividade e inacusatividade. Reflexão teórica e análise do Português dos séculos XIII a XVI. Faculdade de Ciências Sociais e Humanas. Lisboa, Universidade Nova de Lisboa. Frota, S. (2000): Prosody and focus in European Portuguese: phonological phrasing and intonation. New York and London: Garland Publisher. Galves, C. H. Brito et al. (2005): The Change in Clitic Placement from Classical to Modern European Portuguese: Results from the Tycho Brahe Corpus. Journal of Portuguese Linguistics 4,1, (Special Issue on Variation and Change in the Iberian Languages: the Peninsula and Beyond), José Ignácio Hualde (org.). Galves, C. and A. Galves (1995): A Case Study of Prosody Driven Language Change. From Classical to Modern European Portuguese. Download 01012006 from http://www.ime.usp.br/~tycho/papers/agalves_cgalves_1995.pdf.

Prosody, Information Structure and Word Order Changes in Portuguese

159

Galves, C. and M.C.P.d. Sousa (2005): Clitic placement and the position of subjects in the history of Portuguese. Romance Languages and Linguistic Theory. Selected papers from Going Romance 2003. Amsterdam and Philadelphia: John Benjamins, 93-107. Hulk, A. and A. v. Kemenade (1995): Verb Second, Pro-drop, Functional Projections and Language Change. In: A. Battye and I. Roberts (eds.), Clause Structure and Language Change. New York and Oxford: Oxford University Press, 227-256. Kaiser, G. A. (2002): Verbstellung und Verbstellungswandel in den romanischen Sprachen. Tübingen: Niemeyer. Norgård, T. and T. A. Åfarli (1990): Generativ Syntaks. Ei innføring via norsk. Oslo, Novus Forlag. Ribeiro, I. (1995): Evidence for a Verb-Second Phase in Old Portuguese. In: A. Battye and I. Roberts (eds.), Clause Structure and Language Change. New York and Oxford: Oxford University Press, 110-139. Rinke, E. (2007): Syntaktische Variation aus synchronischer und diachronischer Perspektive. Die Entwicklung der Wortstellung im Portugiesischen. Frankfurt am Main: Vervuert Verlag. Salvi, G. (2004): La formazione della struttura di frase romanza: ordine delle parole e clitici dal latino alle lingue romanze antiche. Tübingen: Max Niemeyer. Sousa, M. C. P. d. (2003): Posição de sujeitos e colocação de clíticos em textos portugueses dos séculos 16 a 19. Download 01102006 from http://www.ime.usp.br/~tycho/papers/psousa_2003.pdf. Sousa, M. C. P. d. (2004): íngua barroca: sintaxe e história do português dos seiscentos., unpublished doctoral thesis, State University of Campinas. Torres Morais, M. A. (1996): Aspectos diacrônicos do movimento do verbo, estrutura da frase e caso nominativo no português do Brasil. Português Brasileiro. Uma viagem diacrônica. Homenagem a Fernando Tarallo. I. Roberts and M. A. o. Kato. Campinas SP: Unicamp. Vance, B. (1989): The evolution of pro-drop in Medieval French. In: C. Kirschner and J. DeCesaris (eds.), Studies in Romance linguistics: selected papers from the Seventeenth Linguistic Symposium on Romance Languages (XVII. LSRL), Rutgers University, 27-29 March 1987. Amsterdam and Philadelphia: John Benjamins, 413-441. Vance, B. (1995): On the Decline of Verb Movement to Comp in Old and Middle French. In: A. Battye and I. Roberts (eds.), Clause Structure and Language Change. New York and Oxford: Oxford University Press, 173-199. Vance, B. (1997): Syntactic change in Medieval French. Verb-second and Null Subjects. Dordrecht: Kluwer.

The Development of V-to-C-Movement in the West Germanic and Romance Languages * Melanie Wratil (Düsseldorf)

1 Introduction In the minimalist approaches to the syntax of imperatives, imperative verb movement is, for the most part, seen as a universally determined head movement process that takes its course in all natural languages in a uniform manner (cf. Platzack & Rosengren 1998; Han 2000 among others). The supporting data for this view is usually based only on a few modern Indo-European languages. Among these, mainly modern Germanic and modern Romance languages are the target of investigation. However, it is just the imperative clauses of the modern Germanic and modern Romance languages that exhibit a very special sentence structure which differs from the imperative structures of other language groups in the particular positioning of the main verb. In what follows I will show that the extraordinary structure of the West Germanic and Romance imperatives has to be seen as a result of the emergence of V2 phenomena. Chapter 1 concentrates on the evolution of V2 patterns and the rise of V1 imperatives. Its first section elucidates the canonical imperative verb movement of the Indo-European languages from which, as illustrated in section 1.2., the head movement of the old Germanic and old Romance imperative verbs does not yet usually deviate. Sections 1.3 and 1.4 survey the V2 properties of the modern West Germanic and Romance languages and review some fundamental explanations for their origin. In section 1.5 I show that V2 structures begin to prevail as soon as fronted focus phrases are reanalysed as the scope marking specifier of a new functional projection CP. Having demonstrated that V1 structures are also the structural *

I would like to thank the audience of the workshop “The Role of Information Structure in Language Change” at the DGfS conference (University of Siegen) for constructive comments. Special thanks are due to Marina Stoyanova, Kristine Gunn Eide, Rosemarie Lühr, Marit Westergaard, and Gisella Ferraresi for helpful suggestions. All remaining errors are mine.

162

Melanie Wratil

outcome of a focalization process, I consider in the following section 1.6 why especially the V1 structures of imperative clauses and therefore the imperative verb movement to C has been able to survive in the West Germanic and Romance languages up to this day. In section 1.7 I show in which respect the imperative verb movement of modern Standard English differs from the imperative verb movement of other languages that exhibit residual V2 patterns. Chapter 2 is devoted to the movement of the imperative verb within a decomposed C-projection. In section 2.1 I address some of the newest findings on the internal structure of the C-domain. Against the background of these approaches I discuss in sections 2.2 and 2.3 the syntactic procedures of topicalization and focalization. Based on these considerations, I finally argue in section 2.4 that the imperative verb moves to the CP-internal head Foc°.

1 V2 and the Rise of Imperative V-to-C Movement 1.1 Imperative V-to-I Movement The imperative mood indicates the execution of a directive speech act. Its appearance signals that the speaker imposes or proposes an action which is prospectively realizable by the addressee (cf. Lyons 1977: 745f.). In many languages imperatives exclusively represent deontic strong directives such as commands, orders or instructions and co-occur with other speakeroriented modal markings of the verbal inflection such as prohibitives, admonitives and permissives (cf. Bybee, Perkins & Pagliuca 1994: 210f.). In many other languages, as for instance in the Indo-European languages, the imperative mood indicates directive speech acts on each level of deontic strength (Palmer 1986:108). In those languages its modal category constitutes a genuine directive operator usually represented by a zeroexpression (cf. Bybee 1990). With regard to its covert representation it behaves in the same way as the likewise formally unmarked indicative, which, being the unmarked member in the epistemic domain, holds in its use the entire class of assertives (Palmer 1986: 108). Therefore, the imperative like the indicative has to be defined as the basic mood of its own modal logical system. This is confirmed by the historical development of verbal paradigms. In contrast to other mood markings, the indicative and the imperative have always persisted in the Indo-European conjugational systems (cf. Schwyzer 1939: 303). Imperative and indicative verbs are subject to head movement processes which are originally similar. In the unmarked case both are located in the canonical position of the verbal head. In VSO languages they occupy

The Development of V-to-C in the West Germanic and Romance Languages

163

the regular clause initial position (cf. (1)), in SVO languages they are positioned immediately after the subject (cf. (2)) and in SOV languages they appear in sentence-final position (cf. (3)). (1)

(2)

(3)

Irisch a. Tógann sé an bosca lift.PRES.IND he the box b. Tóg an bosca lift.IMP the box ‘Lift the box’ ‘He is lifting the box’ Bulgarian a. Tanja vidja Marija Tanja see.IMP.3SG Maria ‘Tanja saw Maria’ b. Ti izpej edna pesen you sing.IMP a song ‘You sing a song’ Latin a. Fortes fortuna adiuvat (Terenz, Phormio I 4,203) the-braves fortune favour.IND.PRES.3SG ‘Fortune favours the braves’ b. Sed reliquum vitae cursum videte (Cicero, Oration Philippica Secunda 47) but remaining of-life course look.IMP ‘But look at the remaining course of life’

Like its epistemic counterpart the imperative verb is attracted by the abstract features of the functional head AgrS° before Spell out. Indicating the default mood of the deontic system, it is, in the same way as the indicative verb, exempt from licensing any overt inflectional marking at the head of the Mood projection. However, since it differs from the indicative in that its tense characterization as well as its person agreement is determined in a modal-immanent directive way, it does not check any tense features at a separate functional head. Accordingly, no T-node is contained in the INFL-projection of imperative clauses. These structural properties are in fact reflected in the morphological development of the imperative verb. In the course of morphosyntactic change the latter often allows the affixation of inflectional markings like, for example, aspect morphemes (cf. (4a)), voice morphemes (cf. (4b)) or number agreement morphemes (cf. (4c)). 1 But in order to avoid redun1

Imperatives are also found to enter inflectional paradigms of other moods. The present-day imperative paradigm of the Slavic languages, for example, consists of early optative forms. Therefore, we cannot be sure yet whether it should be regarded as a pure suppletive paradigm.

164

Melanie Wratil

dancy within the modal-inherent prospectivity and reference to an addressee, it never permitted the affixation of morphological tense (cf. (5a)) or person agreement (cf. (5b)) specifications. (4)

(5)

a. Smotri! Posmotri! PERF-see IMP seeIMP ‘See! See! / You have to see’ b. Paídeyei! Paideýoy educate.IMP educate.IMP.PASSIVE ‘Educate" Be aducated!’ c. ¡Tira! ¡Tirad! pull.IMP pull.IMP.PL ‘Pull! ’ ‘Pull!!’ a. Fiorisci! *Fioriscerai! flourish.IMP flourish IMP.FUR ‘Flourish’ b. Spring! *Springst! jump.IMP jump.IMP.2SG ‘Jump!’

(Russian)

(Ancient Greek)

(Spanish)

(Italian)

(German)

The movement of a prototypical imperative verb can be represented by the following simplified structure (cf. (6)) which corresponds to the Russian example (cf. (4a)). (6) AgrSP 3 Spec DPj

Agrs’ 3 Agrs° smotrii

MoodP 3 Spec

Mood’ 3 Mood° ti

VP 3 Spec tj

V’ l V° ti

1.2 Imperative V-to-C Movement in the Modern West Germanic and Romance Languages Especially in the modern West Germanic and Romance languages, imperative verbs canonically appear in the sentence-initial position even if

The Development of V-to-C in the West Germanic and Romance Languages

165

this placement does not match the basic word order of the respective language. Hence, in the West Germanic SOV languages, for example in German (cf. (7a)), as well as in the West Germanic SVO languages, for example in Yiddish (cf. (7b)), imperative verbs normally precede their complements and adjuncts. At least for the West Germanic SOV languages like German, this order allows the unambiguous conclusion that the imperative verb is located above the right peripheral IP. The postverbal positioning of the subject located in SpecI that can be observed in both languages mentioned (cf. (7a, 7b)) indicates that the imperative verbs of the West Germanic SVO languages end up in the C-domain, too. (7)

a. Ruf (du) morgen den Weihnachtsmann an! PRT call (you) tomorrow the Santa Claus ‘You call Santa Claus tomorrow’ b. Gey (du) oyf rekhts! go.IMP (you) on right ‘You go to the right’

(German)

(Yiddish)

In the modern Romance languages there is also strong evidence that the imperative verb obligatorily enters the C-projection. Just like the imperative verb in the West Germanic languages, the imperative verb in the Romance languages canonically appears in sentence initial-position and therefore precedes not only its complements and adjuncts (cf. (8)) but also its nominative subject – even if the latter serves as the specifier of the functional IP (cf. (9)). 2 (8)

(9)

a. ¡Pon un telegrama! send a telegram ‘Send a telegram’ b. Ricordati sempre di prendere il passaporto the passport remember.IMP always to take ‘Always remember to take your passport with you’ a. ¡Vetu de compras! go-you of shopping ‘You go shopping’ b. Taci tu pure! be.IMP.quite you definitely’ ‘You definitely be quiet’

(Spanish)

(Italian)

(Spanish)

(Italian)

It has to be concluded that the imperative verbs of the modern West Germanic and modern Romance languages always end up in C°, thus

2

The assumption that the imperative subject is located in SpecAgrP is supported by imperative clauses like (9 (b)) where the overt imperative subject precedes an adverb which according to Cinque (1999: 4ff.) and Zanuttini (1997: 60ff.) specifies a special IP-internal presuppositive phrase.

166

Melanie Wratil

producing congruity between their mood specification and a corresponding V1 structure. However, it has remained obscure up to this day why – in contrast to the imperative verbs of many other language groups – it is only imperative verbs of the West Germanic and Romance languages that undergo obligatory V-to-C movement. After all, at least in earlier periods of those languages, imperative verbs were not forced to conduct such a movement procedure. As the data in (10) illustrate, the Latin imperative verbs canonically appear in sentence-final position (Leumann, Hofmann & Szantyr 1965: 397ff.). Even in a few Old English texts imperative verbs still follow their subject, their complements and their adjuncts, obeying the underlying SOV word order, as shown in (11) (cf. Mitchell 1985: 371ff.). Since there are neither particles nor auxiliaries in a left or right peripheral position that point to a covert movement to C, it has to be inferred that the imperative verbs of those Old Germanic and Old Romance languages, just like the imperative verbs of the vast majority of the other language groups, still target the head of the functional IP in order to check their mood specification in that position. (10) a. Sed reliquum vitae cursum videte (Latin) but remaining of-life course look.IMP ‘But look at the remaining course of life’ (Cicero, Oration Philippica Secunda 47) b. teque studiis vel otio trade ourself-and studies or leisure devote.IMP ‘And devote yourself to study or sciences’ (Plinius, Epistulae Liber I IX 7) (11) a. Þu þisne middangeard milde geblissa (Old English) you this middle earth please bless.IMP ‘You bless this soil please’ (The Exeter Book, Christ A 249) b. Ge nu hraðe gangað thus now fast go.IMP.PL ‘Thus, go fast now’ (The Vercelli Book, Elene 406)

Thus, it has to be investigated why and how the imperative V-to-C movement could evolve in the West Germanic and Romance languages. The data above seem to justify the assumption that the emergence of this special imperative head movement process is directly connected with the special development of the CP since a C-projection that is responsible for the derivation of V1 imperatives has to possess inflectional V-features besides its operational properties. CPs that meet this criterion are to be found in all those languages which either nowadays belong to the full V2 languages or which used to be full V2 languages in earlier periods but still display V2 residues.

The Development of V-to-C in the West Germanic and Romance Languages

167

1.3 V2 and V1 in the West Germanic and Romance Languages Both the West Germanic and Romance languages have adopted V2 properties in the course of their history (Gerritsen 1984; Weerman 1989; Vance 1989; Rizzi 1990; Ribeiro 1995). 3 The V2 phenomenon as a basic structural constraint requires the head C° of root clauses to bear specific features related to finiteness which attract the finite verb before Spell out and moreover to trigger the appearance of exactly one focus or topic phrase in its specifier position SpecC (den Besten 1983). The specifying XP is represented by a full lexical constituent, an empty pronoun or a null operator (Rizzi 1997). Consequently, the root clauses of full V2 languages on their surface always display a V1 or V2 structure. 4 Whether those V2 properties were preserved for a long time or abandoned after a rather short period depends on the individual morphosyntactic development of the separate languages and dialects of both language areas. As illustrated in (12), German and Rhaeto-Romance, for instance, up to this day belong to the full V2 languages. (12) a. Auf unserem Speicher wohnen on your attic live ‘Bats live in our attic’ b. Gonoot vas-t a ciasa sua often go-you to house his ‘You often go to his plase’

Fledermäuse bats

(German)

(Rhaeto-Romance)

However, English and French, which clearly exhibited characteristics of full V2 languages in the Middle Ages (cf. (13)), 5 today do not permit Vto-C movement in declarative root clauses. Both languages have retained few V2 patterns, though. Such relics of V-to-C movement or so-called V2 or V1 residues (Rizzi 1991) are found, for example, in interrogative sentences of Modern Standard English and Modern French (cf. (14)). 6

3

4 5 6

Kaiser (2002) argues against the view that the entire Romance language area has passed through a V2 period. In his opinion the large amount of occurring V3 and V4 structures in Old French, Old Spanish and Old Portuguese texts is incompatible with an obligatory Vto-C movement in root clauses. In case of left dislocation or insertion of a hanging topic, many full V2 languages expose additional V3 phenomena in root clauses (cf. Benincà 1995). Especially the northern English dialects require head movement to C of the finite verb in declarative root clauses during the entire Middle English period (Kroch & Taylor 1997). According to Kaiser (2002) French and most other Romance languages never belonged to the full V2 languages. In his opinion, V2 residues do not exist in modern Romance. He argues that the 2P-positioning of the finite verb in Romance interrogative clauses has no connection with any V2 characteristics but results from a more general rule.

168

Melanie Wratil

(13) a. Oþir labur sal þai do (Middle English, Northern England) other work shall they do ‘They should do other jobs’ (The Rule of St. Benet 33.20) b. En mi sa voie a Bertran encontré (Old French) on his way have.3.SG Bertran met ‘He has met Bertran on his way’ (Le Charroi de Nîmes, 11.31) (14) a. What did Sally think? (English) b. Voulez vous changer de place? (French) want.2PL you change of seat? ‘Do you want to change your seat?’

1.4 V2 and the Emergence of CP Up to this day the question of how V2 word order patterns and finally V2 languages emerged is strongly debated in linguistic theory (cf. Vennemann (1974) and Weerman (1989) among others). According to Harris & Campbell (1995: 195ff.), the works of Wackernagel (1892) and Delbrück (1878) on Indo-European word order patterns constitute the basis of the theoretical analysis of the V2 phenomenon. Wackernagel (1892) and Delbrück (1878) showed that unaccented words in Proto-Indo-European tend to follow immediately the first stressed constituent in root clauses in order to avoid their own focusing. This characteristic second position placing (2P phenomenon) that can be also observed in many other language families affects not only particles and unaccented weak pronouns but also short and unemphatic inflected full verbs and above all finite auxiliaries whose verbal complements reside in the focus domain at the end of the clause. Harris & Campbell (1995) assume that, due to the proliferation of analytic verb phrases in the early Germanic and Romance languages, the originally prosodic precondition for this 2P positioning of unstressed verbal elements has been reanalysed as a syntactic principle. As a consequence the placing in the second position ceased to depend on some individual prosodic properties of the shifted element but began to correlate with its status as a syntactic category. Therefore, all finite main verbs of the early Germanic and Romance languages moved into the second position regardless of their own stress or accent, thus giving rise to the fixation of the V2 word order in root clauses. Kiparsky (1995) contests the view that some unaccented verbal elements which require a preceding stressed constituent to lean on could really cause the emergence of V2 languages. His criticism is that such a hypothesis is not compatible with the fact that V2 patterns when they are first becoming established are not always regularly found in all root clauses but obviously prefer certain sentence types. Furthermore, the observation that

The Development of V-to-C in the West Germanic and Romance Languages

169

V1 structures were derived even a long time before the final fixation of V2 structures is inconsistent with Harris & Campbell’s (1995) 2P theory. Kiparsky (1995) assumes that rather the development of subordination and consequently the formation of the operational C-projection are the trigger of V-to-C movement and the establishment of V2 languages. 7 However, Kiparsky’s (1995) thesis can also be criticized since it does not allow for any explanation of the fact that not every language which exhibits embedded CP-complements obligatorily passes through a V2 stage. In order to get closer to a possible solution it seems advisable to add Wackernagel’s (1892) and Delbrück’s (1878) observations to Kiparsky’s (1995) approach. The reason is that the reanalysis of root clauses as V2 structures requires as a precondition that, for some reason, the finite auxiliary or lexical verb of a large part of non-embedded sentences already follows an initial focalized or topicalized constituent in the second position. Only if such a linear pattern persistently occurs, root clauses are able to be restructured as those enlarged structures that were formerly introduced in order to make subordination feasible. Thus, the obligatory V-to-C movement of the finite verb in root clauses has to be seen as a result of both the 2P placing of unstressed verbal elements and the rise of subordinated structures introduced by a complementizer. 1.5 V2, V1 and the Rise of imperative V-to-C movement in the Old West Germanic and Romance languages Kiparsky (1995) points out that V2 structures and thus full C-projections come into use and prevail first in those clauses of the Old Germanic and Old Romance languages that display an initial focus phrase, just like in wh-interrogative clauses. The fronted focused constituent of such sentences has been reanalysed as the scope marking specifier of a new functional projection CP while the finite verb which – presumably due to the rules examined by Wackernagel (1892) and Delbrück (1878) – immediately follows in the second position is reinterpreted as the related head C°. Thus, movement of the focused or wh-phrase, respectively, to SpecCP goes hand in hand with head movement of the finite verb to C° in the Germanic (cf. (15a)) and Romance (cf. (15b)) language areas.

7

Lühr (2002) points out that at least in the Italic languages true subordination is possible much earlier than Kiparsky (1995) states.

170

Melanie Wratil

(15) a. Hwœt hœfts þu gedon? (Old English) what have.2SG you done ‘What have you done?’ (The Homilies of Ælfric i.36,28)) b. Quid facient pauci contra tot milia what do.FUT.3PL few against so-many thousand? fortes (Latin) brave-men? ‘What effect will have a few on so many thousand brave men?’ (Ovid, Fast., 2,229)

Even before topicalization also triggers V2 effects, V1 clauses arise. Obviously, they emerge from focalization processes as well. Interrogative clauses that lack a specifying overt wh-phrase belong to the earliest V1 structures. Their finite verb, which represents the head element of the inquired information, has moved on its own into the focus domain and targeted the head C. In this position it establishes a spec-head relationship with the non-overt interrogative operator located in SpecC, forming a yes/noquestion as illustrated by the data of Old English (cf. (16a)) and Old French in (cf. (16b)) (cf. Rizzi 1991 and Roberts 1993: 96). (16) a. OP eart þu se ðe toweard is? (Old English) you the the coming is be.2SG ‘Are you the one who is coming?’ (The Homilies of Ælfric 480)) b. OP est votre sire encore levez? (Old French) be.3SG your master already stood.up? ‘Has your master already got up?’ (Tristan 1.8021)

A similar development can be observed shortly after that in narrative marked declarative clauses (cf. (17)) and in imperative sentences (cf. (18)). The predication represented by the finite verb is introduced as new information and as such is emphasized in both construction types. In declarative clauses this information is an asserted event at the epistemic level whereas in imperative clauses it brings about a directive relation with a future action at the deontic level. Consequently, the imperative verb moves to C° just like the indicative verb of the declarative clause. In contrast to the indicative verb, which follows a declarative null operator, it constitutes a spec-head configuration with an imperative null operator base generated in SpecC. Hence, the imperative verb began to establish its special head movement procedure by analogy with the finite verbs of V1 interrogative and declarative clauses before V2 as basic structural constraint had been solidified. Already in most Old English texts the imperative verb preferably appears in the sentence-initial position (cf. (18a)).

The Development of V-to-C in the West Germanic and Romance Languages

171

(17) a. OP uuârun thô hirtâ in thero lantskeffi? (Old High German) the herdsmen in this landscape? be.IMP.3PL ‘There were herdsmen in this field’ (Tatian, Evangelienharmonie 6,1)) b. OP diremos nós ora, padre que (Old Portuguese) say.FUT.1PL we now father that ‘Father, now we tell you that…’ (Diálogos de São Gregório 1.4.16) (18) a. OP beo ðu on ofeste (Old English) be.IMP you in haste ‘Hurry up’ (Beowulf and the Fight at Finnsburg 386)) b. OP ide vos a bõa ventura! (Old Portuguese) go.IMP.PL you with god luck ‘Go and good luck’ (Diálogos de São Gregório 1.2.44)

1.6 The Fixation of Imperative V-to-C Movement The imperative verb movement to C°, originally motivated by pure information structural requirements, prevails during the increasing stabilization of V2 structures. Notably it resists every syntactic change from V2 to uniform SVO orders. This strong resistance is attributed to the almost complete absence of subordinated imperative clauses on the one hand 8 and on the other hand to the fact that imperative verbs, in contrast to the verbal elements of other mood markings, normally come immediately after a sentence-initial null operator but practically never follow any overt subject phrase. Accordingly, the primary linguistic data that emerge from the external language (Chomsky 1986: 19ff.) of adult speakers and serve as input experience for the next generation of language learners (Lightfoot 1979; Clark 1990) confirm the positioning of the imperative verb within the C-domain but do not provide any evidence of its remaining in I°. Thus, the general unification towards an SVO surface that affects declarative clauses and finally leads to a restructuring of the C-projection as head initial IP is blocked in imperative clauses. Instead, the internal language (Chomsky 1986: 19ff.) acquired by the language learners is increasingly endowed with obligatory imperative V-to-C movement. As a consequence, the head movement to C° of the imperative verb becomes finally fixed in the language areas concerned. 8

According to Schwager (2004) subordinated imperatives are very unusual because imperatives always represent highly performative speech acts that can be embedded only if the corresponding matrix clause provides a compatible directive context. This leads to a redundancy which is felt to be uneconomical. Embedded imperatives nevertheless occur – albeit very rarely – in some early West Germanic languages as for instance in Middle High German (Grimm 1864: 338:ff.). Such occurrences illustrate that the imperative verb of some Middle High German varieties prefers C° as its landing site but is nevertheless still able to check essential features within the I-domain.

172

Melanie Wratil

Hence, the imperative verbs of the Germanic and Romance languages create a unique correlation between their mood marking and their structural positioning. This in turn has a special impact on the organization of their functional IPs. Since the functional head C° became the obligatory landing site for imperative verb movement and therefore appeared to be the inflectional head responsible for imperative mood marking, its maximal projection CP by reanalysis became the imperative inflectional mood projection. 9 As a consequence, the MoodP inside the imperative INFL projection vanished. Thus, the imperative IP of all present and all former West Germanic and Romance V2 languages contains neither a Tprojection nor a Mood-projection. Up to this day, the mood marking of the imperative verb is merely achieved in C° together with its person agreement and tense specification. 10 After checking its number agreement features against the overt or covert addressee subject located in SpecAgrSP the imperative verb moves to the C-head before Spell out where it checks strong verbal features and establishes a spec-head relationship with the imperative null operator. The latter configuration has finally become a definite criterion according to which the imperative verbal head and the corresponding operator obligatorily enter a spec-head agreement relationship within the C-domain. Thus, it has to be concluded that the imperative V-to-C movement is a concomitant of the establishment of V2 patterns. It is a V1 residue in all languages that do not belong to the full V2languages anymore. Figures (19a) and (19b) illustrate the imperative V-to-C movement with examples from the Romance SVO language Spanish which used to be a true V2 language in earlier periods (cf. (19a)) and from modern German, a current full V2 language with a basic SOV word order (cf. (19b)).

9

This development finally blocks the subordination of imperative clauses because their Cprojection has become unable to bear a complementizer. In fact, embedded imperatives do not occur in German texts written after the 13th century (Grimm 1864: 338ff.). 10 That is because the future orientation and the reference to an addressee are always defined by the imperative mood marking itself (cf. section I. 1)

The Development of V-to-C in the West Germanic and Romance Languages

(19) a. ¡Fumad! smoke.IMP.PL ‘Smoke!’

173

(Spanish)

C(Mood)P 3 Spec OPimp

C(Mood)’ 3 C(Mood)°

AgrSP

fuma-di

3 Spec

AgrS’

DPj

3 AgrS°

VP 3

ti'

Spec

V’ l

tj

V° ti

b. Wartet! wait.imp.pl ‘Wait!’

(German)

C(Mood)P 3 Spec

C(Mood)’

OPimp

3

C(Mood)°

AgrSP

warte-ti 3 Spec

AgrS’ 3

DPj

VP

AgrS°

3 Spec tj

ti V’ l V° ti

1.7 Imperative V-to-C Movement at LF: Modern Standard English In Modern Standard English imperative verbs precede their complements and follow their subject phrase in accordance with the underlying SVO word order of the latest English language periods (cf. (20)). So, it seems

174

Melanie Wratil

that the imperative V-to-C movement in English was finally defeated by the expanding rigid SVO word order and that therefore English imperative verbs, unlike the imperative verbs of German and Yiddish (cf. (7)), do not enter the functional C-projection. (20) a. You open the door b. You go home c. *Tell you her always the truth

(English)

However, the syntactic development of the English imperative exposes something completely different. In section 1.5. it was demonstrated that in Old English imperative verbs already strongly tend to move to C° in order to represent a new directive relation in the superficially sentenceinitial position. This head-movement operation becomes standard for imperatives in Middle English. Examples like (21) substantiate this claim. As expected, Middle English imperative verbs precede their complements, adjuncts and even their subjects in the unmarked case. (21) a. Be ye not proud (Middle English) be.IMP you NEG proud ‘Don’t be proud’ (King James Bible 850) b. Helpe þou me help.IMP you me ‘You help me’ (The Earliest English Prose Psalter 150)

As often discussed, Middle English is characterized by incisive inflectional and categorial changes. The inflectional morphology of the verb erodes throughout all paradigms (cf. Mossé 1969: 105; Gray 1985: 495f.). 11 As a consequence, several preterite present verbs take on the function of modals and the causative verb do turns into an aspect marking raising verb (cf. Ellegård 1953; Denison 1985). In the beginning of the Early Modern English period the new modal verbs finally lose their ability to take DPs as their complements while do sheds its semantic content. Accordingly, the modal verbs and do are reanalysed as functional auxiliaries (Roberts 1993: 292 ff.). In Early Modern English do adopts the function of a verbal substitute that indicates the agreement relation between the subject and the full verb as a stand-in element. It represents the basic moods of the two domains of modal logic, hence the indicative and the imperative mood. It is inserted if and only if the indicative or the imperative lexical full verb, 11 First of all, inflectional mood markings disappear (cf. Mossé 1969: 105; Roberts 1993: 316). The erosion of the inflectional agreement markers follows shortly after that. During the 15th century the agreement morphemes of the 1st person singular and all plural forms in the present tense begin to disappear (cf. Gray 1985: 495f.). In Early Modern English all forms in the present tense consist of the verbal stem except for the 2nd and 3rd person singular in the indicative mood. Since the 16th century even the infinitive morpheme does not occur anymore (cf. Roberts 1993: 246ff.).

The Development of V-to-C in the West Germanic and Romance Languages

175

respectively, does not leave the V-projection on its own. In Early Modern English, the structural position which was filled by the raised and inflected lexical full verb in Middle English hosts either a modal or do, both base generated in INFL, or still the moved lexical full verb. Consequently, as illustrated in example (22), either the imperative lexical verb or the imperative substitute do overtly move to C° in the Early New English imperative clauses. (22)

Early Modern English a. Love ye youre enemys (Tindale The Four Gospels Luke 6-35) b. Do thou have mercy on us (Latimer Letters 355)

As Roberts (1985) as well as Kroch (1989) point out, the overt head movement of the finite full verb decreases from the second half of the 16th century on. This can serve as a diagnostic indicator for the weakening of the inflectional V-features and the increase of head movement operations at LF. After the extinction of the overt V-to-I movement and V-to-C movement in the 17th century the obligatory overt representation of the attracting functional head is for the most part abandoned. Only a few operator configurations, like for instance the Wh-Criterion (cf. (23a,b)) (Rizzi 1991) and the Negation and Affirmation Criterion (cf. (23c,d)) (Haegeman & Zanuttini 1991), still demand the overt movement of the verbal head up to this day and consequently trigger the insertion of the verbal substitute do. (23) a. b. c. d.

English What does Nelly intend to buy? OPWH does Nelly intend to buy a Superwoman suit? No, Nelly doesn't intend to buy a Superwoman suit Indeed, Nelly does intend to buy a Superwoman suit

Therefore, according to the principle of least effort, imperative do nowadays only appears in case of negation and emphatic affirmation. It is base generated at the level of the functional projection responsible for polarity marking (cf. Wratil 2005: 191ff.). From its head position Pol° it moves as negative auxiliary don´t (cf. (24a)) or as affirmative auxiliary do (cf. (24b)) to C° before Spell out forming, as illustrated in structure (25), a residual V1 structure. (24)

English a. Don't you laugh b. Do definitely come to our Halloween party

176

Melanie Wratil

(25) C(Mood)P 3 Spec OPimp

C(Mood)’ 3 C(Mood)° don’ti

AgrSP 3

Spec

AgrS’ 3

youj

AgrS° ti'

PolP 3 Spec

Pol’

OPNEG 3 Pol° ti

VP 3 Spec tj

V’ l V° laugh

In non-negated and non-emphasized imperative clauses do does not appear. Since the imperative operator base generated in SpecC requires an imperative verbal element located in C° in order to obtain the necessary spec-head configuration, verb movement to C° nevertheless takes place. Just like the finite full verb of non-interrogative, non-negated and nonemphasized indicative clauses the imperative full verb moves at LF according to the principle of procrastination. It ends up in C° where it licenses its mood specification. For this reason, unlike the imperative substitutes do and don’t (cf. (24)), imperative full verbs have to follow their subjects and clausal adverbs before Spell out (cf. (26)) (cf. Potsdam 1998: 128ff.). Their head movement procedure at LF is illustrated in structure (27). (26)

English a. You laugh b. Someone definitely come to our Halloween party

The Development of V-to-C in the West Germanic and Romance Languages

177

(27) C(Mood)P 3 Spec OPimp

C(Mood)’ 3 C(Mood)°

[+imp]([+2ps][+fut])i

AgrSP 3 Spec

AgrS’

youj

3 AgrS°

VP 3

ti'

Spec tj

V’ l V° laughi

Thus, Modern Standard English does not constitute an exception among the West Germanic languages with respect to imperative verb movement. It still displays imperative head movement to C°. In all non-negated and non-emphasized imperative clauses V-to-C movement applies at LF. In negative and emphasized affirmative imperative clauses do moves to C° before Spell out. In some varieties of English, as for example in the Belfast English language group spoken in the northern and eastern regions of Ireland, elder speakers still utter imperative sentences which clearly show overt imperative V-to-C movement of the lexical full verb. Accordingly, in the nonnegated and non-emphasized imperative clauses of Belfast English, imperative full verbs to this day sometimes precede their subjects located in SpecAgrS (cf. (28)) (cf. Henry 1995: 45ff.; Henry 1997). (28)

Belfast English a. Give you me that letter b. Tell you her always the truth

2 Imperative Verb Movement on the Left Periphery 2.1 The Internal Structure of the C-Projection Since the late nineties the traditional view of the structural properties of the left periphery has been called into question by many linguists (cf. Kiparsky 1995; Rizzi 1997; Haegeman 1997 and Benincà & Poletto 2001, among others). All of them convincingly contest the view that the clausal left periphery consists of only one C-head and one specifier alone. Accord-

178

Melanie Wratil

ing to Rizzi (1997), the C-domain of a sentence encodes different kinds of information and consequently splits up into various functional domains. Those features which establish the connection between the proposition of a sentence and the preceding discourse or the superordinated structure are encoded in the highest functional projection ForceP, whereas those features that are related to the inflectional properties of the INFL domain and its verbal system are associated with the head of the lowest CPinternal phrase FinP. Furthermore, Rizzi (1997) assumes that topicalization as well as focus fronting result in an extension of the C-projection. That is because focused elements as well as topic phrases move inside the C-domain, creating there a spec-head relationship with their corresponding functional heads Foc° or Top° respectively. Hence, according to Rizzi (1997), the left periphery of a sentence displays the structure illustrated in example (29) where the asterisk (*) marks the possible recursion of the functional TopP. (29)

[Force P [TopP* [FocP [TopP* [FinP]]]]]

(Rizzi 1997)

Rizzi (1997) himself as well as Haegeman (1997), Grewendorf (2002), Poletto (2003) and others prove this internal structure to be characteristic at least of Romance and West Germanic C-projections. But they find some differences between the modern Romance SVO languages and the modern West Germanic full V2 languages with regard to their processes of topicalization and focalization. 2.2 Topic Constructions and Wh-Interrogatives in Modern Romance Languages In the Modern Romance SVO languages topicalized constituents move to SpecTopP before Spell out. Inside the I-domain they leave behind a gap which is identified and bound as a trace by an anaphoric operator. Especially in case of object extraction the relevant operator of many Romance languages is represented by a resumptive clitic that is associated with the topic phrase as a correlate (cf. (30)). 12 (30)

Italian a. [TopP Il tuo libroi [FinP ti [IP the your book ‘Your book, I have read it’

loi ho letto ti ]]] it have.1SG read

12 Topicalization of adjuncts does not require the insertion of an anaphoric operator. For more on this issue the interested reader is referred to Rizzi (1997).

The Development of V-to-C in the West Germanic and Romance Languages

179

b. [TopP

Il libroi [TopP a Giannij[FinP ti tj [IP glielo the book to John him-it.CL(RESPRON) daró ti tj senz' altro ]]] give.FUT.1SG without doubt ‘The book, I will give it to John without a doubt’

Unlike topicalization, focusing generally displays quantificational properties. As Lasnik & Stowell (1991) point out, focusing therefore gives rise to weak crossover effects. 13 The focused constituents of the Modern Romance SVO languages move to SpecFocP before Spell out, leaving behind a bound variable in their base position (cf. (31a)). Since wh-fronting still imposes residual V-to-C movement in these languages, the finite verb of wh-interrogatives ends up in Foc°, entering a spec-head relationship with the wh-phrase located in SpecFoc (cf. (31b)). (31)

Italian a. [FocP IL TUO LIBROi [FinP ti [IP ho letto ti ]]] (non il suo) the your book have.1SG read not the his ‘YOUR BOOK I have read, (not his)’ b. [FocP A che cosai pensanoj [FinP tj ti [IP i tuoi amici tj ti ]]] ? the your friends ? about what think.3PL ‘What do your friends think about?’

2.3 Topic Constructions and Wh-Interrogatives in West Germanic V2Languages The C-system of the West Germanic full V2 languages differs from that of the Romance SVO languages in that their functional head Fin° is host to a strong [V]-feature attracting the finite verb before Spell out in root clauses. As Haegeman (1997) points out, Fin° therefore has to be defined as an L-bound head requiring a specifier-XP according to the Extended Projection Principle (EPP). Since the specifying XP solely serves to place the finite verb in the second position, it does not need to be characterized by certain functional or categorial properties. Consequently, a subject-XP (cf. (32a)) or any topic-XP (cf. (32b)) moves to or through the specifier position of FinP, yielding a spec-head relationship with the verb and satisfying the EPP. Hence, in example (32b) the finite verb (mag (“like”)) has moved to Fin° and the object DP (Manfreds Kuchen (“Manfred´s cake”)) to its specifier position SpecFin. In contrast to the gap left behind by a 13 According to Lasnik & Stowell (1991) A’-relations have quantificational properties only if they cause certain co-reference conflicts – the so-called weak-cross-over effect (WCO). Technically speaking, this happens when an operator moves across a constituent which ccommands its variable trace and contains a pronoun itself, which in turn is co-indexed with the variable trace.

180

Melanie Wratil

fronted topic in the Romance SVO languages, the empty category left behind by a topic phrase in the West Germanic full V2 languages does not have to be identified by a special correlate. As Rizzi (1997) claims, the special properties of the functional head Fin° in those full V2 languages endow the specifier SpecFin with the capacity to license a null constant and accordingly to bind the topic trace. (32)

German a. [FinP Ichi magj [IP ti Manfreds Kuchen I like Manfred’s cake ‘I really like Manfred’s cake’ b. [FinP Manfreds Kucheni magj [IP Manfred’s cake like ‘Manfred’s cake I really like’

wirklich really

gerne tj]] gladly

ich ti wirklich gerne tj ]] ? I really gladly

In the wh-interrogative sentences of the West Germanic full V2 languages the strong [V]-feature in Fin° also attracts the finite verb before Spell out (cf. (33a)). As originally discussed in Grewendorf (2002), in the case of short wh-movement the corresponding wh-phrase moves to the specifier position of the FinP, satisfying the EPP and the Wh-Criterion in the spechead relationship to the finite verb before Spell out. Structurally, it behaves as a topic not giving rise to any WCO-effects. Hence, in example (33b) the wh-phrase (wen (“whom”)) and the possessive pronoun (sein (“his”)) can bear the same index without causing any ungrammaticality. The movement of the operational wh-phrase to the specifier position of FocP continues after Spell out. (33)

German a. [FinP Wemi klautj [IP Max die Socken ti tj?]] From-whom steal.3SG Max the socks ‘From whom does Max steal the socks?’ seini bester Freund ti an die Polizei verraten tj?]] b. [FinP Weni hatj [IP Whom have.3SG his best friend to the police betrayed ‘Whom has his best friend betrayed to the police?’

2.4 The Imperative Null Operator in SpecFoc As already been shown in sections I. 5. and I. 6., imperative V-to-C movement is a V1 residue which originates in a focalization process. Thus, the imperative verb, creating a new directive relation with a future action, ends up in the C-domain where it establishes a spec-head configuration with its corresponding null operator. Since it was the functional head Fin° that was equipped with relevant strong verbal features in the C-domain of the Germanic and Romance languages in the time of the fixation of the imperative V-to-C movement, it was solely FinP that could have been

The Development of V-to-C in the West Germanic and Romance Languages

181

reanalysed as the functional projection responsible for imperative mood marking. Accordingly, the imperative verb of all present and all former V2 languages enter the functional FinP before Spell out, licensing its mood specification in its head position Fin(Mood)°. After checking its mood specification the imperative verb of the modern Romance languages moves to Foc° before Spell out where it instantiates a spec-head relationship with the imperative null operator located in SpecFoc as illustrated in the Italian example (34a) and its simplified structure in (34b). (34)

Italian [IP tu pure!]]] a. [FocP OP tacii [FinP ti be.IMP-quit you definitely ‘You definitely be quite’ b. FocP 3 Spec

Foc’

OP

3 Foc° tacii

Fin(Mood)P 3 Spec

Fin(Mood)’ 3

Fin(Mood)° ti'

IP 6 tu ti pure

Therefore, the imperative verb comes after topicalized scene setting adverbs like adesso (“now”) and oggi (“today”), as illustrated by the Italian examples in (35). Furthermore, just as the finite verbs of wh-interrogative clauses (cf. (36a)), it obligatorily precedes focus particles that are generated at the bottom edge of the C-domain (cf. (36b) and (36c)), as shown on the basis of Rhaeto-Romance data. (35)

Italian a. [TopP Adesso [FocP daigli[FinP ti [IP pure ti la notizia! ]]]] now give.IMP.CL(PRON) really the news ‘Definitely tell him the news now!’ b. [TopP Oggi [FocP vaccii [FinP ti [IP tu ti ]]] ? today go.IMP you ‘You go today’

182 (36)

Melanie Wratil

Badiotto (Rhaeto-Romance) o -l [FocPrtP pa [IP ti]]]? a. [FocP Cii what want.3SG -CL(PRON) FOCPRT ‘What does he want?’ [IP dessigÿ ti]]]] b. [FocP Fajéi-l [FinP ti [FocPrtP pa do.IMP.PL-CL(PRON) FOCPRT definitely ‘Definitely do it!’ Fajéi-l [FinP ti [FocPrtP [IP dessigÿ ti]]]] c. *[FocPrt Pa FOCPRT do.IMP.PL-CL(PRON) definitely’

In West Germanic full V2 languages matters are much more complicated. If, according to Haegeman (1997), the EPP has to be satisfied at the level of FinP, the empirical question arises how V1 residues, like V1 imperatives, are structurally represented in these languages. Hypothetically, one could assume that there is a null operator base generated in the specifier position of FinP which establishes a spec-head configuration with the finite verb located in Fin° in order to satisfy the EPP. But this is not plausible at all. On the one hand it appears extremely improbable that a sentence-type operator is base generated in a position that displays A-features like SpecFin. On the other hand it can be observed that, at least in German, the Vorfeld of imperative clauses can be filled by overt lexical material. Accordingly, as illustrated in (37), topicalized DPs and unstressed adverbs that end up in SpecFin precede the imperative verb in German. (37)

German a. [FinP Nun kommj [IP ti schon tj!]] Now come.IMP PRT ‘Come on now!’ ti mal lieber tj !]] b. [FinP Dasi vergissj [IP this forget.IMP PRTbetter ‘You’d better forget this!

The same holds for the wh-phrases of so-called wh-imperatives. The whphrases of these constructions are not able to expand their interrogative scope because the imperative null operator blocks their movement to the specifier position of FocP. Consequently, the fronted wh-phrases of whimperatives behave as topics and therefore reside in the functional FinP. They neither modify the underlying imperative sentence with regard to its illocutionary force (cf. Reis & Rosengren 1992), as illustrated by (38a) and its translation, nor do they give rise to any WCO-effects, as shown in (38b).

The Development of V-to-C in the West Germanic and Romance Languages

(38)

183

German du mal [FinP ti dass [IP Erna wohl ti a. [FinP Wasi rat [IP what guess.IMP you PRT that Erna well kaufen wird!]]]] buy will.3SG ‘Guess what Erna will buy!’ b. [FinP Weni sag [IP mal [FinP ti hat [IP seinei Freundin ti betrogen!]]]] his girlfriend betrayed Who say.imp PRT have.3sg ‘Tell me who has his girlfriend betrayed’

Moreover, even left dislocation is allowed in the imperative constructions of West Germanic full V2 languages. Grewendorf (2002: 76ff.) points out that left dislocated elements in German, such as den Scheck (“the cheque”) or deine Kommentare (“your comments”) in (39), are specifiers of complex DPs that ended up in the specifier position of a TopP, leaving behind their D-heads in the lower unmarked topic position SpecFin. (39)

German a. [TopP Den Schecki [FinP deni behalt [IP ti besser!]]] the cheque it keep.IMP better ‘As for the cheque, you'd better keep it!’ spar [IP dir ti !]]]] b. [TopP Deine Kommentarei [FinP diei Your comments them keep.IMP for-you ‘As for your comments, keep them to yourself! ’

On condition that the EPP had to be met at the level of the FinP in accordance with Haegeman’s (1997) proposal and that consequently the imperative null operator was in fact base generated in SpecFin, the imperative clauses in (37), (38) and (39) should be ungrammatical. The imperative null operator would prevent topicalized phrases from moving to or through SpecFin. However, since these sentences are clearly well formed, it has to be concluded that Haegeman’s (1997) EPP restriction for the root clauses of West Germanic full V2 languages is too strict. The data above rather evince that the EPP can be satisfied not only at the level of the FinP but also at the level of FocP. Accordingly, just like the imperative operator of the modern Romance languages, the imperative operator of the modern West Germanic V2 languages is base generated in the specifier position of FocP. Hence, the imperative verb moves to Fin(Mood)° before Spell out in order to license its mood specification and at the same time its tense and person agreement features (cf. (40a)). After Spell out it targets Foc°, entering a spec-head relationship with its imperative null operator and satisfying the EPP (cf. (40b)). As illustrated in the treestructure (40c), the German imperative verb therefore comes after topicalized phrases. It fulfils the imperative criterion with its corresponding operator located in SpecFoc at LF.

184 (40)

Melanie Wratil

German [IP ti mal lieber tj!]]] a. before Spell out: [FocP OP [FinP dasi vergissj this forget.IMP PRTbetter ‘You'd better forget this’ dasi tj [IP ti mal lieber tj !]]]] b. LF: [FocP OP vergissj [FinP forget.IMP this PRT better ‘You'd better forget this’ c. FocP 3 Spec

Foc’ 3

OP Foc°

[+foc][+imp]j

Fin(Mood)P 3 Spec Fin(Mood)’ 3 dasi Fin(Mood)° IP vergissj 6 ti mal lieber tj

It has to be concluded that the imperative verb of the modern West Germanic and Romance languages establishes a spec-head relationship with the imperative null operator in the CP-internal functional projection FocP.

3 Conclusion The imperative V-to-C movement of the modern West Germanic and Romance languages is anchored in earlier stages of language evolution when focalization began to be accompanied by V2 structures. It has remained stable to this day even in those languages that for the most part have lost V2 patterns. This has to be attributed to the fact that the input data to which the language learner in the language areas concerned has been exposed do not provide any clear evidence, such as numerous embedded or subject initial imperative clauses that point to a canonical remaining of the imperative verb in I°. Therefore, the functional head C° of imperative clauses has been prevented from being reanalysed as left peripheral head I° preceded by a specifying subject phrase. Instead, it has become the inflectional head responsible for imperative mood marking. Up to this day it attracts the imperative verb before Spell out or at LF. Presupposing a decomposed C-projection the imperative verb moves to the CP-internal functional head Fin° where it checks its mood specifica-

The Development of V-to-C in the West Germanic and Romance Languages

185

tion and at the same time its tense and person agreement features. Hereafter it ends up in the head position of the operational FocP there establishing a spec-head configuration with the imperative null operator located in SpecFoc.

References Benincà, P. & C. Poletto (2001): Topic, Focus and V2: Defining the CP Sublayers. Ms. University of Padova, CNR. Bybee, J.L. (1990): The Grammaticization of Zero: Asymmetries in Tense and Aspect Systems. La Trobe Working Papers in Linguistics 3, 1-14. Bybee, J.L., R.D. Perkins & W. Pagliuca (1994): The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago: University of Chicago Press. Chomsky, N. (1986): Knowledge of Language: Its Nature, Origin and Use. New York: Praeger. Cinque, G. (1999): Adverbs and Functional Heads. A Cross-Linguistic Perspective. Oxford: Oxford University Press. Clark, R. (1990): Papers on Learnability and Natural Selection. Technical Reports on Formal and Computational Linguistics 1. Genève: Université de Genève. Delbrück, B. (1878): Die altindische Wortfolge aus dem Catapathabrâ. Halle an der Saale: Buchhandlung des Waisenhauses. den Besten, H. (1983): On the Interaction of Root Transformations and Lexical Deletive Rules. In: W. Abraham (ed.), On the Formal Syntax of the Westgermania. Amsterdam and Philadelphia: John Benjamins, 47-131. Denison, D. (1985): The Origins of Periphrastic Do: Ellegård and Visser Reconsidered. In: R. Eaton et al. (eds.), Papers from the 4th International Conference on Historical Linguistics, Amsterdam, April 10-13, 1985. Amsterdam and Philadelphia: John Benjamins Publishing Company, 45-60. Ellegård, A. (1953): The Auxiliary do: The Establishment and Regulation of its Use in English. Volume II of Gothenburg Studies in English. Stockholm: Goteborg. Gerritsen, M. (1984): Divergent Word Order Developments in the Germanic Languages: A Description and a Tentative Explanation. In: J. Fisiak (ed.), Current Trends in Historical Syntax. The Hague: Mouton, 107-135. Gray, D. (1985): The Oxford Book of Late Medieval Prose and Verse. Oxford: Oxford University Press. Grewendorf, G. (2002): Minimalistische Syntax. Tübingen: Francke. Grimm, J. (1864): Kleinere Schriften. Berlin: Dümmler.

186

Melanie Wratil

Haegeman, L. (1997): Verb Second, the Split CP and Null Subjects in Early Dutch Finite Clauses. GenGenP 4(2), 133-175. Haegeman, L. & R. Zanuttini (1991): Negative Heads and the Neg Criterion. The Linguistic Review 8, 233-252. Hale, K. (1976): The Adjoined Relative Clause in Indo-European. In: R.M.V. Dixon (ed.), Grammatical Categories in Australia. Canberra: Australian Institute of Aboriginal Studies. Han, Ch., H. (2000): The Structure and Interpretation of Imperatives. New York: Garland. Harris, A.C. & L. Campbell (1995): Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Henry, A. (1995): Belfast English and Standard English: Dialect Variation and Parameter Setting. Oxford: Oxford University Press. Henry, A. (1997): Viewing Change in Progress: The Loss of V2 in HibernoEnglish Imperatives. In: A. van Kemenade & N. Vincent (eds.)(1997), 273296. Holland, G.B. (1984): Subordination and Relativization in Early Indo-European. BLS 10, 609-622. Kaiser, G.A. (2002): Verbstellung und Verbstellungswandel in den romanischen Sprachen. Tübingen: Niemeyer. Kemenade, A. van (1997): V2 and Embedded Topicalization in Old and Middle English. In: A. van Kemenade & N. Vincent (eds.), Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press, 326-352. Kiparsky, P. (1995): The Indo-European Origins of Germanic Syntax. In: A. Battye & I. Roberts (ed.), Clause Structure and Language Change. Oxford: Oxford University Press, 140-170 Kroch, A. (1989): Reflexes of Grammar in Patterns of Language Change. Journal of Language Variation and Change 1, 199-244. Kroch, A. & A. Taylor (1997): Verb Movement in Old and Middle English. In: A. van Kemenade & N. Vincent (eds.) (1997): Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press, 297-325. Lasnik, H. & T. Stowell (1991): Weakest Cross-over. Linguistic Inquiry 22, 687720. Leumann, M., J.B. Hofmann & A. Szantyr (1965): Lateinische Grammatik II. Syntax und Stilistik. München: C.H. Beck’sche Verlagsbuchhandlung. Lightfoot, D. (1979): Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Lühr, R. (2002): Merkmale von Nebensätzen indogermanischer Sprachen: Ikonische Syntax. Ms. Universität Jena.

The Development of V-to-C in the West Germanic and Romance Languages

187

Lyons, J. (1977): Semantics. Cambridge: Cambridge University Press. Mitchell, B. (1985): Old English Syntax. Oxford: Clarendon Press. Mossé, F. (1969): Handbuch des Mittelenglischen. München: Max Hueber Verlag. Palmer, F.R. (1986): Mood and Modality. Cambridge: Cambridge University Press. Platzack, C. & I. Rosengren (1998): On the Subject of Imperatives: A Minimalist Account of the Imperative Clause. Journal of Comparative Germanic Linguistics 1, 177-224. Poletto, C. (2003): The left-periphery of V2-Rhaetoromance Dialects: a new View on V2 and V3. Ms. University of Padua. Potsdam, E. (1998): Syntactic Issues in the English Imperative. New York: Garland. Reis, M. & I. Rosengren (1992): What do WH-Imperatives Tell us about WHMovement?. Natural Language and Linguistic Theory 10, 79-118. Ribeiro, I. (1995): Evidence for a Verb-Second Phase in Old-Portuguese. In: A. Battye & I. Roberts (eds.): Clause Structure and Language Change. Oxford: Oxford University Press, 110-139. Rivero, M.-L. & A. Terzi (1995): Imperatives, V-Movement and Logical Mood. Journal of Linguistics 31, 301-322. Rizzi, L. (1990): Speculations on Verb Second. In: J. Mascaró & M. Nespor (eds.), Grammar in Progress. Dordrecht: Foris, 375-386. Rizzi, L. (1991): Residual Verb Second and the Wh-Criterion. University of Geneva Technical Reports in Formal and Computational Linguistics 2. Rizzi, L. (1997): The Fine Structure of the Left Periphery. In: L. Haegeman (ed.), Elements of Grammar, 281-337. Dordrecht: Kluwer. Roberts, I. (1985): Agreement Parameters and the Development of English Modal Auxiliaries. Natural Language and Linguistic Theory 3, 21-58. Roberts, I. (1993): Verbs and Diachronic Syntax. A Comparative History of English and French. Dordrecht: Kluwer. Rögnvaldsson, E. (2003): The Syntax of the Imperative in Old Scandinavian. Ms. University of Iceland. Schwager, M. (2004): Why Imperatives can(not) embed. Poster at GURT 2004, Georgetown University. Schwyzer, E. (1939): Griechische Grammatik I. München: C.H. Beck´sche Verlagsbuchandl. Vance, B. (1989): Null Subjects and Syntactic Change in Medieval French. Ph.D. Dissertation. Cornell University. Vennemann, T. (1974): Topics, Subjects and Word Order: from SXV to SVX via TVX. In: J. Anderson & C. Jones (eds.), Historical Linguistics. Vol.1. Amsterdam: North Holland, 339-376.

188

Melanie Wratil

Wackernagel, J. (1892): Über ein Gesetz der indogermanischen Wortstellung. Indogermanische Forschungen 1, 333-436. Weermann, F. (1989): The V2 Conspiracy. A Synchronic and Diachronic Analysis of Verbal Positions in Germanic Languages. Dordrecht: Foris. Wratil, M. (2005): Die Syntax des Imperativs. Eine strukturelle Analyse zum Westgermanischen und Romanischen. Studia Grammatica 62. Berlin: Akademie Verlag. Zanuttini, R. (1997): Negation and Clausal Structure. A Comparative Study of Romance Languages. Oxford: Oxford University Press.

Evidence for Two Types of Focus Positions in Old High German Svetlana Petrova & Roland Hinterhölzl (Berlin)

1 Word order variation in Early Germanic Standard generative accounts 1 assume a uniform SOV base order in the early Germanic languages. This order shows up in subordinate clauses introduced by an overt complementizer in the head of a functional projection CP. In root clauses, the empty position of the complementizer is filled by the finite verb (Vf) while movement of another constituent to SpecC yields V2 on the surface. Under this analysis, the early Germanic languages share the syntactic structure of asymmetric SOV languages like modern German or Dutch, with some additional properties which explain apparent violations of this basic pattern. One of them concerns the obligatoriness of V-to-C movement in root clauses. While in Old High German (OHG), this kind of operation is well established in root clauses (cf. Axel 2007, for variation between V1 and V2 cf. Hinterhölzl and Petrova 2009), it is argued that in Old English (OE) movement to C is restricted to contexts with syntactic operators (wh- and negation words, sentence adverbials þa and þonne) in SpecC, with the verb targeting a lower projection (e.g. I°) in all remaining types of root clauses (cf. Kemenade 1987, 1997 among others on residual V2 in OE). Another special property of early Germanic syntax that has been in the center of attention in the research is word order variation in the right periphery of clauses. In fact, there is significant evidence for subordinate clauses which display lexical material to the right of Vf in early Germanic. According to van Kemenade (1987, 39–41) and Tomaselli (1995, 350– 351), such surface orders are the result of rightward movement of phrases, i.e. of extraposition, which is attested in the modern SOV languages as well. They also show that rightward movement in early Germanic and in 1

See Lenerz (1984) for Old High German, van Kemenade (1987) for Old English and Erickson (1997) for Old Saxon.

190

Svetlana Petrova & Roland Hinterhölzl

the modern SOV languages affects the same type of constituents, namely i) PPs and other phonologically ‘heavy’ constituents, e.g. CP-complements or heavy DPs, or ii) VPs in clauses with complex verb forms in which Verb Raising (VR) or Verb Projection Raising (VPR) takes place. 2 In all cases, however, Vf is assumed to remain in its basic position at the end of the clause. This account has been challenged on the basis of previously unnoticed evidence from Old English. Pintzuk (1991, 1999) provided examples of embedded clauses in which Vf is followed by types of phrases which do not undergo rightward movement in the modern SOV languages, e.g. pronominal objects, light adverbs or verbal particles. As such orders cannot be derived by extraposition, Pintzuk assumes leftward movement of Vf to the head of a clause-medial functional projection I(nflection)P. Additionally, she found evidence for particles and light elements following the main verb in verb clusters. Such orders can receive no other interpretation than as being instances of basic VO order. This means that against the standard account OE cannot be viewed as uniformly head-final in the base. Instead, word order variation in OE is the result of variation in the head-complement parameter both in VP and IP. This variation, in turn, is said to go back to language contact and to be an instance of grammar competition in the I-language of one and the same speaker (also called Double Base Hypothesis, see Kroch 1989). A refinement of the Double Base Hypothesis is proposed by Fuss and Trips (2002). They account for the fact that one particular order which is potentially possible in the model of Pintzuk (1991, 1999), namely the one with a head-initial V° and a head-final I°, is actually not attested in the Old English records. Additionally, they observe a mismatch between the distributional properties of pronominal subjects and adverbs in main vs. embedded clauses in OE. While in main declarative clauses, adverbs cannot intervene between pronominal subjects and Vf, they are allowed to do so in embedded clauses. On the assumption that i) the position of the pronominal subject marks the left edge of IP (according to Pintzuk 1993, 1999) or TP (according to Fuss 2000) and ii) that Vf in main declarative clauses is in the head I° or T° respectively, leftward movement of Vf in embedded clauses should target another projection below IP/TP. Fuss and Trips assume this projection to be QP, a functional node that is only headinitial and hosts light verbs like auxiliaries and causative verbs. In line with this model, the grammars in competition in OE differ with respect to the 2

VR and VPR refer to some special surface orders in which Vf occurs in medial position and is followed either by the untensed main verb V (VR) or by V and (some of) its complements (VPR). Note that the analysis of VR and VPR as instances of rightward movement reflects the original account proposed by Haegeman and van Riemsdijk (1986).

Evidence for Two Types of Focus Positions in Old High German

191

presence or absence of overt movement of Vf to QP , with additional variation of OV vs. VO in the base. The issues put forward in the discussion on word order variation in OE have been applied to the explanation of this phenomenon in the historical stages of German as well. It is well known that over a long period of time German also displays a great deal of structural variation in subordinate clauses. Traditional descriptions, e.g. Behaghel (1932, 44, § 1457) or Ebert (1978, 39–43), report a high number of subordinate clauses in which Vf occurs in clause-medial position, followed by types of constituents which never appear postverbally in present-day German. Recent work by Weiß (2006) and Schallert (2006) addresses word order variation in the earlier records of German from two rather different perspectives. Weiß discusses word order variation in clauses introduced by dass ‘that’ in the so called ‘minor’ texts of the OHG period. Although he shows that a great number of clauses with verb-medial orders receives a proper interpretation within the standard extraposition account, he identifies some patterns which call for an alternative explanation. First, he provides evidence for embedded V1 and V2 in OHG. He argues that V2orders with a non-subject before Vf are derived via overt movement of Vf to the head of a clause-medial functional (thus left-headed) projection TP which is between CP and P. While the specifier position of P is targeted by subjects, the specifier position of TP can host non-subjects as well, thus giving rise to embedded V2 with non-subjects before Vf in OHG. 3 Second, Weiß provides examples which show properties of VO in the base: these are mainly cases in which a clause-medial Vf is followed by a particle or a non-finite main verb and its object. However, while Weiß is skeptical as to whether basic VO-order can be really claimed for OHG due to scanty evidence in his corpus, Schallert (2006) proposes a model according to which OHG is a language with a mixed OV/VO base order. Our alternative approach is to assume that variation in word order is variation on the surface that is due to the expression of informationstructural categories. Adopting the Universal Base Hypothesis (UBH, cf. Kayne 1994), we assume that there is no variation in the base, deriving different surface orders from a unique order Spec-Head-Complement and leftward movement only (Hinterhölzl 2006). This approach has been applied to the explanation of word order variation and change in Older Icelandic where surface OV and VO orders co-occur for centuries (see Hróardóttir 2000). Hinterhölzl (2004) argues for a similar approach to explaining variation in OE and OHG. The hypothesis that the position of 3

Axel (2007, 97-104) provides an alternative solution to such data. She opts for an analysis of embedded V2 as an instance of CP-recursion which is typical for some modern Germanic languages as well (Vikner 1995).

192

Svetlana Petrova & Roland Hinterhölzl

Vf in OHG correlates with the realization of information-structural categories in the clause is supported by recent empirical investigations showing that Vf separates the less relevant, or presupposed information from the new or asserted information in the clause (see Schlachter 2004, and with some modifications Petrova 2009). In the present study, we want to integrate these empirical facts into a syntactic model of clause structure in OHG.

2 Surface Orders in Dependent Clauses in OHG 2.1 Remarks on the Database In this section, we concentrate on data from the OHG Tatian translation, one of the largest prose texts from the early period of OHG (around 850). We base our observations on clauses in which the OHG text differs in word order from the corresponding Latin original. In applying this method we subscribe to the view that instances of the kind provide evidence for native OHG grammar (cf. Dittmer and Dittmer 1998, Donhauser 1998, Fleischer, Hinterhölzl and Solf 2008). We use a database containing all deviations from the Latin original found in the text parts of the scribes , and H (a total of 1.658 entires) and concentrate on dependent clauses included in it. More precisely, we select clauses which specify a situation variable or fill an argument position of a higher predicate as well as in which the C-domain is filled by an adverbial subordinator, a complementizer or a relative pronoun. In many cases, these structures are used in OHG to translate different kinds of dependent clauses of the Latin original, see (1) below. However, note that in OHG, constructions with the same formal and semantic properties are found even in those cases in which the Latin equivalent is a non-finite clause, e.g. an infinitival clause (2) or participial construction (3). This strongly suggests that the structures in question are instances of genuine OHG syntax: (1)

4

[thô sie thar uuarun] / uuvrđun when they there were PASS.AUX taga gifulte [thaz siu bari] days filled.PL that she gave-birth ‘When they were there, the time came for her to give birth’ (T 35, 22-23) 4 lat. cum essent ibi / Impl&i sunt dies ut parer& Examples are cited by manuscript page and line number according to the diplomatic edition of Masser (1994). A slash represents the end of line both in the Latin and OHG text. Note that in the manuscript, the Latin source and the OHG translation form two juxta-

Evidence for Two Types of Focus Positions in Old High German

(2)

(3)

193

Inphieng thô antwurti […] / [thaz her received then response that he niarsturbi] NEG-die.SBJV ‘It was revealed [to Simeon] that he would not die’ (T 37, 28-29) lat. & responsum acceperat […] / non uisurum sé mortem Thô herod arstarb when Herod died ‘When Herod was dead’ (T 41, 20) lat. Defuncto autem herode

Following traditional classifications, we distinguish between conjunctional and relative clauses in the database. As for the latter, we have to be aware of cases that yield a possible root interpretation as well. See (4) in which the structure in square brackets is ambiguous between a subordinate relative clause with an extraposed object and an asyndetic coordinate root conjunct showing the typical V2 order. 5 But apart from such cases, the database provides numerous examples of relative clauses which can be unambiguously identified as subordinate structures, e.g. headless relative clauses (5a) as well as restrictive relative clauses (5b). Note that Vf in clause-medial position is ungrammatical in such structures in modern German but obviously possible in OHG: (4)

sum tuomo uuas In sumero burgi/ certain judge was in certain town.DAT [ther niforhta got] DEM NEG-feared God (a) ‘There was a judge in a town who did not fear God’ (b) ‘There was a judge in a town. He did not fear God’(T 200, 30-31) lat. Iudex quidam erat In quadam ciuitate/ qui deum non timebat (5) a. salige sint [thiethar sint sibbisame] blessed.PL are who-PRT are peaceful.PL ‘Blessed are the peacemakers’ (T 60, 16) lat. Beati pacifici

5

posed columns. Every line in the OHG text contains exactly the material provided in the corresponding Latin one, with a few exceptions in the entire codex (cf. Masser 1997). Endriss and Gärtner (2005) describe a parallel biclausal construction in modern German and argue for an interpretation of the second clause as a quasi-relative root conjunct with a special topic-marking function. They also outline a number of conditions which allow for V2 in this case. Petrova and Solf (t.a.), Axel (in prep.) point out that the conditions allowing for V2 in the same kind of construction in OHG are identical to those for its modernGerman equivalent. Following this, it is legitimate to assume that not only the formal but also the semantic properties of the construction are identical in OHG and in modern German, which favours a root interpretation of the second conjunct in (4).

194

Svetlana Petrova & Roland Hinterhölzl

b. In thie burg / galileę [thero namo ist nazar&h Into that town [of] Galilee whose name is Nazareth ‘into that town of Galilee whose name is Nazareth’ (T 28, 4-5) lat. In ciuitatem / galileae cui nomen nazar&h

A further specification is needed with respect to causal clauses, cf. Dittmer (1991). It is well-known that the elements introducing them in OHG, mainly uuanta or bithiu uuanta ‘because’, vary between extra-clausal connectives linking two coordinated root clauses (cf. denn in modern German) and subordinating conjunctions placed in C° (cf. da and weil 6 in modern German). Since by virtue of these properties, causal clauses may confuse the picture gained for conjunctional clauses in general, the figures for them are kept separate in the statistics. 2.2 Quantitative Distribution and Clues to Basic Order First, we determine the ratio of surface orders in which Vf is in clause-final position as opposed to those displaying lexical material to the right of Vf. 7 Table 1 provides an overview over the quantitative distribution of these orders in the three groups of dependent clauses distinguished in the database: clause type conjunctional clauses relative clauses causal clauses total number

total 360 203 79 642

clause-final Vf 199 55,28 % 139 68,47 % 29 36,71 % 367 57,17 %

non-final Vf 161 44,72 % 64 31,53 % 50 63,29 % 275 42,83 %

Table 3: Surface orders in dependent clauses in OHG Tatian

6 7

It is well known that in modern German, causal clauses with weil can also display V2 order under special conditions, see Altmann (1997) among others. Postverbal clausal structures, both finite and non-finite ones (infinitives, participial constructions and the like), are not considered in the investigation. It is well known that CPcomplements, in contrast to DP-complements, are regularly extraposed in modern German, i.e. they follow their selecting verb. OHG displays no difference in this respect. See (i): (i) só ír uuoll& [thaz íu man how you want that you.DAT.PL PR.INDEF tuon] / só tuot ír ín selbsama treat the way treat you-PL them.DAT.PL REFL.PR ‘Treat other people the way you want to be treated by them’ (T 65, 16–17) lat. & prout uultis ut faciant uobis homines. / facite illis similiter We consider this type of evidence as irrelevant for the question at issue. Please note that extraposed clausal structures are not included in the figures in Table 1.

Evidence for Two Types of Focus Positions in Old High German

195

At first sight, these figures reveal that clauses with non-final Vf surface order are relatively frequent in OHG. The high proportion of non-final Vf orders among the causal clauses is compatible with the intuition that we probably deal with root conjuncts in part of the data. But among the conjunctional clauses, which do not allow for a paratactic analysis, the proportion of non-final Vf orders is still significantly high. Here, in nearly half of the cases the scribe decided to depart from the structure of the Latin original but ended up in a structure which is different from the assumed basic verb final order. However, the frequency of surface patterns is only one aspect of the phenomenon in question. It has been pointed out in the literature that superficial non-final verb orders are ambiguous between a VO base order and a surface VO order derived by extraposition from an OV base. Recent studies on word order variation in early Germanic therefore consider examples which contain linguistic diagnostics as indications of underlying order (cf. Kroch and Taylor 2000). First, clues to base order are provided by negative evidence from modern SOV languages. In modern SOV languages, extraposition is typical for adjunct-PPs, CPs complements and heavy (modified) DPs, while it is excluded for pronouns, verbal particles, predicative nouns and adjectives, as well as for light (e.g. monosyllabic) adverbs. If constituents of the latter group appear after Vf in a subordinate clause, it is unlikely that they have aquired this position via rightward movement. In case the predicate consists of Vf only, we cannot tell whether we have a surface VO-order derived by leftward movement of Vf, or an instance of basic VO. But, if the predicate consists of a finite auxiliary and a non-finite main verb (V), a light element to the right of V is a clear indication of VO in the base (see, e.g., Kroch and Taylor 2000). Second, there is a class of elements which mark the right edge of the VP in SOV languages. These are non-finite main verbs (6a), nominal parts of predicates (6b), and verbal particles (6c). If in a subordinate clause, an argument XP appears after Vf which follows one of these elements, it is standardly assumed that this is a case of extraposition from a basic OVorder: (6)

a –V–Vf–XP b. –predAdj/N–Vf–XP c. –PRT–Vf–XP

Based on these general observations, we shall turn to the analysis of surface orders in dependent clauses in OHG.

196

Svetlana Petrova & Roland Hinterhölzl

2.3 The Derivation of Orders with clause-medial Vf Table 2 provides an overview over the type and frequency of phrases appearing in postverbal position in dependent clauses in the database. Aditionally, we specify whether the constituent in question is in postverbal position in the Latin original as well (+lat) or whether it is realized after Vf in OHG independently of the Latin order (-lat). clause type conj cl rel cl caus cl total

PP +lat 84 18 12 11 4

-lat -5 3 8

full DP +lat -lat 68 3 12 10 21 2 10 15 1

pron DP +lat -lat 7 -1 -1 2 9 2

adverb +lat -lat 3 -----3 --

predAdj +lat -lat 6 2 3 5 1 -10 7

Inf/Part +lat -lat 1 18 9 4 5 8 15 30

Table 2: Typology of postverbal phrases in dependent clauses in the OHG Tatian database

Let us analyze these orders from the point of view of the diagnostics in (6). It is well-known that PPs are a typical subject of extraposition in modern German. The OHG data fits into this picture. As shown in (7a–c), we find in the examples all types of diagnostics mentioned in (6) above. This seems to support the view that the surface orders attested here are derived from a basic OV order and the PP is moved to the right of Vf: (7)

a. –V–Vf–PP noba imo íz gigeban uuerde fon if not him it given PASS.AUX.SBJV from himile Heaven.DAT ‘unless it be granted to him by Heaven’ (T 57, 7) lat. nisi ei fuerit datum de caelo b. –predAdj–Vf–PP salige sint thiethar arme sint in geiste blessed.PL are who-PRT poor.PL are in mind.DAT ‘blessed are the simple-minded’ (T 60, 3) lat. beati pauperes spiritu c. –PRT–Vf–PP mit thiu hér thó ingieng in capharnaum when he then PRT-went in Capharnaum ‘when he arrived in Capharnaum’ (T 83, 8) lat. Cum introiss& in capharnaum autem

Next to PPs, DPs taking all possible syntactic functions in the clause (i.e. subjects, objects and predicative DPs) may appear after Vf in dependent clauses in OHG. In modern German, this property is attested for heavy

Evidence for Two Types of Focus Positions in Old High German

197

DPs which are able to form an intonational phrase on their own (Hinterhölzl 2004, 152). In the OHG examples, postverbal DPs in part of the data may also be considered as heavy DPs. Additionally, we find a typical diagnostics for extraposition from a basic OV-order in (8a): (8)

a. –V–Vf–DP thaz gibrieuit uuvrdi [al these umbiuuerft] that registered PASS.AUX.SBJV all this universe ‘that all the universe be registered’ (T 35, 9) lat. ut describer&ur uniuersus orbis b. thaz mannes sun / hab& [giuualt in erdu / that man.GEN.SG son has power in earth zifurlazenne sunta] to forgive sins ‘that the Son of Man has the power to forgive sins on Earth’ (T 89, 26-28) lat. quod filius hominis / potestatem hab & in terra / dimittere peccata

However, in OHG, light DPs, e.g. pronouns (9a) and bare nouns (9b) as well as predicative nouns (9c) may also occur after Vf in subordinate clauses. This is a problem for the extraposition account because in modern SOV languages, phrases like these do not undergo rightward movement: (9)

8

a. –PRT–Vf–pronominal DO thaz sie úz uuvrphin sie that they PRT threw them ‘that they expelled them’ (T 76, 2) 8 lat. ut eicerent eos b. –Vf–single DO Inti thiethár hab&un diuual and whoPRT had devil ‘and those who were possessed by the devil’ (T 59, 1) lat. & qui demonia habebant c. –Vf–predN thaz sie hiezzin boanerges that they were called Boanerges ‘that they be called Boanerges’ (T 59, 22) lat. boanerges

One may doubt the validity of (9a) by virtue of the fact that the pronominal object follows the selecting verb in the original as well. In the text sections considered in the database, only causal clauses (being potential root conjuncts) provide examples of pronouns in postverbal position contrary to the Latin, cf. (i) in Fn. 9. However, as has been shown in the literature, examples with postverbal pronominal objects, which are no imitations of the original, also appear in other types of dependent clauses. See Dittmer and Dittmer (1998, 148): (i) thiedar giotmotigot sih who-PRT humiliates REFL.PR ‘who humiliates himself’ (T 195, 19) lat. qui se humiliat

198

Svetlana Petrova & Roland Hinterhölzl

Examples with postverbal DPs as complements of copula verbs of the type heizzan ‘be called’ are especially revealing for further analysis, as they are often passivized and thus contain both an inflected auxiliary and an untensed main verb. On the one hand, patterns like –V–Vf–predN as in (10a) give support to the view that the DP has been moved to the right of Vf from a basic OV order. Under this analysis, we have to hypothesise for OHG a type of rightward movement which is not attested in the modern SOV languages. Nevertheless, this explanation does not hold for all cases involving postverbal predicative DPs. See (10b) where the pattern –Vf–V– predN is best analyzed as an instance of a base order in which all complements are serialized to the right of their heads. Any attempt to provide an alternative derivation from a basic OV order involves a series of complicated and unusual movement operations: (10) a. –V–Vf–predN ther giheizan ist p&rus who called is Peter ‘who is called Peter’ (T 54, 15) lat. qui uocatur p&rus b. –Vf–V–predN thiu uuas ginemnit b&hleem which was called Bethlehem ‘which was called Bethlehem’ (T 35, 17) lat. quae uocatur b&hleem

The same considerations hold for predicative adjectives after Vf. Most remarkable are examples like (11) in which the clause-final placement of Vf in the Latin original is suspended in OHG: (11)

daz sie sin blinte that they be.SBJV blind.PL ‘that they be blind’ (T 224, 6) lat. ut […] caeci fiant

Again, the question arises how to explain these data in OHG. Following the diagnostics in (6a), (12) receives an interpretation according to which the sentence has a basic OV order but the predicative adjective has undergone rightward movement, i.e., we have to assume a type of extraposition which does not exist in modern German: (12)

–V–Vf–predAdj thiu thar ginennit ist unberenta who PRT called is barren-F ‘who is said to be barren’ (T 29, 2) lat. quae uocatur sterilis

Evidence for Two Types of Focus Positions in Old High German

199

Problems also arise when we consider clauses with complex verb forms. It is well known that OHG provides evidence for VR and VPR as also attested in OE and some modern dialects of Geman and Dutch, cf. (13a–c): (13) a

thaz ír nisít fortuomte that you-PL NEG-are judged.PL ‘that you be not judged [as well]’ (T 71, 17) lat. ut non iudicemini b. thó tág uúas giuuortan when day was PASS.AUX ‘when daylight came’ (T 107, 23) lat. cum dies factus ess& c. thiethár uuolle mit thír uuehslon who want.SBJV with you.DAT change ‘who ever wants to change place with you’ (T 65, 12) lat. uolenti mutare a té

Basically, three types of analyses have been put forward in explaining this phenomenon. The standard analysis proposed by Haegeman and van Riemdijk (1986) involves rightward movement of the VP. If the arguments have left the VP prior to VP extraposition, then only the main verb occurs after Vf giving rise to –Vf–V order in the surface (VR); if, however, an argument remained in the VP, it is also moved to the right yielding orders like –Vf–XP–V in the surface (VPR). Pintzuk (1993) identifies cases of VPR that cannot be explained by rightward movement of the VP. Assuming that pronouns cannot be affected by VPR in modern Germanic dialects, she argues that these cases must involve leftward movement of Vf to a clause-medial IP (to ȞP according to Fuss and Trips 2002 and Weiß 2006). However, cases with VPR-ed pronouns can be found in West Flemish (cf. Haegeman 1992). The third type of analysis proposed by den Dikken (1994) assumes that orders in which an untensed main verb follows the inflected auxiliary display VO base order (cf. Hinterhölzl 2006). Let us consider the OHG examples from the point of view of these theories. Leftward movement of Vf is supported by example (14) in which Vf is found to the left of the monosyllabic light adverb giu ‘any longer’ modifying it: (14)

só thaz her nimohta giu/ so that he NEG-might already ougazorht gan in thie burg openly go to this town ‘so that he could not openly enter this town any more’ (T 83, 4-5) lat. Ita ut non poss& manifeste / In ciuitatem iam introire

But at the same time, the first two accounts on VR and VPR run into problems concerning the derivation of examples involving arguments of

200

Svetlana Petrova & Roland Hinterhölzl

copular verbs like heizzan ‘be called’. Let us first look at surface orders of the type –Vf–V–predN as in (15): (15)

thes namo uuas giheizzan simeon whose name was called Simeon ‘whose name was called Simeon’ (T 37, 24) lat. cui nomen simeon

At least three interpretations are possible. (16a–b) involve basic OV order with extraposition of predN provided that nominal predicates actually undergo rightward movement in OHG. (16a) shows the traditional analysis of VR/VPR as VP-extraposition. By contrast, (16b) involves leftward movement of Vf to a higher functional head, say Ȟn° (cf. Fuss and Trips 2002). An alternative solution, though, is (16c) which assumes that the attested order is a base order in which all heads (here Vf and V) select their complement to the right (VO base). In this case, Vf is merged in a functional head above VP. This analysis is supported by the fact that predicative nouns, unlike object DPs, do not have to check case features by leftward movement (to AgrP): (16) a. b. c.

thes namo tVP uuas [VP tj giheizzan ] simeonj thes namo uuasj VP[ti giheizzan tj] simeoni CPthes namo uuas [VP giheizzan simeon] CP CP

Previous analyses of VR and VPR also face a problem with surface orders of the type –predN–Vf–V, as given in (17a). Under the analysis involving overt movement of Vf to Ȟ, the proper landing site of Vf would be to the left of the predicative noun unless it has been moved to the left itself. However, there is no proper motivation for this kind of movement of nominal predicates, e.g., they are not subject to scrambling to the left in modern German (cf. Hinterhölzl 2006, 9). The alternative solution is to assume that Vf is moved to a position immediately following a narrowly focussed constituent, as will be advocated below. This type of derivation is illustrated in (17b): (17)

–predN–Vf–V a. In thero st&i thiu [abilina] uuas heizzan in that town which Abilina was called ‘in that town which was called Abilina’ (T 43, 24) lat. abilinae b. thiuFOC[abilinaj] uuasi tj heizzan ti

2.4 Embedded Root Patterns in OHG? In this section, we want to look at possible indications for embedded root patterns in the OHG data.

Evidence for Two Types of Focus Positions in Old High German

201

Let us first look at data for potential embedded V2. There is plenty of evidence for surface orders in which Vf occupies the second position in a dependent clause. However, in the vast majority of the cases, the position before Vf is occupied by the subject of the clause, see (18a). In order to argue for embedded V2, we need to show that the subject obligatorily follows Vf if a non-subject occupies the preverbal position. But orders with non-subjects before Vf are found rarely in our database. We are able to attest 10 instances of the kind, 6 of which are causal clauses allowing for an interpretation as root conjuncts 9 . The remaining cases are given in (18b–e): (18) a. thaz truhtin mihhilosota sína miltida that Lord showed His mercy ‘that the Lord showed His mercy’ (T 30, 19) lat. quia magnificauit dominus misericordiam suam b. só imo gibot truhtines engil as him.DAT.SG commanded Lord.GEN.SG angel ‘as the angel of God commanded to him’ (T 35, 2) lat. Sicut precepit ei angelus domini c. thô thaz gihorta herodes ther cuning when this heard Herod the king ‘When Herod the kind heard this’ lat. audiens autem herodes rex d. thaz gibrieuit uuvrdi al these umbiuuerft that registered PASS.AUX.SBJV all this universe ‘that all the universe be registered’ (T 35, 9) lat. ut describer&ur uniuersus orbis e. oba in sodomu uuarin gitanu / megin if in Sodom were.SBJV done-PL virtues ‘if in Sodom were preformed virtues’ (T 102, 23-24) lat. si in sodomis facte fuissent uirtutes /

However, none of the examples in (18b–e) provides safe evidence for embedded V2. In (18b–c), we have no diagnostics which could allow us to judge whether the verb has been moved to the left or the full subject has been extraposed. (18d) is ambiguous as well: we cannot tell whether the participle is in the prefield of a V2 clause or whether it is in its basic position in the VP, which in this case contains nothing but the lexical verb and the finite auxiliary, with the subject being extraposed to the right. 9

The examples are T 29, 33; T 35, 27-28; T 39, 29; T 59, 10; T 60, 21 and T 204, 24-25. One of them is given in (i): (i) uuanta bithiu bín ih gisentit because therefore am I sent ‘because that’s why I was sent’ (T 59, 10) lat. quia ideo missus sum

202

Svetlana Petrova & Roland Hinterhölzl

Finally, in (18e), movement of Vf to the second position in the clause would yield a pattern in which the subject constituent megin ‘virtues’ is between Vf and the untensed verb gitaniu ‘performed’. Weiß (2006) also argues for the existence of embedded V1 in OHG in his corpus. In our database, we find subordinate clauses in which Vf precedes all remaining constituents, but an interpretation as a root pattern is not forced. First, we find V1 in examples with complex predicates like those in (19a–b). As was discussed in Section 2.3 above, the –Vf–V–XP orders are interpretable as head-initial base orders: (19) a. –Vfin–V–predN thiethar ist giheizan zelotes who-PRT is called Zelotes ‘who is called Zelotes’ (T 59, 27) lat. qui uocatur zelotes b. –Vfin–V–NP–PP thaz uuvrdi arougit gotes that became shown God.GEN.SG uuerc In imo work in him.DAT.SG ‘that God’s work be shown through him’ (T 220, 20) lat. ut manifest&ur opus dei In illo

At the same time, we hardly find examples which contain diagnostics for a leftward movement of Vf to the position before all constitutents of the clause. A unique case is given in (20) where Vf appears to the left of a reflexive pronoun which is normally situated in the so-called Wackernagel position at the left edge of the middlefield, immediately below C°. However, an interpretation as a coordinate conjunct with V1 order is also possible in this case: (20)

uuanta nahit sih / himilo richi because approaches REFL.PR heaven.GEN.PL kingdom ‘because the kingdom of Heaven is approaching’ (T 44, 1-2) lat. adpropinquauit enim / regnum caelorum

3 An Alternative Approach to Word Order Variation in OHG 3.1 Basic Theoretic Assumptions In the previous section, we presented evidence for surface orders in subordinate clauses which cannot be explained as cases of extraposition from a head-final base order, and which cannot be analyzed as embedded root patterns either. Alternatively, we claim that all attested surface orders in

Evidence for Two Types of Focus Positions in Old High German

203

OV languages can be derived from a universal VO base plus licensing movement of arguments and VP-internal predicates (cf. Zwart 1993 for Dutch, Hinterhölzl 2006 for German), as is illustrated in (21a). In this approach, superficial VO orders are derived by i) spell-out of the lower copies of licensing movement out of the VP (cf. Hinterhölzl 2009) or ii) by remnant movement of the VP (VP-intraposition) across the material extracted from the VP before, as is illustrated in (21b): (21) a. [CP that [AgrO XPi [PRTj [VP V ti tj]]]] b. [CP that [VP V ti] [AgrO XPi ] tVP]

In this paper, we will pursue an approach in terms of option ii) above. Adopting a VO base approach, we can account for heavy and light postverbal constituents alike. In the following, we will provide evidence for the fact that these movement operations are motivated by information structure. The main observation in favour of this view is that we can find a tight correlation between the information-structural value of constituents in the clause and their positional realization with respect to Vf. Basically, we shall investigate the realization of the following information-structural categories: i) the position of background, or presupposed information as opposed to the novel, or asserted information in the clause, and ii) the positional distribution of narrow, e.g. contrastive focus with respect to Vf. 3.2 The Position of Background vs. New Information It is commonly assumed that the background domain of the utterance comprises information which is presupposed in the context. This especially applies to expressions which refer to given, but also to accessible information in the discourse. In OHG subordinate clauses, such expressions show a regular tendency to appear very high in the middlefield, in the position immediately following the C-domain of the clause. This property is well-known for pronominal arguments which regularly appear in the Wackernagel position, i.e. in the middlefield immediately below C°. This is borne out by the data in our database as well. Pronouns are inserted or transposed to the Wackernagel position in 317 cases in conjunctional clauses, in 115 cases in relative clauses and 57 cases in causal clauses; see (22): (22) a. thanne Ir Iz find& when you.PL it.ACC.SG find ‘when you have found it’ (T 40, 4) lat. cum Inueneritis

204

Svetlana Petrova & Roland Hinterhölzl

b. unzan ih thir quede until I you.DAT.SG tell ‘until I tell you’ (T 40, 28) lat. usquedum dicam tibi

There are only 16 examples in which a pronominal constituent does not occupy the Wackernagel position. But we are able to provide an explanation for this fact. First, in 9 of the examples, the pronoun is postverbal in the underlying Latin structure as well, see (23a). Second, in 5 cases, a pronoun is moved out of the postverbal domain in the Latin but its proper realization in the Wackernagel position is blocked by the line break, see (23b). Finally, in the remaining two cases, a root interpretation with Vf in C° is also plausible, see (23c); under this interpretation, the pronoun, which follows Vf immediately, can be analyzed as occupying the Wackernagel position: (23) a. thaz sie fiengin inan that they arrest.SBJV Him ‘that they arrested Him’ (T 119, 9) lat. ut raperent eum b. thaz sie mit Iro hanton / that they with their hands thih nemen you take ‘that they hold you with their hand’ (T 50, 10-11) lat. & in manibus / tollent té c. uuanta / samasó uuîzagon hab&un Inan because like [a] prophet had Him.ACC ‘because they regarded Him as a prophet’ (T 204, 24-25) lat. quoniam / sicut proph&am eum habebant

But apart form pronouns, full lexical phrases are also regularly shifted from the postverbal position in the Latin original to the position below C° in OHG, when they are discourse-anaphoric or accessible. Consider the position of the DP uueralt ‘the world’ in (24a) which takes up a referent already established in the preceding main clause. In (24b), the same position is targeted by the DP gote ‘God-DAT.SG’ whose referent is not explicitly mentioned in the previous context but is part of the common ground of the interlocutors. Full DPs providing anaphoric or inferable information are shifted to the preverbal domain in 38 cases in the database. The reverse transposition, i.e. to shift full DPs in anaphoric use from the preverbal position in the Latin to the postverbal position in the OHG, does not occur in the database.

Evidence for Two Types of Focus Positions in Old High German

205

(24) a. nisanta got sínan sun / In uueralt thaz her NEG-sent God his.ACC son to world that He uueralt tuome / uzouh thaz uuerolt si giheilit world judged.SBJV but that world be.SBJV healed thuruh inan through Him ‘God didn’t send His son into the world that the world be condemned but rather that the world be saved though Him’ (T 197, 30-32) lat. non enim missit deus filium suum / In mundum ut l[sic!]udic& mundum/ sed ut salute&ur mundus per ipsum b. after thiu gifulta uuarun taga / […] brahtun sie Inan after filled-PL were days / […] brought they Him.ACC thô In hierusalem / thaz sie Inan gote then to Jerusalem / that they Him God.DAT giantuuvrtitin presented.SBJV ‘after the days [of her purification] they brought Him to Jerusalem to pre sent Him to the Lord’ (T 37, 11-14) lat. postquam Impl&i sunt dies / […] tulerunt illum In hierusalem / ut sister ent eum domino

A parallel situation is described by Kemenade and Los (2006) and Kemenade (2009) for OE. They observe that discourse-linked material in OE regularly appears in a special syntactic domain situated between the subordinating conjunction and an adverbial þa which functions as a discourse partitioner in the clause. A similar function of the OHG adverbial tho, which is the equivalent of OE þa, can be illustrated in (25). Here, tho is placed after the anaphor thaz ‘this’ referring to the event narrated in the previous sentence, whereas the referent of the subject expression, which is not activated in the preceding context, remains in postverbal position: (25)

mit thiu thaz [tho] gisah simon petrus when this PRT saw Simon Peter ‘when Simon Peter saw this’ (T 55, 29) lat. Quod cum uider& simon p&rus

The current version of our database provides only very few examples with tho in this function in OHG. We assume, however, that the role of a discourse partitioner described for þa in OE is taken by Vf in OHG itself: it is Vf that targets exactly the position which separates the given or presupposed information from the rest of the utterance, which corresponds to the focus domain of the clause. This assumption is borne out when we look at the informationstructural status of constituents realized after Vf in subordinate clauses in OHG. Here, we turn the attention to those examples which contain postverbal material independently of the Latin order. We can distinguish two

206

Svetlana Petrova & Roland Hinterhölzl

basic groups of constituents that appear to the right of Vf in subordinate clauses in OHG: i) complements of main verbs and ii) complements of auxiliary or copula verbs. Complements of main verbs appearing before Vf in the Latin structure but after Vf in the OHG clause are rare in the database. We only find 4 examples of the kind. But it is crucial that in all of them, the postverbal XP represents novel information, which is not pre-established or inferable in the context, cf. (26). At the same time, our database provides no examples in which novel material positioned after Vf in the Latin original is shifted to the position before Vf in OHG: (26)

[Inti bráhtun imo / alle ubil habante =‘and they brought to Him all evil people’] Inti thie thár hab&un diuual and those PRT had devil ‘and those who were possessed by the devil’ (T 59, 1) lat. & qui demonia habebant

In the second group of examples, the constituent that appears after Vf is the non-finite part of a complex predicate, e.g. a predicative adjective (27a), a nominal complement of a copula verb (27b) or the untensed main verb (27c), a total of 52 instances: (27) a. soso thie lihhazara sint gitruobte like the hypocrites are sad.PL ‘like the hypocrites with a sad countenance’ (T 68, 23) lat. sicut hypocrite tristes b. fon theru burgi thiu hiez nazar&h from the.DAT town.DAT which was-called Nazareth ‘from the town which was called Nazareth’ (T 35, 16) lat. de ciuitate nazar&h c. nibi ir uuerdet giuuentite / inti gifremite if you.PL PASS.AUX converted-PL and formed.PL ‘if you do not become like this young boy’ (T 151, 12) lat. nisi conuersi fueritis / & efficiamini sicut paruuli

At first sight, the two groups of postverbal constituents shown in (26) and (27) appear to have little in common. However, from the point of view of information structure, these types of constituents share one important property in the clause: they constitute the domain of new-information, or presentational focus, in the utterance. As for the postverbal arguments like in (26), this interpretation results from their novelty in the discourse. By contrast, the postverbal constituents in (27) do not display properties of discourse referents and therefore cannot be analyzed with respect to the given/new distinction. However, it is obvious that together with Vf, they constitute the asserted part of the proposition, as opposed to the domain of background material which is placed before Vf. This supports the as-

Evidence for Two Types of Focus Positions in Old High German

207

sumption that Vf targets a position in the clause which separates the information-structural domains of given vs. new information in the clause. This generalization, however, shall receive a modification with respect to the placement of contrastive information in the clause. 3.3 The Position of Contrastive Information We also looked at the positional realization of constituents which convey contrastive information in the clause, e.g., they form a complementary pair of alternatives with another entity in the discourse, or are used to express selection, correction, or emphasis. The results of the analysis clearly show that contrastive information is associated with a particular syntactic position in the clause, namely with the position which immediately precedes Vf. This position is targeted by 36 of 43 contrastive phrases in the database, i.e. in 84 % of the cases involving contrast on an XP. It is important to note that this positional realization of contrastive information applies to all types of phrases, so e.g. for arguments (28a–b) but also for adjuncts (28c) and modifiers (28d). The phrases in square brackets either form a contrastive pair with another element explicitly mentioned in the context, or are in the scope of a silent focus operator triggering exhaustive reading (‘this XP and nothing/nobody else’): (28) a. niuuizze íz thin uuinistra/ uuaz [thin zesuua] NEG-know it your left hand what your right hand tuo do ‘your left hand should not know what your right hand is doing’ (T 67, 5) lat. quid faciat dextera tua b. thaz thu [mannon] nisís gisehán / that you people.DAT.PL NEG.be.SBJV seen fastenti úzouh thinemo fater fasting but your.DAT father ‘that you do not appear fasting to the people but to your father’ (T 68, 31) lat. ne uidearis hominibus / ieiunans c. [Ír guhortut thaz then alton giqu&an uúas/ nifursuueri thih=‘you heard that it was told to your predecessors: “You should not swear”’] thane ih quidu íu/ thaz mán PRT I tell you.DAT.PL that INDF ] [zi thuruhslahti] nisuuere at all NEG-swear-SBJV ‘But I tell you that you should not swear at all’ (T 64, 17) lat. non iurare omnino

208

Svetlana Petrova & Roland Hinterhölzl

d.

íogiuuelih gommanbarn thaz uuamba [êrist] Intuot each male-child which wamb first opens ‘each male child which opens the wamb’ (T 37, 17) lat. adaperiens uuluam

In 7 cases (16%), the phrase bearing contrastive focus does not appear in the position immediately before Vf. However, word order in these examples either corresponds to the original as shown for the bare plural rehte ‘the righteous man’ in (29a), or the placement of the contrastive XP to the left of Vf is blocked by the line-per-line principle of the translation as shown for the DP thiu himiliscun ‘the heavenly matters’ in (29b): (29) a. niquam zi thiu thaz ih iladoti NEG-came for the reason that I called [rehte] / ouh untige zi riuue righteous.PL but inful.PL to repentence ‘I haven’t come to call the righteous people but the sinners for repentence’ (T 91, 24) lat. Non enim ueni uocare iustos / sed peccatores in poenitentiam b. [oba ih íu erdlihhu quad=‘when I told to you earty things’] oba ih íu quidu / [thiu himiliscun] if I you.DAT.PL tell the heavenly-PL [things] ‘if I tell you the heavenly things’ (T 197, 14-15) lat. si dixero uobis / caelestia

This picture invokes the conclusion that contrastive information is systematically realized in the position which is left-adjacent to Vf in subordinate clauses in OHG. 3.4 Interim Conclusion: the Structure of the Clause in the OHG Tatian The foregoing investigation revealed that we are able to distinguish three different domains in the structure of the clause in OHG which tightly correlate with the information-structural value of the constituents that occupy them. First, it was shown that presupposed material making up the background domain of the utterance is regularly placed in the position that immediately follows the subordinating conjunction or the relative pronoun; i.e., background material is situated in the so called Wackernagel position at the left edge of the middlefield. Second, constituents conveying the asserted part of the utterance are regularly realized after Vf; in this case, Vf marks the left edge of the domain of new-information or presentational focus in the utterance. Third, phrases bearing a contrastive or exhaustive interpretation regularly appear in the position left adjacent to Vf in the clause.

Evidence for Two Types of Focus Positions in Old High German

209

3.5 Comparison with the OHG Isidor The principles of clause organization described for the OHG Tatian may be detected in other texts of the classical OHG period as well. In this section, we will provide a comparison with another main representative of the OHG corpus, the translation of the theological treatise De fide catholica ex veteri et novo testamento contra Iudaeos written in Latin by Bishop Isidorus of Sevilla (560-636). The OHG translation, called the OHG Isidor, is dated back to the time around 800. The purpose of the text is to vindicate the notion of Trinity against the doctrine of Arianism. By numerous references to statements of the Old Testament, the scribe attempts to verify the divine origin of Christ, who, contrary to the assumptions of the Arianists, was not only adopted but created by God the Father. These facts are crucial to the interpretation of the examples with respect to the information-structural value of the constituents involved. In this text, we continuously come across expressions conveying information related to the divine origin of Christ as well as to the fact that he already came to mankind as a human being. As the aim of the text is to confirm the information in these expressions and to exclude any alternatives to it, it can be taken for granted that these expressions are subject to narrow, or contrastive focus in the clause. Interestingly, the material allowing for such an interpretation regularly appears in the position immediately before Vf, see fona fater ‘by the Father’ in (30a), man ‘a human being’ in (30b), and the contrastive pair iu ‘already’ vs. noh ‘still’ in (30c). At the same time, background information follows the subordinating conjunction, see christ gotes sunu in (30a) as well as the personal pronoun ir-3SG in (30b–c) which refers to the person of Christ, i.e. to the activated topic of elaboration in these passages. By contrast, the domain bearing information that is asserted is opened by Vf: (30) a. [Mit so mihhiles hęrduomes urchundin ist nu so offenliihho armarit=‘By so many facts it is evidently said] dhazs christ gotes sunu [...] [fona fater] that Christ God.GEN Son from Father uuard chiboran PASS.AUX born ‘that Christ, the Son of God, was created by the Father’ (I 96) b. dhazs ir [man] uuardh uuordan that He human AUX become ‘that he became a human being’ (I 393)

210

Svetlana Petrova & Roland Hinterhölzl

c. [Souhhemes auur uuir nu ziidh dhera christes chiburdi=‘But let us explore the time of Christ’s birth’] huuedhar ir [iu] quami odho uuir whether He already came.SBJV or we [noh] sculim siin quhemandes biidan still shall His arrival expect.INF ‘whether He already came or we still have to expect His arrival’(I 434)

More cases involving contrast on a single constituent can be added to this pattern. In (31), the scribe rejects the option that a certain prophecy may apply to someone else rather than to Jesus. Again, the expressions that convey the rejected opposite to the referent of Jesus appear left-adjacent to Vf. Background material like the pronouns ir ‘he’ in (31a) and izs ‘it’ in (31b) follow immediately after the conjunction, and Vf opens the domain of asserted information: (31) a. [Umbi dhesan selbum christ chundida almahtic fater dhurah isaian=‘About the same Christ the almighty Father predicted through Isaiah’] dhoh ir [in cyres nemin] quhadi although he in cyres name told.SBJV ‘although he used the name ‘Cyros’ ’ (I 150) b. so huuer so uuanit dhazs izs [in salomoni] whoever thinks that it in Salomo.DAT uuari al arfullit filu aboho firstandit PASS.AUX.SBJV all fulfilled much incorrectly understood ‘Whoever thinks that this became true in the person of Salomo misunder stood many things’ (I 631)

3.6 Theoretical Implications The previous analysis invokes the conclusion that OHG is discourseconfigurational, i.e., that there is a tight correlation between the information-structural value of sentence constituents and their positional realization in the clause. In this section, we want to address the question of how these empirical observations can be integrated into a syntactic model. The clause structure derived for OHG bears parallels to the situation described by Diesing (1997, 390–396) for Yiddish. According to her, the different syntactic realization of object DPs with respect to the selecting main verb triggers three different types of semantic interpretation. As Hinterhölzl (2004, 154) observes, these interpretations correspond to different categories of information structure. Leftward movement of a DP outside the VP is associated with definitness and specificity, i.e. with background, while postverbal (in-situ) placement yields an existential reading of discourse-new indefinite DPs as instances of new-information, or pre-

Evidence for Two Types of Focus Positions in Old High German

211

sentational focus. Additionally, both definite and indefinite objects in the position left-adjacent to the verb gain a special, marked status which is only possible when contrastive or corrective emphasis is put on them, i.e. when they are contrastively focussed. Hinterhölzl (2004) accounts for this distribution by assuming movement of the verb to the head of a Focus phrase FocP which is situated at the left edge of the middlefield above the Case-checking projections for arguments. While contrastive elements move to SpecFoc, new objects remain in the scope of the focus head where they receive Case, see (32): (32)

[C Backgr [FocPContrF V [AgrP PresF

[VP tV]]]]

In our present account, we retain the idea of a focus phrase FocP which is situated between CP and VP in the structure of the clause. One significant difference to the account in (32) though, remains. From the viewpoint of the empirical study, we have to replace V in (32) by Vf; i.e., the position of contrastive focus is left-adjacent not to the main verb but to Vf in the clause. All patterns attested in OHG subordinate clauses can be derived according to the following operations: (33) a. Vf is moved to the head of focus phrase FocP b. SpecFoc is reserved for contrastively focussed information; this explains the left adjacency of contrastive information to Vf observed in the data 10 c. constituents which convey new information remain in the scope of the fo cus phrase, i.e. after Vf d. given and presupposed constituents leave the scope of the focus phrase by movement to a position outside the VP, e.g. to the Wackernagel position below C.

Let us also address the question why these two focus positions were distinguished in the clause stucture of OHG. Petrova (2009) puts forward two suggestions that provide an explanation for this issue. The first one relates to aspects of the prosodic realization of focus especially in cases of multiple foci. In the OHG Tatian, we often find examples in which two different constituents receive focus interpretation in the clause. Consider (34a–b): (34) a. [nimág ther man Iouuiht intphahén/= ‘a man can receive nothing’] noba imo íz gigeban uuerde fon unless him.DAT it given PASS.AUX from himile Heaven.DAT ‘unless it has been given to him from Heaven’ (T 57, 6-7) lat. Non potest homo quicquam accipere.’/ nisi ei fuerit datum a caelo

10 The movement analysis is strongly supported by the observation that contrastive phrases in the narrow scope of negation are obviously above the negative operator ni, see (27b).

212

Svetlana Petrova & Roland Hinterhölzl

b. [thisu sprahih íu/= ‘these things I have spoken to you’] thaz in mir habet sibba / that in me.DAT have peace In therru uueralti habet ir thrucnessi (T 290, 7-9) In the world have you.PL tribulation ‘that in Me you may have peace; in the world, you will have tribulation. lat. ut in me pacem habeatis / In mundo presuram habebitis

In each of these sentences, two different constituents receive a focus interpretation. One of them is involved in a relation of contrast to another entity in the discourse. In (34a), the participle gigeban ‘given’ refers to the only way to obtain spiritual power, namely by being given it, not by acquiring it oneself. In (34b), the PP in me ‘in me’ forms a contrastive pair with the expression In therru uueralti ‘in this world’. Additionally, there is also material supplying new information to the context. In (34a) this is the source of the spiritual power, namely Heaven, and in (34b), new information is conveyed in the direct object sibba ‘peace’ which is also contrasted to the expression thrucnessi ‘pressure’ in the following conjunct. In the Latin version, in both cases the different types of foci are situated on the same side of Vf, after it in (33a) and before it in (33b). In the OHG text, however, the two different focus constituents are placed in such a way that the contrastive or narrowly focused information is immediately before Vf, while new information focus follows it. Similar patterns can be found in the OHG Isidor translation as well. Consider (35) where each of the subordinate conjuncts contains members of two different pairs of contrastive information: i) chihoric ‘obedient’ vs. furiro ‘superiour’, and ii) gote ‘to God’ vs. andrem gotes chiscaftim ‘to the other creatures of God’. In each of the conjuncts, Vf appears exactly between the two focus expressions: (35)

[Got […] setzida innan in siin paradisis= ‘God placed him [Adam] in His paradise’] dhaz ir chihoric uuari gote endi that he obedient was.SBJV God.DAT and furiro uuari andrem gotes chiscaftim superiour was.SBJV other.DAT.PL God.GEN creatures.DAT.PL ‘that he should be obedient to God and superior to the other creatures of God’ (I 488)

These observations invite the assumption that Vf in OHG is used to avoid the stacking of two different types of focus in one and the same structural domain in the sentence. As focus is prototypically associated with main stress, the placement of Vf between two different types of focus was a means of avoiding a clash of two heavily stressed phrases in the clause. This scenario, however, does not account for the regular association of contrastive vs. presentational focus with a special position in the clause.

Evidence for Two Types of Focus Positions in Old High German

213

However, the different positional realization of focus types allows for the unambiguous interpretation of the pragmatic value of the constituents involved. It guarantees that preverbal focus is interpreted as XP-focus only, excluding the option of focus projection. In this way, OHG avoids a phenomenon known as ‘focus ambiguities’ in modern German (as well as in a number of other non-related languages). In modern German, main accent on the rightmost XP in base order yields both VP- or XP-focus while in scrambled order, the rightmost surface constituent receives an unambiguous contrastive interpretation (see Abraham 1992). By contrast, in the system reconstructed for OHG, phrases belonging to the domains of newinformation focus surface in postverbal position while preverbal focus only triggers the option of XP-focus with additional effects of contrast, emphasis, and exclusion of alternatives.

4 Conclusions In this paper, we addressed word order variation in subordinate clauses as one of the most remarkable properties of OHG syntax. We outlined the main properties of the previous models proposed to handle syntactic variation in Early Germanic and applied them to the analysis of data from the OHG Tatian translation. It turned out that we cannot account for the attested patterns by simply reducing variation to unmotivated extraposition operations. In the same line, we argued that problematic patterns cannot be analyzed as root patterns in subordinate clauses inducing variation in OHG because this is not supported by our data. Alternatively, we adopted the assumption that variation in word order is a correlate of information structure, i.e. that the different word order patterns can be derived form a universal VO base by assuming leftward movement of constituents, which is related to the expression of information-structural categories. It was shown that constituents are mapped according to their information-structural contribution into three different fields of the clause: i) presupposed, or background material is realized adjacent to C, ii) new or asserted information appears post-verbally, while iii) narrowly (contrastively) focussed phrases are realized left-adjacent to Vf. To account for this positional distribution of constituents in the clause, we assume leftward movement of Vf to the head of a functional projection FocP which separates the domains of background and presentational focus in the clause, with SpecFoc being reserved for phrases with a contrastive, narrow-focus interpretation.

214

Svetlana Petrova & Roland Hinterhölzl

Acknowledgements The paper is based on a talk given at the workshop “The role of Information Structure in Language Change” at the 29th Annual Meeting of the German Linguistic Society in Siegen 2007. We thank two anonymous reviewers and the audience in Siegen for helpful discussions and comments.

References Abraham, W. (1992): Clausal focus versus discourse rhema in Geman: a programmatic view. Language and cognition 2, 1-18. Altmann, H. (1997): Verbstellungsprobleme bei subordinierten Sätzen in der deutschen Sprache. In: C. Dürscheid, K. H. Ramers and M. Schwarz (eds.), Sprache im Fokus. Festschrift für Heinz Vater zum 65. Geburtstag, 6984Tübingen: Niemeyer. Axel, K. (2007): Studies in Old High German Syntax: left sentence periphery, verb placement and verb second. Amsterdam and Philadelphia: John Benjamins. Axel, K. (in. prep.): Nicht-kanonische Nebensätze im Deutschen: synchrone und diachrone Aspekte. Habilitation Thesis. Behaghel, O. (1932): Deutsche Syntax. Eine geschichtliche Darstellung. Band IV. Wortstellung. Periodenbau. Heidelberg: Carl Winter. Den Dikken, M. (1994): Minimalist Verb Projection Raising. In: J.-W. Zwart and W. Abraham (eds.), Groninger Arbeiten zur Germanistischen Linguistik 37, 71-88. Diesing, M. (1997): Yiddish VP order and the typology of object movement in Germanic. Natural Language and Linguistic Theory 15.2, 369-427. Dittmer, A. (1991): Topologie der Spannsätze / Nebensätze (NS) der althochdeutschen Tatianübersetzung. In: E. Feldbusch, R. Pogarell, and C. Weiß (eds.), Neue Fragen der Linguistik. Akten des 25. Linguistischen Kolloquiums, Paderborn 1990, 183-191. Band 1: Bestand und Entwicklung. Tübingen: Niemeyer. Dittmer, A. and E. Dittmer (1998): Studien zur Wortstellung - Satzgliedstellung in der althochdeutschen Tatianübersetzung. Göttingen: Vandenhoek & Ruprecht. Donhauser, K. (1998): Negationssyntax im Althochdeutschen. Ein sprachhistorisches Rätsel und der Weg zu seiner Lösung. In: K. Donhauser and L. M. Eichinger (eds.), Deutsche Grammatik. Thema in Variationen, 283-298. Heidelberg: Winter.

Evidence for Two Types of Focus Positions in Old High German

215

Ebert, R. P. (1978): Historische Syntax des Deutschen. Stuttgart: Metzeler. Eggers, H. (ed.) (1964): Der althochdeutsche Isidor. Nach der Pariser Handschrift und den Monseer Fragmenten. Tübingen: Niemeyer. Endriss, C. and H.-M. Gärtner (2005): Relativische Verb-Zweit Sätze und Definitheit. In: F.-J. d’Avis (ed.), Proceedings of the Symposium Deutsche Syntax: Empirie und Theorie, 195-220. Göteborg: Göteborger Germanistische Forschungen. Erickson, J. (1997): Some observations on word order in Old Saxon. In: C. Dürscheid, K. H. Ramers, and M. Schwarz (eds.), Sprache im Fokus. Festschrift für Heinz Vater zum 65. Geburtstag. Tübingen: Niemeyer, 95-105. Fleischer, J., R. Hinterhölzl, and M. Solf (2008): Zum Quellenwert des AHDTatian für die Syntaxforschung: Überlegungen auf der Basis von Wortstellungsphänomenen. Zeitschrift für germanistische Linguistik 36, 210-239. Fuß, E. and C. Trips (2002): Variation and change in Old and Middle English on the validity of the Double Base Hypothesis. Journal of Comparative Germanic Linguistics 4, 171-224. Haegeman, L. (1992): Theory and Description in Generative Syntax: A Case Study in West Flemish. Cambridge: Cambridge University Press. Haegeman, L. and H. v. Riemsdijk (1986): Verb Projection Raising, Scope and the Typology of Rules Affecting Verbs. Linguistic Inquiry 17, 417-466. Hinterhölzl, R. (2004): Language Change versus Grammar Change: What diachronic data reveal about the distinction between core grammar and periphery. In: C. Trips and E. Fuß (eds.), Diachronic Clues to Synchronic Grammar, 131-160. Amsterdam and Philadelphia: John Benjamins. Hinterhölzl, R. (2006): Scrambling, Remnant Movement and Restructuring in West Germanic. Oxford: Oxford University Press. Hinterhölzl, R. (2009): A Phase-Based Comparative Approach to Modification and Word Order in Germanic. Syntax 12 (3). Hinterhölzl, R and S. Petrova (2009): From V1 to V2 in Older Germanic. J. Linguistics (2009), doi: 10.1016/j.lingua.2008.10.007. Hróarsdóttir, T. (2000): Word order change in Icelandic: from OV to VO. Amsterdam and Philadelphia: John Benjamins. Kayne, R. (1994): The Antisymmetry of Syntax. Cambrigde, MA / London: MIT Press. Kemenade, A. v. (1987): Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Kemenade, A. v. (2009): Discourse relations and word order change. In: R. Hinterhölzl and S. Petrova (eds.), Information Structure and Language Change: New Approaches to Word Order Variation in Germanic, 91-118. Berlin: Mouton de Gruyter.

216

Svetlana Petrova & Roland Hinterhölzl

Kemenade, A. v. and B. Los (2006): Discourse Adverbs and Clausal Syntax in Old and Middle English. In: A. v. Kemenade and B. Los (eds.), The Handbook of the History of English, 224-248Malden, MA / Oxford: Blackwell. Kroch, A. (1989): Reflexes of grammar in patterns of language change. Journal of Language Variation and Change 1.3, 199-244. Kroch, A. and A. Taylor (2000): Verb-Object Order in Early Middle English. In: S. Pintzuk, G. Tsoulas, and A. Warner (eds.), Diachronic Syntax. Models and Mechanisms, 132-163. Oxford: Oxford University Press. Lenerz, J. (1984): Syntaktischer Wandel und Grammatiktheorie. Eine Untersuchung an Beispielen aus der Sprachgeschichte des Deutschen. Tübingen: Niemeyer. Masser, A. (ed.) (1994): Die lateinisch-althochdeutsche Tatianbilingue Stiftsbibliothek St. Gallen Cod. 56. Göttingen: Vandenhoek & Ruprecht. Masser. A. (1997): Syntaxprobleme im althochdeutschen Tatian. In: Y. Desportes (ed.), Semantik der syntaktischen Beziehungen. Akten des Pariser Kolloquiums zur Erforschung des Althochdeutschen 1994, 123-140Heidelberg: Winter. Petrova, S. (2009): Information structure and word order variation in the Old High German Tatian. In: R. Hinterhölzl and S. Petrova (eds.), Information Structure and Language Change: New Approaches to Word Order Variation in Germanic, 251-279. Berlin and NewYork: Mouton de Gruyter. Petrova, S. and M. Solf (t.a.): Pronominale Wiederaufnahme im älteren Deutsch. Personal- vs. Demonstrativpronomen im Althochdeutschen. In: A. Ziegler and C. Braun (eds.), Historische Textgrammatik und historische Syntax des Deutschen. Berlin: de Gruyter. Pintzuk, S. (1991): Phrase structures in competition: variation and change in Old English word order. Doctoral dissertation. University of Pennsylvania. Pintzuk, S. (1993): Verb Seconding in Old English: Verb movement to Infl. The Linguistic Review 10, 5-35. Pintzuk, S. (1999): Phrase Structures in Competition. Variation and Change in Old English Word Order. New York and London: Garland. Schallert, O. (2006): Hybride OV / VO-Systeme und syntaktischer Wandel zu OV und VO in den germanischen Sprachen. Diplomarbeit. Universität Salzburg. Schlachter, E. (2004): Satzstruktur im Althochdeutschen. Eine Skizze zur Position des Verbs im Isidor-Traktat des 8. Jahrhunderts. In: K. Pittner, R. J. Pittner, and J. C. Schütte (eds.), Beiträge zu Sprache & Sprachen 4. Vorträge der Bochumer Linguistik-Tage, 179-188. München: Lincom. Tomaselli, A. (1995): Cases of Verb Third in Old High German. In: A. Battye and I. Roberts (eds.), Clause Structure and Language Change. New York / Oxford, 345-369. Oxford University Press.

Evidence for Two Types of Focus Positions in Old High German

217

Vikner, S. (1995): Verb Movement and Expletive Subjects in the Germanic Languages. New York / Oxford: Oxford University Press. Weiß, Helmut (2006): Die rechte Peripherie im Althochdeutschen. Zur Verbstellung in dass-Sätzen, Ms. Zwart, J.-W. (1993): Dutch Syntax: a minimalist approach. Doctoral dissertation. University of Groningen.

Index adjacency 112, 117–124, 211f., 215 admonitive 162 adverbials 8, 29, 30, 53, 71f., 91, 120f., 131, 192, 205 affirmation 175 Ancient Greek 2, 51, 164 animacy 38f., 42, 45, 50f., 57, 74, 92, 149 aspect 44, 63, 163, 174 assertive 162 automatisation 8, 118, 136, 138 background 4, 8, 9, 127, 150f., 203, 206, 208f. 210, 213 Badiotto 182 basic OV 9, 55, 195, 197f., 200 Belfast English 177 Bulgarian 163 case 15, 34, 42 , 53, 55, 67, 69, 97, 211 C-domain 162, 165, 171f., 178, 180f., 192, 203 clitic 6, 27, 64f., 68, 73f., 77, 81, 98, 102, 143, 148, 178 clitic doubling 5, 63f., 66f., 69f., 72, 75, 78, 81f., clitic left dislocation 70, 79, clitic pronoun 6, 43, 67, 72f., 75f., 154f., clitic right dislocation 5f., 64, 74, 78, 82 collocation 8, 118, 132, 138, contrast 4, 46, 207, 210, 212f.

declarative 2, 9, 17f., 43f., 90, 93, 98f., 102, 109, 167, 170, 190 definiteness 3, 5, 37, 56f., 68f., 74, 92, 149 deontic 162f., 170 directive 8, 162f., 170f., 174, 180 discourse frame 121f., 135f. discourse function 8, 121, 128f., 132, 134f., 138 discourse organisation 118, 123, 134f., 138 discourse patterns 143, 152, 156 Double Base Hypothesis 9, 190 Early Modern English 174f. English 9, 25, 87, 90f., 96–100, 106, 109f., 144, 147f., 162, 167, 173f. epistemic 162f., 170 Extended Projection Principle 179 extraposition 9, 189f., 195–200, 202, 213 figura etymologica 48 focalization 9, 80, 162, 170, 178, 180, 184 focus – broad 120, 153, focus reading 153 focus – polarity 126f., 130, 132f., 136f. focus – wide 120, 126-130, 132, , 135f., 138, French 147f., 167f., 170, 109 frequency 124f, 134f., 137, 144, 195f.

220

Index

German 9, 53, 103, 111, 144, 147, 164, 165, 167, 172, 174, 180, 182f., 189, 191, 193f., 196, 198, 200, 203, 213, Germanic 2, 5, 8, 53, 87, 90, 95f., 118, 144, 147, 162, 164170, 172, 177–180, 182f., 189, 195, 199, 213 head movement 8f., 161f.,166f., 169, 171, 174f.,177 historical pragmatics 136 Hixkaryana 55f. humanness 37f., 56 illocution 130, 132, 136f. imperative 43f., 50, 53, 81, 161166, 169-177, 180–186 imperative verb movement 9, 161f., 171f., 177 indicative 8, 44,.162f., 170, 174, 176 Indo-European 2–4, 8, 27, 37, 118, 161f., 168 interrogative 9, 27, 46, 167, 169f., 178-182 intonation phrase 153, 155 inversion 7f., 90, 119, 144f., 147, 152, 154, 156 Italian 181 iteration 26, 47, 57 Latin 163f., 166, 168, 192f., 195f.. 198, 204–206, 209 left dislocation 1, 66, 70, 73, 150, 167, 181 left periphery 2, 119, 121, 177f. LF 9, 173, 175–177, 183f. Middle English 96f., 167f., 174f. Middle High German 96, 111, 171 mixed grammars 6f., 87, 109, 104f., 108, 111f., 133f.

modal 91, 125f., 132, 136, 162f., 174f. mood 44, 50, 53, 162-4, 166, 170, 172–174, 176f., 181, 183f. negation 44, 93–95, 102, 154, 175, 189, 211 null operator 167, 170-2, 180– 185 null subject 111, 148, 156 number agreement 163, 170 numeral 47, 49 Old English (OE) 2, 147, 166, 170f., 174, 189f. Old French 2, 145, 147, 167f., 170 Old Germanic 2, 9, 161, 166, 169 Old High German (OHG) 2, 9, 171, 189 Old Portuguese 167, 171 Old Romance 2, 161, 166, 169 Old Spanish 73, 167 parametric change 137 performative 54, 171 permissive 162 person agreement 163f., 172, 183, 185 Porteño Spanish 69, 81 Prague School 56f., 96 preterite present verb 174 pro (phonetically null pronoun) – pro-drop 3, 65, 120, 122 proclisis 156f. prosodic patterns 143, 150 prosodic phrasing 63 prosodic prominence 20, 52, 143, 151f., 154, 157 prosodic word 81 prosody 1, 64, 80–82, 143f., 156f. Proto-Indo-European 37, 168

Index

reanalysis 74, 78f., 82, 143, 148, 157f., 169, 172 Rhaeto-Romance 167, 181f. Romance languages 2, 8f., 145, 147, 161f., 164-169, 172, 178, 180f., 183f. Russian 96, 104, 109, 117, 164 sandhi 152f. Scandinavian languages 6, 144 scrambling 103, 200 Serbo-Croat 7, 117, 123, 136, 138 SOV 5, 8f., 15, 18, 35, 37, 4952, 54f., 57f., 163, 165f., 172, 189f., 195, 197f. Spanish 6, 63-67, 69-73, 79-82, 164f., 167, 172f., specificity 37, 42, 47, 56f., 210 stress – contrastive stress – focus stress – nuclear stress 65, 81, 119-121, 150, 154f., 157, 168, 212 subject – postverbal subject SVO 8f., 15, 18, 28, 37, 39, 49f., 51f.. 54f., 58, 143f., 147, 152, 154-156, 163, 165, 171-174, 178-180 syntactic change 39, 82, 118, 133, 135, 137, 143, 171 thematic role 68 topic – contrastive topics – dislocated topics – extra clausal topics – topic expression 1f., 20, 32, 35, 37-39, 42, 46f., 53, 56, 63, 66f., 69, 72, 79, 82, 105, 121-123, 135, 143f., 146-148, 150-158, 167, 178-180, 183, 192, 209 topicality 6, 37f., 42, 51f., 54, 56f., 63, 69, 74, 78, 82, 119, 121

221

topicalization 6, 70, 74, 79f., 162, 170, 178f. transitional stage 118, 138 unaccusative verb 8, 143, 145, 148f., 151, 158 Universal Base Hypothesis (UBH) 191 unstressed subject pronouns 148 Vedic 5, 37–41, 44–47, 53-58 verb first (V1) 2, 9, 144, 146148, 154-157, 161f., 166f., 169f., 172, 175, 180, 182, 189, 191, 202 verb fronting 2, 150 verb movement 2, 5, 9, 90, 93, 109f., 161f., 171f., 176f. verb raising (VR) – verb projection raising (VPR) 9, 190 verb second (V2) 2, 7, 9, 87, 8993, 96, 98–102, 104, 109-111. 119, 143f., 145-147, 150, 152, 156f., 161ff., 166-172, 178185, 189, 191, 193f., 201 verb third (V3) 144, 146f., 156f., 167, voice 45, 163 VSO 8, 15, 18, 37, 49f., 52-55, 58, 162 V-to-C-movement 161 Wackernagel position 9, 202-204, 208, 211 weak crossover 179 Wh-Criterion 175, 180 word order change 6–8, 87, 112, 117, 133, 142 word order variation 6f., 9, 88– 92, 96f., 100, 189-191, 195, 202, 213 Yiddish 9, 165, 174, 2

Contributors Armendáriz, Rolando Félix († 2007) PhD at Rice University, Huston, Texas. Research interests: Mexican Uto-Aztecan languages, typology. Eide, Kristine Gunn [email protected] PhD in Portuguese Linguistics, University of Oslo Assistant teacher at the University of Oslo, Research interests: Romance linguistics, syntax and pragmatics, syntactic change, corpus linguistics, medieval philology. Ferraresi, Gisella [email protected] PhD in German Linguistics, University of Stuttgart, Assistant professor at the University of Frankfurt. Research interests: German linguistics, diachronic syntax of Old Germanic languages, language change Gabriel, Christoph [email protected] Professor of Spanish and French Linguistics, University of Hamburg; Research interests: prosody, segmental phonology, and syntax of Romance languages, formal linguistics, Optimality Theory Hinterhölzl, Roland [email protected] assistant professor at the department of German language and linguistics, Humboldt University Berlin; project leader of the Research group “The role of information structure in the development of word order regulaties in Germanic” in Collaborate Research Center “Information stucture: The linguistic means of structuring utterances, sentences and texts” Research interests: comparative syntax, syntax-phonology interface, language change and word order variation, pragmatics

224

Contributors

Lühr, Rosemarie [email protected] Professor of Indo-European Studies. University of Jena Research interests: syntax and semantics of Indo-European languages, language change, pragmatics of Indo-European languages. Matić, Dejan [email protected] Ph.D. University of Cologne 2003; research fellow at Max Planck Institute for Psycholinguistics (MPI) Nijmegen Research interests: information structure, word order, agreement, copular clauses, binding; languages of northern Siberia, Balkan languages. Petrova, Svetlana [email protected] Research assistant of the Research group “The role of information structure in the development of word order regulaties in Germanic” in Collaborate Research Center “Information stucture: The linguistic means of structuring utterances, sentences and texts” Research interests: information structure, word order variation and change in Germanic Rinke, Esther Professor of Portuguese Linguistics, University of Frankfurt [email protected] Research interests: comparative romance linguistics, syntactic theory, diachronic syntax. Viti, Carlotta Assistant professor of Latin and Greek Linguistics, University of Zurich [email protected] Research interests: phonology and morphology of Latin, Ancient Greek and Old Indic, Syntax of Old Indo-European languages. Westergaard, Marit [email protected] Professor of Linguistics, CASTL – Center for Advanced Study in Theoretical Linguistics, University of Tromsø. Research interests: First and second language acquisition, diachrony, syntax and information structure, word order. Wratil, Melani [email protected] Dr. phil.; Department of Linguistics, University of Düsseldorf; Research interests: historical linguistics; syntactic theory; Creole languages