Lemmata Linguistica Latina: Volume I Words and Sounds 9783110647587, 9783110645187

Citation preview

Lemmata Linguistica Latina

Lemmata Linguistica Latina Volume I: Words and Sounds Edited by Nigel Holmes, Marijke Ottink, Josine Schrickx and Maria Selig

ISBN 978-3-11-064518-7 e-ISBN (PDF) 978-3-11-064758-7 e-ISBN (EPUB) 978-3-11-064572-9 Library of Congress Control Number: 2019946357 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2019 Walter de Gruyter GmbH, Berlin/Boston Cover image: ‚Zettelarchiv‘ of the Thesaurus Linguae Latinae, Munich Typesetting: Integra Software Services, Pvt. Ltd Printing and binding: CPI books GmbH, Leck www.degruyter.com

Preface The field of Latin Linguistics encompasses phonological, morphological, syntactic, semantic, lexicographical, metrical and pragmatic research. Approaches include the study of texts and (digital) text corpora, comparison of data sets from various stages of the Latin language and from different languages, in-depth analyses of contextual factors and the definition or refinement of parameters and theoretical concepts. In all this variation, the aim is to describe and explain linguistic phenomena, ranging from sounds and words to clauses and complete discourses. Since their start in 1981, the biennial International Colloquia on Latin Linguistics have attracted an ever-growing number of Latinists and stimulated cross-fertilization of the various subfields and research traditions. These meetings have also formed an international community of researchers which continues to welcome new scholars. Recent innovations to further this aim include a pre-conference PhD course and the establishment of the Machteld Bolkestein Award for the best presentation of a young researcher. With this award, which in 2017 was won by Stefan Höfler, the community keeps alive the encouraging, enquiring and stimulating presence of one of the founding members of the International Colloquium on Latin Linguistics. The 19th meeting of this colloquium in 2017, April 24–28, brought together more than 150 researchers at the Thesaurus Linguae Latinae in Munich. The conference was organised jointly by the Thesaurus Linguae Latinae and the University of Amsterdam. More than 120 papers were given in plenary talks, thematic sessions and workshops. A selection of them are presented in two separate volumes, thematically organized around Words and Sounds, and Clause and Discourse. The research published in the present two volumes can be seen as a collection of lemmata (in the Latin sense of ‘topics’) in the heterogeneous field of Latin linguistics. The title Lemmata Linguistica Latina is, of course, also a reference to the herculean lexicographical task taken up by the Thesaurus Linguae Latinae in 1894 to give a full diachronic picture of all Latin words in all their uses. In the first volume, the part Words starts off with lexicological studies, followed by studies on word formation and etymology, chapters on language contact, and studies on ancient linguistic terminology. The volume ends with chapters on phonology and prosody under the heading Sounds. The second volume Clause and Discourse contains studies usually listed under headings as syntax, semantics, pragmatics and text linguistics. The part on the Clause starts with a chapter which illustrates how syntactic research and lexicographical work at the Thesaurus may mutually inform and inspire each other. The following part contains investigations of syntactic classes and constructions of the semantic-syntactic features of specific verbal expressions and https://doi.org/10.1515/9783110647587-201

VI

Preface

of semantic concepts. The part on Discourse also has two main foci: first, the temporal organization of discourse, and second, communicative interaction, more specifically identity construction and politeness phenomena. In 1987, Robert Coleman wrote in his introduction to the Acta of the Fourth International Colloquium on Latin Linguistics that ‘in the study of language there are no sharply drawn frontiers’ (1987: x). He welcomed the integration of adjacent disciplines into Latin Linguistics, and we hope that the present volumes illustrate how this open attitude may yield fruitful results. One of the new directions represented at the conference, but not included in these volumes, is the field of Didactics and Linguistics. The workshop organized by Suzanne Adema during the Colloquium has been a successful initiative, and seven articles on this topic have been published separately, in the Journal of Latin Linguistics 17, 2 (2018), 279–365. Citations of Latin texts use the format and abbreviations of the Thesaurus Linguae Latinae as set out in the Index (1990). A version of this index, significantly updated with the material that has been published since the date of the printed version, is available online at the website of the Thesaurus Linguae Latinae (www.thesaurus.badw.de/hilfsmittel-fuer-benutzer/index.html). The conference, as well as the publication of the ensuing volumes, have been generously sponsored by the Bayerische Akademie der Wissenschaften, the Deutsche Forschungsgemeinschaft, the Amsterdam School of Historical Studies and the Amsterdam Center for Language and Communication of the University of Amsterdam, and the Munich Graduate School for Ancient Studies Distant Worlds. The editors have been greatly helped in their work by the readiness of anonymous reviewers to give their advice on papers and by the invaluable support of Elianne Bruin and Ramon Selles at the University of Amsterdam, who efficiently prepared the manuscripts for publication. Furthermore, we want to thank our publisher, De Gruyter, for the careful publication of these proceedings, and Torben Behm for his critical redaction of the text. Finally, we want to thank the authors of the chapters for their efforts to respect the publication standards and their patience during the editing process. The editors, Nigel Holmes Marijke Ottink Josine Schrickx Maria Selig Munich April 

Lidewij van Gils Caroline Kroon Rodie Risselada Amsterdam

Contents Volume One Preface

V

Words Lexicography and semantics Bruno Rochette Constitutus = καϑεστώς : un aspect du supplétisme du verbe « être » Leonardo Costantini Exploring the semantic complexity of the voces mediae: magus, magicus, and magia 21 Cristina Martín Puente & Matilde Conde Salazar La rivalidad entre praeses y praesidens 36 Laurent Moonens Les emplois des « quantifieurs » quantus, quam multus, quot et de leurs corrélatifs en latin classique 50 Krzysztof Nowak Tempus mutatur: analysing collocations of tempus ‘time’ with distributional semantic models 69 María Carmen Arias Abellán Análisis y causas de la diversidad semántica y lexicológica del adjetivo uiridis 86 Nigel Holmes Authorial error and the dictionary

101

Sophie Van Laer Les verbes de déplacement en latin : préverbation et arguments

110

3

VIII

Contents

Word formation and etymology Laurent Gavoille Auctor adjectif verbal dans la périphrase auctor sum ‘conseiller’ Romain Garnier Secondary forms reanalysed as primary ones: the crētus-class in Latin 149 Benjamín García-Hernández Origen y significado de dēsīderāre: nostalgia y deseo desde Plauto 161 Luca Rigobianco Latin dubenus ‘dominus’: an attempt at etymology Kanehiro Nishimura Fors and fortūna: linguistic and cultural aspects

178

192

Simona Georgescu The world as a yawning gap: new insights into the etymology of Lat. mundus ‘world’ 206 Eleonora Litta & Marco Passarotti (When) inflection needs derivation: a word formation lexicon for Latin 224 Marco Budassi, Eleonora Litta & Marco Passarotti What’s beyond ‘inchoatives’? Derivation types on the basis of -sc- verbs 240 Davide Bertocci & Francesco Pinzin Towards a morpho-syntactic analysis of -ī-scō and -ā-scō verbs 258

135

Contents

Linguistic systems in contact Theodor Georgescu Les mots latins d’origine grecque avec diffusion panromane : comment et pourquoi ? 277 Chiara Fedriani, Maria Napoli & Nadia Rosso Forms and functions of Greek words in Late Latin literary texts: a corpus-based approach 294 Robert Maltby Greek in Donatus’ Terence commentaries

312

Šime Demo & Nathalie Tassotti Verb conjugation selection in macaronic Latin: a corpus-based analysis 329

Ancient discourse on language Cécile Conduché Clausula : un terme technique de la métrique à la morphologie

351

Wolfgang David Cirilo de Melo & Panagiotis Filos Linguistic terminology in Varro 368 Anna Novokhatko The typology of linguistic metaphor in first-century CE Roman thought 384

Sounds Francesca Cotugno The use of initial h- in the writing-tablets from Roman Britain

401

IX

X

Contents

Éloïse Lemay The merger of /ō/ and /ŭ/: a computational take on the inscriptions of Gaul (330–730 CE) 419 Giovanna Marotta & Irene De Felice Patterns of prosodic distribution of Latin long vowels Index

436

453

Volume Two Preface

V

Clause Constructions Marijke Ottink & Josine Schrickx Linguistic theory in daily lexicographical practice: dealing with arguments and satellites in the entries of regnare and nectere in the Thesaurus Linguae Latinae 3 Esperanza Torrego The expression of knowledge in Latin: cognosco, nosco, scio, nescio and ignoro 20 Olga Álvarez Huerta Los verbos latinos timeo y metuo: sintaxis, semántica y pragmática Adriana M. Manfredini Potest + passive infinitives: auxiliary or impersonal verb? Antonio María Martín Rodríguez Ruinam dare : les complexités d’une construction latine à verbe support 79

67

48

Contents

Giovanbattista Galdi & Jasper Vangaever On the use of the ablative of the gerund and the nominative of the present participle in Latin technical literature 96 Concepción Cabrillana Praedicativum and subject complement: a question revisited in light of the Latin verb sto 116 Satoko Hisatsugi Der lateinische Dativ: neue Wege in Transitivität und funktionaler Semantik 134 Daniela Urbanová Between syntax and magic: some peculiarities of nominal syntax in Latin curse tablets 155 Lyliane Sznajder Les complétives en quoniam : étude á partir du latin biblique

174

Semantics Elena Zheltova Animacy in Latin: explaining some peripheral phenomena

199

Andrea Nuti Gallia est omnis divisa in partes tres: Sapir’s typology and different perspectives on totality 219 Elisabetta Magni General extenders in Latin

241

Anna Orlandini & Paolo Poccetti Les parcours sémantiques vers l’adversatif : une approche typologique des langues anciennes 259

XI

XII

Contents

Emilia Calaresu & Silvia Pieroni The diffraction of iam: contextual effects in interpretation

280

Michèle Fruyt Le système latin de la déixis et de l’endophore : l’évolution linguistique chez Sénèque 296

Discourse Tense and discourse Jesús de la Villa On the expression of relative time in Latin narrations

319

Sándor Kiss Progression thématique et types de séquences chez quelques historiens romains 339 Lidewij van Gils & Caroline Kroon Engaging the audience: an intersubjectivity approach to the historic present tense in Latin 351 Stefan Höfler ‘I hereby present the use of the Latin first-person perfect indicative as a performative’ 374

Politeness and identity Luis Unceta Gómez Expressing happiness as a manifestation of positive politeness in Roman comedy 393 Sophie Roesch Impoliteness in Plautus’ comedies

413

Contents

XIII

Federica Iurescia How to assess politeness in response to impoliteness: some examples from Latin comedy 431 Merlijn Breunesse Cicero vs. Mark Antony: identity construction and ingroup/outgroup formation in Philippics One and Three 448 Roman Müller Ego sum Amphitruo: Selbstidentifikation in der römischen Komödie Index

479

464

Words Lexicography and semantics

Bruno Rochette

Constitutus = καϑεστώς : un aspect du supplétisme du verbe « être » Abstract: The past participle constitutus (and also positus) has been considered as a substitute for the present participle of the verb esse (*ens, entis) which is missing in Latin. This opinion has been caused by the fact that in Justinian’s Novellae constitutus is a translation of the Greek καϑεστώς which has an active sense. From this fact, it has been inferred that it could be a particularity of the language of the jurists which appears for the first time in Julianus, Scaevola and Ulpianus. However, this use, which seems not to be attested before Seneca, is not specific to juridical Latin. It can also be found in Imperial Latin writers (pagan and Christian) of Africa such as Apuleius or Cyprian of Carthage and especially in Late Latin. Therefore, this specific feature involves the question of the specificity of juridical language and the influence of Greek on this technical language. The great number of the particularities in juridical Latin, considered by some scholars as grecisms, are in fact specificities of Late Latin and have parallels in authors like Apuleius, Tertullianus, Cyprianus or Ammianus Marcellinus. I study here the uses of constitutus as supposed substitute of the present participle of esse in a broad corpus of texts. To explain this specificity it is not necessary to refer to Greek. As for other phenomena, this is an evolution of the Latin which probably felt the need to fill the gap made by the lack of a present participle of esse. Keywords: Latin linguistics, juridical Latin, Grecism, Africanism, verb “to be”

1 Introduction Dans le livre XVIII de son Ars, Priscien fait remarquer que les Grecs usent du participe présent comme d’un substantif, alors que les Latins n’ont pas cette possibilité. À l’en croire, César (anal. frg. 28 Funaioli = F31 Garcea = PRISC. gramm. III 239, 5–11) aurait proposé, selon le principe de l’analogie, un

Bruno Rochette, Université de Liège https://doi.org/10.1515/9783110647587-001

4

Bruno Rochette

participe *ens, entis, forgé sur le modèle potens (< posse).1 La forme *ens est toutefois restée théorique, puisque César n’a jamais utilisé cette création. Dans la lettre 58 consacrée à la façon de rendre en latin des notions platoniciennes relatives à l’être, Sénèque apparaît résigné devant la difficulté de traduire le τὸ ὄν des philosophes grecs. Il n’a guère qu’une périphrase insatisfaisante à proposer : quod est. Dans un chapitre du livre VIII consacré à la légitimité des néologismes, Quintilien rappelle que deux mots nouveaux ne sont pas admis dans l’usage : ens et essentia – qui est savant et tardif. Aucun écrivain chrétien n’utilisera le participe *ens. Chalcidius, dans sa traduction du Timée (fin du IIIe s.–début du IVe s.), traduit ὄν par existens (52d). Il faudra attendre Boèce, à la fin du Ve s., pour voir ens apparaître comme un mot accepté, mais dans un emploi substantivé.2 Peut-on en conclure que les Latins se sont résignés à ce que leur langue soit privée de participe présent du verbe être ? Tel est sans doute le cas de Virgile, lorsqu’il dit (Aen., 3, 398–399) : Cape dona externa tuorum, o mihi sola mei super Astyanactis imago ‘prends les derniers présents qui te viendront des tiens, toi, seule image qui me reste de mon Astyanax’.3 Super équivaut à qui super-es pour le grec περι-ών.4 À l’instar d’autres langues indo-européennes, le latin a toutefois pratiqué le supplétisme verbal.5 Le participe présent de substo, qui n’apparaît guère que chez le médecin Celse (2, 7, 25 uel nullo dolore substante somnus ereptus [2, 12, 2 C ; 3, 24, 1 ; 6, 10, 3]),6 toujours à l’ablatif, serait un essai pour pallier l’absence de participe présent du verbe être. Les écrivains tardifs et les traducteurs de textes grecs, religieux ou non, ont profité de l’affaiblissement sémantique de certains verbes, comme stare et ses composés (surtout exsto ou exsisto, mais aussi consisto), pour suppléer à certaines déficiences du latin, notamment à l’absence de participe du verbe être.7 Ces verbes ont formé, avec esse, les verbes d’existence au sens plein, mais ont aussi fonctionné

1 La forme, sans doute récente, *sens, n’est attestée que dans quelques composés comme absens, prae-sens . . ., à défaut de *essens, qui n’existe pas, mais que saint Augustin envisage comme forme théorique (AVG. loc. hept. 3, 32). Voir Ernout (1954 : 152 et n. 3). 2 BOETH. in Porph. comm. sec. 3, 7. 3 Sauf indication contraire, les traductions sont celles de la CUF. 4 Leumann (1977 : 523). Les grammairiens ne semblent pas embarrassés par l’absence de forme équivalente au ὤν ὄντος du grec. Voir EXC. Bob. gramm. I 534, 39–535, 2 (= p. 4, 12–13 De Nonno ; cf. DOSITH. gramm. 19 p. 33 Tolkiehn = VII 394 Keil, app. cr.) quarto cum, ut † Asprus rettulit, Latinum eloquium in quodam verbo deficit, velut in illo ὄντος οὔσης ὄντων οὐσῶν. Schöpsdau (1992 : 131). 5 Ernout (1954 : 152–153), Ernout (1957), Brachet (1998). 6 OLD s.v. substo (2). 7 Brachet (1998).

Constitutus = καϑεστώς

5

comme copule. Dans certains cas, un participe parfait passif bien présent en latin peut avoir permis de combler la lacune. Il s’agit de constitutus (parfois positus),8 qui pourrait revêtir un sens proche du participe présent actif ‘étant, se trouvant’.9 Comment ce sens s’est-il imposé ? Je propose de réexaminer le dossier, qui semble ne pas avoir été rouvert depuis l’étude de Schrijnen et Mohrmann (1936–1937 : I, 6–11), en commençant par s’intéresser à l’évolution sémantique du verbe constituere.

2 Le verbe constituere : évolution sémantique Les premières attestations du verbe (on en compte cinq) se trouvent chez Plaute. L’occurrence probablement la plus ancienne apparaît dans un vers du Mil. (917), où le verbe a le sens de ‘fixer matériellement’ ou ‘placer les fondements’. Bien que le texte soit altéré, le sens du verbe, qui est employé avec fundare, ne fait toutefois aucun doute. Dans les autres passages plautiniens, constituere n’a plus le sens matériel ‘fixer ensemble’, mais celui de ‘décider ensemble’.10 Le sens du verbe évolue chez Térence, qui l’emploie sept fois. Le champ sémantique s’étend : le verbe revêt presque exclusivement le sens de ‘décider à travers un accord’ et ‘établir’ (Hec. 195 ; Andr. 269). L’évolution se poursuit à la fin de la période républicaine, comme on peut le voir dans l’œuvre de Cicéron, qui fait un usage très important du verbe (720 occurrences). Si le sens ‘établir ensemble’ est toujours attesté (constituimus inter nos ; constitui cum hominibus), le verbe prend d’autres sens, plus larges et plus spécialisés, en particulier une acception juridique ‘fixer la peine’ (poenam constituere), ‘fixer le prix’ (pretium constituere). Avec l’évolution du sens du verbe, le suffixe cum, qui mettait l’accent sur l’idée d’union, ne semble plus avoir de fonction particulière. Nous arrivons à un sens « classique », qui est ‘établir’ avec un accent placé sur le sujet (senatus, maiores). Nous pouvons appeler ce sens le « sens fort » de constituere. À la fin de l’époque républicaine, constituere a atteint son champ sémantique maximal. À partir de l’époque impériale, le processus

8 TLL vol. X, 1, p. 2633, 16–17 : quod haud raro vicem agit part. praes. verbi q.e. esse (v. e.g. p. 2640, 3 sqq. 2664, 19 sqq.). 9 La plus grande partie des exemples viennent des ouvrages suivants : TLL (vol. IV, p. 468, 72–82 ; p. 523, 45–524, 21 [Gudeman 1906–1909]) ; OLD s.v. constitutus (2) ; Svennung (1935 : 647) ; Giodice-Sabbatelli (1981). 10 Pseud. 549.

6

Bruno Rochette

d’amplification progressive des sens du verbe semble terminé. Dans la prose impériale, constituere conserve le sens de ‘décider à travers un accord’. C’est chez Frontin (aq. 99, 3–4) que l’on trouve le premier exemple de constituere avec pour sujet l’empereur. Le verbe exprime donc une décision forte.11 Il est parfois utilisé avec d’autres verbes de volonté, comme imperare.12 Chez les juristes de la fin de la République et de l’époque augustéenne, le sens de ‘décider’ ou ‘établir ensemble’ ou ‘établir’ reste essentiel. Si nous prenons comme référence le manuel de Gaius, où constituere apparaît 57 fois, la polysémie du verbe apparaît de façon évidente. Il semble que, parallèlement à ce processus qui conduit constituere vers un sens prégnant (‘décider’), se soit développé un sens faible, qui découle de la signification ancienne du verbe. Les écrivains de l’Empire ont peut-être repris le sens ancien du mot, lequel avait disparu de la langue littéraire de l’époque classique. Le verbe constituere est dans ce cas plutôt pris dans son sens étymologique, avec comme élément central le verbe stare dans un sens qui le distingue peu de esse. Ce sens n’est pas sans rappeler le verbe grec ἵστημι dans sa valeur intransitive.13 Les langues romanes ont utilisé certains formes de stare (a.fr. ester) pour constituer la conjugaison du verbe être. Ainsi ce verbe a fourni aux langues romanes le participe présent (*estant), un imparfait stabam, qui a supplanté eram, et la forme io sono stato de l’italien. Enfin, l’infinitif « être » vient d’une forme essere, une sorte de fusion entre esse et la forme de l’ancien français ester, qui vient de stare. Un lien fort lie donc esse et stare.

3 Les juristes Le Vocabularium Iurisprudentiae Romanae de Gradenwitz, Kuebler et Schulze (1903 : I, 952) enregiste l’emploi spécifique de constitutus dans le sens de ‘étant’, que W. Kalb (18882 [1961] : 97 ; 1890 [1975] : 98 ; 1912 [1961] : 126) considère comme un hellénisme. Ce serait la traduction du grec καϑεστώς, comme on le voit dans les Novellae de Justinien, où ce participe apparaît de façon récurrente. Cet usage se trouve pour la première fois chez l’Africain Salvius Julianus (IIe s.) et aurait été diffusé plus largement par Q. Cervidius

11 Ce sens se trouve aussi chez Sénèque (Med. 444 ; Phaedr. 265–266) et chez Tacite (ann. 14, 3, 1 ; avec un infinitif). 12 CAES. Gall. 2, 44, 3. 13 Comme on peut peut-être le voir dans une traduction (mot-à-mot) de la grammaire de Dosithée (gramm. 1, 2 = VII 376, 6 Keil) : constat partibus quattuor est traduit par συνέστηκεν μέρεσιν τέτρασιν.

Constitutus = καϑεστώς

7

Scaevola, sous Marc Aurèle, et ses élèves, comme Aemilius Papinianus, actif sous Septime Sévère. L’usage de positus est identique. W. Kalb (1890 [1975] : 98) estime que cet emploi de constitutus est spécifique aux juristes, puisqu’il ne cite qu’une seule occurrence en dehors de la prose juridique.14 La présence d’hellénismes dans la langue des juristes romains a été fortement exagérée.15 Aujourd’hui, plusieurs chercheurs expliquent autrement que par le grec certaines particularités de la langue juridique considérées comme des influences du grec.16 Même si la présence d’hellénismes dans un texte latin n’est pas nécessairement liée à une origine grecque de son auteur, la biographie de ces juristes est mal connue. Certains seraient originaires du monde grec, mais rien n’est certain.17 Seul Domitius Ulpianus cite Tyr comme sa ville d’origine.18 La même objection vaut pour les africanismes. Q. Cervidius Scaevola a été considéré tantôt comme un Grec, tantôt comme un Africain à cause de prétendus hellénismes ou africanismes que certains chercheurs croyaient découvrir dans sa langue. En réalité, les deux hypothèses sont vaines, d’autant plus que les digesta et les responsa ne sont pas des ouvrages originaux du juriste, mais des résumés postclassiques venant d’une publication posthume de ses avis.19 Le même problème se pose pour Aemilius Papinianus, qui ne serait pas un Romain de naissance, mais originaire de province, de Syrie ou d’Afrique. À propos de l’origine de Julius Paulus, nous n’avons aucun élément certain, même si la plupart des spécialistes ne le tiennent en général pas comme étant d’origine grecque ou orientale.20 Voici les exemples les plus significatifs (1)

In aegritudine constitutus. (EPIT. Gai 1, 2, 3) ‘Se trouvant dans la maladie.’21

(2)

Romae constitutus. (IVLIAN. dig. 46, 1, 16, 2) ‘Se trouvant à Rome.’

14 Voir infra (24). 15 Löfstedt (1933 : 431–432). 16 Déjà Kunkel (2001 [1967²] : 205–213) réfutait la présence de prétendus hellénismes dans la langue de Gaius expliquant ces constructions par l’évolution du latin. Nelson (1981 413–417) a confirmé cette vue. 17 Kunkel (2001 [1967²] : 227 n. 476, 259 n. 552). 18 Kunkel (2001 [1967²] : 247). 19 Kunkel (2001 [1967²] : 217). 20 Kunkel (2001 [1967²] : 245 n. 507). 21 Traductions personnelles.

8

Bruno Rochette

(3) Res extra dotem constitutas. (SCAEV. dig. 35, 2, 95) ‘Des biens se trouvant en dehors de la dot.’ In discrimine vitae constitutum. (SCAEV. dig. 32, 39, 1) ‘Se trouvant à un moment critique de sa vie.’22 (4) Impubere filia constituta. (PAPIN. dig. 35, 1, 77, 3) ‘Sa fille étant impubère.’ In extremis vitae constitutus. (PAPIN. dig. 39, 6, 42, 1) ‘Étant à la fin de sa vie.’ (5) Apud hostes constitutus. (VLP. dig. 3, 5, 19) ‘Se trouvant chez les ennemis.’ In saevissimo furore muliere constituta. (VLP. dig. 24, 3, 22, 8) ‘La femme se trouvant dans une folie très violente.’ In spe constitutus successionis. (VLP. dig. 25, 4, 1, 13) ‘Se trouvant dans l’espoir d’une succession.’ (6) In bona valetudine posito. (PAVL. dig. 39, 6, 35, 4) ‘Se trouvant en bonne santé.’ On ne peut exclure que la prédilection des juristes pour le participe constitutus vienne du fait que constitutum est un terme technique de la langue juridique,23 employé pour la première fois par Varron (rust. 2, 5, 1 qui tam sero venisset ad constitutum c’est-à-dire ‘à l’heure convenue’). On trouve peut-être ce sens chez Pétrone dans un récit d’Encolpe (15, 5 ceterum apparebat nihil aliud quaeri nisi ut . . . nos metu criminis non veniremus ad constitutum). On peut sans doute sous-entendre diem, comme en 57, 5 (langue vulgaire) constitutum habui numquam, mais constitutum est bien un terme de la langue du droit, comme on le voit chez Sénèque (dial. 10, 16, 3 cum . . . exspectatur constitutum).24 Il est courant de trouver chez les juristes le verbe constituere dans le sens de ‘établir’.25 La forme passive constitutum est est banale dans la langue juridique de l’époque impériale.26 Le souci de précision de la langue

22 Voir aussi SCAEV. dig. 4, 4, 39, 1 curatoribus eius idoneis constitutis. 23 TLL vol. IV, p. 522, 35 et p. 524, 47–60. 24 TLL vol. IV, p. 524, 47–60 et p. 525, 7–13 et Petersmann (1977 : 121). Le substantif constitutum a en outre une acception technique dans la langue juridique, à savoir ‘promesse de payer (sans stipulation)’. 25 Voir supra point 2. Voir TLL vol. IV, p. 522, 35–68. 26 Klingenberg (1999 : 283–314).

Constitutus = καϑεστώς

9

juridique a peut-être aussi conduit les juristes à ne laisser subsister aucune ambiguïté dans leurs propos et à utiliser ce même verbe dans un sens plus faible, à savoir ‘se trouver, être’. Ce passage d’une signification forte vers un sens affaibli a sans doute été favorisé par le fait que constituere a un emploi au passif qui le rapproche de fieri. Dans les exemples rassemblés, on remarque que constitutus s’emploie régulièrement avec un complément prépositionnel placé avant, le plus souvent introduit par la préposition in (mais aussi extra, apud) et généralement constitué d’un terme abstrait désignant un état ou une circonstance. On trouve un cas (voir 2) où constitutus est accompagné d’un locatif (Romae). Enfin, constitutus s’emploie aussi dans un ablatif absolu avec un complément attribut (voir 4), ce qui montre que le participe peut fonctionner comme prédicat.27

4 Les auteurs païens Le participe constitutus se rencontre dans la prose impériale de Sénèque le Jeune à Martianus Capella.28 (7) Et natura partes suas velut in ponderibus constitutas exanimat, ne portionum aequitate turbata mundus praeponderet. (SEN. nat. 3, 10, 3) ‘La nature place les parties dont elle est formée comme dans une balance et les contrôle pour empêcher que les justes proportions ne soient troublées et que le monde ne perde son équilibre.’29 (8) Tunc nos in ancipiti periculo constituti vel opprimendi nostri vel deserendi socii remedium e re nata validum eo volente comminiscimus. (APVL. met. 4, 11, 1) ‘Dans la périlleuse alternative ou de nous perdre ou d’abandonner notre compagnon, et pressés dans les circonstances, nous nous avisons, avec l’assentiment de notre chef, d’un remède énergique.’

27 Enfin, l’exemple (6) montre que positus peut aussi avoir le même sens que constitutus, mais son emploi semble moins fréquent. Positus revêt certainement un sens proche du grec κεῖμαι. 28 On peut ajouter : FRONTO p. 28, 8 ; DECL. in Catil. (IIe/IIIe s.) 50 ; SOL. 38, 2 ; PS. APVL. Ascl. 32. Voir TLL vol. IV, p. 523, 45–524, 21. 29 Trad. légèrement modifiée.

10

Bruno Rochette

(9) Nec ab illa tamen paterna gloria vel mea uirtute descivi, quanquam semitrepidus iuxta mucrones Martios constitutus, sed habitus alieni fallacia tectus villas seu castella solus adgrediens viaticulum mihi conrasi. (APVL. met. 7, 8, 2) ‘Je n’ai démenti néanmoins ni la gloire paternelle ni ma propre valeur à demi rassuré seulement au voisinage des glaives belliqueux, j’ai, sous le fallacieux couvert de mon déguisement, attaqué tout seul des métairies ou des bourgs et ramassé ainsi une petite réserve pour la route.’ (10) Ac dum ludicris scaenicorum choreis primitiae spectaculi dedicantur, tantisper ante portam constitutus pabulum laetissimi graminis . . . (APVL. met. 10, 29, 3) ‘Pendant le prélude de la représentation, consacré à des ballets dansés par des professionnelles, je prenais plaisir, momentanément arrêté devant la porte, à allonger le cou vers le riche gazon . . .’ (11) Quippe cum transactis vitae temporibus iam in ipso finitae lucis limine constitutos, quis tamen tuto possint magna religionis committi silentia. (APVL. met. 11, 21, 7) ‘Les mortels qui, parvenus au terme de l’existence, foulent le seuil où finit la lumière, et à la condition que l’on puisse leur confier sans crainte les augustes secrets de la religion.’ (12) Haec vero proclivius, si ratio in regni sui solio constituta adpetitus e[s]t iracundias semper [in]domitas et in frenis habet ipsaeque ita oboediunt, ut tranquillo ministerio fungantur. (APVL. Plat. 2, 5 p. 227). ‘Avantages d’accès plus facile lorsque la raison, fermement établie sur le trône de sa royauté, a su dompter et tient toujours en bride les appétits et l’irascibilité et si ceux-ci de leur côté lui obéissent en s’acquittant tranquillement de leur service.’ (13) In summo monte constitutum. (PORPH. Hor. Carm. 3, 4, 14)30 ‘Placée au sommet de la montagne.’ (14) Unum cum Hortensio, alterum in urbe constitutum de pecuniis repetundis, tertium Siciliae praetorem. (PS. ASCON. Verr. p. 214, 25–27).

30 Landgraf (1896 : 560) donne d’autres exemples d’emplois équivalents dans le commentaire de Porphyrion.

Constitutus = καϑεστώς

11

‘L’un consul avec Hortensius, l’autre se trouve en ville pour un procès de concussion, le troisième préteur en Sicile.’ (15) Constituti in terra, spiritus cuiusdam interni motu suspensa, rerumque magnitudini instar exigui subdita puncti, nunc caelo infixas suspicere stellas, quarum ordo est sempiternus, aliquitiens, humana visione languente, discedere suis sedibus arbitrari. (AMM. 20, 3, 12) ‘Nous trouvant sur la terre, suspendue par l’impulsion d’une sorte de souffle intérieur et située comme un point minuscule sous l’immensité de l’univers, tantôt nous voyons fixées au-dessus de nous dans le ciel les étoiles dont l’ordre est éternel, et parfois, par une défaillance de la vue humaine, nous croyons qu’elles quittent leur séjour.’ (16) Apud Savariam eodem adhuc constituto, bubo culminibus regii lavacri insidens occentansque funebria, nulla iacentium sagittas et lapides contemplabili dextera cadere potuit, certatim licet ardenti studio petebatur. (AMM. 30, 5, 16) ‘Et à Savaria, quand Valentinien y était encore installé, un hibou, perché sur le toit des thermes impériaux, poussant des cris funèbres, sans qu’on pût le faire tomber, ni par les flèches ni par les pierres décochées par des mains adroites, bien qu’on rivalisât d’un zèle ardent pour l’atteindre.’ Le participe constitutus est presque toujours employé, comme chez les juristes, avec un complément prépositionnel. La préposition la plus fréquente est in et la précision apportée a trait au lieu où l’on se trouve (concret) ou à la situation dans laquelle on se trouve (abstrait). Seules les scholies du Pseudo-Asconius (voir 14) utilisent constitutus dans un ablatif absolu avec l’adverbe ita, une formule figée qui semble bien être un hellénisme (div. in Caec. p. 185, 16 ; p. 186, 9 ; Verr. p. 236, 7 ; p. 247, 16 his ita constitutis = τούτων οὕτως ἐχόντων).31 Le participe constitutus semble fréquent dans la prose technique du commentateur Porphyrion (voir 13). L’exemple de Sénèque (voir 7) pose question.32 Dans ce passage, Sénèque parle de la transmutation. Il développe l’idée qu’il existe entre les éléments des retours réciproques ce que l’un a perdu, l’autre le gagne. Le philosophe emploie une image (velut),33 celle de la balance dans laquelle on place les parties de la nature. L’affaiblissement sémantique de constitutus n’est probablement pas encore entièrement réalisé.

31 Stangl (1909 : 85), qui reconstitute l’équivalent grec. 32 Skahill (1934 : 175) « Sen. nat. 3, 10, 3, seems to be the earliest certain example ». Voir aussi Svennung (1935 : 647). 33 Voir l’exemple (24) de Lactance.

12

Bruno Rochette

Le nombre d’occurrences de constitutus dans le sens qui nous occupe chez Apulée est proportionnellement important. Louis Callebat (1968 156) signale cet usage comme un trait de la langue populaire.34 Il faut toutefois se demander si tous ces emplois de constitutus ont la même valeur. Dans l’exemple (8), qui décrit la situation tendue dans laquelle sont placés les voleurs à la suite de l’action de Chryseos, la valeur de constitutus est assez claire il s’agit bien d’un sens affaibli. Le participe a mis le traducteur de la CUF dans l’embarras, puisqu’il ajoute un ‘pressés par les circonstances’ qui paraît inutile. La traduction des Groningen commentaries on Apuleius (GCA) est plus satisfaisante ‘At that point, faced with the dilemma of either being caught or deserting a comrade, we devised . . . ’. Dans le passage du livre VII (voir 9), constitutus a un complément un peu inattendu qui ne laisse planer aucun doute sur le sens du participe, rendu en anglais par (GCA) ‘so close to those martial swords’. La phrase du livre XI (voir 11) est fort intéressante. Dans ce passage, qui décrit Lucius aux portes des enfers, il faut probablement tenir compte d’un sens concret de constituere ‘placer quelqu’un quelque part’ que nous trouvons ailleurs dans ce même livre (11, 23, 2 (me) ante ipsa deae vestigia constituit et 11, 24, 4 (me) in vicem simulacri constituto). Il faut aussi prendre en considération le sens de limen, à la fois ‘seuil des enfers’ (concret) et ‘seuil de la mort’ (imagé). Apulée pourrait jouer sur deux sens de constituere le sens fort ‘placer, établir’ et le sens faible ‘se trouver, être’. Les hommes se trouvent à la fin de leur vie au seuil de la mort (« sens faible »), mais sont aussi placés (« sens fort ») par la déesse devant l’entrée des enfers où règne Isis/Proserpine. Dans l’exemple (10), il est possible aussi d’hésiter entre un sens fort, comme le fait le traducteur (‘arrêté devant la porte’) et un sens faible (‘se trouvant devant la porte’). Le traducteur du De Platone (voir 12) a donné un sens fort à constitutus. On peut se demander si ce choix est justifié. Le chapitre concerne la vertu selon Platon. Nous sommes dans l’abstraction, mais aussi dans le langage imagé. La raison est placée sur le trône de son royaume. La même question vaut pour l’extrait d’Ammien Marcellin (voir 15) appartenant à un excursus sur les éclipses de soleil et sur les diverses phases et formes de la lune. À la fin de ce passage, il est question de ce que voient les hommes qui se trouvent sur la terre. Il n’y a aucune raison de donner ici un sens fort à constituti. Le participe donne simplement une indication de lieu sans plus, comme dans l’exemple suivant (voir 16).35

34 Constitutus, employé avec un complément prépositionnel qui le précède, est signalé par les Groningen Commentaries on Apuleius comme « a substitute for the present participle of esse ». 35 C’est probablement aussi le cas dans un passage de Martianus Capella 6, 621 Willis.

Constitutus = καϑεστώς

13

5 Les auteurs chrétiens Le participe constitutus (parfois positus, voir 18 et 23) est très souvent employé par les auteurs chrétiens. Plus on avance dans le temps plus les emplois deviennent fréquents, à tel point qu’il faut renoncer à vouloir établir une liste exhaustive des occurrences. La tradition apologétique de langue latine, essentiellement africaine, commence avec Tertullien (voir 17) vers 197, est reprise par Minucius Félix (voir 19) et Cyprien de Carthage (voir 20), l’un et l’autre lecteurs de Tertullien, se poursuit avec Arnobe (voir 21) et Lactance (voir 24), vers 300–310, et s’achève, un siècle plus tard, avec Augustin (voir 25), Sidoine (voir 26) et Orose.36 Nous trouvons aussi deux exemples dans le De errore religionum profanarum de Firmicus Maternus (voir 22), de Syracuse, composé entre 343 et 350. Nous pouvons suivre les emplois de constitutus selon cette ligne, qui se prolonge jusqu’à Cassiodore,37 où constitutus (et positus) apparaît régulièrement dans les titres et documents des Variae,38 et même jusqu’à Bède le Vénérable.39 (17) Quia contingentium veritatem illam arcanam et supernam et apud pleroma constitutam, haereticarum idearum sacramenta. (TERT. anim. 18, 4) ‘Parce qu’elles découvrent cette vérité mystérieuse, supérieure à notre monde, et placée dans le Plérôme, sacrement des idées hérétiques.’40 (18) Ut ille recogitabis eo tempore de deo, positus illic ubi nihil de deo ? (TERT. spect. 25, 1) ‘Cet homme, je le demande, songera-t-il en même temps à Dieu : avec une majuscule, installé là où il n’y a rien de Dieu ?’41 (19) Ipsa praecipue formae nostrae pulchritudo Deum fatetur artificem status rigidus, vultus erectus, oculi in summo velut in specula constituti et omnes ceteri sensus velut in arce compositi. (MIN. FEL. 17, 11)42

36 OROS. hist. 1, 2, 105 ; 3, 23, 11 ; 7, 15, 12. Svennung (1922 : 141). 37 TLL vol. IV, p. 524, 15–16. 38 CASSIOD. var. 3, 45, 1 ; 4, 35, 2. Voir Skahill (1934 : 174–175). 39 BEDA hist. 4, 22. Voir Druhan (1938 : 142). 40 Trad. de Genoude (1852). 41 Trad. de Turcan (1986). 42 Wölfflin (1892). Voir aussi 32, 7.

14

Bruno Rochette

‘C’est surtout la beauté même de notre forme qui manifeste un dieu artisan : la posture verticale, le visage levé, les yeux placés au sommet du corps comme sur un observatoire et tous les autres organes des sens rassemblés comme dans une citadelle.’ (20) . . . Feminas quae vobiscum sunt in eadem confessionis gloria constitutae. (CYPR. epist. 6, 3, 1)43 ‘. . . De bienheureuses femmes, qui partagent la gloire de votre confession.’ (21) Antequam sunt irati et perciti an cum fuerint moti ipsisque in indignatibus constituti ? (ARNOB. nat. 7, 6, 2) ‘Avant que les dieux soient courroucés et irrités ou lorsqu’ils se sont mis en colère et sont au beau milieu de leur indignation ?’44 (22) Solem etiam Apollonem dicunt, quia sol cottidie in occasu constitutus splendorem luminis perdat. (FIRM. err. 17, 3)45 ‘Certains dénomment aussi le soleil ‘Apollon’, avec l’idée qu’il ‘perd’ chaque jour au coucher l’éclat de la lumière.’ (23) In primis librorum partibus, id est in ipsis auspiciis positus, dixit. (FIRM. err. 13, 4) ‘Dans les premières pages du livre (c’est-à-dire dans l’introduction même) il a déclaré.’ (24) Iuvat igitur velut in aliqua sublimi specula constitutum, unde universi exaudire possint, Persianum illud proclamare o curvae in terris animae et caelestium inanes ! (LACT. inst. 2, 2, 18) ‘Il plaît donc de proclamer, comme placés dans un lieu d’observation céleste, de manière que tout le monde puisse nous entendre, ce vers de Perse : ‘Pourquoi les hommes rampent-ils toujours sur la terre, et pourquoi s’éloignent-ils si fort du ciel’.’46

43 Goetz (1896). Un grand nombre d’occurrences sont relevées par Schrijnen et Mohrmann (1936–1937 : I, 6–9). 44 Voir aussi 7, 44 : Aesculapius . . . Tiberina in insula constitutus. 45 Pastorino (1956 : 152–153). Voir aussi 13, 6. 46 Trad. personnelle.

Constitutus = καϑεστώς

15

(25) Iam vero in Africa constitutus, scripsi duos libros De genesi contra Manichaeos. (AVG. retract. 1, 10, 1) ‘Une fois établi en Afrique, j’ai écrit deux livres Sur la Genèse contre les Manichéens.’47 (26) Postremo iste est ille, carissimi, cui in tenebris ergastularibus constituto multipliciter obseruata barbarici carceris divinitus claustra patuerunt. (SIDON. epist. 7, 9, 20)48 ‘C’est lui enfin, mes très chers frères, qui, plongé dans les ténèbres de la prison, vit s’ouvrir devant lui, par la volonté du ciel, les portes du cachot barbare verrouillées de mille manières.’ Nous retrouvons le même schéma que chez les auteurs païens, c’est-à-dire constitutus accompagné (le plus souvent précédé) d’un complément prépositionnel indiquant le lieu. L’exemple du De anima de Tertullien (voir 17) est intéressant parce que constitutus y est employé avec un terme abstrait (veritas)49 et a comme complément un vocable technique (apud pleroma), la plénitude, mot de la langue théologique paulinienne. Un tel emploi donne la preuve que constitutus n’a pas son sens fort et qu’il ne peut guère se traduire que par ‘étant’. Il en va de même pour le passage de Minucius Felix (voir 19). Je ne traduirais pas, comme Beaujeu, ‘établi au ciel’, mais ‘résidant au ciel’. Au IIIe s., Arnobe fait un usage important de constitutus dans le sens de ὤν.50 Il utilise même le néologisme essentia avec constitutus dans un passage relatif à la notion de réalité où l’on trouve aussi le terme substantia.51 Essentia, qui n’apparaît ni chez Minucius Félix, ni chez Tertullien,52 ni chez Lactance, ni chez Cyprien, désigne ici non le fait d’être (comme chez Quintilien), mais la nature de ce qui est.

47 Trad. de Bardy (1950). 48 Mossberg (1934 : 69). Voir aussi 8, 13, 2. 49 Waszink (1947 : 259). Voir aussi TERT. pudic. 9. 50 Voir l’index de l’édition de Reifferscheid (1875 : 315) de l’Adversus nationes. 51 2, 70 certum est utique Mineruam non fuisse neque rerum in numero aut ulla esse in substantia computatam, sed ex quo capite Iovis enata est, et esse res coepit et nonulla in essentia constituta . . . 52 Les manuscrits de TERT. adv. Val. 30 p. 206, 3 ont essentia (= 30, 1 Fredouille, qui accepte la conjecture de Braun/Riley) et TERT. carn. 2, 1.

16

Bruno Rochette

6 La littérature de traduction La littérature de traduction est fort intéressante, car elle confirme que le participe constitutus est bien ressenti comme un équivalent du participe grec ὤν (ou équivalents). Bible : (27) Civitas super montem constituta. (VET. LAT. Matth. 5, 14 [cod 1], gr. κειμένη, Vulgate posita) (28) In figura Dei constitutus. (VET. LAT. Phil. 2, 6 [Cyr. testim. 2, 13. 3, 39], gr. ὑπάρχων, Vulgate cum . . . esset) Écrits médicaux : (29) Color totius vultus viridis et niger constitutus . . . Καὶ τὸ χρῶμα τοῦ ξύμπαντος προσώπου χλωρόν τε ἢ καὶ μέλαν ἐὸν . . . (HIPPOCR. progn. 2, 2) ‘Teint de l’ensemble du visage jaune ou même noir . . .’ (30) Deinde eriguntur quidam ipsorum salvi constituti, alii vero non ἔπειτα ἀνεγείρον ται, οἱ μέν τινες ὑγιέες ἐόντες, οἱ δ᾽ οὔ. (HIPPOCR. aer. 22, 5) ‘Puis ils se réveillent, les uns étant guéris, les autres pas.’ Dans la Vetus Latina, e(x)stans sert à traduire ὤν (VET. LAT. Hebr. 3, 2 [Mutian. Chrysost. hom. 5, 2 p. 273in.]) : considerate . . . Christum fidelem exstantem [gr. ὄντα], alors que la Vulgate utilise une relative [qui fidelis est]).53 Les exemples (voir 27 et 28) montrent que la Vulgate fait d’autres choix que la Vetus Latina, même si constitutus dans le sens de ὤν se trouve dans la Vulgate.54 En outre, on trouve au moins deux occurrences de constitutus (voir 29 ; 30) pour rendre ὤν dans une traduction latine anonyme d’Hippocrate datant très probablement du Ve s. La correspondance est parfaite.

53 TLL vol. V, 2, p. 1935, 15. 54 TLL vol. IV, p. 523, 77.

Constitutus = καϑεστώς

17

7 Les inscriptions Le sens « faible » de constitutus apparaît également dans le latin des inscriptions. On peut citer un texte épigraphique de Rome (voir 31) et un autre d’Afrique (voir 32).55 Dans l’inscription de Rome, qui concerne un soldat qui a servi dans l’officium des castra peregrina, constitutus est accompagné d’un adverbe indiquant une localisation (peregre ‘à l’étranger’), tandis que dans celui d’Afrique, que l’on peut dater du IIIe s., d’après le formulaire et la structure onomastique, on trouve le complément prépositionnel habituel. (31) Genio sancto / castrorum / peregrinorum / Aur Alexander / canaliclarius / quod peregre / constitutus vovit / aedil castrorum / votum libens solvit. (CIL VI 231 = ILS 2215 [Rome, IIIe s.])56 (32) Situs hic Alfius Felix Flavianus est, / qui partam nimia frugalitate substantiam / in diem mortis servavit incolumem sibi. / Hanc (i.e. substantiam) constitutus vitae in confinio / filis distribuit, consulens concordiae, / satisque animo dum valet fecit suo . . . (CIL VIII 27587 [CE 1869], Afrique [Sicca Veneria], IIIe s.)57 Cette épitaphe (voir 32) soignée (en sénaires iambiques avec quelques anomalies), fort bien conservée, est assez longue (16 l.). L’expression que l’on y trouve constitutus vitae in confinio (‘se trouvant au terme de sa vie’)58 pourrait être rapprochée de celle qui apparaît dans le livre XI des Métamorphoses d’Apulée à propos du seuil des enfers (voir 11).

8 Conclusion Le participe constitutus comme substitut du participe présent du verbe être apparaît à partir du IIe s. apr. J.-C. et devient fréquent, surtout chez les auteurs

55 Le TLL signale CIL III 10591 (vers 200–270) Marciano in flore aetatis constituto hominibus erepto, inscription sur le sarcophage de Marcus Aurelius Marcianus (Panonnia Superior). La consultation de la base de données db.edcs.eu/epigr donne quelques exemples supplémentaires : in pericu/lis constitutus (Année Épigr. 1993, 422 [Latium]) ; in acie constitutus (Année Épigr. 2008, 1145 [Pannonia Inferior]). 56 Sur le terme canaliclarius, Gilliam (1976). 57 Hamboune (2011 : n. 63, 118–120). 58 Hamboune (2011 : 119) traduit ‘arrivé à ses derniers moments’.

18

Bruno Rochette

chrétiens, à l’époque tardive. Les inscriptions confirment cette chronologie. L’affaiblissement sémantique de constituere date en effet de l’époque impériale. Constitutus fait partie des mots qui ont pris un sens nouveau par l’évolution naturelle de la langue. On peut peut-être le considérer comme un trait de la langue populaire.59 Il faut noter toutefois que cet affaiblissement est resté relatif, car l’emploi de constitutus comme copule construit avec un attribut s’est peu développé. Constituere a sans doute joué en latin le rôle que remplissent dans les langues germaniques, comme l’allemand, des verbes indiquant une position (all. stehen, latin stare). Le sens de ces verbes indiquant une position s’est affaibli au point de devenir des verbes d’existence ‘se trouvant, être’. Il est possible qu’il faille interpréter l’emploi de positus, que l’on trouve parfois dans le même sens que constitutus, comme un verbe de position correspondant au grec κεῖμαι. Dans les exemples réunis, c’est en général le complément de lieu qui porte l’information essentielle, le sémantisme du verbe étant relégué à l’arrièreplan. Doit-on s’étonner de trouver un participe parfait (qui est en principe passif) dans le sens d’un participe présent ? Non, car constitutus est un parfait résultatif intransitif. Il correspond donc bien au grec καϑεστώς. Cet emploi de constitutus est-il un africanisme, comme le pensait Wölfflin (1892) ?60 Il est vrai que beaucoup d’auteurs africains emploient constitutus dans le sens de ‘étant’, mais on trouve l’emploi ailleurs. Est-ce un hellénisme ? C’est possible, même s’il ne faut pas exagérer l’influence du grec. Les traductions de la Bible ont peut-être joué un rôle. Faut-il y voir une influence de la langue juridique ? C’est probablement le cas, même si on peut remarquer que constitutus est plus fréquent chez Cyprien que chez Tertullien, plus habitué au vocabulaire juridique. La langue juridique a peut-être aussi influencé Apulée, familier du barreau. Minucius Félix, avocat distingué, emprunte des mots à la langue des tribunaux. Quoi qu’il en soit, constitutus semble être le seul participe parfait qui ait servi à suppléer un participe présent. Les autres verbes qui servent au supplétisme sont des formes en -ens, comme existens.61

Bibliographie CUF = Collection des Universités de France. GCA = Groningen commentaries on Apuleius (1977–). Groningen: Bouma’s Boekhuis.

59 Callebat (1968 : 156). 60 Brock (1911 : 249–250). Voir toutefois les objections de Bayard (1902 : 327). 61 TLL vol. V, 2, p. 1875, 32.

Constitutus = καϑεστώς

19

OLD = Glare, P. G. W. (Ed.) (1968–1982). Oxford Latin Dictionary. Oxford: Oxford University Press. TLL = Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. Bardy, G. (1950). Œuvres de saint Augustin 1re série: Opuscules vol. 12. Les Révisions (Retractiones). Paris: Desclée de Brouwer. Bayard, L. (1902). Le latin de saint Cyprien (diss.). Paris: Hachette. Brachet, J.-P. (1998). Les verbes “être” du latin et le supplétisme roman. Revue de Philologie, 72, 175–187. Brock, M. D. (1911). Studies in Fronto and his age. Cambridge: Cambridge University Press. Callebat, L. (1968). Sermo cotidianus dans les Métamorphoses d’Apulée. Caen: Association des Publications de la Faculté des Lettres et Sciences Humaines de l’Université de Caen. Druhan, D. R. (1938). The syntax of Bede’s Historia Ecclesiastica. Washington: The Catholic University of America. Ernout, A. (1954). Aspects du vocabulaire latin. Paris: Klincksieck. Ernout, A. (1957). Exsto et les composés latins en ex-. Dans A. Ernout, Philologica II (pp. 198–207). Paris: Klincksieck. Genoude, E.-A. (1852). Tertullien De l’âme. www.tertullian.org/french/g2_02_de_anima.htm (dernier accès: 22 février 2019). Gilliam, J. F. (1976). Canaliclarius and Kananiklarios (P.Oxy. XL 2925). The Bulletin of the American Society of Papyrologists, 13, 49–52. Giodice-Sabbatelli, V. (1981). Constituere, dato semantico e valore giuridico. Labeo, 27, 338–357. Goetz, C. (1896). Constitutus = καϑεστώς, ὤν bei Cyprian. Archiv für lateinische Lexicographie, 9, 307–308. Gradenwitz, O., Kuebler, B., Schulze, E. Th. (1903). Vocabularium iurisprudentiae Romanae. Vol. I. Berlin: Reimer Verlag. Gudeman, A. (1906–1909). Art. Constituo. Thesaurus Linguae Latinae, IV, 510–524. Hamboune, C. (2011). Vie, mort et poésie dans l’Afrique romaine d’après un choix de Carmina Latina Epigraphica. Bruxelles: Latomus. Hijmans, B. J., Paardt, R. T. van der, Smits, E. R., Westendorp Boerma, R. E. H., Westerbrink, A. G. (1977). Apuleius Madaurensis Metamorphoses. Book IV, 1–27 (Groningen Commentaries on Apuleius). Groningen: Bouma’s Boekhuis. Kalb, W. (18882 [1961]). Das Juristenlatein. Versuch einer Charakteristik auf Grundlage der Digesten. Nürnberg: H. Ballhorn [Aalen 1961]. Kalb, W. (1890 [1975]). Roms Juristen nach ihrer Sprache dargestellt. Leipzig: Teubner [Aalen 1975]. Kalb, W. (1912 [1961]). Wegweiser in die römische Rechtssprache für Absolventen des humanistischen Gymnasiums. Leipzig: Otto Nemnich Verlag [Aalen 1961]. Klingenberg, G. (1999). Constitutum est in D. 47,2,14,4. Revue internationale des Droits de l’Antiquité, 46, 243–314. Kunkel, W. (2001 [19672]). Die römischen Juristen: Herkunft und soziale Stellung. Köln-WeimarWien: Böhlau. Landgraf, G. (1896). Über die Latinität des Horazscholiasten Porphyrion. Archiv für lateinische Lexicographie, 9, 560. Leumann, M. (1977). Lateinische Laut- und Formenlehre. Munich: Beck.

20

Bruno Rochette

Löfstedt, E. (1933). Zur Frage der Gräzismen. Dans E. Löfstedt, Syntactica. Studien und Beiträge zur historischen Syntax des Lateins II (pp. 406–457). Lund: C.W.K. Gleerup. Mossberg, K.-A. (1934). Studia Sidoniana critica et semasiologica (diss.). Uppsala: Almqvist & Wiksell. Nelson, H. L. W. (1981). Überlieferung, Aufbau und Stil von Gai Institutiones (Studia Gaiana, 6). Leiden: Brill. Pastorino, A. (1956). Iuli Firmici Materni de errore profanarum religionum. Firenze: La Nuova Italia Editrice. Petersmann, H. (1977). Petrons urbane Prosa. Untersuchungen zu Sprache und Text (Syntax). Wien: Verlag der österreichischen Akademie der Wissenschaften (SBWW, 323). Reifferscheid, A. (1875). Arnobii adversus nationes libri VII (Corpus scriptorum ecclesiasticorum latinorum IV). Wien. Schöpsdau, K. (1992). Vergleiche zwischen Lateinisch und Griechisch in der antiken Sprachwissenschaft. Dans C. W. Müller, K. Sier & J. Werner (Eds.), Zum Umgang mit fremden Sprachen in der griechisch-römischen Antike (pp. 115–136). Stuttgart: Franz Steiner Verlag. Schrijnen, J. & Mohrmann, C. (1936–1937). Studien zur Syntax der Briefe des hl. Cyprian (2 vol.). Nijmegen: Dekker & van de Vegt. Skahill, B. H. (1934). The Syntax of the Variae of Cassiodorus (diss.). Washington: The Catholic University of America. Stangl, T. (1909). Pseudoasconiana. Textgestaltung und Sprache der anonymen Scholien zu Ciceros vier ersten Verrinen auf Grund der erstmals verwerteten ältesten Handschriften. Paderborn: Ferdiand Schöningh. Svennung, J. (1922). Orosiana, syntaktische, semasiologische und kritische Studien zu Orosius (diss.). Uppsala: Akademiska Bokhandeln. Svennung, J. (1935). Untersuchungen zu Palladius und zur lateinischen Fach- und Volkssprache. Uppsala: Almqvist & Wiksells. Turcan, M. (1986). Les spectacles (De spectaculis). Paris: Éditions du Cerf. Vocabularium iurisprudentiae Romanae (1903/1910), 5 vols. Berlin: Reimer Verlag. Waszink, J. H. (1947). Quinti Septimi Florentis Tertulliani De anima. Edited with introduction and commentary. Amsterdam: J. M. Meulenhoff. Wölfflin, E. (1892). Minucius Felix. Archiv für lateinische Lexicographie, 7, 481.

Leonardo Costantini

Exploring the semantic complexity of the voces mediae: magus, magicus, and magia Abstract: This chapter presents a new semantic taxonomy of the Latin terms magus, magicus, and magia. Through analysing their occurrences, it will be argued that the twofold interpretation of magus, commonly accepted in current scholarship as ‘priest of the Persians’ or ‘wicked enchanter’, does not enable us to fully understand the semantic spectrum of this set of terms. I propose a distinction between three main connotations of magus and related terms: the first (philosophical-religious magic) occurs when magus indicates the Zoroastrian wise men. The second connotation (goetic magic) refers to the real magical practitioners and practices condemned by the Lex Cornelia de sicariis et ueneficis. The third connotation introduced here (literary magic) designates the dramatized descriptions of goetic magic in poetry and in prose fiction. This new semantic taxonomy will make it possible to cast more light on the synchronic variation in this set of terms and on their contextual meaning in the passages in which they occur. Keywords: Latin linguistics, magia, magicus, magus

1 Introduction Magic seems to have an ever-growing popularity in both the academic and nonacademic communities: from contemporary fiction to the fields of religious studies and anthropology, the term has become so frequently employed that, at times, there seems to be little awareness of its classical origin.1 Thus, alongside this general interest in magic, it is equally important to assess the meaning of the Latin magus, from which the English ‘magic’ stems,2 bearing in mind that current views on magic do not precisely match those of ancient times. The purpose of this study is to throw new light on the semantic complexity and development of the Latin terms magus – attested in Latin literature since the first

1 There are, of course, exceptions: for references to ancient magic in modern arts, cf. Carlà and Berti (2015). For a study of magic in anthropology, cf. Otto and Stausberg (2013: 68–262). 2 Barnhart (1988: 622, s.v. magic). Leonardo Costantini, Albert-Ludwigs-Universität Freiburg https://doi.org/10.1515/9783110647587-002

22

Leonardo Costantini

century BC – magicus, and magia, a set of terms deriving from the Greek μάγος, μαγικός, μαγεία.3 The concept of magic adopted in this study differs from that of Frazer, Mauss, and later anthropologists, who consider ‘magic’ a transcultural label applied to various supernatural beliefs and to evidence which is unconnected to magus, μάγος, and related terms. I propose, instead, to focus on evidence that we need not interpret as magical, but that was in fact already labelled as such in Greco-Roman times. A fundamental step towards the development of this methodological approach is the thought-provoking view offered by Bremmer.4 He challenges the idea supported by Versnel that a scholarly discourse should always be etic,5 and explains that the old-fashioned pattern of magic versus religion inherited from earlier scholarship should not be applied to the classical world, since magus was not opposed to priest,6 at least before late antiquity. Thus, as Bremmer puts it, “in order to be workable, the etic definition of a concept should always be as close as possible to the actors’ point of view: if not, it will soon cease to be a useful definition”,7 a stance that is very close to the so-called emic approach.8 I argue that a lexicographical enquiry, especially in the case of the Thesaurus Linguae Latinae, bears substantial comparison to the emic methodology, which is consequently particularly fitting to the analysis of the term magus and its cognates presented in this study.

3 For the occurrences of magus in Cicero and Catullus, cf. n. 35 and 36 below. The term magicus was first found in VERG. ecl. 8, 66 (cf. also n. 69), while the earliest occurrence of magia is in Apuleius’ Apologia. That the term magus is a Greek coinage is also evident from observing the employment of the Greek genitive form magices (a substantivized use of the adjective magicus) in PLIN. nat. 28, 188; 30, 8; 30, 11, the accusative form magian at APVL. apol. 27, 12; 28, 4; 29, 2; 31, 1; 31, 4; TERT. anim. 57, 2; AVG. civ. 10, 9 p. 415, 10; AMM. 23, 6, 32, and from the fact that the ī in magia remains long, as in PRVD. c. Symm. 1, 89. Greek loanwords do not always respect the rule according to which uocalis ante uocalem breuis est, cf. De Climent (1981: 109–110); Meiser (1998: 76), who give examples of how such borrowings preserve their original Greek accent. For a discussion of Greek loans and phonetic interference in Latin, cf. Adams (2003: 443–447). 4 Bremmer (2008: 235–247; 347–352). 5 Versnel (1991: 177–197). 6 Bremmer (2008: 347–352). 7 Bremmer (2008: 348). 8 The terms ‘emic’ and ‘etic’ were coined by Pike from the linguistic terms ‘phonetic’ and ‘phonemic’. Pike explains that the emic approach is based on the analysis of a specific language and its sociocultural values; thus, it seeks to reconstruct the conceptualizations, beliefs, and customs of the culture being analysed. The etic approach instead studies a cultural behaviour from the viewpoint of an external observer, who does not share the same customs of the culture under examination; cf. Pike (1967: 37–72). The importance of applying the emic approach to ancient magic has also been acknowledged by Dickie (2001: 19) and Stratton (2007: 1–38).

Exploring the semantic complexity of magus, magicus, and magia

23

A linguistic concern for ancient magic characterizes the studies by Bremmer, Dickie, and Rives, who agree that the Greek μάγος and the Latin magus had a double meaning: on the one hand, they could refer to the priest of the Persians, followers of Zarathustra; on the other hand, they could indicate a fearsome ‘enchanter’ (γόης, φαρμακεύς, ueneficus, maleficus) in a word a μάγος-magus, a skilled practitioner who could use his uncanny powers on human and non-human beings alike.9 Although these studies are methodologically valuable, since they dismiss modern views on magic and focus on the original terminology, these conclusions – partially anticipated by Nock, and Bidez and Cumont10 – are not definitive. The problem with this twofold distinction between priest of the Persians and enchanter, which is also found in the entry magus in the TLL by Dietzfelbinger,11 consists in the fact that much evidence comes from literary sources which do not aim to give a detailed and realistic account of the actual practices of contemporary magi. To unfold the semantic polyvalence of magus one needs to take into account more than the two aforementioned meanings. I shall, therefore, distinguish three types of magic.12 The first two kinds mirror the aforementioned twofold division: the first and oldest type is what I define as philosophical-religious magic, which occurs when Magus-Μάγος indicates the priest of the Persians, a wise man retaining a superior lore sought out by Greeks philosophers. In this case, I use the capital letter since the term indicates the ethnonym from which the religious sect derives, not the wicked enchanters. The second type

9 Dickie (2001: 18–46; 124–141); Bremmer (2008: 235–248); Rives (2009: 119–132, 2010: 53–77). 10 Cf. Nock (1933: 164–188); Bidez and Cumont (1938: 10–11). 11 Cf. TLL, vol. VIII, s.v. magus, p. 149–152, with a reference to the commentary by Pease (1920–1923: 175) on CIC. div. 1, 46. Dietzfelbinger structures the entry by focusing first on the Persian Magi, cf. group A: there he unites the ethnic with the religious connotation of magic, cf. subgroup A1 (nomen est gentis Persarum, unde sacerdotes eligebantur), to which he thereby devotes a specific subgroup, cf. A2 (de ipsis sacerdotibus Persarum). A2 alongside group B (de sacerdotibus uel sapientibus aliarum gentium) conform to the conventional twofold distinction, contrasting philosophical-religious magic with goetic magic, cf. group C (de quodlibet homine occultis artibus dedito). 12 In absolute terms, one could identify another type of magic, namely an ‘ethnic magic’, where the term μάγος indicates specifically the member of a Median tribe, as in HERODOT. 1, 101 and STRABO 15, 3, 1. This meaning can already be seen in the Old Persian maguš (‘priest’) which originally meant ‘member of a tribe’, cf. De Jong (1997: 387). This ethnic connotation – from which that of ‘Persian priest’ derives – does not occur in the Latin sources that I examine, since magus is always associated to the priests of the Persians; thus, it is unnecessary to apply this further distinction in this study.

24

Leonardo Costantini

of magic, which I call goetic magic,13 refers to the real goetic practices and practitioners condemned by the Lex Cornelia de sicariis et ueneficis – originally promulgated by Sulla in 81 BC – as re-issued in the second century AD.14 The third kind of magic that I introduce is called literary magic. This connotation, which stems from the goetic type of magic and may be considered its subgroup, designates the dramatized descriptions of goetic magic in poetry and prose fiction. A remarkable source to glimpse the semantic complexity of magus is Apuleius’ Pro Se de Magia or Apologia, a speech delivered by Apuleius in AD 158–159 before the proconsul of Africa, Claudius Maximus, to defend himself from the allegation of practicing magic and having used his uncanny skills to win over the wealthy widow Aemilia Pudentilla.15 The Apologia, which is also the starting point for Rives’ discussion of magus,16 is the richest source to observe the employment of the term magus and its cognates in Latin literature (magus occurs 40 times,17 magia 39 times,18 and magicus 22 times19), and it is also the source in which the full semantic spectrum of this set of terms is used, since Apuleius plays with the three meanings of magus in order to weaken the accusations brought against him.20 This threefold distinction that I propose is not meant to constitute a rigid grid, but rather to

13 Unlike μάγος, γόης had only the negative meaning of wicked ‘enchanter’ in Classical Greek, which probably developed from the original connection of this term with the sphere of death (Johnston 2008: 14–20). Therefore, I consider the adjective ‘goetic’ as the most suitable to specify the harmful type of magic by using a terminology which mirrors that adopted in classical antiquity. 14 On this law, cf. the detailed discussion by Pellecchi (2012: 266–277). 15 I follow the edition of the Apologia by Martos (2015) and his subdivision of the chapters. For a study of magic in Apuleius’ speech, cf. Abt (1908) and Costantini (2019). 16 Rives (2010: 54–58). 17 APVL. apol. 9, 3; 25, 8; 25, 9; 26, 3; 26, 6; 26, 9; 27, 2; 28, 4; 30, 1; 30, 2; 31, 9; 32, 2; 32, 5; 40, 3; 43, 1; 43, 2; 43, 8; 43, 10; 45, 5; 48, 2; 51, 10; 54, 7; 66, 3; 78, 2; 79, 1; 79, 2; 79, 4; 79, 6; 81, 1; 82, 1; 82, 6; 84, 4; 90, 1; 90, 6; 91, 1. We also find the Greek forms μάγος and μαγεύω at 82, 2; 83, 1; 84, 2. 18 APVL. apol. 2, 2; 9, 5; 25, 5; 25, 10; 26, 1; 27, 9; 27, 12; 28, 4; 29, 1; 29, 2; 29, 6; 29, 9; 30, 5; 31, 1; 31, 2; 31, 4; 47, 1; 47, 3; 53, 2; 54, 6; 58, 5; 62, 3; 63, 2; 64, 8; 67, 1; 67, 3; 70, 3; 78, 5; 80, 5; 81, 1; 82, 4; 83, 5; 84, 3; 84, 4; 87, 2; 90, 4; 96, 2; 102, 1; 102, 2. 19 APVL. apol. 1, 5; 9, 2; 17, 3; 32, 2; 34, 5; 36, 7; 38, 7; 41, 5; 42, 2; 42, 6; 47, 2; 47, 5; 53, 4; 53, 9; 53, 12; 54, 1; 54, 8; 61, 2; 63, 6; 69, 4; 80, 1; 102, 7. 20 For example, at apol. 25, 9–26, 5; 30, 2–4 Apuleius refers to the philosophical-religious connotation of magus. At apol. 26, 6–9 and 47, 3–4 he alludes to goetic magic, while at apol. 30, 6–8; 30, 11–13; 31, 5–7 he draws on literary magic.

Exploring the semantic complexity of magus, magicus, and magia

25

restore three important semantic tendencies to which this set of terms was subject in Greco-Roman times. Fowler rightly argues for the semantic fluidity of magus and its cognates,21 and the following discussion will show how the Romans could pick a meaning within these fluctuating boundaries, but they could also reinterpret the term magus negatively, to the detriment of the philosophical-religious connotation, according to their own sensibility and understanding of these figures and their practices. Bearing these considerations in mind, I aim to throw new light on the synchronic meanings of magus and related terms, and on the influence that the Greek term μάγος had in shaping these connotations.

2 Philosophical-religious magic I shall first focus on the positive connotation of the term magus, arguing that it derives from a long-lasting belief that the wisdom of the Magi had been the source from which many philosophers from Pythagoras onwards gained their knowledge. A brief diachronic overview of evidence in Greek is required, since Latin sources attesting the term magus and its cognates were influenced by earlier Greek writings. This philosophical-religious connotation of magus is likely to be that from which the other connotations derive, and was probably already attested in the lost Lydiaká by Xanthus the Lydian (ca. 450 BC), a part of which was devoted to the customs of these priests.22 The earliest nonfragmentary source on the Mάγοι is Herodotus, who describes them as a Median tribe with priestly functions within the Persian Empire.23 As Bremmer suggests, Herodotus assumes that his readership is already familiar with the Mάγοι24: this is unsurprising since the Greeks living in Asia Minor would have known these priests since Cyrus’ conquest in the sixth century,25 while those in Greece could have seen or heard of the Magi who accompanied

21 Fowler (1995: 19–22, 2005: 283–284). 22 CLEM. ALEX. strom. 3, 11, 1 and DIOG. LAERT. 1, 2 may also belong to this section of the Lydiaká. 23 HERODOT. 1, 107–108; 1, 120; 1, 128; 1, 131–132; 7, 19. 24 Bremmer (2008: 240). 25 Xenophon explains that Cyrus the Great first established the priestly college of the Magi (XEN. Cyr. 8, 1, 23), who accompany him throughout his conquest (XEN. Cyr. 4, 5, 14; 4, 6, 12; 5, 3, 4).

26

Leonardo Costantini

Xerxes during his invasion.26 Yet, Herodotus – probably because of his bias against the Persians27 – does not express admiration for these priestly figures: on the contrary, he describes the Magi as ‘employing enchantments’ (φαρμακεύσαντες) when sacrificing white horses to cross the river Strymon.28 Xenophon of Athens considers the Mάγοι as priestly figures,29 but to observe a truly positive connotation of this term we need to look at the writings of Hellenistic philosophers. It is Sotion, author of the Successions of Philosophers and a source for Diogenes Laertius, who specifies that philosophy arose from the wisdom of the Magi in Persia, of the Chaldeans in Babylon, the Gymnosophists in India, and the Druids in Gaul and Britain.30 This interest in the Magi can already be seen in Heraclides Ponticus, a pupil of Plato and the author of a dialogue entitled Zoroaster,31 and probably in the Perì Mágōn by the biographer and grammarian Hermippus of Smyrna.32 In addition, in the Magikós, ascribed to Aristotle by Diogenes Laertius, it was made clear that the Magi did not know the goetic type of magic.33 Later, in the second and first century BC, Greek intellectuals such as Bolus of Mendes and the physician Cleemporus also seem to have fostered the idea that the earlier Greek philosophers travelled eastwards to meet with the Magi.34 These accounts and these views contributed to shape the philosophicalreligious connotation of magus, which is first attested in Latin in the first century BC. In Cicero, Magus indicates the Persian priests and diviners,35 and so in Catullus, although he regards them with contempt especially for their incestuous conjugal customs.36 To observe a positive regard for the Magi, one must turn to Apuleius’ works: the idea that the Magi were the bearers of a higher wisdom worth studying and pursuing underlies the passages in apol. 25, 9–26, 5; Socr. 6 p. 133; flor. 15, 14; Plat. 1, 3 p. 186. Although the religious and philosophical esteem for

26 HERODOT. 7, 114; 7, 191. Centuries later, Pliny the Elder (PLIN. nat. 30, 8) reports that Ostanes was the high priest in Xerxes’ entourage, responsible for sowing the seeds of this lore in the Greek world. 27 Cf. Hall (1989: 76–100) and Bremmer (2008: 243–244). 28 Bremmer (2008: 240); Dickie (2001: 34); Collins (2008: 57). 29 XEN. Cyr. 5, 4, 14; 4, 5, 51; 4, 6, 11; 5, 3, 4; 7, 3, 1; 7, 5, 35; 7, 5, 57; 8, 1, 23; 8, 3, 11; 8, 3, 25. 30 DIOG. LAERT. 1, 1, where Diogenes refers to book 23 of the Successions of Philosophers. 31 Wehrli (1969: frg. 68–70). 32 Wehrli (1974: frg. 2–4 and the comment at pp. 45–47). 33 DIOG. LAERT. 1, 8. The authorship of the treatise is debated. Cf. Rives (2004: 35–54) who argues for a spurious attribution. 34 Cf. PLIN. nat. 24, 159; 24, 160; COLVM. 7, 5, 17, and the discussion in Dickie (2001: 119–121). 35 CIC. div. 1, 46; 1, 47; 1, 90–91; fin. 5, 87; leg. 2, 26; nat. deor. 1, 43; Tusc. 1, 108. 36 CATVLL. 90, 1–6.

Exploring the semantic complexity of magus, magicus, and magia

27

the Magi survives into later time, as shown by Ammianus Marcellinus,37 this was not predominant. From the third century AD, the reputation of magus and μάγος were in decline: Philostratus endeavours to dissociate the Pythagorean Apollonius of Tyana from the μάγοι in his Life of Apollonius.38 The Platonic philosopher Plotinus considers μαγεία as a form of γοήτεια and biases his readers against it,39 and – with even more contempt – Augustine irrevocably equates magia with goetia.40 This derogatory connotation was, however, far from being new: I shall now examine the origin and the development of this negative meaning of magus.

3 Goetic magic Most of the occurrences of magus in Latin literature refer to the detrimental meaning of this term. It is necessary, therefore, to shed more light on the origin and development of this negative understanding of magic, attested by the issuing of a severe law against the magi, namely the Lex Cornelia de sicariis et ueneficis as reported in the Sententiae by Iulius Paulus.41 In order to do so, I will discuss the goetic connotation of magus and the Greek μάγος when employed to designate people to whom supernatural abilities were attributed. If we exclude Greek drama, for the moment, and a much-disputed fragment from Heraclitus preserved by Clement of Alexandria,42 the first derogatory

37 AMM. 23, 6, 32 (In his tractibus Magorum agri sunt fertiles, super quorum secta studiisque, quoniam huc incidimus, pauca conueniet expediri. Magian opinionum insignium auctor amplissimus Plato hagistiam esse uerbo mystico docet, diuinorum incorruptissimum cultum, cuius scientiae saeculis priscis multa ex Chaldaeorum arcanis Bactrianus addidit Zoroastres, deinde Hystaspes rex prudentissimus, Darei pater). On this and the following passages at 23, 6, 32–36, cf. Boeft et al. (1998: 168–170). 38 PHILOSTR. vita Apoll. 1, 2. 39 PLOTIN. 1, 4, 9; 2, 9, 14; 4, 3, 14; 4, 4, 26; 4, 9, 3. 40 AVG. civ. 10, 9 p. 415, 10 sq. For Augustine’s derogatory interpretation of magic, cf. Graf (2002: 87–103). 41 PAVL. sent. 5, 23, 15–18: Qui sacra impia nocturnaue, ut quem obcantarent, defigerent, obligarent, fecerint faciendaue curauerint, aut cruci suffiguntur aut bestiis obiciuntur. Qui hominem immolauerit exue eius sanguine litauerint, fanum templumue polluerint, bestiis obiciuntur uel, si honestiores sint, capite puniuntur. Magicae artis conscios summo supplicio adfici placuit, id est bestiis obici aut cruci suffigi. Ipsi autem magi uiui exuruntur. Libros magicae artis apud se neminem habere licet; et penes quoscumque reperti sint, bonis ademptis ambustis his publice in insulam deportantur, humiliores capite puniuntur. Non tantum huius artis professio, sed etiam scientia prohibita est. 42 CLEM. ALEX. protr. 2, 22, 2–3, its authenticity is largely disputed: cf. Bremmer (2008: 236, and n. 9 with additional bibliography).

28

Leonardo Costantini

references to the μάγοι as people with unearthly skills dates to the fifth century BC. In the Hippocratic treatise De Morbo Sacro the μάγοι are scorned for being pseudo-physicians who pretend to cure epilepsy with incantations and purifications.43 In Gorgias’ Helenae Encomium the connection between μαγεία and γοητεία is made explicit,44 and Aeschines employs μάγος καὶ γóης as synonyms to insult Demosthenes.45 But while the Greeks could employ both γόης and μάγος, the Romans only transliterated the latter into Latin, thus they had one word – already ambiguous in Greek – to indicate the goetic practitioners and the Magi. As already noted, the first occurrences of magus in Cicero and Catullus refer to the priests of the Persians. With Vergil, however, we find magicus employed with the derogatory meaning of something related to goetic magic, although in a literary dimension: this is a connotation that becomes conventional in Latin literature, as discussed in Section 4 below. But did any goetic magi actually exist in the Roman world? As Dickie argues, the Late Republic and the Early Empire is the time in which professional goetic practitioners appear in Italy:46 historical sources refer to the expulsion of magi and seers (Chaldaei and mathematici) during the first century AD,47 and to their supposed involvement in practices harming the emperors’ lives.48 The presence of such goetic magi aroused mixed feelings: Pliny the Elder expresses an open contempt for them and does not seem to wish to distinguish between goetic practitioners and Persian sages.49 The Naturalis Historia reflects, in fact, a stage in which magia had fully encompassed the older term ueneficium to indicate goetic practices. This must have been deduced from the idea that the practices of the magi and the uenefici overlapped,50 and this is

43 HIPPOCR. morb. sacr. 2–4; 21. 44 Cf. GORG. Hel. enc. 10 and the discussion by Dickie (2001: 34–35). 45 AESCHIN. Ctes. 137, on which cf. Carey (2000: 210 n. 152). With the exception of these sources, the evidence for this negative use of μάγος and its cognates is scarce up until the first century AD. This may be due to the loss of a substantial amount of Hellenistic literature and, as Bremmer (2008: 247) explains, to the fact that the form γόης was more popular, probably because it was thought to be ‘more Attic’ (ἀττικώτερον, cf. PHRYN. Praep. Soph. 56, 8). Some occurrences of this derogatory connotation of μάγος to describe goetic practitioners are found in PLATO rep. 572e; polit. 280e, on which cf. Dickie (2001: 62–65) and Bremmer (2008: 239). A significant occurrence is in the Derveni Papyrus col. VI, 8–9, on which cf. Piano (2016: 240–241). 46 Dickie (2001: 192–201). 47 CASS. DIO 49, 43, 5; TAC. ann. 2, 32. 48 TAC. ann. 3, 22; 6, 29; 12, 22; 12, 52. 49 Cf. in particular PLIN. nat. 30, 17: Proinde ita persuasum sit, intestabilem, inritam, inanem esse, habentem tamen quasdam ueritatis umbras, sed in his ueneficas artes pollere, non magicas. 50 Plautus is the first author in which we find ueneficus as goetic practitioner, cf. PLAVT. Amph. 1043; Epid. 221; Most. 218; Persa 278; Pseud. 872; Rud. 987; 1112; Truc. 762.

Exploring the semantic complexity of magus, magicus, and magia

29

probably the reason why the Lex Cornelia de sicariis et ueneficis gradually encompassed the crimen magiae: the presence of people believed to have supernatural skills and who thereby endangered the community led to the inclusion of magia amongst the prosecutable crimes.51 This is also why, in the second century AD, Apuleius employs magus, ueneficus,52 as well as maleficus53 interchangeably to designate the practitioner of goetic magic in the Apologia. It is worth noting that in Christian sources the general attitude towards magic is negative: a detrimental understanding of magus and its cognates is clearly apparent in most Christian literature in Latin.54 This attitude might be seen as a counterreaction to non-Christian intellectuals, who already in the second century AD claimed that Christ was a goetic practitioner and performed his miracles through his magical craft.55 The only few exceptions are the references to the three Magi in the Vulgata, which can also be found in Prudentius.56 This linguistic overview shows how the disparaging employment of magus and its cognates reflects the reactions to real goetic practitioners in GrecoRoman world. This analysis has enabled us to look into sources about goetic practitioners across the span of several centuries, pointing to the development and progressive dissemination of this detrimental connotation of magus and its cognates.

4 The literary dimension of magic The popularity of goetic practitioners and the impact that their activities had on the collective imagination was such as to leave a deep mark on classical literature since the fifth century BC.57 Magic in ancient literature – mostly poetry, but also literary prose fiction of the Imperial age – is characterized by dramatic

51 PAVL. sent. 5, 23, 15–18. 52 APVL. apol. 78, 2. 53 APVL. apol. 51, 10. In the speech we also find maleficium associated with goetic magic (magica maleficia). Maleficus, similarly to ueneficus, is first attested in Plautus, cf. PLAVT. Bacch. 280; Cas. 783; Pseud. 195a; 938. Cf. TLL, vol. VIII, s.v. maleficium, p. 176. 54 E.g. TERT. apol. 23, 1; 35, 12; adv. Marc. 1, 13; ARNOB. nat. 1, 52; 4, 12; CYPR. idol. 6; MIN. FEL. 26, 10; 26, 11; SIDON. carm. 2, 83; LACT. inst. 2, 14, 10; 7, 13, 7; AVG. civ. 8, 16; 8, 19; epist. 137, 13; FIRM. math. 3, 2, 18; err. 5, 1. 55 IVSTIN. apol. 1, 30; ORIG. c. Cels. 1, 6; 1, 68; 1, 71; 2, 48. On this cf. also Smith 1978: 81–139. 56 VVLG. Matth. 2, 1–17; PRVD. apoth. 646; cath. 12, 28; 12, 61; 12, 182; tituli 105. 57 Although methodologically outdated, cf. the comprehensive discussion of magic as a literary topos in Eitrem (1941: 39–83).

30

Leonardo Costantini

descriptions of goetic performers and their uncanny skills, which were meant to impress and entertain the readership. The purpose of this literary type of magic may be compared to the manner in which people nowadays enjoy horror, fantasy, and sci-fi. The difference lies, however, in the fact that in the GrecoRoman world goetic practitioners existed and were commonly believed to possess fearsome powers. While Gordon suggests that such literary accounts, in the specific case of Augustan literature, had very little to do with the practice of contemporary goetic magi,58 I argue, instead, that there were connections between real and literary magic. As Ruiz-Montero observes, dividing literary from real magic is problematic59: authors often enrich their dramatized descriptions of magic with details taken from contemporary goetic practices.60 Vice versa, the fortune of literary magic influenced the fashion of a first-century AD curse tablet from Rome, which contains references to the mythical figures of the Sirens, Geryon, and Circe and her transformation of Odysseus’ companions into animals.61 Besides the interconnections between fiction and reality, what is important for the present enquiry is to point out that the knowledge of magic as depicted in literary sources diverges from the actual knowledge of goetic practices because of one substantial reason: while the former simply indicates one’s erudition, the latter is, instead, a punishable crime under the Lex Cornelia de sicariis et ueneficis.62 This literary type of magic, which may be regarded as a subgroup of the goetic kind of magic, appears first in Greek drama: as Photius records, tragedians wrote about magic,63 and there are various occurrences of μάγος and its cognates in the plays of Aeschylus, Sophocles, and Euripides. There are uncertainties about the meaning of Μᾶγος Ἄραβος in a list of deceased Persian commanders in Aeschylus’ Persae 31864; yet, in Sophocles and Euripides μάγος and its cognates already

58 Gordon (2009: 209–228). 59 Ruiz-Montero (2007: 38–39). 60 Graf (1997: 175–204) discusses the possible connections between Theocritus and Lucan, and the Greek Magical Papyri (PGM). The monograph by Reif (2016) provides a comprehensive discussion of the parallels between literary magic and the PGM from the Hellenistic period up until Lucan’s Bellum Ciuile. 61 Thesaurus Defixionum Magdeburgensis 517, 4; 517, 58; 517, 60–61. Cf. www.thedema.ovgu. de/thedema.php?tafel=517&seite=1 (accessed: 26 May 2019). Cf. INSCR. Année Épigr. (2007 n. 260). 62 PAVL. sent. 5, 23, 17–18. 63 PHOT. lex. s.v. μάγους. 64 This has been interpreted both as ‘Magos the Arab’ and ‘Arabos the Magian’, cf. Bremmer (2008: 238) who argues that the combination of names simply shows that Aeschylus was clueless about the Persian Magi.

Exploring the semantic complexity of magus, magicus, and magia

31

indicate the goetic enchanters.65 The occurrence of this derogatory connotation in fifth-century Athens could be ascribed to the conflictual relationships with the Persian Empire.66 In Athenian drama we also find reference to the uncanny powers of Thessalian women,67 a trope which became mainstream in following centuries and which may have inspired the depictions of Medea by Ovid, Seneca, and Lucan, who explicitly associate her supernatural powers with magia.68 The reference to these authors brings us to the appearance of goetic magic in Latin literature. Vergil employs the adjective magicus to refer to goetic magic for purposes of seduction in his eighth Eclogue, where we find not a female – as Simaetha in Theocritus’ second Idyll – but a male character dabbling in love-magic.69 The adjective magicus is attested in the Aeneid,70 and both this and the eighth Eclogue are cited in Apuleius’ Apologia as exemplary sources on literary magic.71 Apuleius’ passage is significant: although he was defending himself from the accusation of being a magus, he freely displayed his knowledge of literary magic in his speech, since both the judge and the audience were well aware that this type of magic belonged to the realm of fiction and could not be incriminating evidence of his purported dabbling in goetic magic. The literary use of magus/maga and magicus referring to practitioners of magic and their rites is common in Horace,72

65 SOPH. Oed. tyr. 387; EVR. Suppl. 1110; Iph. Taur. 1338; Or. 1497. 66 On this, cf. Hall (1989: 56–62). 67 Although neither these female practitioners nor their arts are described with μάγος and its cognates until later times (AESOP. fab. 56; PS. LVCIAN. asin. 4), the strong connections between φάρμακον and goetic practices (GORG. Hel. enc. 14, on which cf. Dickie 2001: 35) made the association between φάρμακον and the derogatory meaning of μάγος possible. This is fundamental to understanding how the Latin magia could encompass ueneficium and how maga could become a synonym of uenefica (cf. OV. am. 1, 8, 5; SEN. Med. 36; Herc. O. 523–527; APVL. met. 2, 5, 4), in the same way in which magus and ueneficus are employed interchangeably by Apuleius (APVL. apol. 78, 2). On the terminology of female goetic practitioners, cf. Paule (2014: 745–757). 68 OV. met. 7, 1–403 (e.g. 194–196: tuque, triceps Hecate, quae coeptis conscia nostris / adiutrixque uenis cantusque artisque magorum, / quaeque magos, Tellus, pollentibus instruis herbis; 330: magicae linguae); SEN. Med. 670–848 (e.g. 685: magicis cantibus); LVCAN. 4, 552–556 (Phasidos et campis insomni dente creati / terrigenae missa magicis e cantibus ira / cognato tantos inplerunt sanguine sulcos, / ipsaque inexpertis quod primum fecerat herbis / expauit Medea nefas). 69 VERG. ecl. 8, 66 (magicis sacris). On magic in this poem, cf. Tupet (1976: 223–232); Clausen (1994: 233–239, 255–265); Luck (1999: 121). 70 VERG. Aen. 4, 493 (magicae artes). 71 APVL. apol. 30, 6–8. 72 HOR. carm. 1, 27, 21–22 (quae saga, quis te soluere Thessalis / magus uenenis, quis poterit deus?); epist. 2, 1, 212–213 (falsis terroribus inplet, / ut magus); 2, 2, 208–209 (somnia, terrores magicos, miracula, sagas, / nocturnos lemures portentaque Thessala rides?).

32

Leonardo Costantini

Propertius,73 Tibullus,74 Ovid,75 Juvenal,76 and Apuleius’ Metamorphoses, where we find the Thessalian saga Pamphile depicted as a lustful woman using her supernatural powers on her victims,77 while a male goetic practitioner (magus) is at the centre of a declamation falsely attributed to Quintilian, known as the Sepulcrum Incantatum.78

5 Conclusion This study shows that through adopting this new semantic taxonomy it is possible to gain a fuller understanding of the meanings of magus, magicus, and magia, and to shed more light on the sociocultural factors that contributed to shaping the semantic development of these terms. The sources analysed highlight that the three connotations of magus were often employed interchangeably. The choice to employ a broader or a more specific connotation depends on two factors, namely the genre and the author’s views on magic. On the one hand, the presence of a well-established literary tradition, in which goetic magic plays a fundamental role, would have induced authors of fictional and dramatized accounts to present the derogatory connotation of magic: this is the case, for example, of the first three books of Apuleius’ Metamorphoses. On the other hand, a key factor would have been the authorial esteem or indeed contempt for the figures of the magi. As noted, Apuleius holds a high regard of magic in his philosophical production, while writers such as Pliny or Christian authors seem to outright despise magic. In conclusion, by bearing in mind the

73 PROP. 1, 1, 20 (magicis focis); 2, 28, 35 (sub magico carmine); 4, 1b, 106 (magicis aquis); 4, 4, 51 (magicae Musae). Cf. Tupet (1976: 348–378); La Penna (1977: 192–195); Ogden (2009: 127). 74 TIB. 1, 2, 44 (magico ministerio); 49 (magico stridore); 64 (magicos deos); 1, 5, 12 (carmine magico); 1, 8, 5 (magico nodo); 24 (magicis auxiliis). Cf. Tupet (1976: 337–348); Maltby (2002: 165–166). 75 OV. am. 3, 7, 35 (quid uetat et neruos magicas torpere per artes?). Cf. Ogden (2009: 126). 76 IVV. 6, 610 (hic magicos adfert cantus); 15, 5 (dimidio magicae resonant ubi Memnone chordae), on which cf. Courtney (2013: 298; 525). 77 APVL. met. 2, 5, 4 (maga primi nominis); 3, 16, 2 (inlecebras magiae); 3, 18, 3 (inexpugnabili magicae disciplinae potestate). 78 PS. QVINT. decl. 10, especially 10, 4 (iam magum misera quaerebat, ut umbra euocaretur); 10, 7 (sic magum protinus nescia matre, cuius horrido murmure imperiosisque uerbis dii superi manesque torquentur); 10, 8 (o natura crudelis, plus magum posse quam matrem!); 10, 13 (effecit magus, ne filium uideres). Cf. Schneider (2013: 13–46).

Exploring the semantic complexity of magus, magicus, and magia

33

distinction here proposed it is not only possible to gain a better understanding of magus, its cognates and their semantic development, but also to clarify the cultural factors behind an author’s use of magus and its intended connotation.

Bibliography TLL = Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. Abt, A. (1908). Die Apologie des Apuleius von Madaura und die antike Zauberei. Gießen: Töpelmann. Adams, J. N. (2003). Bilingualism and the Latin language. Cambridge: Cambridge University Press. Barnhart, R. K. (1988). Chambers dictionary of etymology. London: Chambers Harrap. Bidez, J. & Cumont, F. (1938). Les mages hellénisés. Tome I. Introduction. Paris: Les Belles Lettres. Boeft, J. den, Drijvers, J. W., Hengst, D. den, & Teitler, H. C. (1998). Philological and historical commentary on Ammianus Marcellinus XXIII. Groningen: Egbert Forsten. Bremmer, J. N. (2008). Greek religion and culture, the Bible and the ancient Near East. Leiden/ Boston: Brill. Carey, C. (2000). Aeschines. Austin: University of Texas Press. Carlà, F. & Berti, I. (2015). Ancient magic and the supernatural in the modern visual and performing arts. London: Bloomsbury. Clausen, W. (1994). A commentary on Virgil. Eclogues. Oxford: Oxford University Press. Collins, D. (2008). Magic in the ancient Greek world. Malden/Oxford/Victoria: Blackwell. Costantini, L. (2019) Magic in Apuleius’ Apologia. Understanding the charges and the forensic strategies in Apuleius’ speech. Berlin/Boston: De Gruyter. Courtney, E. (2013). A commentary on the Satires of Juvenal. Berkeley: California Classical Studies. De Climent, M. (1981). Fonetica Latina. Madrid: Consejo Superior de Investigaciones Científicas. De Jong, A. (1997). Traditions of the Magi. Zoroastrianism in Greek and Latin literature. Leiden/New York/Cologne: Brill. Dickie, M. W. (2001). Magic and magicians in the Greco-Roman world. London/New York: Routledge. Eitrem, S. (1941). La magie comme motif littéraire chez les Grecs et les Romains. Symbolae Osloenses, 21, 39–83. Fowler, R. L. (1995). Greek magic, Greek religion. Illinois Classical Studies, 20, 1–22. Fowler, R. L. (2005). Magische Rituale. In Thesaurus Cultus et Rituum Antiquorum (TheCRA) (vol. 3, pp. 286–287). Los Angeles: The J. Paul Getty Museum. Gordon, R. (2009). Magic as a topos in Augustan poetry: Discourse, reality and distance. Archiv für Religionswissenschaft, 11, 209–228. Graf, F. (1997). Magic in the ancient world (translated by Franklin Philip). Cambridge, MA: Harvard University Press.

34

Leonardo Costantini

Graf, F. (2002). Augustine and Magic. In J. N. Bremmer & J. R. Veenstra (Eds.), The metamorphosis of magic from late antiquity to the early modern period (pp. 87–103). Leuven: Peeters. Hall, E. (1989). Inventing the barbarian. Greek self-definition through tragedy. Oxford: Oxford University Press. Johnston, S. I. (2008). Magic and the dead in ancient Greece. In J. C. B. Petropoulos (Ed.), Greek magic: Ancient, medieval and modern (pp. 14–20). London/New York: Routledge. La Penna, A. (1977). L’integrazione difficile. Un profilo di Properzio. Turin: Einaudi. Luck, G. (1999). Witches and sorcerers in classical literature. In V. Flint, R. Gordon, G. Luck & D. Ogden (Eds.), Witchcraft and magic in Europe. Ancient Greece and Rome (pp. 91–158). London: Athlone. Maltby, R. (2002). Tibullus, Elegies: Text, introduction and commentary. Cambridge: Francis Cairns. Martos, J. (2015). Apuleyo de Madauros. Apología o Discurso sobre la magia en defensa propia. Floridas. [Prólogo de El dios de Sócrates]. Madrid: Consejo Superior de Investigaciones Científicas. Meiser, G. (1998). Historische Laut- und Formenlehre der lateinischen Sprache. Darmstadt: WBG. Nock, A. D. (1933). Paulus and the Magus. In F. J. Foakes Jackson & L. Kirsopp (Eds.), The beginning of Christianity (pp. 164–188). London: Macmillan. Ogden, D. (2009). Magic, witchcraft, and ghosts in the Greek and Roman worlds. A sourcebook. Oxford: Oxford University Press. Otto, B.-C. & Stausberg, M. (2013). Defining magic: A reader. Sheffield/Bristol: Equinox. Paule, M. T. (2014). Quae saga, quis magus: On the vocabulary of the Roman witch. Classical Quarterly, 64, 745–757. Pease, A. S. (1920–1923). M. Tulli Ciceronis. De divinatione. Libri II. Urbana: University of Illinois. Pellecchi, L. (2012). Innocentia eloquentia est. Analisi giuridica dell’Apologia di Apuleio. Como: New Press. Piano, V. (2016). Il papiro di Derveni tra religione e filosofia. Florence: Olschki. Pike, K. L. (1967). Language in relation to a unified theory of the structure of human behaviour. The Hague: Mouton. Reif. M. (2016). De arte magorum. Erklärung und Deutung ausgewählter Hexenszenen bei Theokrit, Vergil, Horaz, Ovid, Seneca und Lucan unter Berücksichtigung des Ritualaufbaus und der Relation zu den Zauberpapyri. Göttingen: Vandenhoeck & Ruprecht. Rives, J. B. (2004). Aristotle, Antisthenes of Rhodes, and the Magikos. Rheinisches Museum für Philologie, 147, 35–54. Rives, J. B. (2009). Apion Perì Mágōn and the meaning of the word Mágos. Mene, 9, 119–132. Rives, J. B. (2010). Magus and its cognates in classical Latin. In R. L. Gordon & F. Marco Simón (Eds.), Magical practice in the Latin West (pp. 53–77). Leiden/ Boston: Brill. Ruiz-Montero, C. (2007). Magic in the ancient novel. In M. Paschalis, S. Frangoulidis, S. Harrison & M. Zimmerman (Eds.), The Greek and Roman novel. Parallel readings (pp. 38–56). Groningen: Barkhuis. Schneider, C. (2013). [Quintilien]. Le tombeau ensorcelé. (Grandes déclamations, 10). Cassino: Edizioni Università di Cassino. Smith, M. (1978). Jesus the magician. London: Gollancz.

Exploring the semantic complexity of magus, magicus, and magia

35

Stratton, K. B. (2007). Naming the witch. Magic, ideology, and stereotype in the ancient world. New York/Chichester: Columbia University Press. Tupet, A.-M. (1976). La magie dans la poésie latine. Paris: Les Belles Lettres. Versnel, H. S. (1991). Some Reflections on the Relationship Magic-Religion. Numen, 38, 177–197. Wehrli, F. (1969). Die Schule des Aristoteles. Texte und Kommentar. Herakleides Pontikos. Basel/Stuttgart: Schwabe. Wehrli, F. (1974). Die Schule des Aristoteles. Texte und Kommentar. Hermippos der Kallimacheer. Basel/Stuttgart: Schwabe.

Cristina Martín Puente & Matilde Conde Salazar

La rivalidad entre praeses y praesidens Abstract: The verb praesum (derived from sum) has, among others, the meanings of ‘to be in command’, ‘to direct’ and ‘to be governor (of a province)’. At the same time, praesideo is a regular verb with the meanings of ‘to protect’, ‘to preside’, ‘to be in command’, ‘to direct’ and so on. Both were used throughout Latin, but only praesideo has survived, even in non-Romance languages. Related to the verb praesideo is the noun praeses, which has the meaning of ‘guardian’, ‘protector’, ‘who presides’ and, at least since the time of Tacitus, denominates the governor of a province with civil functions, as opposed to dux. The noun praesidens also was soon used with the sense of ‘the one in charge’ and, with a more technical sense, designates the ‘governor of a province’. Both forms coexisted for a long time, but, while praeses has left no trace in modern languages, praesidens has survived with great vigour even in non-Romance languages. Keywords: Latin linguistics, praesum, praesideo, praesidens, praeses

1 Introducción Desde hace tiempo estamos llevando a cabo un estudio sobre las preposiciones que evolucionan a prefijos en latín (ex, uice, pro), especialmente centrado en la denominación de cargos o magistraturas. Esta investigación ha dado ya algunos resultados y está en la línea del artículo de García Hernández (2000) sobre sub-. Por otro lado, estamos estudiando el vocabulario que surge para denominar las nuevas magistraturas y el acceso a ellas (consiliarius, dux, comes, minister, camerarius, maior domus, caesar, princeps, regulus, ordinare, eligere, etc.).1 En el

1 Cf. Conde Salazar y Martín Puente (2006: 315–332); Martín Puente y Conde Salazar (2014a: 151–161, 2014b: 577–594). Agradecimiento: Este trabajo, que queremos dedicar a Michèle Fruyt, se enmarca en el Proyecto de Investigación FFI2014–51910-P. Queremos manifestar nuestro agradecimiento a las personas que lo han revisado en profundidad, cuyas valiosas aportaciones lo han mejorado de forma muy notable sin duda. Cristina Martín Puente, Universidad Complutense de Madrid Matilde Conde Salazar, Consejo Superior de Investigaciones Científicas https://doi.org/10.1515/9783110647587-003

La rivalidad entre praeses y praesidens

37

presente trabajo confluyen ambas líneas de investigación, ya que abordamos el caso del prefijo prae-, tanto en los verbos praesum y praesideo, como en los nombres praeses y praesidens, que presentan respectivamente sentidos y construcciones sintácticas muy similares. El verbo praesum es un derivado de sum que tiene, entre otras, las acepciones de ‘estar al mando’, ‘dirigir’ y ‘ser gobernador (de una provincia)’; por otro lado, praesideo es un verbo regular con los sentidos de ‘proteger’, ‘presidir’, ‘estar al mando’, ‘dirigir’, etc. Ambos se usaron durante toda la latinidad. Relacionado con el verbo praesideo está el sustantivo praeses, que tiene el sentido de ‘guardián’, ‘protector’, ‘quien preside’ y, al menos desde época de Tácito, denomina al gobernador de una provincia con funciones civiles, frente al dux. Finalmente, praesidens se usó pronto con el sentido de “el que está al mando” y, con un sentido más técnico, denomina al ‘gobernador de una provincia’. Ambas formas coexistieron durante muchos siglos. Ahora bien, mientras que actualmente términos como “presidente”, “president”, “président”, “Präsident”, etc., formados sobre praesidens, continúan vigentes en muchas lenguas modernas; el término praeses no ha dejado huella en ellas. Llama la atención que a lo largo de muchos siglos hayan coexistido y sean frecuentes los dos verbos mencionados y los dos sustantivos derivados de uno de ellos. Fruyt (2011: 149–154) sostiene que en todos los periodos, el latín aumentó su número de palabras combinando elementos preexistentes por medio de sufijación o prefijación y dio así lugar a nuevos significados con nuevos designados, aunque en este caso lo que vemos es una coexistencia de vocablos muy similares en significante y significado durante mucho tiempo. Analizaremos si existe algún tipo de distribución complementaria entre ellos. De hecho, según Fruyt (2009: 31), “Una palabra nueva puede aparecer aún cuando su existencia no era una necesidad como consecuencia de la búsqueda consciente y deliberada de ciertas resonancias asociativas, de connotaciones culturales, etc.”. Eso es lo que creemos que ocurre aquí.

2 Verbos praesum y praesideo La preverbación, como dice Fruyt (2011: 171), es un procedimiento muy productivo en todos los períodos del latín. Desde una perspectiva sincrónica, los preverbios son lexemas o morfemas libres (preposiciones) que adheridos a los verbos quedaron congelados y, en su nuevo estatus, evolucionaron más rápido que las preposiciones. Entre los más productivos está el preverbio espacio-temporal prae (‘en la parte delantera de’), que, según García Hernández (1980: 184–185), dota

38

Cristina Martín Puente & Matilde Conde Salazar

tanto a praesum como a praesideo de una función clasemática “complementaria” (‘presidir’, ‘tener el mando’) y, como afirman Martín García y Varela Ortega (1999: 5012, 5018) para el prefijo “pre-” en español, proporciona el sentido de primacía, que deriva del espacial.

2.1 Verbo praesum (prae-sum) El verbo praesum es irregular – razón por la cual los numerosos compuestos de sum tendieron a desaparecer a lo largo de la latinidad – y tiene al menos tres construcciones sintáctico-semánticas (cf. Tabla 1).2 Tabla 1: marcos predicativos de praesum. praesum complementación

sentido

+ dativo + predicativo

‘guardar, proteger, estar al frente’

+ sintagma preposicional + locativo + adverbio

‘ser gobernador’

construcción absoluta

‘estar presente’

2.1.1 praesum [+ dativo / + predicativo] con los sentidos de ‘guardar’, ‘proteger’, ‘estar al frente de’ Construido con dativo – ver de (1) a (6) – o con predicativo – ver (7) –, el verbo praesum tiene los sentidos de ‘guardar, proteger, estar al frente de’. Esta construcción se documenta durante toda la latinidad. (1)

scitis concessum et datum / mi esse . . . nuntiis praesim. (PLAVT. Amph. 11–12) ‘Sabéis que se me ha concedido y asignado proteger [estar al frente de] las noticias.’

2 En algunos pasajes se puede recuperar el lexema del complemento, pero no el caso en el que aparecería, por ejemplo en: legatis suis . . . negotium . . . dedit; itaque praefuerunt. (CIC. Font. 18, ‘encargó el asunto a sus legados; de esta forma se pusieron al mando’); sicut angeli praesunt, ita et hi qui uitam meruerint angelorum. (AMBR. in Luc. 8, 96, ‘del mismo modo que los ángeles la presiden a ella, así también los que merecerían la vida de los ángeles’).

La rivalidad entre praeses y praesidens

(2)

39

quae [sc. natura] quoniam talis est ut et praesit omnibus. (CIC. nat. deor. 2, 36) ‘puesto que tal es la naturaleza que incluso está al frente de todo.’3

(3) conuocatis . . . principibus . . . in his . . . Lisco qui summo magistratui praeerat. (CAES. Gall. 1, 16, 5) ‘convocó a los nobles . . . entre ellos . . . Lisco, que estaba al frente de la magistratura suprema.’ (4) omnibus Druidibus praeest unus, qui summam inter eos habet auctoritatem. (CAES. Gall. 6, 13, 8) ‘al frente de todos los druidas hay uno, que tiene entre ellos la autoridad suprema.’ (5) Torquatus qui iussu Pompei oppido praeerat. (CAES. civ. 3, 11, 3) ‘Torcuato, quien por orden de Pompeyo estaba al frente de la ciudad.’ (6) capto C. . . . Regino equite Romano Scipionis familiarissimo qui ei oppido praeerat. (BELL. Afr. 68, 4) ‘habiendo hecho prisionero a C. . . . Regino, caballero romano muy amigo de Escipión, que tenía el mando de la plaza.’ (7) qui gregi praesunt. (TERT. fug. 11, 4) ‘los que guardan el rebaño.’ (8) ipsi [dii] . . . medii praefuerunt. (AVG. civ. 3, 20 p. 130, 20) ‘ellos mismos [los dioses] estuvieron al frente como mediadores.’

2.1.2 praesum [+ SPrep / + locativo / + adverbio] con el sentido de ‘ser gobernador’ Con SPrep – ver (9) –, con locativo – ver (10) – o con adverbio – ver (11) – el verbo praesum tiene el sentido de ‘ser gobernador’ ya desde el latín clásico. (9) siquis est forte ex ea prouincia in qua tu triennium praefuisti, qui te nolit perisse. (CIC. Verr. II 3, 180)

3 Las traducciones, salvo que se diga lo contrario, son nuestras.

40

Cristina Martín Puente & Matilde Conde Salazar

‘si por casualidad hay alguien de aquella provincia en la que fuiste gobernador tres años que no quiera tu perdición.’ (10) Otacilius Crassus qui Lissi praeerat expugnare [scil. naues] parabat. (CAES. civ. 3, 28, 2) ‘Otacilio Craso, que era gobernador en Liso, se disponía a atacar las naves.’ (11) L. Staberius, qui ibi praeerat (CAES. civ. 3, 12, 1) ‘Lucio Estaberio, que era gobernador allí’

2.1.3 Construcción absoluta con el sentido de ‘estar presente’ Desde época tardía y en autores cristianos praesum en forma personal se construye como un verbo de una sola valencia y tiene el sentido de ‘estar presente’ – ver (12) y (13). (12) praeest diaboli laqueus manifestus, uenenum portans letalem serpentis. (PS. CYPR. aleat. 5, 4) ‘está presente el lazo del demonio llevando el mortal veneno de la serpiente.’ (13) et nos autem sermoni atque rationi itemque uirtuti, per quae omnia molitum deum ediximus, propriam substantiam spiritum inscribimus, cui et sermo insit pronuntianti et ratio adsit disponenti et uirtus praesit perficienti. (TERT. apol. 21, 11) ‘Nosotros, en cambio, a la palabra, al pensamiento y al poder por medio de los cuales afirmamos que Dios lo ha creado todo, les atribuimos una sustancia propia, espíritu, en que se inserta la palabra cuando pronuncia, reside el pensamiento cuando ordena y está presente el poder cuando concluye algo.’ Desde el latín preclásico la forma praesens, por oposición a absens (de absum), se emplea solamente con el sentido de ‘el que está presente’, nunca con el de ‘guardián’ o ‘gobernador’; así que no compite con praeses ni con praesidens.4 (14) ita quam antea absens habuerat auctoritatem eam omnem praesens dimiserat. (BELL. Afr. 48, 3)

4 Sobre el estatuto de estas formas, cf. Fruyt (1992: 30–31).

La rivalidad entre praeses y praesidens

41

‘así pues, la autoridad de la que había gozado ausente, la había perdido entera estando presente.’

2.2 Verbo praesideo (prae-sĕdĕo) El verbo regular praesideo, formado a partir del intransivo sedeo,5 tiene solamente una construcción con tres tipos de complemento: con dativo, con sintagma preposicional y, al menos desde Tácito, con acusativo (cf. Tabla 2).6 Este es un ejemplo de cómo un prefijo puede contribuir a la modificación de la estructura argumental de la base exigiendo la presencia del argumento seleccionado (Martín García y Varela Ortega 1999: 4999) ‘estar sentado frente a’ o afectar a la semántica del predicado ‘gobernar’.

Tabla 2: marco predicativo de praesideo. praesideo complementación

sentido

+ dativo + sintagma preposicional + acusativo (Tácito)

‘guardar, proteger, estar al frente, ser gobernador’

2.2.1 praesideo + dativo / + SPrep / + acusativo con el sentido ‘guardar, proteger, estar al frente, ser gobernador’ Construido con dativo – ver (15) a (17) –, con sintagma preposicional – ver (18) – o con acusativo – ver (19) y (20) –, el verbo praesideo significa ‘guardar, proteger, estar al frente’. Cuando el complemento designa una provincia, adquiere el sentido más técnico de ‘ser gobernador’. (15) hic ego tibi praesidebo, ne interbitat quaestio. (PLAVT. Most. 1096) ‘yo te protegeré a ti aquí, para que no se agote la cuestión.’

5 La relación entre sum (‘ser’) y sedeo (‘estar sentado’) es tan estrecha, que, por ejemplo, el infinitivo ‘ser’ en español procede de sedere, infinitivo de sedeo. 6 Algo similar sucede con praecurro, que se construye con dativo, lo que no sucede con curro.

42

Cristina Martín Puente & Matilde Conde Salazar

(16) Mars Latio uenerandus erat, quia praesidet armis. (OV. fast. 3, 85) ‘Marte debe ser adorado en el Lacio porque protege la guerra.’ (17) Mucianus . . . cum praesideret Lyciae. (PLIN. nat. 13, 88) ‘Muciano . . . cuando estaba al frente / era gobernador de Licia.’ (18) Metellus Celer cum tribus legionibus in agro Piceno praesidebat (SALL. Catil. 57, 2) ‘Metelo Céler estaba al frente con tres legiones en términos del Piceno’ (19) proximumque Galliae litus rostratae naues praesidebant. (TAC. ann. 4, 5, 1) ‘la costa más cercana de la Galia la guardaban las naves de guerra.’ (20) Atellio, qui Pannoniam praesidebat. (TAC. ann. 12, 29, 2) ‘Atelio, que estaba al frente / era gobernador de Panonia.’ A partir de pasajes como (17) y (20) el verbo praesideo fue adquiriendo el sentido de ‘ser gobernador’ construido con dativo o, en un proceso de transitivación del verbo, con acusativo.

3 Neologismos nominales praeses y praesidens Relacionados con el verbo praesideo existen dos lexemas, praeses y praesidens, que pertenecen a otra categoría. La siguiente afirmación de Fruyt (1992: 30–31) respecto al estatuto ambiguo de los “adjetivos verbales” es perfectamente aplicable a praesidens: “Anciennes formes nominales suffixées (*-nt-, *-to- ou *-no-, *-mno-. . .) intégrées dans la flexion verbale de diverses langues – tout en gardant souvent une existence nominale parallèle –, les « adjectifs verbaux », à l’origine des participes, offrent aussi parfois des statuts ambigus. Certaines formes adjectivales reconnues comme participiales, intégrées dans le paradigme verbal, manifestent des indices de rôle nominal . . .”

Y lo que dice a continuación, a praeses: “Inversement, certaines formes adjectivales non intégrées dans le paradigme verbal manifestent des indices de verbalisation et jouent un rôle quasi-participial.”7

7 Cf. también Franz Rainer (1999: 4597) respecto a los denominados “adjetivos activos puros” y Amador Rodríguez (2009: 35, 37–38) sobre los nombres que derivan de verbos y las formas nominales del verbo.

La rivalidad entre praeses y praesidens

43

3.1 praeses, -idis El vocablo praeses8 quizá no se percibía claramente ligado etimológicamente a praesideo – al igual que ocurre con obses, -idis (‘rehén, fiador, garante, fianza’) y obsideo o reses, -idis (‘inactivo’) y resideo –, como se desprende de este texto de Isidoro de Sevilla, en el que se vincula a praesum: (21) praesides rectores prouinciae sunt, dicti quod praesunt. (ISID. orig. 9, 3, 28) ‘los praesides son los gobernadores de una provincia, denominados así porque están al frente.’ Este término está muy lexicalizado como nombre y por eso lleva más frecuentemente un complemento en genitivo – ver (23) y (24) –, que en dativo9 – ver (25) –, que sería el complemento heredado del verbo (Pena 1999: 4312).

3.1.1 praeses con el sentido de ‘guardián’, ‘protector’, ‘el que preside’ (22) Ite, inquam, domum ambo nunciam ex praesidio praesides. (PLAVT. Rud. 1051) ‘Os digo, guardianes, que os marchéis ambos inmediatamente del puesto de guardia a casa.’ (23) armipotens, praeses belli, Tritonia uirgo. (VERG. Aen. 11, 483) ‘armipotente, patrona de la guerra, doncella tritonia.’ (24) sacrificio Mineruae, praesidi arcis, in urbe facto profectus Corinthum altero die peruenit. (LIV. 45, 28, 1) ‘ofrecido un sacrificio a Minerva, patrona de la acrópolis, en la ciudadela, partió hacia Corinto y llegó al día siguiente.’

8 Tobalina (2009: 189, 201, 203), aunque no incluye praeses entre las magistraturas, nos recuerda que, tras la instauración del imperio, los gobernadores de rango senatorial de muchas provincias se encargarían de las funciones civiles, mientras los caballeros (praesides) harían lo propio con las militares, pero con el tiempo los primeros fueron reemplazados por los del ordo ecuestre. 9 El pasaje LIV. 10, 22, 7 urbi ac foro praesides habendos . . . esse (‘habría que nombrarlos protectores de la ciudad y del foro’) plantea problemas de transmisión que afectan justamente al caso gramatical de los nombres que dependen de praeses (Oakley 2006: 243). No obstante, hay otros ejemplos con dativo: AVG. cons. euang. 1, 25, 38.

44

Cristina Martín Puente & Matilde Conde Salazar

(25) Phoebe potens numeris, praeses Tritonia bellis. (AVSON. 20, 1 (1 S.), 1). ‘Febo, que gobiernas los metros, Tritonia, patrona de las guerras.’

3.1.2 praeses con el sentido de ‘gobernador de provincia’ Aunque en latín designan a un gobernador de provincia los términos praefectus, proconsul, procurator, propraetor, rector, corrector provinciae, etc., Tácito usa praeses con este nuevo significado – la adquisición de nuevas acepciones es un fenómeno muy frecuente en latín (Fruyt 2011: 149–154). Más tarde, con Constantino, el cargo de praeses se especializa en tareas judiciales (Bravo 1981–1985; Christol 1997). Los estudios de los historiadores sobre el cursus honorum, las entradas de los diccionarios y las traducciones parecen estar en desacuerdo.10 (26) M. Trebellius legatus, a Vitellio praeside Syriae . . . missus. (TAC. ann. 6, 41, 1) ‘Marco Trebelio legado enviado por Vitelio, gobernador de Siria.’ (27) praesidibus onerandas tributo prouincias suadentibus rescripsit. (SVET. Tib. 2, 32) ‘escribió a unos gobernadores que le aconsejaban que aumentase los tributos a las provincias.’ (28) praesidis nomen generale est eoque et proconsules et legati Caesaris et omnes prouincias regentes, licet senatores sint, praesides appellantur: proconsulis appellatio specialis est. (MACER dig. 1, 18, 1) ‘la denominación general de praeses se debe a que se denomina praesides a los procónsules, a los legados del emperador y a los gobernadores de todas las provincias. La designación del ‘procónsul’ es especial.’ (29) praesides uocabantur a praesidendo. (CHRIST. STAB. Matth. p. 1348C)11 ‘se les denomina praesides [‘gobernadores’] a partir de praesideo [‘presidir’].’

10 Además de estudios como el de Tobalina (2009), se echan en falta otros en la línea de los que encontramos en los lemas “corrector” y “praeses” en Paulys Realencyclopädie der Classischen Altertumswissenschaft. Tomo IV, Halbband 8, Corniscae-Demodoros. pp. 1645–1655. es.wikipedia.org/wiki/Corrector_romano. 11 Cf. (21) praesides rectores prouinciae sunt, dicti quod praesunt (ISID. orig. 9, 3, 28).

La rivalidad entre praeses y praesidens

45

4 praesidens, -ntis La forma praesidens, con una entrada específica como sustantivo masculino en el diccionario de Gaffiot y un apartado especial en TLL – lo que no sucede con obsidens o residens –, tiene los mismos sentidos que praeses, pero mantiene la estructura argumental típica de praesideo con dativo12 y además es muy usada por los autores cristianos como un término técnico relacionado con la jerarquía eclesiástica. Mientras que praeses no ha dejado huella en las lenguas modernas, praesidens ha sobrevivido con gran pujanza incluso en las no romances. Así en español presidente puede referirse a un cargo o puesto (“lo nombraron presidente”) o designar al que lo ejerce (Nueva Gramática Española 2009: 482).

4.1 praesidens como sinónimo de praeses praesidens se usó pronto con los mismos sentidos que praeses, (‘guardián’, ‘protector’, ‘el que está al frente’): (30) uobis non modo inspectantibus sed etiam armatis et huic iudicio praesidentibus. (CIC. Mil. 101) ‘a vosotros no solo presentes, sino también armados y protegiendo este tribunal.’ (31) legati ab Orico ad M. Valerium praetorem uenerunt, praesidentem classi. (LIV. 24, 40, 2) ‘Embajadores de Orico se presentaron al pretor M. Valerio, que estaba al frente de la flota.’ (32) quem praesidentes studiis deae propius audirent? (QVINT. inst. 10, 1, 91) ‘¿A quién escucharían con más cercanía las diosas que protegen las artes?’

12 Las formas nominales derivadas suelen heredar la estructura argumental de la base léxica de la que proceden (cf. para el caso español Pena 1999: 4312, 4394).

46

Cristina Martín Puente & Matilde Conde Salazar

4.2 praesidens como término que designa a un ‘gobernador de provincia’ También puede designar a un tipo de ‘gobernador de una provincia’, sin complemento – ver (33) –, con un complemento en dativo – ver (34) y (35) – o con un adverbio de lugar – ver (36): (33) per conciliabula . . . disserebant de . . . saeuitia ac superbia praesidentium. (TAC. ann. 3, 40, 3) ‘en asambleas disertaban sobre . . . la crueldad y la soberbia de los gobernadores.’ (34) Gaius Iulius Maximinus, praesidens Trebellicae, primus e militaribus, litterarum fere rudis, potentiam cepit suffragiis legionum. (AVR. VICT. Caes. 25, 1) ‘G. Julio Maximino, gobernador de Trebellica, el primero de los militares, casi analfabeto, obtuvo poder por el sufragio de las legiones.’ (35) Vitrodorus . . . et Agilimundus . . . uariis populis praesidentes . . . iussa fecerunt. (AMM. 17, 12, 21) ‘Vitrodoro . . . y Agilimundo, que gobernaban varios pueblos . . . cumplieron las órdenes.’ (36) dux Cassianus praesidentesque ibi proceres alii . . . Craugasium incessebant. (AMM. 19, 9, 6) ‘el general Casiano y los próceres que gobernaban allí acusaban a Craugasio.’

4.3 Sentidos de praesidens en autores cristianos En autores cristianos, en paralelo a lo que ocurre en los paganos, praesidens, que es un término muy agentivo,13 denomina, bien al que está al frente de una sesión, bien al que está al mando de una diócesis. En el pasaje (39) se contrapone a subditos, tanto por el significante como por el significado (‘gobernado’, ‘súbdito’)14: 13 En español los sustantivos deverbales en ‘(-a/-e)nte’ son nombres de agente y muchos designan profesiones, como “cantante” (Bajo Pérez 1997: 19–29; Amador Rodríguez 2009: 28). 14 El lexema subditus también está a caballo entre la flexión verbal y la nominal (Fruyt 1992: 30), pero en cualquier caso es claramente pasivo.

La rivalidad entre praeses y praesidens

47

(37) ecclesiae praesidentem Beatum Innocentium si audire uoluisses. (AVG. c. Iulian. 1, 4, 13) ‘si hubieras querido escuchar al Bienaventurado Inocencio que estaba al frente de esa iglesia.’ (38) praesidente Aurelio, episcopo ecclesiae Carthaginiensis. (CONC. Carth. a. 411, 1, 55 l. 13) ‘estando al frente Aurelio, obispo de la Iglesia de Cartago.’ (39) quarta tentatio quae et ad praesidentes et ad subditos pertinet. (PROSP. in psalm. 106, 91) ‘la cuarta tentación, la que atañe a los que están al mando y a los subordinados.’ (40) ad praesidentis genua prouoluti, qui adstiterant, precabantur. (HIER. epist. 22, 30) ‘postrados a los pies del que estaba al mando, los que estaban presentes le rogaban.’ (41) habet ecclesia ordines consummativos qui sunt tres scilicet ordo praesidentium ordo magistrantium ordo regularium. (BONAVENTVRA Collationes in Hexaemeron, visio 4 coll. 3, 9) ‘la Iglesia tiene en conjunto tres órdenes, a saber, el orden de los que están al mando, el orden de los que enseñan y el orden de los regulares.’

5 Conclusiones Los verbos praesum y praesideo, con construcciones y sentidos muy similares, entre los que está el de ‘ser gobernador (de una provincia)’, conviven a lo largo de toda la latinidad. Ahora bien, mientras que el verbo praesideo sigue vigente en muchas lenguas modernas, incluso no romances, praesum, aunque, como possum, tuvo bastante más éxito que otros derivados de sum, no ha dejado huella. Relacionado con praesideo, praeses, que tiene los sentidos de ‘guardián, protector, el que está al frente’, adquiere en época de Tácito el sentido técnico de ‘gobernador de provincia’, y aparece con un complemento en genitivo o, menos frecuentemente, en dativo. Praesidens tiene un estatuto a caballo entre la categoría verbal y la nominal, funciona de forma similar a otros sustantivos

48

Cristina Martín Puente & Matilde Conde Salazar

deverbativos y suele construirse, como praesideo, con dativo. A partir de Tácito designa también al gobernador de una provincia y en los autores cristianos a un cargo eclesiástico. Este lexema ha sobrevivido con gran pujanza en muchas lenguas modernas; mientras que praeses desapareció sin dejar rastro.

Bibliografía Amador Rodríguez, L. A. (2009). La derivación nominal en español: Nombres de agente, instrumento, lugar y acción. Frankfurt am Main: Peter Lang. Bravo, G. (1981–1985). El praeses de Diocleciano: Título y función. Hispania antiqua, 11–12, 37–80. Christol, M. (1997). M. Simplicinius Genialis: Ses fonctions (vir perfectissimus agens vice praesidis). Cahiers G. Glotz, 8, 231–241. Conde Salazar, M. & Martín Puente, C. (2006). Ex de preposición a sustantivo. In M. Villayandre Llamazares (Ed.), Actas del XXXV Simposio Internacional de la Sociedad Española de Lingüística (pp. 315–332). León: Universidad de León. Conde Salazar, M. & Martín Puente, C. (2016). Expresión de la noción de ‘sustituir’ y ‘sustitución’ en latín. In Paolo Poccetti (Ed.), Latinitatis rationes: Descriptive and historical accounts for the Latin language (pp. 331–351). Berlin/Boston: De Gruyter. Demonte, V. & Bosque, I. (Coords.), Gramática descriptiva de la lengua española. Madrid: Espasa. Fruyt, M. (1992). Le paradigme verbal: Un ensemble flou. In C. Moussy & S. Mellet (Eds.), La validité des catégories attachées au verbe: Table ronde de Morigny, 29 (pp. 21–36). Paris: Presses Paris Sorbonne. Fruyt, M. (2009). La creación lexica: Consideraciones generales y su aplicación a la lengua latina. Estudios Clásicos, 136, 7–54 (trad. E. Nieto). Fruyt, M. (2011), Word-Formation in Classical Latin. In J. Clackson (Ed.), Companion to the Latin Language (pp. 157–175). Malden, MA: Wiley-Blackwell. García-Hernández, B. (1980). Semántica estructural y lexemática del verbo. Reus: Avesta. García Hernández, B. (1998). Sème et classème dans la structure du champ sémantique de video: Revisions de quelques opinions. In Structures lexicales du latin: actes de la table ronde du VIIe colloque international de linguistique latine, Jérusalem, 20 avril 1993 (pp. 9–36). Paris: Presses de l’Université de Paris-Sorbonne. García Hernández, B. (2000). Los resultados del prefijo latino “sub-” en español. In GarcíaHernández, B. (Ed.), Latín vulgar y tardío: Homenaje a Veikko Väänänen (1905–1997) (pp. 63–69). Madrid: Ediciones Clásicas. Martín García, J. & Varela Ortega, S. (1999). La prefijación. In V. Demonte & I. Bosque (Coords.) (pp. 4993–5040). Martín Puente, C. & Conde Salazar, M. (2014a). El paso de la preposición latina pro a prefijo: una gramaticalización poco productiva. Latomus, 73(3), 577–594. Martín Puente, C. & Conde Salazar, M. (2014b). Vice de sustantivo a prefijo (pasando por adposición). In P. Molinelli, P. Cuzzolin & Ch. Fedriani (Eds.), Proceedings of the LVLT 10 International Conference, Bergamo 2012, I (pp. 151–161). Bergamo: University Presssestante edizioni.

La rivalidad entre praeses y praesidens

49

Nueva Gramática española (2009). Madrid: Real Academia Española, Espasa. Oakley, S. P. (2006). A commentary on Livy. Books VI-X. vol. 4, Book X. Oxford/New York: Clarendon Press/Oxford University Press. Pena Seijas, J. (1999). Partes de la morfología: Las unidades del análisis morfológico. In V. Demonte & I. Bosque (Coords.) (pp. 4305–4365). Rainer, F. (1999). La derivación adjetival. In V. Demonte & I. Bosque (Coords.), pp. 4595–4644. Paulys Realencyclopädie der klassischen Altertumswissenschaft (1901). Tomus IV, Halbband 8, Corniscae-Demodoros (pp. 1645–1655). Stuttgart: J. B. Metzler. Tobalina Oraá, E. (2009). El cursus honorum. In J. Andreu Pintado (Coord.), Fundamentos de epigrafía latina (pp. 175–234). Madrid: Liceus, Servicios de Gestión y Comunicación.

Laurent Moonens

Les emplois des « quantifieurs » quantus, quam multus, quot et de leurs corrélatifs en latin classique Abstract: This chapter examines the meaning and the part played in quantification of the correlative pairs tantus-quantus and tot-quot within the nominal syntagm in Classical Latin. Particular attention has been given to the question of the participation of the pair tanti-quanti in countable quantification, for which a famous example taken from the Elegies of Propertius is often cited. Keywords: Latin linguistics, quantification, tantus, quantus, tot, quot

1 Introduction Dans un précédent travail (voir Moonens 2016), nous avons, à la suite de Spevak (2014), et dans le souci de disposer d’une définition de « quantifieur » interne à la langue latine, baptisé quantifieur tout modifieur d’un nom N au sein d’un syntagme nominal, faisant de ce dernier une réponse possible à la question quot N ? ou quantus N ? Cette définition met ainsi au centre du champ de la « quantification » ainsi définie les mots quot ? et quantus ? Une étude précise des emplois de ces derniers termes est donc indiquée si l’on souhaite déterminer ensuite dans quelle mesure l’usage des « quantifieurs » ainsi définis participe aux phénomènes de « quantification » tels que décrits par Wilmet (1986). C’est cet objet que nous nous sommes fixé pour la présente étude, consacrée aux expressions quantus, quot et à leurs corrélatifs tantus et tot en latin classique. Le présent travail aura donc deux lectures : il s’agira pour commencer d’une étude sémantique et syntaxique des emplois de quot et quantus (et de leurs corrélatifs) en syntagme nominal, mais aussi d’une façon de mettre à l’épreuve, et de délimiter la portée et les conséquences de la définition choisie des « quantifieurs ». Remerciement : Nous souhaitons remercier ici la Professeure Olga Spevak, qui a accepté de diriger le travail doctoral dont ce texte fait plus largement partie, et de nous formuler des suggestions et critiques très utiles. Merci également aux deux rapporteurs qui ont relu le manuscrit dont cet article est issu, et dont les suggestions, précises et pertinentes, ont permis l’amélioration. Laurent Moonens, Université de Toulouse 2 – Jean Jaurès / É.H.É.S.S. https://doi.org/10.1515/9783110647587-004

Les emplois des « quantifieurs »

51

Dans la mesure où les « quantifieurs » sont souvent, dans les travaux de linguistique latine, étudiés en tant que sous-catégorie d’une catégorie plus générale de modifieurs du nom (les déterminants ou « Apports de Classe Fermée » (ACF), comme chez Fugier 1983, ou encore les indéfinis, comme chez Touratier 1994), il nous semble en effet important de contribuer à une étude des quantifieurs pour eux-mêmes. On situera donc dans ce cadre la présente étude. En outre, le choix des paires quantus/tantus et quot/tot s’explique par l’abondance de leurs occurrences en latin classique, et par le fait que leur participation à la quantification dénombrable pose des problèmes délicats au pluriel, notamment dans une perspective diachronique. Nous articulerons cette étude de la façon suivante : après avoir détaillé la terminologie utilisée, et avoir rappelé quelques concepts tenant à la quantification, nous détaillerons le corpus de textes et la méthode choisis. Nous étudierons ensuite le cas de la paire quot/tot, avant d’en venir aux cas de quantus/tantus (sg.) et quanti/tanti (pl.) – ainsi qu’à celui, plus marginal, de quam multi. Au fil de cette étude, nous tâcherons notamment de dégager quel(s) paramètre(s) ont pu influencer la participation du pluriel quanti/tanti à la quantification dénombrable.

2 Quantifieurs et quantification Nous avons déjà rappelé notre choix de définition du terme quantifieur ; cette dernière, qui fait référence exclusivement aux questions quot ? et quantus ?, semble donc a priori indépendante du phénomène linguistique de quantification à l’œuvre dans les « expressions de quantité ». Ainsi conviendra-t-il toujours, dans la suite, de déterminer comment un quantifieur participe à ce phénomène de quantification. À cet effet, il est probablement utile de revenir sur quelques concepts utiles à la description de la quantification, déjà évoqués dans notre précédent travail sur le sujet (voir Moonens 2016) ; nous ne retiendrons toutefois de notre présentation antérieure que les éléments essentiels à notre étude, et y renvoyons le lecteur pour plus de précision. La première distinction importante est celle de quantification dénombrable vs non-dénombrable.

2.1 Quantification (non)-dénombrable La quantification dénombrable et la quantification non-dénombrable font écho, en dernier ressort, à deux « représentations du monde » bien décrites par Wilmet

52

Laurent Moonens

(1986). Pour le grammairien belge, en effet, l’existence de deux représentations : dénombrable (Wilmet dit « numérative ») vs massive, est parallèle à la répartition des « objets du monde » sous les catégories : discret ou dense, lesquelles catégories se révèlent « à l’épreuve de la division (p. ex. un fragment de cheval n’est plus un cheval » – ce qui fait du cheval un « objet du monde » discret –, « tandis qu’une goutte même infime du vin reste du vin, au moins dans la pensée commune » (Wilmet 1986 : 50) – ce qui fait du vin un « objet du monde » dense). Le caractère dénombrable (numératif) ou massif, poursuit le grammairien belge, « répond au critère de la réunion (suivant que l’addition de deux quantités q, p. ex. de cheval ou de vin, laisse subsister entre elles une limite ou entraîne un amalgame) ». La quantification dénombrable est bien décrite par la présentation de Wilmet (1986) ; nous lui emprunterons quelques concepts. C’est afin de distinguer l’applicabilité d’un mot « hors contexte » et « en contexte » que M. Wilmet choisit de distinguer extensionalité et extension. La première concerne l’applicabilité hors contexte ou « en dehors de tout énoncé » (Wilmet 1986 : 43) ; la seconde, au contraire, a trait à l’applicabilité en contexte (Wilmet 1986 : 43). De façon plus explicite, l’extensionalité d’un substantif ou d’un syntagme nominal désigne le nombre d’entités auxquels il est susceptible de s’appliquer, en dehors de tout énoncé ; l’extension le nombre d’entités auxquels il est susceptible de s’appliquer en énoncé. L’applicabilité d’un mot doit, naturellement, être distinguée de son application effective ; c’est à cet effet que Wilmet (1986 : 47) introduit l’extensité, dont la paternité revient à Guillaume (1973), et qui tient compte de l’application effective. L’extensité d’un substantif ou d’un syntagme nominal est donc le nombre d’entités auxquels il est effectivement appliqué dans un énoncé donné. Ainsi, dans l’énoncé « 51 % des français ont voté Mitterrand en 1981 », l’extensité de « 51 % des français » rend-elle compte du nombre des français ayant réellement participé au scrutin, et déposé un bulletin Mitterrand dans l’urne à cette occasion, et non des nombres que représentent 51 % de la population française en âge de voter (mieux décrit par l’extension) ou de la population française sans restriction (mieux décrit par l’extensionalité). Wilmet (1986) introduit également d’autres concepts utiles à une description complète de la quantification dénombrable ; étant donné que nous ne nous en servirons pas ici, nous renvoyons le lecteur à son bel ouvrage pour plus de détails. Dans la mesure, aussi, où les concepts introduits par Wilmet décrivent stricto sensu la quantification dénombrable, il faudrait pour bien faire parler d’extensionalité, extension et extensités dénombrables. Passons à présent à un prototype de quantification non-dénombrable : la quantification massive. Pour donner un cadre à une description de la quantification massive, Kleiber (1989 : 77–78) note que « l’extension d’un N (ou de N+Modificateur) intrinsèquement massif, tel que sable, par exemple, est, au

Les emplois des « quantifieurs »

53

contraire, perçue comme homogène : elle rassemble des occurrences non différenciables et ne peut être assimilée par là-même à une classe. À quoi correspond-elle alors ? Nous postulerons qu’il s’agit d’une entité individuelle constituée de l’ensemble de ses occurrences non discernables : . . . sable a pour extension le tout formé par la réunion des occurrences (non distinguables) qui sont du sable ». On pourrait parler (c’est le choix que nous avons fait dans notre mémoire de Master (2ème année) dont ce texte est une suite, voir Moonens 2013) d’extensionalité, d’extension et d’extensité volumiques et étendre de cette façon aux noms massifs la description proposée par Wilmet (1986) de la quantification.

2.2 La quantification en latin Pour le cas spécifique du latin, plusieurs auteurs ont proposé des descriptions de phénomènes liés à la quantification. La quantification « logique » (existentielle et universelle) a bénéficié d’une étude très complète par Bertocchi, Maraldi et Orlandini (2010) ; souvent laissée de côté par les grammaires traditionnelles, la catégorie des « quantifiants indéfinis » (dont relèvent les expressions que nous nous proposons d’étudier) est traitée par Touratier (1994 : 55–72) au sein d’une section consacrée aux pronoms indéfinis. Dans son étude des déterminants (ACF), Fugier (1983) traite également des « quantitatifs » et de leurs propriétés de compatibilité – description à laquelle Risselada (1984) a apporté plusieurs modifications affectant aussi la délimitation des quantifieurs. Finalement, dans son étude récente du syntagme nominal, Spevak (2014) est revenue sur les présentations de Fugier (1983) et Risselada (1984), et a proposé un schéma plus complexe de la combinatoire des déterminants, dont les quantifieurs forment une sous-catégorie : ce sont des « modifieurs » du nom dont il est possible de dresser une typologie. Nous nous appuierons, dans la suite, sur les travaux de Wilmet (1986), Kleiber (1989) et Spevak (2014) pour déterminer dans quelle mesure les termes concernés par notre étude (quantus, quot, quam multi et leurs corrélatifs) participent à la quantification.

3 Choix méthodologiques S’agissant de la méthode, nous avons souhaité travailler sur un corpus de textes allant de Plaute à Sénèque, afin notamment de mesurer l’importance éventuelle

54

Laurent Moonens

de l’évolution de la langue sur les emplois des expressions étudiées. Plus précisément, nous avons utilisé, pour Plaute, César, Cicéron et Sénèque (le Jeune), des textes issus de la base de données du LASLA Opera Latina1 ; pour Tite-Live, nous avons utilisé la Latin Library of Texts (LLT-A) de Brepols,2 en nous arrêtant au livre XXXIX pour le singulier. Les traductions des passages cités (nous adoptons les conventions de numérotation du Thesaurus Linguae Latinae) sont par ailleurs, sauf mention explicite du contraire, tirées de la Collection des Universités de France (CUF). Ce corpus représente, pour information, environ 800 occurrences du sg. tantus/quantus, et environ 500 occurrences du pl. quanti-quot/tanti-tot. Dans la mesure où l’étude de la paire quot/tot pose moins de problèmes, commençons par celle-ci l’examen de notre corpus.

4 La paire quot/tot et la quantification numérique ; le statut de quam multi Dans ses emplois, la paire tot/quot semble n’imposer qu’une seule restriction au nom qu’elle accompagne en syntagme nominal : celle d’apparaître au pluriel en représentation dénombrable. Comme en témoignent les exemples suivants, elle ne fait pas de distinction, au sein de ces substantifs, entre les noms concrets, abstraits, animés ou inanimés : (1)

neque enim . . . alia de causa tot tantasque classis paratas. (CAES. civ. 1, 85, 6) ‘point d’autres raisons, en effet, . . . à l’armement de tant de flottes si considérables.’ (trad. Fabre)

(2)

Erat noua et inusitata belli ratio cum tot castellorum numero tantoque spatio et tantis munitionibus et toto obsidionis genere tum etiam reliquis rebus. (CAES. civ. 3, 47, 7) ‘C’était une guerre toute nouvelle d’un aspect inusité, tant à cause du grand nombre de fortins, de l’étendue de la ligne, de l’importance des

1 La base de données Opera Latina du LASLA est consultable sur la page qui lui est dédiée sur le site web de l’Université de Liège cipl93.philo.ulg.ac.be/OperaLatina/. 2 La base de données Library of Latin Texts – Series A éditée par Brepols est consultable en ligne via la plupart des accès institutionnels ; voir about.brepolis.net/databases/latin/libraryof-latin-texts/.

Les emplois des « quantifieurs »

55

retranchements, de toute l’allure du siège que de ses autres caractères.’ (trad. Fabre) (3) Intellexistis quot res et quantas xuiri legis permissu uendituri sint. (CIC. leg. agr. 2, 47) ‘vous connaissez le nombre et l’importance des domaines que vendront les décemvirs en vertu de la loi.’ (trad. Boulanger) (4) capras et oues quot quisque haberet, dicere posse ; amicos quot haberet, non posse dicere. (CIC. Lael. 62) ‘chacun . . . peut dire combien il a de chèvres ou de brebis ; il ne peut dire combien il a d’amis.’ (trad. Laurand) Le paramètre qu’interroge quot et sur lequel agit tot, est l’extensité de Wilmet (1986). À la lumière toutefois de l’exemple (4) ci-dessus, et au vu des vers suivants tiré des Métamorphoses d’Ovide : (5) Hoc pecus omne meum est ; multae quoque uallibus errant, / Nec, si forte roges, possim tibi dicere, quot sint. (OV. met. 13, 821–823) ‘Tout ce troupeau m’appartient et je possède bien d’autres bêtes qui errent dans les vallées . . . ; si tu m’en demandais le nombre, je ne saurais te le dire.’ (trad. Lafaye) on aurait envie de dire que la paire quot/tot agit plus précisément sur l’extensité numérique, c’est-à-dire sur le nombre (précis) d’entités auxquelles le syntagme qu’il forme avec un substantif est appliqué en énoncé ; c’est ce que semble également indiquer le syntagme tot castellorum numero cité ci-dessus (voir l’exemple 2), où le nombre est explicitement mentionné. Cette thèse peut être renforcée par un fait linguistique qu’il nous semble pertinent de mentionner ici. À côté de quot/tot, on rencontre en effet une autre paire de modifieurs du nom (la paire quam/tam multi) agissant sur l’extensité d’une façon plus vague ; on pourrait dire, suivant la terminologie de Bertocchi et Maraldi (2010), que quam multi appelle, et que tam multi agit comme, un mid-scalar quantifier. Cette paire, qui se rencontre dans notre corpus presqu’exclusivement au pluriel,3 ne fait pas non plus de distinction entre entités concrètes, abstraites, animées ou inanimées :

3 On rencontre quelquefois quam/tam multum avec multum adverbial.

56

Laurent Moonens

(6) Ere, si ego taceam seu loquar, scio scire te / Quam multas te cum miserias mulcauerim. (PLAVT. Stich. 419–20) ‘Maître, que je me taise ou que je parle, je sais que tu sais combien j’ai enduré avec toi de misères.’ (trad. Ernout) (7) quam multis istum putatis hominibus honestis de digitis annulos abstulisse ? (CIC. Verr. II 4, 57) ‘À combien d’honnêtes gens croyez-vous que notre homme ait arraché des doigts des anneaux d’or ?’ (trad. Rabaud) Elle paraît donc combler l’espace laissé vide par quot/tot dans l’interrogation et l’expression de la grande quantité vague en représentation dénombrable ; peut-être faudrait-il, de ce fait, intégrer quam multi ? parmi les questions définissant la notion de quantifieur – au risque, sinon, d’empêcher les mid-scalar quantifiers de Bertocchi et Maraldi (2010) de figurer parmi nos quantifieurs. Venons-en à présent à l’étude de la paire tantus/quantus, en commençant par ses occurrences au singulier.

5 Le cas du singulier quantus et de son corrélatif tantus L’examen des occurrences de quantus et tantus en syntagme nominal dans notre corpus suggère une classification de leurs emplois en un petit nombre de catégories (inégales en importance numérique). La première, qui compte très peu d’exemples au sein de notre corpus, contient des occurrences de la paire tantus/quantus à côté de noms « massifs » en représentation non-dénombrable (tels aurum). Dans une seconde catégorie d’emplois, on rencontre tantus/quantus à côté de substantifs « physiques » faisant référence à des entités du monde dont on peut mesurer la grandeur ou l’étendue (la longeur, la surface ou le volume) – à ces emplois se rattachent naturellement les occurrences de quantus/tantus à côté de substantifs indicateurs d’étendue (tels magnitudo, spatium, molis, etc.). La troisième catégorie rassemble les occurrences de quantus/tantus accompagnant les substantifs que nous avons appelés dans un précédent travail « noms de patrimoine » (et que l’on peut à certains égards compter parmi les noms massifs, avec lesquels ils partagent des caractéristiques sémantiques et grammaticales), parmi lesquels figurent, à côté du nom-modèle pecunia, des

Les emplois des « quantifieurs »

57

substantifs tels aes alienum, merx, hereditas, praeda, etc. Dans la dernière catégorie, de loin la plus importante en nombre d’occurrences, on rencontre quantus/tantus à côté de noms abstraits.

5.1 Les emplois de quantus/tantus à côté de noms massifs en représentation non-dénombrable Important pour autoriser le classement parmi les quantifieurs de certains modifieurs d’un nom massif (rappelons que, pour nous, un modifieur n’est appelé quantifieur que s’il est susceptible de répondre à la question quot ? ou quantus ?), l’emploi de quantus (et tantus) à côté de noms massifs est plutôt rare dans notre corpus ; marginal donc, on le rencontre par exemple à côté du nom massif prototypique aurum, comme dans l’extrait suivant tiré du corpus plautinien : (8) Si me illoc auro tanto circumduxerit. (PLAVT. Bacch. 311) ‘s’il me soufflait tout cet or-là.’ (trad. Ernout) Dans ces emplois, le paramètre de la description linguistique que quantus interroge (resp. sur lequel tantus agit) dans le substantif qu’il accompagne, est l’extensité volumique ; à cet égard, quantus et tantus dans ce type d’emplois participent donc d’une quantification au sens de Kleiber (1989) ou de l’adaptation au contexte massif que nous avions proposée du cadre théorique de Wilmet (1986).

5.2 Les emplois de quantus/tantus à côté de substantifs « physiques » On peut décrire l’emploi « physique » de quantus/tantus comme accompagnant un substantif concret, généralement en représentation dénombrable (qu’il apparaisse ou non au pluriel) et occupant une certaine étendue (grandeur, taille, surface, volume, longueur), que l’on interroge (pour quantus) ou qualifie (pour tantus), ou par un substantif abstrait indicateur d’une telle étendue (ex. : interuallum, spatium, molis, magnitudo, etc.). Un exemple manifeste de cet emploi de quantus se rencontre dans l’extrait suivant de l’Heautontimoroumenos, où quantam fenestram fait référence à la taille d’une brèche (imagée) :

58

Laurent Moonens

(9) Huic quantam fenestram ad nequitiem patefeceris ! (TER. Haut. 481) ‘Quelle brèche tu lui auras ouverte vers la dissipation !’ (trad. Marouzeau) C’est aussi cette catégorie d’emplois de quantus/tantus qui est en situation dans notre corpus à côté de noms collectifs, et notamment dans l’extrait suivant de la guerre des Gaules où l’importance de la cavalerie revient à qualifier l’espace qu’elle est susceptible d’occuper : (10) ac saepe in eum locum uentum est tanto in omnis partis diuiso equitatu, ut . . . (CAES. Gall. 6, 43, 4) ‘souvent, il y arriva avec une si nombreuse cavalerie battant le pays dans tous les sens, que . . .’ (trad. d’après Constans) on la rencontre encore dans l’extrait suivant tiré de l’Histoire romaine de TiteLive : (11) quantas autem Hannibal copias duceret aut quanto exercitu ad Capuam obsidendam opus esset, ipsos scire. (LIV. 26, 8, 6) ‘mais combien de soldats conduisait Hannibal ou de quels effectifs ils avaient besoin pour assiéger Capoue, c’est eux qui le savaient.’ (trad. Jal) L’exemple du collectif populus montre bien comment on peut passer, avec ce type de substantifs, d’une indication d’étendue à celle d’une importance plus imagée : (12) Tam paruus animus tanti populi est, ut . . . (LIV. 6, 18, 10) ‘Un si grand peuple a donc si peu de cœur, qu’il . . .’ (trad. Bayet) On la trouve également dans l’extrait suivant où quantus accompagne un substantif indicateur d’étendue : (13) uidete quantum interuallum sit interiectum inter maiorum nostrorum consilia et inter istorum hominum dementiam ! (CIC. leg. agr. 2, 89) ‘voyez quelle distance sépare la sagesse de nos ancêtres de la démence de ces hommes !’ (trad. Boulanger) On rattachera finalement à ces emplois de la paire quantus/tantus quelques occurrences à côté de noms abstraits tels tempus ; ces derniers, par leur nature, se rattachent en effet dans ces emplois à des indicateurs d’étendue (le temps est ainsi mesuré) :

Les emplois des « quantifieurs »

59

(14) Non id ago nunc hac diligentia, ut ostendam, quantum tempus apud grammaticum perdiderim. (SEN. epist. 58, 5) ‘Si j’apporte une telle minutie à ces citations, ce n’est pas que je cherche à montrer tout ce que j’ai perdu de temps chez le grammairien.’ (trad. Noblot) Un substantif concret en représentation dénombrable très particulier, uir, se rencontre également qualifié par quantus/tantus (surtout dans la partie la plus tardive de notre corpus, i.e. chez Tite-Live et Sénèque) ; il relève plutôt, selon nous, de cette catégorie d’emplois, même s’il y occupe une place singulière puisque l’on ne peut pas, à l’inverse des cas précédents, arguer que c’est uia sa taille ou son étendue strictement physiques qu’un homme est appréhendé par quantus/tantus ; il s’agit plutôt ici d’insister sur l’épaisseur intellectuelle ou symbolique (quantus agissant ici comme qualificatif) : (15) quantus uir fuit, qui in admirationem sui adduxit hominem etiam Catoni suo mirandum ! (SEN. dial. 11, 9, 5) ‘quelle grandeur que celle d’un homme . . . qui forçait l’admiration d’un héros lui-même admiré de son cher Caton !’ (trad. Waltz) Dans toutes ces occurrences, quantus n’interroge pas (et tantus n’agit pas sur) les paramètres de la quantification au sens de Wilmet (1986) ; il agit sur un paramètre « physique » (l’étendue) de la réalité à laquelle réfère le substantif qu’il accompagne. Si on ne peut pas, de ce fait, parler de quantification en ce sens, il faut – si l’on se tient à notre définition – accepter pourtant de classer parmi les quantifieurs, les modifieurs d’un nom de ce type qui agissent sur ce paramètre (et répondent donc à la question quantus ?) ; ainsi magnus, indicateur par excellence d’une grande étendue (vague) avec un nom « physique » en représentation dénombrable, est-il quantifieur dans les exemples suivants : (16) cum prima legio in castra uenisset reliquaeque legiones magnum spatium abessent. (CAES. Gall. 2, 17, 2) ‘lorsque la première légion arriverait au camp et que les autres en seraient loin.’ (trad. Ozanam) (17) ipse ille C. Marius cuius diuina atque eximia uirtus magnis populi Romani luctibus funeribusque subuenit. (CIC. prov. 32) ‘le grand C. Marius lui-même, dont la valeur divine et supérieure sut remédier aux grands malheurs et aux grandes pertes du peuple romain.’ (trad. Cousin)

60

Laurent Moonens

(18) et spolia plurima Gallica fuere, aurei torques armillaeque, magnus numerus. (LIV. 24, 42, 8) ‘quant aux dépouilles, elles furent pour la plupart gauloises, colliers et bracelets d’or, en grand nombre.’ (trad. Jal)

5.3 Les emplois de quantus/tantus à côté des « noms de patrimoine » Nous avons déjà insisté dans une précédente étude (Moonens 2016) sur l’importance d’une étude spécifique des noms de patrimoine (tels pecunia, hereditas, etc.) et des substantifs qui s’y rattachent naturellement (tels pretium) dans le phénomène de la quantification. Présentant la particularité d’être généralement accompagnés par magnus plutôt que par multus dans l’expression de la grande quantité vague, ils se distinguent ainsi des noms massifs avec lesquels ils partagent pourtant les traits de désigner des entités concrètes et de relever de la représentation non-dénombrable ; à cet égard, on peut donc considérer les « noms de patrimoine » comme une sous-catégorie des « noms massifs ». Eu égard au phénomène de quantification cependant, ils s’en distinguent par le fait que, si l’apparition de ces substantifs avec quantus (et tantus) atteste la possibilité de les voir accompagnés d’un quantifieur, l’expression de la grande quantité « vague » se fait (quasiment) exclusivement à l’aide de magnus et non de multus : au sein de ce phénomène de quantification particulier, cette catégorie de noms se rapproche donc d’une catégorie « intermédiaire » entre noms massifs (tels aurum, pour lesquels aucune échelle d’étendue n’est envisageable) et noms « physiques » (porteurs d’une échelle d’étendue intrinsèque). C’est pour cette raison que nous choisissons de traiter la traiter séparément ici. Très bien représentée dans notre corpus, à toutes les époques, nous renvoyons à (Moonens 2016), où nous l’avons étudiée avec plus de détails.

5.4 Les emplois de quantus/tantus à côté de noms abstraits Les emplois de la paire quantus/tantus à côté de noms abstraits sont les plus représentés dans notre corpus – faisant de fait entrer un certain nombre de modifieurs des substantifs abstraits en question parmi les quantifieurs, bien qu’il soit difficile de faire relever leur action sur ces substantifs d’un phénomène de quantification tel que celui décrit par Wilmet (1986) ou Kleiber (1989).

Les emplois des « quantifieurs »

61

Lexicalement, les noms abstraits que l’on rencontre à côté de la paire quantus/tantus relèvent de quelques catégories. La première regroupe des sentiments ou impressions et épreuves que l’on peut ramener aux sens, et est illustrée par les cas, très fréquents, de terror (dolor, etc.) : (19) causam huius tanti terroris occultas. (CIC. Verr. II 1, 80) ‘la cause d’une situation si terrible, tu la caches.’ (trad. de la Ville de Mirmont) La seconde catégorie comprend les bonnes ou mauvaises actions d’un individu (du crime au bienfait) et leurs conséquences, en rapport avec la notion de jugement moral ; elle est abondante en occurrences de noms tels scelus, facinus, bonum, malum, honos, etc. (20) nomen quidem populi Romani tanto scelere contaminauit ut id nulla re possit nisi ipsius supplicio expiari. (CIC. har. resp. 35) ‘du moins a-t-il souillé le nom du peuple romain par un crime si grand que seul son propre supplice pouvait l’expier.’ (trad. Wuilleumier et Tupet) (21) doce me, quid sit pudicitia et quantum in ea bonum, in corpore an in animo posita sit. (SEN. epist. 88, 8) ‘Enseigne-moi ce que c’est que la chasteté et tout le prix de cette vertu, si elle tient au corps ou à l’âme.’ (trad. Noblot) Les noms de la troisième catégorie désignent des capacités (intellectuelles ou physiques) ou dispositions d’un individu ou d’un objet ; le nom uis y occupe une place centrale : (22) illae celeriter procucurrerunt infestisque signis tanta ui in Pompei equites impetum fecerunt ut . . . (CAES. civ. 3, 93, 6) ‘Ces troupes s’élancèrent aussitôt en avant et firent en colonnes d’attaque une charge si vigoureuse contre les cavaliers de Pompée, que . . .’ (trad. Fabre) (23) Vide quanta uis honesti sit. (SEN. epist. 76, 27) ‘vois jusqu’où l’honnête étend sa puissance.’ (trad. Noblot)

62

Laurent Moonens

Enfin, on trouve quantus/tantus à côté de substantifs exprimant un état ou un événement (naturel, social, etc.) susceptible d’être ou de survenir. À côté de pax, bellum, tumultu, etc., on y rencontre un très grand nombre d’occurrences de periclum : (24) Quanto in periclo et quanta in pernicie siet. (PLAVT. Bacch. 827) ‘en quel péril, en quel danger mortel [il] se trouve.’ (trad. Ernout) (25) quantum periculum ab illis populis fuerit . . . (LIV. 3, 19, 12) ‘ce que nous avions à redouter de ces deux peuples . . .’ (trad. Baillet) Ces catégories – dont la délimitation est forcément quelque peu artificielle – épuisent l’essentiel des emplois de quantus/tantus avec des noms abstraits. On notera qu’au-delà de la classification proposée de ces emplois de quantus/tantus à côté de noms abstraits, la propriété caractéristique des substantifs abstraits qu’ils accompagnent est de désigner une entité abstraite d’intensité variable ; c’est ce paramètre susceptible de gradation qui, dans les substantifs appartenant aux catégories précédentes, est alors interrogé ou qualifié par quantus/tantus dans les syntagmes qu’ils forment avec ces deux termes. On notera au passage qu’à l’instar de ce qui se produit avec les « noms de patrimoine », c’est en très grande majorité par magnus (et non par multus) que ces substantifs abstraits sont modifiés, lorsqu’il s’agit d’exprimer que le paramètre d’intensité est élevé dans les entités abstraites auxquelles ils font référence ; par définition, ces emplois de magnus (susceptibles d’être interrogés par quantus, comme nous l’avons vu), sont donc quantifieurs : (26) quem si consulem illa prouincia putaret neque eum reciperet, magno scelere se adstringeret. (CIC. Phil. 4, 9) ‘si cette province, tout en le considérant comme un consul, refusait de l’accueillir, elle se rendrait coupable d’un grand crime.’ (trad. Boulanger et Wuilleumier) (27) contione dimissa magnus terror Vrbe tota fuit. (LIV. 39, 17, 4) ‘après la dispersion de l’assemblée, une grande panique envahit toute la ville.’ (trad. Adam) (28) in his magno periculo erramus. (SEN. epist. 45, 7) ‘est en pareille matière . . . il y a grand péril à se tromper.’ (trad. Noblot)

Les emplois des « quantifieurs »

63

5.5 Deux emplois difficile à classer ; une action sur l’extensité dénombrable Il reste deux emplois fréquents difficiles à classer du singulier quantus/tantus ; il s’agit des occurrences apparaissant à côté du nom à large spectre sémantique res (qui peut jouer le rôle d’un nom de patrimoine, ou, comme ci-dessous, faire figure de substantif abstrait), et de l’emploi – principalement, dans notre corpus, chez Tite-Live et Sénèque – de quantus/tantus comme génitif de prix ou d’estimation : (29) magna uis frumenti ex Sicilia aduecta, agitatumque in senatu quanti plebi daretur. (LIV. 2, 34, 7) ‘il y eut un gros arrivage de blé de Sicile, et l’on examina au sénat à quel prix on le céderait à la plèbe.’ (trad. Baillet) (30) est tanti uulnus suum premere et patibulo pendere districtum . . . ? est tanti habere animam ut agam ? (SEN. epist. 101, 12) ‘est-ce la peine de peser sur sa plaie, de pendre au gibet les bras étendus . . . ? – “est-ce la peine que je garde le souffle pour le rendre ?”’ (trad. Noblot) Il faut également mentionner la participation, lexicalement conditionnée, de quantus/tantus à la quantification dénombrable, en combinaison avec le substantif numerus et le génitif pluriel d’un substantif en représentation dénombrable : (31) Tum meretricum numerus tantus quantum in urbe omni fuit. (PLAVT. Epid. 213) ‘et puis toutes les courtisanes que la ville peut compter couraient.’ (trad. Ernout) Ce phénomène d’expression de l’extensité dénombrable avec un génitif pluriel est susceptible de se rencontrer en présence d’autres substantifs, tels multitudo, ou uis dans certains de ses emplois, comme en témoigne l’exemple suivant, où magna uis telorum indique sans ambiguïté une action sur l’extensité dénombrable : (32) simul ex minoribus nauibus magna uis eminus missa telorum. (CAES. civ. 2, 6, 3) ‘en même temps, les bâtiments les plus petits tiraient de loin une grand quantité de projectiles.’ (trad. Fabre)

64

Laurent Moonens

On notera au passage qu’au singulier, on rencontre également l’expression uis + génitif (sg.) pour exprimer la quantification massive – ce que montre l’extrait (29) ci-dessus (magna uis frumenti). Il convient à présent de passer quelque temps sur la paire plurielle quanti/ tanti, que nous avons jusqu’ici laissée de côté.

6 La paire plurielle quanti/tanti Parmi les emplois du pluriel quanti/tanti, il convient de distinguer ceux qui accompagnent des pluralia tantum (parmi lesquels on trouve des noms de patrimoine tels diuitiae, manubiae, etc., des noms abstraits tels minae et dont on rapprochera aussi le pluriel copiae, agissant comme nom « physique », ou encore le pluriel opes agissant comme « nom de patrimoine »). Dans ces occurrences, en effet, on peut analyser l’action de la paire quanti/tanti comme parallèle à l’action exercée par quantus/tantus sur les noms singuliers issus des catégories correspondantes, et déjà analysée ci-dessus (voir la Section 4) ; c’est de cette façon qu’on analysera l’exemple (11) ainsi que ceux-ci : (33) Caesaris copiae nequaquam erant tantae, ut . . . (CAES. civ. 3, 109, 2) ‘les troupes de César n’étaient pas encore du tout assez nombreuses pour . . .’ (trad. Fabre) (34) quantum nomen illius fuerit, quantae opes, quanta in omni genere bellorum gloria, quanti honores populi Romani, quanti senatus, quanti tui, quis ignorat ? (CIC. Deiot. 12) ‘la grandeur de son nom, sa puissance, ses succès militaires de toutes sortes, les distinctions qu’il reçut du peuple romain, du Sénat, de toi-même, qui ne les connaît ?’ (trad. Lob) (35) Quae ferarum immanitas . . . tantis . . . feruebit minis ? (SEN. Med. 408–10) ‘Quels fauves cruels . . . bouillonneront de si violentes menaces ?’ (trad. Chaumartin) Dans un cas comme le suivant : (36) Quantae autem copiae peditum equitumque additae sint, adeo et numero et genere copiarum uariant auctores . . . (LIV. 22, 36, 1)

Les emplois des « quantifieurs »

65

‘mais, sur les chiffres et la nature des contingents supplémentaires pour l’infanterie et la cavalerie, il y a divergence chez les historiens . . .’ (trad. d’après Flobert) on peut même soutenir que l’on a affaire à un substantif en représentation dénombrable, avec action de quantae sur le paramètre d’extensité (c’est, en effet, ce que semble indiquer numero copiarum dans la suite de la phrase). L’analyse des occurrences de quanti/tanti à côté de substantifs pluriels en représentation dénombrable (hors des pluralia tantum) requiert un soin particulier ; on mentionne souvent, en effet, l’exemple du syntagme quanta milia chez Properce : (37) Quod si forte tuis non est contraria uotis, / at tibi curarum milia quanta dabit ! (PROP. 1, 5, 9–10) ‘Et si d’aventure elle n’est pas opposée à tes vœux, cependant combien de milliers de soucis elle te donnera !’ (trad. Viarre) comme annonçant la tendance, aboutie dans les langues romanes, de quanti/ tanti à agir sur l’extensité de Wilmet (1986), et donc à participer à la quantification dénombrable. Il semble donc utile d’examiner les occurrences du pluriel quanti/tanti dans notre corpus, à l’aune de la question suivante : est-il possible d’y déceler une participation de ces termes à la quantification dénombrable, et donc à l’action sur l’extensité ? À cet égard, le corpus choisi est plutôt de nature à nous inviter à la prudence : au-delà du fait qu’aucune des occurrences du pluriel quanti/tanti que l’on y rencontre ne peut s’interpréter positivement comme agissant exclusivement sur l’extensité du substantif accompagné, plusieurs raisons grammaticales semblent par ailleurs s’opposer à une interprétation quantitative du pluriel quanti/tanti avec un substantif en représentation dénombrable. On observera pour commencer que l’on rencontre très souvent la paire quot/tot coordonnée avec la paire quanti/tanti, montrant par là que quanti/tanti n’agit vraisemblablement pas, dans ces occurrences, sur l’extensité du substantif accompagné (reprenant simplement, dans un contexte pluriel, l’action réalisée par quantus/tantus au singulier dans le syntagme nominal correspondant, et déjà décrite ci-dessus) ; c’est le cas dans tot tantasque classis et dans quot et quantas res ci-dessus (voir les exemples (1) et (3) ci-dessus) ; on trouve ce phénomène à de nombreuses autres reprises : (38) Pro his tantis totque uictoriis uerum esse grates deis immortalibus agi haberique. (LIV. 23, 11, 12)

66

Laurent Moonens

‘Pour ces victoires, si importantes et si nombreuses, il est juste de témoigner aux dieux immortels de la reconnaissance.’ (trad. Jal) (39) Nam quae iste in Apollonio fecit, ea primum breuiter cognoscite quot et quanta sint, deinde haec expendite atque estimate pecunia. (CIC. Verr. II 5, 23) ‘Car les abus qu’il a commis au sujet d’Apollonius, apprenez d’abord en peu de mots quelle en est la quantité et l’importance ; puis pesez ces faits, évaluez en argent.’ (trad. Rabaud) L’exemple (39) en particulier est intéressant, en ce qu’il montre que, s’agissant d’un pillage (nom abstrait d’intensité graduable), on peut répondre à la question quantus ? en termes d’une somme d’argent (pecunia). Un examen attentif de l’ensemble des substantifs modifiés par quanti/tanti dans notre corpus montre aussi que, hormis le cas des pluralia tantum déjà mentionné, ceux-ci appartiennent aux mêmes catégories lexicales que les substantifs en représentation dénombrable accompagnés par quantus/tantus au singulier ; nombreux sont d’ailleurs ceux qui apparaissent fréquemment au singulier et au pluriel ; c’est notamment le cas dans les exemples suivants : (40) Neque istic in tantis periclis umquam committam ut siet. (PLAVT. Aul. 450) ‘Je ne veux plus la laisser exposée ici à de si grands dangers.’ (trad. Ernout) (41) uiri, quantas pecunias ab uxoribus dotis nomine acceperunt, tantas ex suis bonis aestimatione facta cum dotibus communicant. (CAES. Gall. 6, 19, 1) ‘les hommes, en se mariant, mettent en communauté une part de leurs biens équivalant, d’après estimation, à la somme d’argent apportée en dot par les femmes.’ (trad. Constans) (42) quo propius spem uenerant tantis duobus ducibus exercitibusque caesis debellatum ibi ac pulsos inde Romanos esse . . . (LIV. 26, 37, 8) ‘plus ils s’étaient rapprochés de l’espoir de voir, après le massacre de deux si grands généraux et de leurs armées, la guerre terminée là-bas et les Romains chassés du pays, plus . . .’ (trad. Jal) (43) refert quantas uires quamque integras habeat. (SEN. dial. 5, 1, 1) ‘il faut voir si son énergie est grande et intacte.’ (trad. Bourgery) Ces exemples plaident donc pour une interprétation où quanti/tanti n’agit pas sur le paramètre d’extensité des substantifs qu’il accompagne ; il s’agit simplement de

Les emplois des « quantifieurs »

67

la mise au pluriel du syntagme singulier quantus/tantus N correspondant, où le rôle de quantus s’analyse donc de la même façon que dans le syntagme singulier. On notera au passage que, dans l’exemple (42) ci-dessus, tantis peut difficilement s’interpréter comme porteur d’une action sur l’extensité dans le syntagme tantis duobus ducibus : on imagine mal comment pourrait se justifier un indicateur de grande extensité à côté du nombre deux ! Il faut donc conclure, en tout cas pour ce qui concerne notre corpus, que prêter à quanti/tanti, une action sur l’extensité (dénombrable) du substantif qu’il accompagne, est très difficile à soutenir hors du cadre des pluralia tantum ; l’exemple (37) ci-dessus, tiré des Élégies de Properce, semble donc devoir s’analyser à l’instar de ces derniers, et ne pas permettre d’en dégager une tendance de quanti à participer à la quantification dénombrable.

7 Conclusion Nous avons essayé, dans le présent texte, de dégager une classification des principaux emplois de quantus, quot et quam multi (et de leurs corrélatifs tantus, tot et tam multi) en syntagme nominal dans un corpus de textes latins classiques allant de Plaute à Sénèque – étude que notre définition du terme « quantifieur » rendait indispendable. Dans chacun de ces types d’emploi, nous nous sommes efforcé de déterminer dans quelle mesure ces termes agissent sur les paramètres de la description linguistique ayant trait à la quantification. Au-delà des cas où cette action est effective, nous avons noté que la grande majorité des occurrences de quantus/ quot (resp. tantus/tot) accompagne des substantifs abstraits en représentation graduable, avec action de ces termes sur l’échelle d’intensité correspondante. Il est aussi apparu, au fil de cette étude, que l’emploi du pluriel quanti/tanti pour exprimer une quantification dénombrable, attesté chez Properce dans un vers souvent cité, ne peut pas être positivement attesté dans notre campus hors du cas des pluralia tantum.

Bibliographie CUF = Collection des Universités de France: www.lesbelleslettres.com/les-budes (dernier accès: 22 février 2019). LASLA = Laboratoire d’analyse statistique des langues anciennes, Opera latina: cipl93.philo. ulg.ac.be/OperaLatina/ (dernier accès: 22 février 2019). LLT-A = Library of Latin texts – Series A, Brepolis: about.brepolis.net/databases/latin/libraryof-latin-texts (dernier accès: 22 février 2019).

68

Laurent Moonens

Bertocchi, A. & Maraldi, M. (2010). Mid-scalar quantifiers in Latin. Dans P. Baldi & P. Cuzzolin (Eds.), New perspectives on historical Latin syntax: vol. 3, constituent syntax (pp. 19–173). Berlin/New York: Mouton de Gruyter. Bertocchi, A., Maraldi, M. & Orlandini, A. (2010). Quantification. Dans P. Anreiter & M. Kienpointner (Eds.), Latin linguistics today: Akten des 15. Internationalen Kolloquiums zur Lateinischen Linguistik, Innsbruck, 4.–9. April 2009 (pp. 193–206). Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck. Fugier, H. (1983). Le syntagme nominal en latin classique. Dans W. Haase (Ed.), Aufstieg und Niedergang der römischen Welt (vol. 2, 29, 1, pp. 212–269). Berlin/Boston: De Gruyter. Guillaume, G. (1973). Principes de linguistique théorique de Gustave Guillaume. Paris/ Québec: Klinksieck/Presses de l’Université de Laval. Kleiber, G. (1989). L’article LE générique. La généricité sur le mode massif. Genève/Paris: Droz. Moonens, L. (2013). De quelques expressions de quantité dans la comédie latine. Les cas de multus et magnus. Mémoire de master 2ème année, Université de Toulouse 2, le Mirail (dir. O. Spevak). Moonens, L. (2016). Expressions de quantité chez Plaute et Térence: les « quantifieurs » multus et magnus. Dans O. Spevak (Ed.), Études de linguistique latine I, Pallas 102 (pp. 25–33). Toulouse: Presses universitaires du Midi. Risselada, R. (1984). Coordination and juxtaposition of adjectives in the Latin NP. Glotta 62, 202–231. Spevak, O. (2014). The Noun Phrase in Classical Latin Prose. Amsterdam: Brill. Touratier, C. (1994). Syntaxe latine (Bibliothèque des cahiers de l’Institut de Linguistique de Louvain 80). Louvain-la-Neuve: Peeters. Wilmet, M. (1986). La détermination nominale. Quantification et caractérisation (Linguistique Nouvelle). Paris: Presses Universitaires de France.

Krzysztof Nowak

Tempus mutatur: analysing collocations of tempus ‘time’ with distributional semantic models Abstract: In recent years, distributional semantic models (DSM) built from large corpora have become an important tool in modelling word meaning. They not only have proved to be efficient in a variety of traditional sense-related tasks, but have also been successfully employed in diachronic semantics (Heylen et al. 2008). Although, in order to yield meaningful results, the method does not require painstakingly annotated corpora, it does not seem to have gained major popularity in Classical or Medieval Latin research so far. In the present chapter, we employ DSMs to cluster collocations of the lexeme tempus ‘time’ in two corpora: the Classical Latin Texts collection composed of mostly old and Classical Latin texts (ca. 8M tokens) and the first 60 volumes of the Patrologia Latina (ca. 22M tokens), which comprise Christian writings from the early centuries. In a series of analyses, we show how DSM-based clustering may help in the diachronic study of vocabulary. The analysis shows that, in Latin, time was consequently conceptualised as space, spatium temporis, as was recently showed by Thomas (2011) for Classical Latin. What varies, though, across the two corpora is mostly the depth of the spatial mapping and the nature of the time reference. Keywords: Latin linguistics, diachronic corpus semantics, distributional semantics, conceptual metaphor, ‘Time is space’ metaphor

1 Introduction Investigating lexical phenomena, whether they are single words or multi-word expressions, inevitably involves analysing large quantities of corpus data. However, nowadays, lexicologists and lexicographers can easily use tools that enable the efficient handling of even extensive linguistic evidence. Apart from Acknowledgment: I would like to thank the two anonymous reviewers of this chapter for their helpful comments. Krzysztof Nowak, Institute of Polish Language (Polish Academy of Sciences) https://doi.org/10.1515/9783110647587-005

70

Krzysztof Nowak

traditional forms of corpus inspection (e.g. concordance or on-the-fly generated word lists), collocation candidates and keyword lists, as well as such proprietary tools as Word Sketches (Kilgarriff and Tugwell 2001), are now extensively used. All these tools are particularly useful when a need to summarise raw corpus information arises, and, as may be expected, they are especially beneficial in handling high-frequency phenomena. On the most basic level, the in-built features of popular corpus query systems improve access speed, help to manipulate, sort and reorder linguistic evidence, and provide less or more sophisticated means of quantification of occurrences. However, what seems to be external to linguistic analysis, in fact, contributes to the objectivisation of description, which is a crucial feature in dealing with more subjectivity-prone tasks, such as meaning description or sense clustering, and affects its foundations. If we limit ourselves to vocabulary study only, corpus-based methods, for instance, have drawn our attention to the contextual determinants of meaning.1 Furthermore, as we are able to generalise over a single occurrence, we would be more prone to search for what is universal in language rather than what is anomalous, with the latter becoming sometimes a too important interest of historical lexicography. In this way, both manual and automatic corpus analyses enable the emergence of those language use patterns that would remain hidden in linear analysis. Consequently, they contribute to the redefinition of linguistic concepts, requiring from us, at the same time, to make explicit and to formalise our assumptions. In this study, we demonstrate the preliminary results of an ongoing corpusbased research into the evolution of the metaphors of time between antiquity and Middle Ages. This chapter is concerned with the stability of the collocational profile of the high-frequency lexeme tempus ‘time’ within and between two corpora, namely the general corpus of Latin literature from the second century BC to the second century AD and the collection of patristic authors of the first four centuries of the Christian era. More than focusing on the differences concerning single lemmas, we examine the changes that affected entire collocational clusters. To make the semantic relations between co-occurring words explicit, we use distributional semantic models, which proved efficient in representing and modelling word meaning. In this way, we hope to detect the emergence of a new means to conceptualise time.

1 This impact was perhaps less perceived in traditional historical lexicography and lexicology which, as Geeraerts (2009: 169) rightly observes, have always been inherently empirical and sensitive to context.

Tempus mutatur

71

2 Distributional semantics and Latin corpora 2.1 From word’s company to meaning with distributional semantic models The Firthian conception of the word’s sense turned out to be a major inspiration for NLP work in meaning modelling. The distributional hypothesis (Sahlgren 2008), which is often illustrated with Firth’s famous adage You shall know a word by the company it keeps, boils down to the claim that the linguistic context of a word is indicative of its meaning. It is central to what is baptised distributional semantics and its implementations, often referred to as distributional semantic models (DSM), which have gained major popularity in recent years. The latter umbrella term, in fact, covers an impressive variety of computational methods and techniques. They are all based on the same assumption, namely that from a sufficiently large textual corpus we can build a representation of its vocabulary by quantifying those distributional properties of the words that we believe to account best for their meaning. Therefore, whether we call them word spaces, distributional semantic models, or any other name, the underpinning assumption remains the same,2 and what changes are more or less crucial details of implementation. In the case of the word’s meaning3 that we believe to be determined and recoverable from the co-text, we may wish, for instance, to register for each word in the corpus how many times it co-occurred with any other word within some pre-defined textual window. However, the context may be operationalised not only as a single word but also as a paragraph or an entire document. The words we count may be purely empirical collocations (Evert 2009), or we may require that they stand in some syntactic relation to the node word. Whatever be our choice, this procedure results in the collection of a large matrix of word vectors (cf. Table 1).

2 Sahlgren (2006: 17) acutely observes that “[t]he propensity for term-coining in this area of research is not only a major source of confusion, but a symptom of the theoretical poverty that permeates it. The single most important reason why researchers do not agree upon the terminology is because they fail to appreciate the fact that it is the same underlying ideas behind all their implementations.” 3 We follow Harispe et al. (2015: Ch. 2) who offer a recent survey of similarity measures (incl. knowledge-based methods). In this paragraph, we do not provide a rigorous description of the method: for an advanced introduction, see Turney and Pantel (2010). A more accessible introduction to the theory that stands behind the research in distributional semantics may be found in Widdows’ (2004) book.

72

Krzysztof Nowak

Table 1: Sample frequency matrix built from a corpus: words (in rows) are characterised in terms of their co-occurrence with all or selected words in the same corpus (in columns).

liber uolumen bellum pugna

sacer

lego

ciuilis

gero

infero

   

   

   

   

   

It should next be subject to several manipulations which handle two main problems that frequency matrices extracted from corpora usually suffer from. The first is their sparsity (most cells would contain zero) and the second (related to the former) is the poor informativeness of such data (high-frequency function words would usually be less informative about the word’s sense than relatively less frequent content terms). Numerous techniques were proposed to reduce the sparsity as well as to weight raw frequency counts. Once normalised, linear algebra provides us now with a set of measures that allow us to gauge the distance between word vectors which, to make a long story short, is believed to be a proxy for semantic relatedness and, more specifically, the semantic similarity of two or more terms (cf. Figure 1).4

uolumen liber

bellum

pugna

Figure 1: Representing words as numeric vectors allows to reformulate the problem of their semantic similarity in terms of the ‘mathematical’ distance between vectors.

4 See, e.g., Baroni and Lenci (2010: 674) for the distinction between attributional and relational similarity.

Tempus mutatur

73

The distributional semantic models have many well-known advantages, which we cannot discuss here in more detail. They have proved successful in such senserelated tasks as word sense discrimination, collocation clustering, word similarity evaluation, and many others (Harispe et al. 2015). Less frequently, though, the distributional semantic models have been applied to tracking semantic change, a task of crucial interest for diachronic study of the lexicon. Important work on both the theoretical and practical level was presented in a series of studies by Heylen, Peirsman, Speelman, and Geeraerts (Heylen et al. 2008; Heylen et al. 2012). However, the DSMs still remain a less known technique, probably because until recently their compilation from corpora and the subsequent computation of distances between word vectors have been all but a trivial task. This is expected to change, however, with the advent of such easy-to-use software tools and packages as Wordspace (Evert 2014).

2.2 Corpora and method For the purposes of the present study two pre-existing corpora were adopted: 1) The PHI corpus comprising of texts composed between the second century BC and the sixth century AD and included in the Classical Latin Texts collection prepared by the Packard Humanities Institute (2017). The corpus is significantly skewed since it contains mostly pre–second century AD texts. Later Latin is represented, among others, by Historia Augusta, Servius, and The Digest. 2) The PL corpus covering the period from the second to early sixth century AD was built from the first sixty volumes of the first series of the Migne’s (1844–1855) Patrologia Latina,5 from which texts of uncertain origin, those by unknown authors, and those that could not be dated were excluded.6 The two corpora differ in many respects (cf. Table 2). Firstly, the PL corpus is almost three times larger than the PHI. Secondly, the PHI corpus consists of a variety of texts composed by mostly pagan authors, while the PL corpus is representative of predominantly Christian writing spanning roughly four centuries. Thirdly, the quality of the editions in the PHI corpus is considerably higher

5 The texts were generously provided to me by Alain Guerreau to whom I offer my thanks. 6 The decision to base analyses on separate corpora rather than on one built by integrating both collections might be questioned. However, it arose as a consequence of the challenges this integration would pose due to the differences between the two corpora in the quality of editions, size, type of texts included, and many other factors.

74

Krzysztof Nowak

Table 2: Corpora used in the study.

No. of tokens No. of lemmas / word forms Time coverage Frequency of tempus

PHI

PL –

≈ M ≈k / k Second century BC–second century AD (cf. above) ≈ k occurrences

≈ M ≈ k / k Second century AD–early sixth century AD ≈ k occurrences

than that of the 19th-century editions of the PL. At this point, though, no modern corpus of Christian Latin literature exists that would lend itself to more advanced statistical analysis. The decision to base our analyses on the comparison of separate corpora might be questioned. In an ideal scenario, we would rather investigate a wellbalanced, representative diachronic corpus of Latin from its beginnings to the Middle Ages. Yet it seems that, owing partly to the insufficient availability of the texts, we are still far from building an equivalent of what the Helsinki Corpus of English Texts is to English studies (Rissanen and Tyrkkö 2013).7 In the present paper, we have therefore decided instead to compare two corpora that seem to embody not only different phases of the history of the Latin language but also different textual genres and topical varieties. Their mechanical integration, apart from concealing their obvious differences, would by no means result in what could be considered a balanced corpus. Texts in both corpora were lemmatised and PoS-tagged using the TreeTagger (Schmid 1994). To that purpose, the set of tagging parameters provided by Gabriele Brandolini (TreeTagger 2017) was employed. The lemmatisation process was fully automatised, a fact that is sometimes revealed by the non-standard lemma strings the reader will certainly notice in some of the following figures. Labels, such as bellis/bellum, were generated due to a partial homophony between both words that could not be resolved by the lemmatiser with certainty. This is, however, less serious than it might seem at first glance because ambiguity would often arise between pairs of high- and low-frequency words: in this specific case, it is bellum (‘war’) that should be the lemma. Since the goal of the present study was by no means to develop new, or improve existing, methods, a basic approach was adopted in both constructing

7 Considerable effort has been expended in recent years to build open corpora for the study of Medieval Latin.

Tempus mutatur

75

and deploying the DSMs. Following loosely the framework adopted by Gulordava and Baroni (2011), we started by generating two word-to-word cooccurrence matrices, each taken from a list of bigrams that had been extracted from the respective corpus. For both matrices, raw frequency counts were weighted with a single log-likelihood measure and log-transformed. The size of the two matrices obtained in this way was next reduced to 300 dimensions by applying the SVD algorithm (Evert 2014). In the present study, similarity between word vectors was computed using the maximum distance score which yielded best results when informally tested against other metrics. Regarding the tools, the bigram lists were extracted with a command line tool, cwb-corpus-scan, which is a part of the Corpus Workbench software (Evert and Hardie 2011). Matrix calculation and transformation and similarity computation were performed with the Wordspace R package developed by Evert (2014). Plots were generated with the ggplot2 R package (Wickham 2016). Lists of collocation candidates were extracted by applying the logDice score (Rychlý 2008) as implemented in the NoSketch Engine (Rychlý 2007) to a wide window of five terms to the left and five to the right of the node word. In the following, the series of analyses that were performed is presented in more detail. First, the relatively stable collocations of the word tempus were extracted from the PHI and then clustered using the DSM generated from the same corpus. Secondly, the analysis was extended to a larger set of nominal collocates of the word. Finally, the results were compared with those obtained from the PL corpus.

3 The PHI: the collocations of tempus 3.1 The PHI corpus: stable collocations For each century that the PHI corpus covers, namely the second century BC to the sixth century AD, a separate list of the 50 strongest collocates of the word tempus in the period was extracted. Once the stop words were filtered out,8 the lists were checked for the extent to which they overlapped. It was found that the number of common collocates (cf. Table 3) was particularly high for the period from first century BC to first century AD, in which 18 of 50 words overlapped; significant overlap (11 out of 50 collocates) was also observed between the first and second centuries AD. These figures may be interpreted as

8 Stop words excluded were function words such as sum, idem, per, post, etc.

76

Krzysztof Nowak

Table 3: Overlap between the lists of 50 collocates (PHI corpus). Overlapping periods

Number of shared collocations

 BC and  AD  AD and  AD  BC and  AD  AD and  AD  AD and / AD

    

indicating the relative stability of the collocational profile and thus of the meaning of the word tempus in the given periods. However, one should be also aware that the mathematical properties of the association measures applied here do not always allow for direct comparisons between results from two or more corpora. Moreover, the lists of collocations for the fourth to sixth centuries should be interpreted with caution: as we have noticed before, they were retrieved from a limited number of texts or even single authors (Table 4). Table 4: 10 out of 50 strongest collocations retrieved for each period (PHI corpus; in bold: overlapping collocations).  BC

 BC

 AD

 AD

 AD

/ AD

 AD

occasio venio loquor annus quaero terra dies video

brevis aetas certus publicus superus reliquus annus Caesar persona spatium

longus annus nox brevis certus tero locus dies nascor exiguus

spatium brevis longus momentum diverto adverbium syllaba aestivus vicis disciplina

vernus nox medicus hora verus orbis annus duco urbs nomen

adverbium praesens futurus particeps praeteritus perfectus vernus significo autumnalis praetereo

mors testamentum medius certus specto fructus possessio pubertas condicio manumitto

Turning back to the results, the only collocation that occurs in the majority of centuries is uernus ‘springtime, of springtime’. Its relative prominence is due to the fact that its use is indeed strongly associated with time: the phrase uernum tempus accounts for the majority of the occurrences of the word uernus in the PHI corpus (76 occurrences), with the next most frequently used pattern, uernum aequinoctium, being attested significantly less, namely 46 times. The adjective is missing from the lists in two centuries only, namely in the second

Tempus mutatur

77

century BC and the fifth century AD. However, these are also periods for which the PHI collection is not representative, be it due to the number of texts in the collection or their insufficient variation. Furthermore, the fact that the word is absent from the collocation candidates list is by no means equivalent to its disappearance from language use. In our case, it only means that the strength of its association with the node word was relatively weaker at that time. Indeed, a quick corpus query reveals three occurrences of uernus in the work of Plautus, four in the work of Cato and two in Digest. All in all, apart from the adjective uernus, eight words recur as the strongest collocates of tempus in four centuries (annus, breuis, certus, dies, differo, longus, praesens, spatium) and four in three centuries (bellum, medius, persona, uenio).

3.2 The PHI corpus: clustering stable collocations The 13 collocates that had been found to be attested in at least three periods were further examined. The goal of the analysis was to reveal their distributional similarity and thus to make explicit their semantic relatedness. Firstly, similarity was computed for all pairs of the terms just listed, based on the distributional model built from the entire PHI corpus. Computed distances between word vectors were normalised, so that the words could be represented as points in a two-dimensional space: the closer the words are plotted the more distributional and thus semantic similarity is to be assumed. Quick inspection of the plot (cf. Figure 2) reveals not only the most eminent word clusters that might be worth further investigation but also important properties of the DSM-based results. A relatively coherent semantic cluster Figure 2 seems to be formed by adjectives, such as longus ‘long’, brevis ‘short’, and medius ‘middle’, and by nouns, such as spatium ‘space’. All the words are obviously semantically related, with longus and brevis being direct antonyms, and they contribute to the Space → Time mapping that is constitutive of the conceptual metaphors TIME IS SPACE and DURATION IS LENGTH. This result is hardly surprising and is perfectly consistent with Thomas’ (2011) recent study which showed that time in Latin is consequently conceptualised in terms of space.9 Another apparent cluster emerging from the inspection of the similarity plot is one consisting of the terms certus ‘some (time)’ and praesens ‘present (time)’. Both

9 The same holds true for, e.g., Polish medieval Latin (Nowak 2016) and also for the vast majority of world languages (Evans 2003, Evans 2013).

78

Krzysztof Nowak

vernus

certus medius praesens

brevis

differo

longus

venio veneo

spatium bellis bellum persona

annus dies

Figure 2: Tempus: 13 stable collocations (PHI corpus).

words are not semantically, but rather syntagmatically, related, since they are general adjectives used frequently or even exclusively (as is the case of praesens) to talk about time. They do not correspond to any particular metaphorical mapping, but rather are indicative of the Moment sense of the word (Evans 2003). Cohyponyms of the word tempus are represented by the neat cluster of annus and dies. Semantically related, two motion verbs, namely uenire and diferre, differ as to the function the word tempus plays in their argument structure. Consequently, the conceptual metaphors that they could be assigned to differ as well. More detailed discussion will be provided in one of my forthcoming papers, but let us mention only that the standard use of the verb uenio in such constructions as tempus uenit may be regarded as a realisation of the conceptual metaphor TIME IS A MOVING OBJECT. In contrast, the lexeme differre (found in such expressions as, e.g., in aliud tempus differre ‘to postpone to another time’) construes the time as fixed and the events as moving towards the given moment.

Tempus mutatur

79

3.3 The PHI corpus: clustering nominal collocations Not only both verbs were clustered as a result of their semantic similarity, but their group is also clearly separated from noun and adjective collocates. This performance of the DSMs in rough part-of-speech discrimination was confirmed when we extended our analysis to a longer list of terms co-occurring with tempus in the entire PHI corpus, regardless of their stability. From the original list, only function words, such as sum or neque, were removed and the list was reduced to comprise over 70 nouns, adjectives, and verbs. Again, the distance between word vectors was computed and then represented in form of a two-dimensional plot (cf. Figure 3) that revealed neat part-of-speech distinctions.

causa res ratio tempus lex modus condicio vita pater facio fio publicus animus bonus vis Caesar aetas dies consul

volo debeo soleo

nascor capio

pars annus nox spatium

facio habeo video fio dico do utor ago accipio fero quaero differo

alter certus omnis tres brevis medius

liber

Figure 3: Tempus: clustering larger set of collocations (PHI corpus).

Since in that form the plot does not lend itself to easy inspection, we focused on nominal collocates. Extending the word set results in slightly less neat clusters (cf. Figure 4) than those in the Figure 2. Likewise, their interpretation has now become slightly more challenging. In part, it was to be expected: a wide collocational window (five terms to the left and five terms to the right of the

80

Krzysztof Nowak

causa

ratio persona lex

tempus nomen

modus verbum

locum locus condicio

vita

pater

mors animus populus Caesar bellis bellum aetas dies

vis testamentum consul

pars annus spatium

nox

Figure 4: Tempus: noun collocations (PHI corpus).

node word) will inevitably contain both terms that stand in close syntactic relation to the node word as well as those that indicate rather the topic of a paragraph or of an entire text. One cluster that clearly stands out from the plot contains the co-hyponyms of the node word, such as annus, dies, nox, and aetas. The apparent closeness of the word bellum is probably a result of its distributional rather than semantic similarity to the words just listed, as it is often employed in the same tempus + genitive construction. Another large cluster, albeit not as clear-cut as the previous one, contains terms referring in the most general sense to the sociopolitical life of the Romans, such as populus, Caesar, or consul.10 These terms usually stand in no close syntactic relation to the node word, but instead correspond to the topics of the texts or paragraphs in which the word tempus was used. As further analysis shows, it is exactly such heterogeneous collocational clusters that should be expected to change significantly with time and across the types of texts in the corpus.

10 Law-related terms, e.g., vis, lex, condicio, or testamentum, originate almost exclusively from the DIG.

81

Tempus mutatur

Note that slightly different collocations and thus sense clusters will emerge if we turn to the noun co-occurrences of the genitive form temporis ‘of time’ in the same corpus. Again, one easily observable cluster (cf. Figure 5) contains words related to space. Apart from the ubiquitous spatium, few other terms appear, such as intervallum ‘interval’, angustia ‘narrowness’, and brevitas ‘shortness’. Another group that stands out clearly from the graphic consists of words referring to the value and the use of time, which suggests the metaphors TIME IS MONEY or TIME IS A RESOURCE. These terms include: fructus, aestimatio, and usura, and all are attested in DIG. As before, the method seems to also do systematically well as far as hyponym clustering is concerned, which set has been now enriched with hora ‘hour’ and seculum ‘century’.

spatium medium medius

punctum

intervallum angustia

momentum

brevitas accessio

necessitas ratio

hora opportunitas praescriptio adverbium

mora morum

fructus

particeps aestimatio usura

participium

Figure 5: Temporis: noun collocations (PHI corpus).

4 The PL corpus: comparing clusters between corpora In the final part of this study, the collocations of tempus extracted from the PHI corpus were compared to those computed from the corpus of contemporary and slightly later Christian literature. As the Christian corpus is nearly thrice the size

82

Krzysztof Nowak

of the PHI, the list of collocations has been significantly enriched. However, sense clustering helps to generalise over single lexemes and focus on the evolution that affected entire word groups. Again, the method performs well in clustering the hyponyms of the node word, whose set has now been extended to contain, e.g., hora and seculum (cf. Figure 6). Spatial mapping that is central to time conceptualisation is still at work here; this can be seen from the cluster of such terms as spatium ‘space’ or intervallum ‘interval’. However, there are also clear signs that the times had changed, and that they had become Christiana tempora. Times start with initium temporis, go through ordo temporis and end with finis temporum. The idea of completion of time is present in such expressions as plenitudo temporis. The content of the large heterogenous group of collocates reminds us of the changing nature of time reference in the patristic literature. The word tempus is here employed to fix events of both historical and symbolical nature in biblical chronology (tempus persecutionis, captiuitatis, and tempus passionis) and eschatological perspective (tempus iudicii, tribulationis, and resurrectionis).

propheta prophetes adventus passio signum resurrectio iudicium judicium ratio Christianus locum locus tribulatio necessitas motus captivitas tempus dispensatio persecutio seculum ordo fructus messis numerus nox etas plenitudo dies hora finis initium sus momentum

mensis

annus

vitis spatium

necessarius

brevis

Figure 6: Tempus: noun collocations (PL corpus).

intervallum

Tempus mutatur

83

5 Discussion and conclusions The present study is a preliminary one and aims at diagnosing problems that will be addressed in further work. In particular, the readability of the clusters seems to be a major challenge. As our setup was basic on purpose, there is obviously much room for further improvement. Firstly, the set of features we built the DSMs from may be enriched ad infinitum, for example, by introducing syntactical constraints or manipulating the width of the window we obtained the features from Harispe et al. (2015) and Turney and Pantel (2010). Secondly, in building and using distributional semantic models, we may experiment with different weighting scores, distance metrics, and so on. In this study, however, we focused more on the interpretability of the results. Undoubtedly, it could be significantly improved if the clusters were automatically highlighted and attributed to semantic classes. In the present chapter, such semantically coherent groups were systematically marked by the author with circles. However, it is quite paradoxical that the method that was expected to produce a more objective sense classification still requires subjective analysis.11 The problems we encountered are to some extent related to the word under investigation. The Latin tempus is an abstract and polysemous term that occurs in almost every single text in both corpora. Therefore, it is hardly surprising that its collocations would be extremely varied. As we observed, in juristic and theological writing the word tempus would co-occur with abstract terms whose attribution to a specific semantic class might pose a problem even to an experienced lexicographer. Having said all that, although their use almost always requires concordance inspection, the distributional semantic models turn out to be an efficient tool for clustering collocations and extracting semantic knowledge from a corpus. Firstly, our analysis confirmed the relative stability of spatial mapping in the history of time conceptualisation, and secondly, it revealed the evolution of time reference between two corpora. The technique functioned surprisingly well even if the models had been built from scarce linguistic information and from only lightly annotated corpora. This makes their application even more interesting in the study of under-resourced languages, such as Classical or Medieval Latin.

11 This very functionality has been implemented in the Guerreau (2018) R scripts for statistical semantic analysis. In his approach, however, the semantic knowledge is retrieved from secondorder collocations rather than DSMs.

84

Krzysztof Nowak

Wider adoption of the distributional semantic methods will not be possible, however, without dedicated tools becoming a standard part of corpus query systems. Easy-to-use software makes this more realistic today than ever. What should now be supplied to the community of Latin lexicographers and linguists in general are pre-compiled models that they could use in their research scenarios and a testing environment in which they could experiment with parameters (Nowak 2017).

Abbreviations DSM PHI

= =

PL

=

Distributional semantics model The Classical Latin Texts Database by Packard Humanities Institute latin.packhum. org (accessed: 1 November 2017) Patrologia Latina database

Bibliography Baroni, M. & Lenci, A. (2010). Distributional memory: A general framework for corpus-based semantics. Computational Linguistics, 36(4), 673–721. Evans, V. (2003). The structure of time: Language, meaning, and temporal cognition. Amsterdam /Philadelphia: John Benjamins. Evans, V. (2013). Language and time: A cognitive linguistics approach. Cambridge: Cambridge University Press. Evert, S. (2009). Corpora and collocations. In A. Lüdeling & M. Kytö (Eds.), Corpus linguistics: An international handbook (vol. 2, pp. 1212–1248). Berlin/New York: Mouton de Gruyter. Evert, S. (2014). Distributional semantics in R with the wordspace package. In Proceedings of COLING 2014, the 25th international conference on computational linguistics: System demonstrations (pp. 110–114). Dublin: Dublin City University and Association for Computational Linguistics. Evert, S. & Hardie, A. (2011). Twenty-first century corpus workbench: Updating a query architecture for the new millennium. In Proceedings of the corpus linguistics 2011 conference. Birmingham: University of Birmingham. Geeraerts, D. (2009). Theories of lexical semantics. Oxford: Oxford University Press. Guerreau, A. (2018). Coocr. Retrieved from: github.com/medialatinitas/varia (accessed: 20 July 2018). Gulordava, K. & Baroni, M. (2011). A distributional similarity approach to the detection of semantic change in the Google Books Ngram Corpus. In Proceedings of the GEMS 2011 workshop on GEometrical Models of natural language semantics (pp. 67–71). Stroudsburg: Association for Computational Linguistics. Harispe, S., Ranwez, S., Janaqi, S. & Montmain, J. (2015). Semantic similarity from natural language and ontology analysis. In Synthesis lectures on human language technologies. San Rafael: Morgan & Claypool.

Tempus mutatur

85

Heylen, K., Peirsman, Y., Geeraerts, D. & Speelman, D. (2008). Modelling word similarity: An evaluation of automatic synonymy extraction algorithms. In Proceedings of the sixth international language resources and evaluation (LREC’08) (pp. 3243–3249). European language resources association. Heylen, K., Speelman, D. & Geeraerts, D. (2012). Looking at word meaning: An interactive visualization of semantic vector spaces for Dutch synsets. In Proceedings of the EACL 2012 joint workshop of LINGVIS & UNCLH (pp. 16–24). Stroudsburg: Association for Computational Linguistics. Kilgarriff, A. & Tugwell, D. (2001). Word Sketch: Extraction and display of significant collocations for lexicography. In Proceedings of the ACL workshop on COLLOCATION: Computational extraction, analysis and exploitation (pp. 32–38). Toulouse: Association for Computational Linguistics. Migne, J.-P. (1844–1855). Patrologiae cursus completus. Series Latina (vol. 1–211). Paris. Nowak, K. (2016). Tempus in Polish Medieval Latin: A conceptual metaphor approach. Archivum Latinitatis Medii Aevi, 74, 111–125. Nowak, K. (2017). Voces. An R-based dashboard for lexical semantics. In Digital humanities 2017. Conference abstracts. Montréal: McGill University & Université de Montréal. Rissanen, M. & Tyrkkö, J. (2013). The Helsinki corpus of English texts. Studies in variation, contacts and change in English, 14: Principles and practices for the digital editing and annotation of diachronic data. Retrieved from www.helsinki.fi/varieng/series/volumes/ 14/rissanen_tyrkko (accessed: 1 November 2017). Rychlý, P. (2007). Manatee / Bonito. A modular corpus manager. In 1st workshop on recent advances in Slavonic natural language processing (pp. 65–70). Brno: Masaryk University. Rychlý, P. (2008). A lexicographer-friendly association score. In P. Sojka & A. Horák (Eds.), RASLAN 2008. Recent advances in Slavonic natural language processing (pp. 6–9). Brno: Masaryk University. Sahlgren, M. (2006). The word-space model: Using distributional analysis to represent syntagmatic and paradigmatic relations between words in high-dimensional vector spaces. Stockholm: Stockholm University. Sahlgren, M. (2008). The distributional hypothesis. Rivista Di Linguistica, 20, 33–53. Schmid, H. (1994). Probabilistic part-of-speech tagging using decision trees. In Proceedings of the international conference on new methods in language processing (vol. 12, pp. 44–49). Manchester. Thomas, J.-F. (2011). Problèmes de polysémie et de synonymie dans la lexicalisation de l’espace et du temps en latin. In C. Moussy (Ed.), Espace et temps en latin (pp. 47–60). Paris: Presses de l’université Paris-Sorbonne. TreeTagger – a part-of-speech tagger for many languages. (2017). Retrieved from www.cis.uni-muenchen.de/~schmid/tools/TreeTagger (accessed: 1 November 2017). Turney, P. D. & Pantel, P. (2010). Journal of Artificial Intelligence Research, 37(1), 141–188. Wickham, H. (2016). ggplot2. Elegant graphics for data analysis (2nd ed.). Cham: Springer. Widdows, D. (2004). Geometry and meaning. Stanford: CSLI Publications.

María Carmen Arias Abellán

Análisis y causas de la diversidad semántica y lexicológica del adjetivo uiridis Abstract: The purpose of this work is to investigate the underlying causes of the diversity of meanings in the adjective uiridis, which I do following the theoretical and methodological principles of structural semantics. We know that the predominant meaning of uiridis, the chromatic one, occurs basically when the adjective determines the vegetable nature – where it is associated with the freshness and vigor of it, while the signaling of “non-chromatic” states corresponds, on the contrary, to the qualification of nouns indicative of other domains (cf. uiridi caelo). From an exhaustive analysis – philological and linguistic – of a broad textual corpus, I propose: a) to outline the causes of the emergence of the “non-chromatic” meanings of uiridis and describe these non-chromatic meanings; b) to clarify the mechanisms of sliding, through these non-chromatic meanings, to other semantic spheres (equally non-chromatic); and c) to carry out, therefore, an integral delimitation of all the meanings of uiridis, all from a “structural” perspective, perspective that explains in a “relational and causal” way the production and origin of all meanings. Keywords: Latin linguistics, Latin lexicology, Latin semantics, Latin lexicography, uiridis

1 Introducción Aunque resulte un poco heterodoxo, quiero comenzar poniendo de relieve la emoción que me embarga ante esta vuelta mía al Thesaurus (con motivo del Congreso en el que se inscribe la presente contribución), Instituto en el que se desarrolló una estancia mía de investigadora invitada como colaboradora en la redacción de palabras de este diccionario (en 1993).

María Carmen Arias Abellán, Universidad de Sevilla https://doi.org/10.1515/9783110647587-006

Análisis y causas de la diversidad semántica

87

Por otra parte, el tema elegido, el adjetivo uiridis forma parte de mis comienzos investigadores sobre el campo léxico del color, de modo que como componente de este campo léxico conservo con este adjetivo una relación también antigua. A lo largo de los años y como filóloga en continuo contacto con los textos he ido descubriendo diversas y curiosísimas acepciones de este adjetivo y dada mi consideración del vocabulario como un código de comunicación regido (como todo el sistema lingüístico al que pertenece) necesariamente por principios económicos y constantes de organización, he decidido investigar las razones subyacentes e inherentes a esta aparente dispersión significativa de uiridis, cuestión que llevo a cabo – como ya lo he señalado en el Abtract – siguiendo los principios teóricos y metodológicos de la semántica estructural1 y que paso a abordar sin más dilación. Según es de todos conocido, la alusión cromática de “verde” responde básicamente a la calificación de sustantivos de la naturaleza vegetal (uiridis hortulus, uiridis herba, uiridis ramus, etc.) y de ella está plagada la literatura Latina,2 mientras que el señalamiento de otras acepciones “no cromáticas” se correspondería con la calificación a otro tipo de sustantivos. La cuestión, como vamos a ver, obedece a un trasvase de las ideas meliorativas asociadas al mundo vegetal cuando posee este color “verde” (“vigor”, “florecimiento”, etc.) a diversos ciclos de la existencia en su conjunto dotados de estas características positivas.3 El desvío a los sentidos metafóricos de uiridis se produciría, según esto, desde el ámbito vegetal y seguiría una senda que iría desde lo tangible (el verdor de la naturaleza y los rasgos unidos a este verdor), junto con otros conjuntos también tangibles, hasta otros conjuntos inmateriales (como, por ejemplo, la valoración de determinadas fases del ciclo de la existencia del hombre). Es aconsejable, pues, investigar detenidamente el manejo de nuestro término en el corpus más representativo de este dominio vegetal y donde uiridis es

1 La concepción del estructuralismo en este plano del vocabulario, concepción que viene a explicar el léxico –como el resto de los planos lingüísticos – por principios oposicionales subyacentes de “economía y no de dispersión ni atomización”, es un hecho que doy por conocido (al ser ésta una metodología antigua y asentada) y en el que no podría, por otra parte, detenerme aquí. Remito, por ello, a mi obra (Arias Abellán 1994: 17–36) y a la amplia bibliografía allí citada. 2 Para su ejemplificación, cf. Oxford Latin Dictionary, s.u. 3 Esta conexión del vigor de uiridis entre la naturaleza (cuando ésta proyecta su plenitud) y las etapas más florecientes de la vida del hombre, es una conexión, en efecto, destacada ya desde antiguo: en los diccionarios (cf. Forcellini, s.u.; Ernout-Meillet, s.u.) y en obras como la fundamental de André (1949: 186) o la mía propia (Arias Abellán 1994: 129, Arias Abellán 2000).

88

Carmen Arias Abellán

frecuentísimo, que es “la langue paysanne”,4 y cuya fuente documental latina más representativa la constituyen los tratadistas de Agricultura; podremos indagar y perfilar en este corpus (según vamos a tener ocasión de constatarlo) las causas del surgimiento de sus significados “no cromáticos” y las fases necesarias de dicho surgimiento, cuestión fundamental en cualquier concepción estructural de “sistematicidad de la lengua”. Dichas fases comportarían el examen de: a) razones u origen del surgimiento de sus significados “no cromáticos”; b) descripción pormenorizada de dichos significados; c) mecanismos del deslizamiento, a través de estos significados no cromáticos, a otras esferas semánticas; d) delimitación completa de estas otras esferas, que, como vamos a ver, sobrepasan, con mucho, a la más conocida de la vida humana.5

2 Análisis textual 2.1 Calificación de sustantivos concretos relativos básicamente al ámbito de la naturaleza vegetal 2.1.1 Las obras de los agrónomos documentan,6 en efecto, el funcionamiento de uiridis actualizando meramente la cualidad de color de la vegetación en textos del tipo siguiente:

4 Cf. André (1949: 186); cf. asimismo, Ernout-Meillet, s.u.: “emploi fréquent dans la langue rustique”. La lengua de la Agricultura latina es, en efecto, en lo que concierne al uso de los adjetivos de color, la más representativa de las lenguas ténicas latinas. 5 Según vamos a comprobarlo, los significados “simbólicos”, “abstractos”, etc., se deben, anticipándolo ahora resumidamente, a la translación del estado de excelencia de la naturaleza vegetal (y a su “color verde” cuando se encuentra en este estado) a la excelencia también en muchas otras esferas temáticas, produciéndose así diversas metáforas (de indicación todas ellas del grado óptimo de cualquiee ciclo evolutivo). Una vez explicado este “origen” de los valores metafóricos (cf. nota 3), estaremos en situación de observar su presencia en todo tipo de textos (prosa y verso); en los poéticos, sobre todo, como se observa en los CE que analizamos seguidamente, y también en todos los géneros poéticos que pueblan las citas de las notas de nuestro estudio; géneros, por cierto, y como lo advierte con total acierto André (1949: 266) sobre los que no se puede trazar ninguna tipología distintiva – entre unos y otros – en lo que concierne al uso de nuestro adjetivo. 6 El corpus considerado comprende: De agri cultura (Catón); De re rustica (Varrón); De re rustica (excepto libro X) y De arboribus (Columela); Opus Agriculturae (Paladio), y también – por

Análisis y causas de la diversidad semántica

(1)

89

Folio lato, colore uiridi gignit quod quidam prason uocant. (PLIN. nat. 13, 135) ‘produce largas hojas verdes y algunos autores lo llaman prasos.’

2.1.2 Es más frecuente, no obstante, la incidencia, antes que en la mera cualidad cromática, en acepciones alusivas a los estados (o condiciones transitorias) que acompañan a la naturaleza en sus puntos de “vigor”, frente a las etapas finales de “sequedad” y “muerte” de la misma. Estos hechos son observables en los grupos textuales que siguen: 2.1.2.1 En primer lugar, la condición de “frescor” de “no sequedad” que aparece en las oposiciones, frecuentísimas, del tipo uiridis / aridus (peraridus) mostradas por los subsiguientes textos y de entre las que hay que destacar el par uiride / aridum pabulum de una frecuencia extraordinaria: (2)

Bracchia . . . uiridia seruato, arida . . . falce amputato. (COLVM. 4, 24, 7) ‘conserve los brazos [de la vid] . . . verdes, corte con la podadera los secos.’7

(3) ouibus . . . octo mensibus uiridi eo pabulo uti et postea arido possis. (COLVM. 5, 12, 1) ‘para las ovejas . . . puedes aprovecharlo [el cítiso] como forraje verde durante ocho meses, y después, como pasto seco.’ (4) est autem modus in siccando, ut neque peraridum neque rursus uiride conligatur. (COLVM. 2, 18, 1)8 ‘en el punto de sequedad hay que guardar un término medio, de forma que ni se recolecte muy seco (el heno) ni, por el contrario, aún verde.’

sus concomitancias temáticas – los libros XII–XIX de la Naturalis Historia de Plinio el Viejo (cf. Arias Abellán 1994: 23). 7 En las versiones de los textos de Columela y Paladio citados a partir de este momento, sigo, en general, las obras de Holgado Redondo (1988) y Moure Casas (1990). 8 Para el elenco total de citas en los tratadistas de Agricultura representativas de esta oposición en la que cobra protagonismo especial, como ya lo he señalado, el sintagma uiride pabulum / aridum pabulum – cf. Arias Abellán (1994: 125–126).

90

Carmen Arias Abellán

Esta misma condición de “frescor” de “no sequedad” puede adoptar como órgano de expresión las oposiciones de uiridis / siccus (siccatus) legibles en los ejemplos siguientes: (5) . . . praebeas . . . siccam uel uiridem medicam. (COLVM. 7, 4, 2) ‘. . . que le des . . . alfalfa seca o verde.’ (6) rufa acrior quam candida, et sicca quam uiridis. (PLIN. nat. 19, 105) ‘[la cebolla] roja es más fuerte que la blanca y la seca más que la verde.’ (7) Vineis anno siccata utilior quam uiridis. (PLIN. nat. 17, 146) ‘Para las viñas es mejor dejarla secar un año que verde.’ E, igualmente, dicho estado de “frescor” puede actualizarse mediante la confrontación de uiridis con nociones verbales como arescere (inarescere) y uiescere, etc. documentada en estos otros casos: (8) de lupino nihil dubito atque etiam de pabulari uicia, si tamen eam uiridem desectam confestim aratrum subsequatur et, quod falx reliquerit, prius quam inarescat, uomis rescindat atque obruat. (COLVM. 2, 13, 1) ‘Ninguna duda tengo del altramuz y tampoco de la veza para el forraje, si, segada aún verde, se mete inmediatamente el arado y, lo que deja la hoz, antes de que se seque, lo corta y lo entierra el arado.’ (9) alii pinguissimam quamque uiridium ficorum eligunt, et harundine uel digitis diuisam dilatant, atque ita in sole uiescere patiuntur. (COLVM. 12, 15, 5) ‘Otros eligen la clase más gorda de los higos verdes y, tras haberlos partido con una caña o con los dedos, los estiran y de esta manera los dejan secar al sol.’ (10) uiridem etiam gallinis, aut si inaruerit, madefactum . . . (PLIN. nat. 13, 131) ‘[El cítiso] verde se les da incluso a las gallinas o, en el caso de que se haya secado, humedecido . . .’ (11) lupinus et uicia pabularis, si uirides succidantur et statim supra sectas eorum radices aretur, stercoris similitudine agros fecundant; quae si exaruerint ante quam proscindas, in his terrae succus aufertur. (PALLAD. 1, 6, 14) ‘el altramuz y la veza forrajera, si se cortan verdes y se ara por encima de sus raíces inmediatamente después de seccionadas, fertilizan los campos

Análisis y causas de la diversidad semántica

91

a semejanza del estiércol, pero si se secan por sí mismos antes de que los cortes, el jugo de la tierra se va con ellos.’ (12) putanda persicus in eo est, ut arida et putria tantum uirgulta tollantur; nam si quid uiride resecemus, arescit. (PALLAD. 12, 7, 3)9 ‘el melocotonero hay que podarlo de forma que se le saquen exclusivamente las ramas marchitas y podridas, ya que si se les corta alguna verde, seca.’ 2.1.2.2 En segundo lugar, tras la condición de “frescor” que acabamos de examinar,10 se pueden sumar las características, de “dúctil, flexible, tierno”, que arrojan los sintagmas de coordinación a tener o mollis que se recogen a continuación: (13) = (2) Bracchia tenera et uiridia seruato . . . (COLVM. 4, 24, 7) ‘conserve los brazos [de la vid] tiernos y verdes . . .’ (14) tum etiam uitem minus cicatricosam reddit, quoniam id, ex quo uiride et tenerum decerptum est, celeriter consanescit. (COLVM. 4, 27, 3) ‘y asimismo deja una vid con menos cicatrices, puesto que el punto de donde se arrancó un brote verde y tierno sana rápidamente. (15) tum etiam uitis minus cicatricosa fit, quoniam quod uiride et tenerum decerpitur, protinus conualescit. (COLVM. arb. 11, 1) ‘y además, la vid queda con menos cicatrices, porque al cortar lo que aún está verde y tierno, la herida se restablece con prontitud.’ (16) Teneram et uiridem uitem ferri acie ne recidas. (PALLAD. 1, 6, 9) ‘No cortes la vid tierna y verde a filo de hierro.’ (17) uiridemque frondem uel aliud molle pabulum. (COLVM. 6, 14, 2)11 ‘hojas verdes u otros forrajes tiernos.’

9 Cf. igualmente, Arias Abellán (1994: 125). 10 En este terreno de “frescor” de uiridis, como contrario a “árido”, “seco”, “caluroso” me viene a la memoria el sintagma tan poético de uiridis umbra ( ‘refrescante sombra’, cf. VERG. ecl. 9, 20), que Virgilio llega a igualar con dulcis umbra (cf. VERG. Aen. 1, 694). 11 El registro completo de citas puede leerse en Arias Abellán (1994: 127).

92

Carmen Arias Abellán

2.1.2.3 En tercer lugar, y para terminar con la documentación de este apartado (iniciado en 2.1.2), debe destacarse el señalamiento de “permanencia (de lo vegetal)” en las fases previas a la “sazón o vejez” que proyectan las contraposiciones a uetus o las coordinaciones a iuuenis o inmaturus mostradas seguidamente: (18) = (13) = (2) Bracchia tenera et uiridia seruato, arida et uetera falce amputato. (COLVM. 4, 24, 7) ‘conserve los brazos [de la vid] . . . verdes, corte con la podadera los . . . viejos.’ (19) . . . quoniam et de longissimo aeuo arborum diximus – ad Neronis principis incendia cultu uirides iuuenesque, ni princeps ille adcelerasset etiam arborum mortem. (PLIN. nat. 17, 5) ‘. . . puesto que hemos hablado también de la larga longevidad de los árboles, con el cuidado correspondiente se habrían conservado verdes y jóvenes hasta el incendio del emperador Nerón, si aquel no hubiese acelerado la muerte de dichos árboles.’ (20) alterum, quia sauciat, quod in uiridi et adhuc stirpe inmatura fieri noxium est. (COLVM. 4, 7, 1)12 ‘otra, porque la hiere, lo que es nocivo hacer en una planta verde y todavía en formación.’ A través de las cualidades que uiridis actualiza en la realidad material y concreta del mundo “vegetal” vistas hasta aquí, es decir: “fresco” (por oposición a aridus, siccus, etc.), “tierno”, “flexible”, “dúctil” (por coordinación a tener, mollis, etc.), “no viejo” (por oposición a uetus, etc.), cualidades contrapuestas a las de “seco”, “marchito”, “viejo”, etc., y que hemos observado – junto con todas sus correspondencias y oposiciones léxicas – en los textos examinados hasta este momento, a través de estas cualidades – digo –, cualidades que son las que “caracterizan a las manifestaciones más fértiles del crecimiento en el mundo vegetal”, podemos concluir y comprender la actuación de nuestro adjetivo cuando describe (igual y paralelamente a como lo hace en el ámbito vegetal, su ámbito más propio, cf. nota 4) las etapas de mejor calidad de cualquier otro ciclo evolutivo (aunque no sea el vegetal).

12 Cf. todos los ejemplos en Arias Abellán (1994: 126–127).

Análisis y causas de la diversidad semántica

93

2.1.3 Es eso lo que ocurre con “líquidos”, sobre todo con oleum, donde uiridis – coordinado a optimus, egregius y contrapuesto a corrumpere – apunta a las nociones de “reciente”, “fresco”, “en buen estado”, que pueden leerse en las siguientes citas: (21) oleum uiridius et melius fiet. (CATO agr. 3, 3) ‘el aceite resultará más verde y mejor.’ (22) Posiae tamen oleum saporis egregii, dum uiride est. (COLVM. 5, 8, 4) ‘el aceite del olivo “posio” es de egregio sabor mientras está verde.’ (23) Pausia tamen oleum quod reddit, dum uiride est, optimum, sed cito uetustate corrumpitur. (PALLAD. 3, 18, 4) ‘el aceite que da la posia, mientras está verde y fresco, es excelente, pero enseguida se pone rancio con el tiempo.’13 2.1.3.1 Eso es lo que sucede asimismo con el “sabor reciente o fresco” del caseus o del oliuarum sapor: (24) Casei quoque faciendi non erit omittenda cura . . . Is porro si tenui liquore conficitur, quam celerrime uendendus est, dum adhuc uiridis succum retinet . . . (COLVM. 7, 8, 1) ‘No debe ser omitido tampoco ningún cuidado a la hora de hacer el queso . . . Si éste [el queso] se hace con leche poco grasa, hay que venderlo lo más rápido posible, mientras está aún reciente y retiene el jugo . . .’ (25) Oliuae Pauseae, cum iam decolorantur, antequam mitescant, peciolis leguntur, et in oleo quam optimo seruantur. Haec maxime nota etiam post annum repraesentat uiridem saporem oliuarum. (COLVM. 12, 49, 8) ‘Se cogen aceitunas posias con sus rabillos cuando ya cambian de color, antes de que se pongan blandas, y se conservan en aceite del mejor; esta clase reproduce muy bien, incluso después de un año, el sabor fresco de las aceitunas.’14

13 Para este aceite en los agrónomos, donde a veces se constata expresamente la oposición de uiride (“fraiche”) a “rance” (conditum), véase igualmente Christol (2016: 306). 14 De este sabor de las aceitunas, hay referencia también en Bradley (2009: 7).

94

Carmen Arias Abellán

2.2 Calificación de sustantivos concretos de otros ámbitos diferentes a la naturaleza vegetal (caelum, sonus, etc.) Y así se le lee también – dentro de su capacidad de designación de la excelencia, cuyo origen se ubica en el mundo vegetal – y fuera de este ámbito, ¡hasta con términos como caelum!, con el que indica un cielo todavía “no alterado”, es decir, “fresco”, “despejado” (“sin vientos”): (26) . . . uiridi caelo . . . ac sine aura. (PLIN. nat. 17, 74) ‘que le ciel soit serein et le vent nul’ (‘con el cielo sereno y sin viento’15) Todas estas acepciones no cromáticas de uiridis, referidas a las fases de más valor en el desarrollo de campos tangibles como el de la vegetación, de los líquidos, de los alimentos, ¡del tiempo atmosférico incluso!, etc., nos explican igualmente la incidencia en sustantivos – tan extraños al color – como sonus, mostrando uiridis – en su línea de señalamiento de la excelencia – la “intensidad” o “fuerza” de la “sonoridad”, que vemos en la cita de Aulo Gelio: (27) H litteram, siue illam spiritum magis quam litteram dici oportet, inserebant eam ueteres nostri plerisque uocibus uerborum firmandis . . . ut sonus earum esset uiridior uegetiorque. (GELL. 2, 3, 1) ‘La letra h o la aspiración . . . los antiguos la insertaban para sostener . . . la sonoridad de muchas palabras a fin de que se entendiesen de manera más vigorosa y poderosa.’16 E, incluso, nos aclaran su referencia al propio color, denotando, como lo remarca Forcellini s.u. (dicitur de colore uiuido et uegeto) su viveza y contraposición a lo desvaído, que puede observarse en Quintiliano: (28) non ille pallens nec acerbo languore consumptus, . . . sed uiridis et sane pulcher. (QVINT. decl. 10, 5) ‘aquel no pálido ni con una languidez desvaída, . . . sino vigoroso y muy bello.’

15 Versión de André (Les Belles Lettres, 1964); Bradley (2009: 7, nota 27) cita también este texto en nuestra misma línea: ‘uiride caelum, not literally “green” of course, but “clear and fresh and conducive to germination”’. 16 Cf. igualmente Gellius. 13, 21, 13 uiridior sonus (Bradley 2009: 7).

Análisis y causas de la diversidad semántica

95

2.3 Calificación de sustantivos abstractos del ámbito de la vida y las fases de su transcurso (aetas, iuuenta, etc.) Partiendo de todo este ámbito – instalado todavía en el campo de lo tangible – el adjetivo es capaz aún de dar un salto hacia lo animado en sustantivos como aetas,17 situación en la que uiridis, desde su recurrencia en la indicación de los niveles óptimos y más intensos de cualquier proceso evolutivo y en un grado mayor de abstracción,18 dada la naturaleza del nombre calificado (aetas), actualiza, entre otras, las nociones de “belleza” y de “vitalidad” (cf. pulchritudinem, robore) – de animales y hombres – que señalan nuestras dos manifestaciones de la prosa de la Agricultura: (29) Nec uero patiendus est oryx, aut aper, aliusue quis ferus ultra quadrimatum senescere. Nam usque in hoc tempus capiunt incrementa, postea macescunt senectute. Quare dum uiridis aetas pulchritudinem corporis conseruat, aere mutandi sunt. (COLVM. 9, 1, 7) ‘No se debe consentir que la gacela o el jabalí o cualquier otro animal salvaje sobrepase los cuatro años, pues hasta esa edad están en proceso de desarrollo, pero después empiezan a ponerse flacos por obra de la vejez; por ello deben ser objeto de venta mientras el vigor de su edad les mantiene la belleza del cuerpo.’ (30) siue fundum locuples mercatus est, e turba pedisequorum lecticariorumque defectissimum annis e uiribus in agrum relegat, cum istud opus non solum scientiam, sed et uiridem aetatem cum robore corporis ad labores sufferendos desideret. (COLVM. 1 praef. 12) ‘si un rico ha comprado una hacienda, de entre la muchedumbre de lacayos y esclavos de litera relega al campo al de más años y al más menguado de fuerzas, cuando este trabajo requiere no sólo conocimientos, sino también la juventud y la fuerza corporal necesarias para tolerar las fatigas.’

17 Para la documentación completa – literaria y epigráfica – del sintagma uiridis aetas puede consultarse el Thesaurus (s.u. aetas) (véase igualmente nuestra selección de nota 19); sobre este empleo en los agrónomos, cf. Arias Abellán (1994: 129); acerca de su presencia epigráfica, cf. Plessis (1905: 158); y, especialmente, el amplio comentario de Massaro (1992: 155–156). 18 La excelencia e intensidad se reproducen también en el campo de lo abstracto y de lo referido al “vigor o intensidad” de las cualidades humanas tanto positivas como negativas retratadas en los subsiguientes empleos: (consilio uiridis, sed belli seruus ‘vigoroso en sus decisiones, pero lento en su acción’, SIL. 3, 255; cruda mente et uiridissimus irae, ‘de mente insensible y lleno de vigor en su ira’, SIL. 5, 569).

96

Carmen Arias Abellán

Sintagma éste, el de uiridis aetas para cuya ejemplificación poética – aun a sabiendas de que se documenta en toda la literatura latina19 – elijo a propósito la epigráfica, poesía que – por su propia temática – fija su mirada en el transcurso de la vida y presta una especial atención al esplendor de la juventud, tal y como se constata en los carmina que transcribimos a continuación20: uiridis aetas: (31) Heic uiridis aetas cum floreret artibus crescente et aeuo gloriam conscenderet, properauit hora tristis fatalis mea et denegauit ultra ueitae spiritum. ... bis hic septeni mecum natales dies tenebris tenentur Ditis aeterna domu. (CE 55, 5–8 y 18–19) ‘cuando mi edad llena de vigor florecía en diversas artes y, al crecer yo, se elevaba mi gloria, se aceleró mi triste hora fatal y denegó por más tiempo la energía a mi existencia . . . Catorce días de cumpleaños están retenidos aquí conmigo en las oscuridades, en la casa imperecedera de Dite.’21 19 Como ejemplos de la uiridis aetas, de la uiridis iuuentus en la literatura no epigráfica, recordamos citas como VERG. Aen. 5, 295 Euryalus forma insignis uiridique iuuenta . . . “Euríalo, célebre por su belleza y su lozana juventud”; OV. Pont. 4, 12, 29 haec uiridi concordia coepta iuuenta . . . “esta concordia nacida en la floreciente juventud”; CVRT. 10, 5, 10 uoces exaudiebantur, tam uiridem et in flore aetatis fortunaeque inuidia deum ereptum esse rebus humanis, “tan joven y en la flor de la juventud . . . ”; SEN. Oed. 775 Primone in aeuo uiridis an fracto occidit? “¿murió joven en el principal momento de su vida o en la edad ya achacosa?”; MART. 7, 40, 5 occidit . . . uiridi . . . iuuenta, “murió en la floreciente juventud”; STAT. silv. 1, 2, 276 . . . longe uiridis sic flore iuuentae / perdurent uultus tardeque haec forma senescat “que tu rostro quede mucho tiempo en la flor de la juventud y que tus encantos sean lentos en envejecer”; STAT. silv. 3, 1, 161 si dulce decus uiridesque resumeret annos “si ella pudiese recobrar el dulce honor de los años vigorosos de la juventud”; STAT. silv. 3, 3, 125: si longa dies, si cernere uultus / natorum uiridisque genas tibi iusta dedissent / stamina “si los hilos de las Parcas, basándose en la justicia, te hubieran concedido ver los rostros de tus hijos y sus florecientes mejillas”. 20 Esta calificación (en los CE) de los adjetivos de color en las edades de la vida humana y sus causas, recibió una atención precedente mía en 2014 (cf. Arias Abellán 2014), aunque allí, además de tratarse del binomio aureus-uiridis, la atención básica del análisis era el funcionamiento de los procedimientos semánticos y literarios de la “metáfora” en sí misma, dando como resultado, lógicamente, un examen de diferente perspectiva y mucho más reducido y sucinto de ambos adjetivos. 21 Sobre la datación, fluctuante entre fines de época republicana y fines del s. I o s. II, de este poema, dedicado a una muchacha de nombre Eucharis, cf. Massaro (1992: 116 y 118–119); sobre el carmen en su conjunto, cf. el excelente comentario de este autor (Massaro 1992: 115–195) e igualmente, Courtney (1995: 238–239).

Análisis y causas de la diversidad semántica

97

uirides annos: (32) in lucem po]stquam [te fudit celsa u]oluntas annos et uir]ides postquam doctrina repleuit, tunc lacrimat]us abis, tunc gaudia larga parentum spesque omnes] mesto tecum sub puluere condis. at super astra] deus, qui te formauit et aufert, clemens accipi]at seruetque ad gaudia membrum. (CE 752) ‘Tras haberte traído a la vida el designio de Dios y haber llenado la formación tus florecientes años, te marchas ahora, motivando las lágrimas de los tuyos, y las intensas alegrías de tus padres y todas las expectativas puestas en ti las entierras bajo funesto suelo. Pero ojalá que Dios, que te creó y te aleja de la vida, te reciba lleno de clemencia en el cielo y te preserve hasta el momento gozoso de la resurrección de la carne.’22 con iuuenta: (33) occidis infelix uiridi sub flore iuuentae, purpureusque nitor, quam breuis, euanuit. uiximus et mutuum quodcumque exegimus ae[u]o, nam tecum, Rhodine, gaudia nostra iacent [. . .] aeternam in pacem, unde erat, et rediit. hunc tumulum fecit Dionysius, “iste quis ad me”? si quaeris, coniux. cetera uiue uale. (CE 1431, 7–14) ‘Has muerto, desgraciada, en la vigorosa flor de la juventud y tu saludable23 esplendor se ha desvanecido de manera brevísima. Vivimos y lo llevamos todo con reciprocidad en nuestra vida, y por ello mis alegrías están ahora, Ródine, contigo . . . al descanso eterno, de donde era, regresó. Este sepulcro lo construyó Dionisio. Si preguntas qué significa éste en mi vida, es mi marido. Por lo demás, vive y adiós.’24

22 Para la cronología de este epígrafe, cf. Massaro (1992: 155): “ . . . 1431, 7, cristiano del sec. V . . ., e il coevo 752”. De otro lado, Massaro, uno de los mejores especialistas de este carmen, da por segura la conjetura de Bücheler: uir]ides, integrada, además, en un tópico epigráfico (y literario) frecuentísimo (Lier 1903: 445–447; Lier 1904: 54–65; Lattimore 1942: 195–198, entre otros). 23 Purpureus califica en efecto en ciertas ocasiones a miembros corporales o al rostro humano cuando están llenos de salud (cf. André 1949: 98). 24 Sobre la datación de este epígrafe, cf. Massaro (1992: 155): “cristiano del sec. V”.

98

Carmen Arias Abellán

Esplendor de la vida, por cierto, que se adjudica a veces también a la vejez, pero sólo cuando ésta mantiene el vigor de la juventud: (34) iam senior, sed cruda deo uiridisque senectus. (VERG. Aen. 6, 304) ‘Ya bastante viejo, pero su vejez es la vejez intacta y vigorosa de un dios.’ (35) . . . senem, sed mehercules uiridem animo ac uigentem. (SEN. epist. 66, 1) ‘. . . un viejo, pero, en todo caso, vigoroso y lozano de espíritu.’ (36) uiridique ad dura laborum / bellator senio iam castra mouebat in hostem. (SIL. 7, 3–4) ‘guerrero en su vigorosa vejez frente a las duras pruebas, entraba en campaña contra el enemigo.’25

3 Conclusiones El examen detenido (que acabamos de realizar) de los datos aportados por la lengua de la Agricultura y otras documentaciones, nos hace factible describir el cuadro significativo completo26 de nuestro adjetivo, y, lo que es más importante,

25 Como ejemplos del mismo tipo (de uiridis en la edad avanzada de la vejez) pueden añadirse, entre otros, SEN. Oed. 297; SIL. 5, 569; VAL. FL. 1, 77, etc. 26 Si he hablado en repetidas ocasiones de que esta investigción se basa en el cuadro “completo y total” de los significados de uiridis, es, en primer lugar, porque en ella se da una explicación semántica unitaria de nuestro adjetivo (“color de excelencia de la vegetación, desde donde puede transferirse a la excelencia en cualquier ciclo evolutivo”), explicación capaz de abrazar y aclarar cualquier presencia textual del mismo; y, en segundo lugar, porque además de ofrecer esta citada visión semántica unitaria del adjetivo, he analizado todas las situaciones semánticas recogidas en los diccionarios que he venido citando, el más variado y completo de los cuales es, hasta ahora, el Oxford Latin Dictionary. Se citan en él (cf. s.u.) dos conjuntos textuales que repiten el valor cromático sobradamente analizado en este trabajo: a) su manejo, con la “vegetación” – tan estudiada por nosotros – del fondo del mar (de ahí su empleo, más bien escaso, con el mar, ríos, etc., y las divinidades relacionadas con los mismos (cf. Bradley 2009: 7); b) su manejo, también minoritario, en las gamas claras del “verde”, es decir, como un “verde claro, pálido” y, por ende, próximo a lo “amarillento”, situación ésta que se fundamenta en ser uiridis término genérico con la capacidad – estructural – de aludir a los matices “claros y oscuros del verde”. Igualmente cromática (“verde”) es la descripción de piedras y animales con la piel de tal color, situación que Trinquier (2002: 98–114) ha analizado, pero desde un punto de vista extralingüístico, centrándose en sus propiedades terapéuticas y en los ámbitos ritual e iconográfico egipcios, cuestión que (como advierte

Análisis y causas de la diversidad semántica

99

captar las causas y relaciones subyacentes al despliegue – en diferentes acepciones – de dicho cuadro: un significado originario “cromático” de aplicación preferente para el mundo “vegetal”, espacio que presenta dicho color en el punto álgido de su desarrollo y donde se asocia a las notas “meliorativas” de “fresco”, “tierno”, “no viejo”; desde esa zona meliorativa queda capacitado para señalar las cualidades mejores de otros ciclos de evolución: el de los líquidos, el de los alimentos (“frescos”, recientes”, “no alterados”), el de los estados atmosféricos (“despejados”, “sin turbulencias”), el de los sonidos y los propios colores (“vívidos”, “no apagados”), etc., y, avanzando hacia la abstracción y siempre con el mismo tono valorativo, el del ciclo de la vida animal y el más reseñado de la plenitud y juventud de la vida humana. Acabo ya: con el examen precedente de todo el panorama significativo de nuestro adjetivo, y, lo que es más importante, de las bases y de las relaciones subyacentes que son las que provocan este panorama que se ha ido desplegando en diferentes acepciones, he querido resaltar la causalidad de las acepciones (no su enumeración), causalidad que es, en definitiva, la que explica lingüísticamente la diversidad significativa de las palabras en sí mismas, o las diferencias significativas de las palabras entre sí (dentro de los llamados “campos semánticos o léxicos”). Este requisito de la investigación de los orígenes y relaciones que subyacen a y que motivan todas las acepciones de uiridis me parece la aportación fundamental de mi trabajo y me parece también de ineludible incorporación en la redacción de los lemas de los diccionarios (en los que suele dominar un dibujo más bien “numerador”). Me agrada pensar, por tanto, que esta investigación mía sirva de pequeña contribución, cuando llegue el momento, a la redacción de este lema en el Thesaurus Linguae Latinae.

Bibliografía André, J. (1949). Étude sur les termes de couleur dans la langue latine, París: Klincksieck. Arias Abellán, C. (1994). Estructura semántica de los adjetivos de color en los tratadistas latinos de Agricultura y parte de la enciclopedia de Plinio. Sevilla: Secretariado de Publicaciones. Universidad de Sevilla. Arias Abellán, C. (2000). Peruiridis. Thesaurus Linguae Latinae, vol. X, 1, p. 1877, 60–63. Leipzig: Teubner.

acertadamente Bradley 2009: 9) se aparta y no tiene nada que ver con la lingüística latina de uiridis, que es la que aquí nos ha ocupado.

100

Carmen Arias Abellán

Arias Abellán, C. (2014). Metáforas adjetivas de color para la designación epigráfica del esplendor de la vida. In P. Molinelli, P. Cuzzolin y C. Fedriani (Eds.), Latin vulgaire et tardif X (pp. 397–414). Bergamo: Bergamo University Press. Bücheler, F. (1982 [1895–1897]). Anthologia latina II. 1–2: Carmina latina epigraphica, Stuttgart: Teubner. Bradley, M. (2009). Colour and meaning in ancient Rome. Cambridge: Cambridge University Press. Christol, A. (2016). Le latin des cuisiniers: L’alimentation végétale, étude lexicale. Broché: Presses Universitaires Paris-Sorbonne. Courtney, E. (1995). Musa lapidaria: A selection of Latin verse inscriptions. American Classical Studies, 36, Atlanta. Ernout, A. & Meillet, A. (19944 [19321]). Dictionnaire étymologique de la langue latine: Histoire des mots. París: Klincksieck. Forcellini, A. (1965). Lexicon totius latinitatis (cur. F. Corredini et J. Perin, secund. impr. anast. confecta quartae edit. a.a. 1864–87). Padua. Glare, P. G. W. (Ed.) (2012=1982). Oxford Latin dictionary. Oxford: Oxford University Press. Holgado Redondo, A. (1988). De los trabajos del campo: Lucio Junio Moderato Columela. Madrid: Ministerio de Agricultura. Lattimore, R. (1942). Themes in Greek and Latin epitaphs. Urbana: The University of Illinois Press. Lier, B. (1903). Topica carminum sepulcralium latinorum I-II. Philologus, 62, 445–477. Lier, B. (1904). Topica carminum sepulcralium latinorum III. Philologus, 63, 54–65. Lommatzsch, E. (1982 [1926]). Anthologia latina II.3: Carmina latina epigraphica, supplementum. Stuttgart: Teubner. Massaro, M. (1992). Epigrafia metrica latina di età repubblicana. Bari: Università di Bari. Moure Casas, A. (1990). Paladio. Tratado de Agricultura. Medicina. Veterinaria. Poema de los injertos. Madrid: Gredos. Plessis, F. (1905). Poésie latine: Epitaphes. París: Frédéric Édouard. Sanders, G. (1991). Lapides Memores. Païens et chrétiens face à la mort: le témoinage de lʹépigraphie funéraire latine. Faenza: Fratelli Lega. Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. Trinquier, J. (2002). Les vertus magiques et hygiéniques du vert dans l’antiquité. In L. Villard (Ed.), Couleurs et visions dans l’antiquité classique (pp. 97–128). Rouen: Publications de l’Université de Rouen.

Nigel Holmes

Authorial error and the dictionary Abstract: Both editors of ancient texts and lexicographers are sometimes too ready to correct what may be errors of ancient authors. Where an author is taking over unfamiliar material from another source, a simple misreading or misunderstanding can conjure up words that never existed. For example, several errors in Pliny’s Natural History are best attributed to Pliny or his sources, since the corruptions that must be supposed would take place more easily in a Greek text than in a Latin one, and the same reading is found not only in Pliny’s exposition but also in the index that he created for the work. The transmission of knowledge can also introduce errors that create a new meaning for an existent word. Here too editors are sometimes too ready to make changes, a problem illustrated by Lindsay’s treatment of GLOSS.L I Ansil. PE 805. The correct treatment of such isolated words and meanings remains a problem for comprehensive dictionaries. Keywords: Latin linguistics, lexicography, textual criticism, glossaries

1 Introduction My subject is perhaps one of the linguistic phenomena least interesting to linguists, words and meanings that only exist because of a language user’s mistake. Some errors of this kind may be interesting for their role in language development, but my instances cannot even claim that. They are single errors that left no echo in later usage beyond the texts of the writers who made the original mistake. What should a comprehensive historical dictionary do in such cases? In my view (as one of the editors of the Thesaurus Linguae Latinae), this is a point where we have unusual obligations. Every usage needs to be registered, whatever the source or linguistic competence of the user.

2 Words created through a mistake The editors of ancient texts reasonably assume that their authors have not written nonsense, but where an author is taking over unfamiliar material from another Nigel Holmes, Thesaurus Linguae Latinae https://doi.org/10.1515/9783110647587-007

102

Nigel Holmes

source, a simple misreading or misunderstanding can conjure up words that never existed. The kind of material where you could expect such errors to be particularly common is technical vocabulary, such as the names of plants or places. At PLIN. nat. 22, 40 editors print the following text: leucacantham alii phyllon, alii ischada, alii polygonaton appellant (ʻSome people call leucacantha phyllon, others ischas, others polygonatosʼ). In this phyllon is a conjecture of Hermolaus Barbarus; the manuscripts have pymon, phimon, phinion and phimum. In a sense phyllon is certainly correct; the related passage at DIOSC. 3, 19 has λευκάκανθα· οἱ δὲ πολυγόνατον, οἱ δὲ ἰσχιάδα, οἱ δὲ φύλλον, οἱ δὲ λάδανον 〈καλοῦσι〉 (ʻSome people call leucacantha polygonaton, some ischias, some phyllon, some ladanonʼ1). But the conjecture should be rejected for two reasons: firstly, phymos or phimos is also the reading of the manuscripts in Plinyʼs index of chapters (1, 22, 17); secondly, the corruption of ΛΛ to Μ is easier than the corruption of LL to M, so that it seems likely that Pliny either misread a Greek manuscript or used a Greek manuscript that was already corrupt.2 The misunderstanding then belongs to Pliny or his source and should not be removed by conjecture. His nephew’s account of his working methods perhaps makes such a mistake unsurprising: memini quendam ex amicis, cum lector quaedam perperam pronuntiasset, revocasse et repeti coegisse; huic avunculum meum dixisse: ʻIntellexeras nempe?ʼ Cum ille adnuisset, ʻCur ergo revocabas? decem amplius versus hac tua interpellatione perdidimus.ʼ Tanta erat parsimonia temporis. (PLIN. epist. 3, 5, 12) ‘I remember one of his friends, when the reader mispronounced some words, stopped him and made him say them again; my uncle said to him: “You understood, didn’t you?” When he nodded, “Well why did you stop him? We’ve missed more than ten lines by your interruption.” That’s how sparing he was with his time.’

Editors of Pliny have sometimes allowed forms which misrepresent the Greek original to stand in Plinyʼs text, where it seems plausible that the error may have been in Plinyʼs original words,3 but clearly there are instances where they

1 There are some who think that Pliny made use of Dioscorides (e.g. Ernout 1959: 7–8), but most think that the sometimes striking similarities between the two texts are attributable to a common source, Sextius Niger, a point convincingly argued by Wellmann (1889). 2 For this point it makes no difference whether Plinyʼs source was Dioscorides or Sextius Niger, since Sextius Niger also wrote in Greek, see the list of authors in Plinyʼs index of chapters at 1, 12: Sextio Nigro qui graece de medicina scripsit (ʻSextius Niger, who wrote in Greek on medicineʼ; the same or similar description recurs in other lists throughout Book 1, e.g. 1, 13; 1, 20). 3 For instance, at PLIN. nat. 9, 98 both Mayhoff (1892–1909) and De Saint-Denis (1955) leave pinotheras of the manuscripts rather than replacing it with pinoteres to accord with Greek πινοτήρης (ARIST. hist. anim. 547b28 has nominative plural πιννοτῆραι, but the manuscripts

Authorial error and the dictionary

103

have tried to make Plinyʼs text more correct than Pliny himself had made it. At 13, 135 where the manuscripts have grason (M) or grasson (cett.), editors take Hermolaus Barbarusʼ conjecture prason, to accord with THEOPHR. hist. plant. 4, 6, 2 πράσον. Again the agreement of the index of chapters (at 1, 13, 52) speaks against this: the manuscripts there have grascon (M), grasson or crasson. Here too the misreading would be easier in Greek, from a Π with a short second leg to a Γ, and this time we also have the support of an ancient manuscript (M). At 21, 53, the manuscripts give thryganum, thriganum or similar, editors print the conjecture origanum to accord with THEOPHR. hist. plant. 6, 1, 4 ὀρίγανον. In the index of chapters (1, 21, 29), the manuscripts have thricanum, tricanum or triganum, and the conjecture supposes a more difficult corruption than in Greek from Ο to Θ. The evidence for retaining the manuscripts’ readings in these cases would be weaker if there was evidence that there had been scribal interference in the manuscript tradition, correcting either the index or the books it references to make the two agree. If such interference had taken place, it should have resulted in passages where the main text and the index agree in errors which cannot easily be attributed to Pliny himself. A thorough search for such evidence would need better and clearer manuscript reporting than current editions offer,4 and I have not attempted it. It is not hard to find such interference in individual manuscripts,5 but corrections of this sort at archetype level are at least rare.6

have πιννοθηραι or πινοθηραι). Unfortunately, the corresponding entries in Book 1 have conjectural pinotere for transmitted pinothera (1, 9, 51). At 13, 114 both Mayhoff (1892–1909) and Ernout (1956) reject Barbolausʼ conjecture of elloborinen for embolinen in Plinyʼs manuscripts to accord with DIOSC. 4, 108 ἐλλεβορίνην (the same confusion of ΛΛ and Μ as in phymos above). Where Pliny reworks the same material at 27, 76 (and 1, 27, 52) he seems to have correctly read elleborine. 4 For the manuscript tradition of Pliny, see most recently Reeve (2006, 2007). 5 For instance, at 1, 21, 52 the manuscripts dT add de cichorio referencing the text of 21, 88. T is a close relative of d, possibly a descendant (Reeve 2006: 155). At 1, 21, 38 most manuscripts have melanthium or similar, but d has meliantum to correspond with the text in 21, 65 where the manuscripts have melianthum or similar. The correct text is probably the index text, closer to the Greek source (THEOPHR. hist. plant. 6, 8, 1 τὸ μέλαν ἴον), and André (1969) was probably right to restore melanthium at 21, 65. In some passages where dT have a reading in the index which agrees, against the other manuscripts, with the reading in the main text, the wrong grammatical case has also been taken over: e.g. at 1, 22, 43 where most manuscripts have scolybo or scolibo, dT have scolymum like 22, 86; at 1, 22, 45 d’s chondrion, against other manuscripts’ harisc(h)ondro, reflects 22, 91 condrion. 6 One candidate would be 20, 117 where editors correct the manuscripts’ reading, hortoselinum or similar, to oreoselinum. In the index to the passage 1, 20, 46, the manuscripts have ortoselinum or similar, which editors likewise correct.

104

Nigel Holmes

It seems clear enough what an editor of Pliny’s text should do in such cases. Their business is not to improve Pliny, but to restore what he wrote. What should we want to find in a dictionary like ours, based on recorded usage? I think that again our business is to report the words that ancient language users have left us, even where they are only the result of a mistake. The Thesaurus Linguae Latinae has not always upheld this principle. We have an entry for phymos with the passages from Pliny and a brief explanation of our reason,7 but the entries for prason and origanum simply follow the editors and their conjectural text.8 At least in the case of prason we supply the evidence that this conjecture is a mistake. With origanum we do not even manage that. Perhaps we can make some amends to the reading thriganum when we get on to the letter T.

3 Meanings created through a mistake The Oxford English Dictionary cites among the meanings of the word jejune, (3b) ‘Puerile, childish; also, naïve’, noting at the same time that a false etymology (from Latin iuvenis or French jeune) apparently lies behind the use. Misunderstandings of a word found in reading can establish themselves as meanings in their own right, and a descriptive dictionary will need to find a place for them. Some errors, though, are not destined to find more than one victim. They too will have a place in a comprehensive dictionary, but it may be more grudging. In the Victorian poet Robert Browning’s work Pippa Passes, the epilogue has a puzzling passage: Then, owls and bats, Cowls and twats, Monks and nuns, in a cloister’s moods, Adjourn to the oak-stump pantry

A late 19th-century edition (Rolfe and Hersey 1887: 195) commented:

7 TLL vol. X, 1, p. 2062, 14–15 (phymos): ʻphyll- edd., cf. p. 2061, 59sqq.; sed error vix codd. lat. tribuendus estʼ. 8 TLL vol. X, 2, p. 1131, 39–43 (prason): ‘nomen gr. plantarum: 1 fruticis cuiusdam marini: PLIN. nat. 13, 135 (1, 13, 52 -on) orientis oceanus folio lato, colore viridi gignit quod quidam -on vocant, alii zostera (grasson sim. codd. utroque loco, corrigitur sec. Theophr. hist. plant. 4, 6, 2).ʼ TLL vol. IX, 2, p. 978, 24–25 (origanum): ʻ21, 53 in coronamenta folio venere melotrum, spiraea, -um eqs. (inter coronamenta etiam: 21, 55).ʼ

Authorial error and the dictionary

105

Twats is in no dictionary. We now have it from the poet (through Dr. Furnivall) that he got the word from the Royalist rhymes entitled ‘Vanity of Vanities,’ on Sir Harry Vane’s picture. Vane is charged with being a Jesuit. ‘’Tis said they will give him a cardinal’s hat: They sooner will give him an old nun’s twat.’ ‘The word struck me,’ says Browning, ‘as a distinctive part of a nun’s attire that might fitly pair off with the cowl appropriated to a monk.’

This, then, is a case where Browning guessed the meaning of a word encountered in his reading. In fact a ‘twat’ is not part of a nun’s attire, but a word for a woman’s genitals. The word is still in use today, but coarse enough that a well-brought-up Victorian might not have encountered it. The Oxford English Dictionary entry records Browning’s error. In this case though, unlike the case with the equally erroneous jejune, the dictionary brings the passage between the heading and the main body of the article (“Erroneously used (after quot. 1660) by Browning Pippa Passes iv. ii. 96 under the impression that it denoted some part of a nun’s attire”), as if to get it out of the way. Our largest source of errors of this kind are glossaries and translations, and they bring certain special problems with them, which we can approach by looking at a gloss from the Liber Glossarum. This work, sometimes known as the glossary of Ansileubus, is a compilation of the late seventh or early eighth century (Goetz 1891: 286–287).9 That date would normally make it too late for us to cite it in the Thesaurus, but glossaries are a special case for us. The individual glosses in a glossary may be older than the compilation, going back either to an older glossary or to a gloss in the margin of some ancient text. This compilatory character of many glossaries makes it often more useful to edit the glosses than to edit the glossaries. This is the approach of Goetz (1888–1923) in the Corpus Glossariorum Latinorum. The various glossaries are edited not as conventional critical editions, but as transcriptions of a single manuscript, retaining even the most evident errors. An apparatus reports variants from other manuscripts or early editions, but Goetz reserved the work of figuring out what was the true reading behind any gloss for the index of the whole collection. Goetz’s rival Lindsay is much more enthusiastic in his efforts to reconstruct the text, but I am not always sure just what text he was trying to reconstruct. A puzzling gloss on Virgil gives us a good illustration of the problem. In Lindsay’s (1926) edition the gloss (GLOSS.L I Ansil. PE 805) looks like this:

9 Lindsay dated the work even later, seeing it as a product of Charlemagne’s revival of learning (Lindsay 1926: 8). His view is still influential, but Conduché (2016) among others supports Goetz’s view that the latest certain source used in the glossary is Julian of Toledo.

106

Nigel Holmes

-ifrastici (-ce) dies iberni ‘: periphrastically winter days’

The gloss is given a marginal source indication as coming from Virgil in the manuscripts, and Lindsay identifies this as VERG. Aen. 1, 745–746: quid tantum Oceano properent se tingere soles hiberni ‘why winter suns hasten so much to submerge themselves in Ocean’

He compares a later gloss in the same glossary (SO 125–126): soles: dies. soles hiberni: hibernales dies ‘suns: days. winter suns: winter days’

So there is little doubt that his ideas on the genesis of our gloss are basically correct. The original gloss explained the expression ‘winter suns’ with the comment ‘periphrastically winter days’. But ‘soles hiberni’ got detached from the gloss, and as a result ‘periphrastically’ was reinterpreted as the word needing a definition. Lindsay changes perifrastici here to perifrastice to make an adverb. This change is paleographically easy, but it is probably wrong. In surviving Latin texts the word for ‘periphrastically’ kept the Greek adverbial ending periphrasticos (περιφραστικῶς). 10 Now if a copyist saw a gloss periphrasticos: dies hiberni, they might reasonably think that an excerpter had taken a gloss from a text with periphrastici in the accusative and failed to put it properly into the nominative form. Servius, whose commentary has many parallels with other glosses in the Liber Glossarum,11 comments at VERG. Aen. 1, 745 soles . . . pro diebus

10 TLL vol. X, 1, p. 1490, 19–27. We find the same pattern with metonymicos. Although Thesaurus headings often use the word metonymice, the ancient sources only offer metonymicos (μετωνυμικῶς: TLL vol. VIII, p. 891, 31–41). Some similar Greek derived adverbial terms occur in both forms, e.g. Latin ironice (TLL vol. VII, 2, p. 382, 82–383, 3) and (more common) Greek ironicos (εἰρωνικῶς: TLL vol. VII, 2, p. 383, 4–23), metaphorice (TLL vol. VIII, p. 876, 53–57) and (more common) Greek metaphoricos (μεταφορικῶς: TLL vol. VIII, p. 876, 39–52), Latin hyperbolice (TLL vol. VI, 3, p. 3149, 84–3150, 5) and Greek hyperbolicos (ὑπερβολικῶς: TLL vol. VI, 3, p. 3149, 77–84). 11 Servius, Servius auctus and other ancient commentators are reflected in the Virgil glosses (Lindsay 1937; Gorla 2016). The Harvard Servius does not note our gloss in its apparatus of parallels. Lindsay (1937: 2) had a list of glosses prepared for the Servius editors, but they never received them (Rand et al. 1946: xii). Even so, the cited parallels show how close the relationship is between the two works. Within a couple of lines of our passage they mention

Authorial error and the dictionary

107

posuit, and I wonder whether our gloss is not a hint that a periphrasticos has dropped out of the text of Servius here. Compare SERV. georg. 1, 111 gravidas aristas periphrasticos pro spicis posuit (ʻhe put laden beards periphrastically for earsʼ).12 The Thesaurus entry for periphrasticos not only brings the passage as a possible example for periphrastice but also plays with the idea of keeping periphrastici, so that the original gloss would then have been ‘winter suns: periphrastic winter days’ (i.e. ‘periphrastically described winter daysʼ).13 It just about makes sense, but it is very hard to believe that someone would actually choose to write it. But whether the original gloss had periphrastice, periphrasticos or periphrastici, it was not the same gloss as what the creator of the Liber Glossarum produced. The gloss as edited by Lindsay starts with ‘SO’ but it is in the PE section of the glossary. For the author of the glossary periphrastici dies meant ‘winter days’, and what Lindsay has restored is not the Liber Glossarum. We can leave the question of what Lindsay was trying to do for the next editor of the glossary.14 What should the Thesaurus do with passages like this? There are two basic alternatives: to treat it like other meanings of the word and give it its own section (with a warning perhaps that the meaning in question is based on a misunderstanding), or to quarantine it off in some way (the same variety of treatment that we saw in the Oxford English Dictionary at the beginning of this section). The second alternative might involve citing the passage in the head of the article, perhaps in the section which gives ancient definitions of a word, perhaps in the section on textual errors. If the error arose in explanation of a passage quoted in the article, we might append the gloss to the passage in question, noting that it was a mistaken explanation.15

GLOSS.L I Ansil. GE 49 and SE 101 on SERV auct. Aen. 1, 744 and 747 respectively (Rand et al. 1946: 306–307). 12 Serviusʼ preferred expression, however, is per periphrasin. Servius auctus favours periphrasticos. Since the other text to offer numerous instances of periphrasticos is Donatusʼ commentary on Terence, it might be that the source of the word in both Servius and Servius auctus is Donatus. If so, it could be that the gloss also reflects not Servius, but his source Donatus. 13 TLL vol. X, 1, p. 1490, 25–26. 14 The new online edition of the glossary (liber-glossarum.huma-num.fr/index.html) cites Lindsay’s explanation and edits the text as follows: PE805 Perifrastici [perifrastice] Virgili: Perifrastici [perifrastice] dies iberni. 15 For instance, at TLL vol. VII, 2, p. 1863, 13–14 (luridus): ʻEVAGR. vita Anton. 40 -am faciem serpentis agnovi (aliter gr.; inepte GLOSS. ad l.: luto sordidam. luto pollutam)ʼ.

108

Nigel Holmes

In general the Thesaurus treats differently ancient passages that talk about the usage of a word and actual instances of a usage. The sections in the main body of the article normally contain the instances of usage, not what grammarians or glossaries say about them. But if the only evidence for an ancient usage is the testimony of grammarians and glossaries, we will probably create a section with just this. Sometimes we do this even when the meaning is based on a mistake. So in the entry for pelicanus most passages concern the pelican, as you might expect, but there are glosses that claim that the word means porphyrio (which is elsewhere the purple gallinule) and these get their own section.16 We attribute the error to passages in the Bible that mention the two birds together. I am not sure what made the writer of pelicānus decide to take this approach. Possibly the meaning is simply a little more plausible than the other cases. The names of animals and plants often get transferred when people enter a new environment armed with an already formed vocabulary. Still it is hard to feel confident what the creator of the gloss, who had very likely never seen a pelican or a purple gallinule, meant to say.17 Consistency is very important for reference works and we try hard to maintain it. But cases like these often have no perfect solution. Different principles push us in different directions, and a slight difference in the weighting one gives them may lead to a quite different treatment of cases that on the surface seem quite similar.

Bibliography TLL = Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. André, J. (1969). Pline l’Ancien, Histoire Naturelle, Livre XXI. Paris: Les Belles Lettres. Conduché, C. (2016). Présence de Julien de Tolède dans le Liber glossarum. In Dossiers d’HEL, SHESL, 2016, Le Liber glossarum (s. VII–VIII): Composition, sources, réception, (pp. 141–157). htl.linguist.univ-paris-diderot.fr/hel/dossiers/numero10 (accessed: 2 January 2019). De Saint-Denis, E. (1955). Pline L’Ancien, Histoire Naturelle, Livre IX. Paris: Les Belles Lettres. Ernout, A. (1956). Pline L’Ancien, Histoire Naturelle, Livre XIII. Paris: Les Belles Lettres.

16 TLL vol. X, 1, p. 996, 9 (pelicanus): ʻ2 i. q. porphyrio, sc. per confusionem ex p. 995, 61 ortam: GLOSS. IV 557, 61 porphyrionem: -um (item V 332, 10. 380, 13)ʼ. 17 The problem becomes sharper when we see that after using the gloss to show that pelicanus can mean purple gallinule in the Thesaurus entry pelicanus (n. 17 above), the same gloss is cited in the Thesaurus entry porphyrio to show that porphyrio can mean pelican: TLL vol. X, 1, p. 2750, 54–55.

Authorial error and the dictionary

109

Ernout, A. (1959). Pline L’Ancien, Histoire Naturelle, Livre XXVII. Paris: Les Belles Lettres. Goetz, G. (1888–1923). Corpus Glossariorum Latinorum. Leipzig: Teubner. Goetz, G. (1891). Der Liber Glossarum. Leipzig: Hirzel. Gorla, S. (2016) Per una definizione delle glosse virgiliane contenute nel Liber glossarvm con indicazione Virgili. In Problemi e prospettive. Dossiers d’HEL, SHESL, 2016, Le Liber glossarum (s. VII–VIII): composition, courses, réception (pp. 209–224). htl.linguist.univparis-diderot.fr/hel/dossiers/numero10 (accessed: 2 January 2019). Lindsay, W. M. (1926). Glossaria Latina vol. I. Glossarium Ansileubi sive Librum Glossarum. Paris: Les Belles Lettres. Lindsay, W. M. (1937). Vergil Scholia in the Ansileubus Glossary. American Journal of Philology, 58, 1–6. Mayhoff, C. (1892–1909) C. Plini Secundi Naturalis Historiae Libri XXXVII. Leipzig: Teubner. Rand, E. K., Stocker, A. F. & Hartman, A. (1946). Servianorum in Vergilii Carmina Commentariorum (vol. II). Lancaster: American Philological Association. Reeve, M. D. (2006). Manuscripts of Plinyʼs Natural History in Spain. Exemplaria Classica, 10, 151–186. Reeve, M. D. (2007). The Editing of Plinyʼs Natural History. Revue d’histoire des textes, 2, 107–179. Rolfe, W. J. & Hersey H. E. (1887). Select Poems of Robert Browning. New York: Harper and Brothers. Wellmann, M. (1889). Sextius Niger. Hermes, 24, 530–569.

Sophie Van Laer

Les verbes de déplacement en latin : préverbation et arguments Abstract: This chapter concerns the argument structure of motion verbs with preverbs. A motion verb takes at least two arguments, the subject and a complement. A preverb, being a relator, also takes two arguments, the locatum and the locator. The question that arises is: how does the locator connect to the argument structure of the verb? This study shows that the relator is very often identified with the complement of the verb (a prepositional phrase whose head does not necessarily correspond to the preverb). Relators can also remain implicit, in particular in situations where they would introduce an argument that goes beyond what is permitted by the argument structure of the verb. There are only two cases that seem to be closely linked to the presence of a preverb: (non-prepositional) accusative and dative. Keywords: Latin linguistics, motion verb, argument structure, preverb, spatial complement, argument

1 Introduction La problématique de cet article porte sur l’articulation entre deux structures argumentales, celle du verbe de déplacement et celle du préverbe. 1)

(1)

Un “verbe de localisation met en jeu une relation locative entre deux arguments au moins” (Boons 1987 : 5–6). Lorsqu’il s’agit d’un verbe dénotant le déplacement de l’argument sujet, ces deux arguments sont : – l’entité mobile (argument sujet) – le complément de lieu1

Paul va à Munich.

1 Sur les questions que soulève le statut d’argument du complément de lieu, voir Cabrillana (1997a : 328 et 1997b : 118–119). Le cas le plus complexe est sans doute celui des verbes de déplacement dépourvus d’orientation (erro, ambulo) : voir Section 3.3 et n. 23. Sophie Van Laer, Université de Nantes https://doi.org/10.1515/9783110647587-008

Les verbes de déplacement en latin : préverbation et arguments

111

Paul est l’entité mobile, à Munich le complément de lieu, indiquant ici le lieu final du déplacement. 2)

Les préverbes, tout comme les prépositions, sont des relateurs dont le rôle est de mettre en relation deux entités, l’une étant située par rapport à l’autre. Nous appellerons ces deux entités2 : – “l’entité située” (ES) – “l’entité-repère” (ER)

(2a) Paul est dans un hôtel. Le relateur dans localise Paul (ES) à l’intérieur d’un repère : un hôtel (ER). La localisation est ici prédiquée par le verbe est. (2b) L’homme sur le toit installe l’antenne. Le relateur sur localise l’homme (ES) à la surface supérieure d’un repère : le toit (ER). La localisation est effectuée directement par la préposition sur, sans être prédiquée par un verbe. (2a) et (2b) font intervenir une préposition. Mais le préverbe, en latin, se soude au verbe pour former une unité lexicale.3 L’enjeu est donc d’articuler deux structures argumentales, celle du verbe de déplacement et celle du préverbe. L’étude ne portera que sur les verbes dénotant le déplacement de l’argument sujet. En ce cas, on peut supposer que l’ES du préverbe coïncide avec le sujet grammatical de la base verbale. Mais une question reste en suspens : comment le repère (ER) du préverbe va-t-il s’intégrer à la structure argumentale du verbe ?4 Pour conduire l’étude, nous prendrons appui sur la classification des verbes locatifs proposée par Boons (1987), qui s’organise autour de deux critères.

2 Terminologie de Van Laer (2010 : 24). Différents termes sont employés pour désigner le couple ES/ER : locatum/relatum (Lehmann 1983), lieu/corrélat de lieu (Guillet et Leclère 1992), cible/site (Vandeloise 1986), trajector/landmark (Langacker 1987), figura/fondo (Palachi 2010), ces deux termes reprenant le couple Figure/Ground utilisé par Talmy (1975 : 419). Dans le résumé anglais, les termes utilisés sont locatum et locator. 3 Voir Serbat (2001 [1997]) pour l’incidence syntaxique du préverbe. 4 Problématique déjà soulevée par Lehmann (1983).

112

Sophie Van Laer

Le critère de “l’orientation inhérente au procès” (Van Laer 2010 : 32–37)5 prend en compte le caractère orienté ou non du déplacement : le déplacement s’organise-t-il à partir d’un point de départ et d’un point d’arrivée distincts, ou s’agit-il d’un déplacement dans l’espace qui ne s’oriente pas vers un lieu final ? Ce premier critère permet de définir 3 sous-classes de verbes de déplacement,6 que nous appellerons : 1) verbes de déplacement (VDD : orientés) : eo ‘aller’, uenio ‘venir’. 2) verbes de mode de déplacement (VMD : orientés ou non orientés), faisant intervenir une modalité physique de déplacement : curro ‘courir’, salio ‘sauter’, uolo ‘voler’. 3) verbes de déplacement intrinsèque (VDI : non orientés) : ambulo ‘se promener’, erro ‘errer’. Les bases données comme exemple, prototypiques de chacune de ces sousclasses, sont les seules que nous étudierons. Le deuxième critère évalue la compatibilité des verbes avec les différentes catégories de lieu. Il fait donc intervenir les 3 catégories de lieu en lien avec un déplacement : 1) le “lieu initial” (L1) 2) le “lieu final” (L2) 3) le “lieu médian” (L3) qui correspond au lieu par où l’on passe. Il convient de distinguer du “lieu médian” (L3) le “lieu du procès” (L4), propre à localiser n’importe quel procès (Guillet et Leclère 1992 : 218). (3a) Il se promène dans le jardin. (L3 avec un VDI) (3b) Il lit dans le jardin. (L4 : le verbe ne dénote pas un déplacement) L’étude se fera sur corpus.7 Le corpus retenu associe poètes, historiens et autres prosateurs. On peut en effet supposer que la poésie, plus économe

5 Correspondant à ce que Boons (1987 : 19–22) appelle critère de “l’amplitude aspectuelle”. 6 La typologie de Palachi (2010) ne distingue que deux sous classes, les Verbos de Movimiento + Trayecto (qui correspondent aux VDD) et les Verbos de Movimiento + Manera (qui regroupent les VMD et les VDI). 7 Détail du corpus : poètes (CATVLL., VERG. Aen., OV. fast.), historiens (SALL. Catil., SALL. Iug., CAES. Gall., LIV. I–IV), autres prosateurs (CIC. Verr. II, CIC. fin., SEN. epist. 1–88). À ce corpus ont parfois été ajoutés des occurrences significatives issues de la consultation du Thesaurus

113

Les verbes de déplacement en latin : préverbation et arguments

de prépositions,8 offre un traitement syntaxique particulier du complément d’un verbe préverbé, d’où l’intérêt de mettre en regard des œuvres relevant d’une prose stylistiquement relevée. N’ont été retenues que les occurrences dénotant un déplacement dans l’espace d’un animé, généralement humain.9 Nous envisagerons successivement les préverbes supposés, au vu de leur signifié, entretenir des liens privilégiés avec l’un des lieux du déplacement : ab-, de- et ex- pour L1 (Section 2) ad-, in- et ob- pour L2 (Section 3) per-, trans- et praeter- pour L3 (Section 4) Nous étudierons comme contrepoint com-, a priori sans rapport avec les lieux du déplacement (section 5).

2 Les préverbes en lien avec le lieu initial : ab-, ex- et deTab. 1: Nombre total d’occurrences du corpus (inclut les emplois où le verbe ne dénote pas le déplacement dans l’espace d’une entité animée). Les cases sont grisées lorsque le verbe préverbé n’est pas attesté (pas d’entrée dans le TLL). VDD eo ab-



ex-



detotal



VMD uenio

curro

salio

VDI uolo

ambulo



total erro 





































On constate dans le tableau 1 une inégalité numérique entre les préverbes (deest sensiblement moins représenté) et une disparité dans la répartition des préverbes selon les bases qui guidera notre étude.

Linguae Latinae. Le relevé a été fait à partir de la base de données LLT-A. Sauf indication contraire, les traductions proposées sont celles de la CUF. 8 Hypothèse corroborée par une étude sur corpus (voir McGillivray 2013 : 125–127). 9 Cet article présente donc une étude qui se veut systématique d’un ensemble homogène de verbes dans un corpus contrastif (prose et poésie) étendu à tous les préverbes supposés être en lien avec un déplacement dans l’espace.

114

Sophie Van Laer

2.1 Le cas des VDD10 : abeo et exeo L’angle d’approche utilisé sera celui de la catégorie de lieu des compléments exprimés. Les VDD préverbés en ab- et en ex- manifestent-ils une affinité avec L1 ? Tab. 2: Catégorie de lieu des compléments exprimés. L

L

L + L

abeo







exeo





total







ø

total

total 























L

Lecture du tableau (Tableaux 2 et 4) : – ‘ø’ : pas de complément exprimé ; – ‘total’ : total des occurrences dénotant un déplacement dans l’espace d’une entité animée ; – ‘total 2’ : total des occurrences du corpus. Le repère du déplacement (voir tableau 2) est souvent effacé, en particulier pour abeo (60,5% des occurrences). D’autre part, abeo s’avère presque également compatible avec L1 et L2, contrairement à ce que l’on pourrait attendre. Doit-on considérer que le préverbe n’a aucune incidence ? Ab- marque l’éloignement d’un repère afférent déictique (cf. Palachi 2010 : 465), souvent restituable contextuellement, ce qui permet son effacement : (4) Si periturus abis, et nos rape in omnia tecum. (VERG. Aen. 2, 675) ‘Si tu (= Énée) nous quittes pour mourir, emmène-nous également avec toi, à toute aventure.’ Il convient donc de distinguer le plan sémantico-logique, où ab- réalise un apport, et le plan syntaxique, où cet apport ne se réalise pas nécessairement par la présence d’un complément. Dans le cas d’exeo, l’expression de L1 est plus nettement privilégiée, mais le préverbe apporte des restrictions sur le choix d’ER. En effet, ex- s’applique à

10 Seuls ab- et ex- peuvent être mis en regard pour les VDD. Euenio prédique toujours, dans notre corpus, l’arrivée d’un événement.

Les verbes de déplacement en latin : préverbation et arguments

115

la sortie d’un repère formalisé comme une limite double (Pottier 1962 : 280), c’est-à-dire qui comporte un espace intérieur délimité par cette limite double. (5) Caesar . . . milites ex oppido exire iussit. (CAES. Gall. 2, 33, 1) ‘César ordonna que . . . les soldats sortissent de la ville.’ Mais exeo est également le seul à recourir à L3 pour préciser par où s’opère la sortie prédiquée par le verbe. En ce cas, le lieu initial, repère du préverbe, peut rester implicite : (6) Illa (= porta Carmentis) fama refert Fabios exisse trecentos. (OV. fast. 2, 203) ‘C’est par elle (= la porte Carmentale) que sont sortis, selon la tradition, les trois cents Fabius.’ Le préverbe n’ouvre donc pas nécessairement une place de complément. Il se greffe sur la structure argumentale de la base en apportant des précisions, donc des restrictions, sur le choix de L1 (exprimé ou implicite), en vertu de sa valeur sémantique. La forme prise par le complément du verbe, s’il relève de L1, n’est pas figée. Lorsqu’il s’agit d’un groupe prépositionnel, la préposition n’est pas nécessairement identique au préverbe : (7) profugus ex patria abierat. (SALL. Iug. 35, 1) ‘ avait dû . . . s’enfuir et s’exiler de sa patrie.’ (8a) ciuitati persuasit ut de finibus suis cum omnibus copiis exirent. (CAES. Gall. 1, 2, 1) ‘ persuada à ses concitoyens de sortir de leur pays avec la population entière.’ (8b) Post eius mortem nihilo minus Heluetii id quod constituerant facere conantur, ut e finibus suis exeant. (CAES. Gall. 1, 5, 1) ‘Après la mort d’Orgétorix, les Helvètes n’en persévèrent pas moins dans le dessein qu’ils avaient formé de quitter leur pays.’

116

Sophie Van Laer

2.2 Le cas des VMD : decurro Nous n’étudierons que de- qui se caractérise par la diversité du repérage. Si desilio ne fait intervenir que L1 et L2, dans un déplacement vertical,11 decurro fait intervenir un autre type de complément, dénotant le “lieu parcouru” et prenant la forme d’un accusatif non prépositionnel. (9) Cum lecti iuuenes, Argiuae robora pubis auratam optantes Colchis auertere pellem aussi sunt uada salsa cita decurrere puppi. (CATVLL. 64, 4–6) ‘Quand des jeunes gens d’élite, force de la jeunesse argienne, voulant ravir à la Colchide la toison d’or, osèrent lancer sur les flots salés leur nef rapide.’ Cet accusatif peut être interprété comme un accusatif d’extension spatiale qui, dans la catégorie du ‘lieu médian’, est ordinairement introduit par la préposition per-.12 Cette forme non prépositionnelle du complément semble directement liée à la présence d’un préverbe, même si les témoignages des grammairiens anciens envisagent essentiellement ce cas lorsque le préverbe est per- : (10) Nos dicimus ‘per undas labitur’, illi dicebant ‘perlabitur undas’. Item ‘per forum curro’ et ‘percurro forum’. (SERV. Aen. 1, 147) ‘Nous, nous disons per undas labitur, ‘voler au-dessus des eaux’, lorsqu’ils (= les Anciens) disaient perlabitur undas, ‘survoler les eaux’. Il en est de même pour per forum curro, ‘courir par le forum’, et percurro forum, ‘parcourir le forum’.’13 S’il semble sémantiquement lié à L3, cet accusatif d’extension spatiale est traité comme un objet grammatical et peut devenir sujet en cas de transformation passive, ou nom support d’un participe parfait passif :

11 Lorsque L1 est exprimé par un groupe prépositionnel, la préposition n’est pas non plus toujours identique au préverbe : de nauibus desiliendum (CAES. Gall. 4,24,2) ‘ devaient sauter à bas des navires’ face à uniuersi ex naui desiluerunt (CAES. Gall. 4,25,5) ‘ils sautèrent ensemble hors du vaisseau’. 12 Tandis que l’accusatif de durée (= d’extension temporelle) se réalise usuellement sous forme non prépositionnelle. 13 Traduction de Baudou et Clément-Tarantino (2015).

Les verbes de déplacement en latin : préverbation et arguments

(11)

117

Nec uero uelim quasi decurso spatio ad carceres a calce reuocari. (CIC. Cato 83) ‘Et je ne voudrais certes pas, parvenu, pour ainsi dire, au bout de ma carrière, être ramené de la ligne aux écuries.’

Ovide étend audacieusement cette possibilité à l’accusatif directif qui relève de L2, tandis qu’en ce cas, Cicéron maintient la construction prépositionnelle, conforme à l’usage du latin : (12a) Dum notat haec hospes, decursa nouissima meta est. (OV. met. 10, 597) ‘Pendant que l’étranger (= Hippomène) observe ce spectacle, la borne a été franchie à l’extrémité de la carrière.’ (12b) in uita nunc uideo calcem, ad quam cum sit decursum, nihil sit praeterea extimescendum. (CIC. Tusc. 1, 15) ‘À présent je vois dans l’existence un terme au-delà duquel nous n’aurons plus rien à redouter, quand une fois nous l’aurons atteint.’

2.3 Le cas des VDI : deerro La compatibilité des préverbes avec les VDI est faible, sans doute du fait qu’ils sont dépourvus d’orientation, c’est-à-dire non directionnels. Une occurrence de Virgile (hors corpus, cf. n. 7) nous permettra d’envisager le cas de de-. De-, lorsqu’il est associé à la base -uenio, est systématiquement accompagné, dans notre corpus, d’un complément relevant de L2. De fait, le VDD uenio envisage l’atteinte de L2.14 La compatibilité du préverbe semble alors s’expliquer par la valeur directive que Brachet (2000 : 34) prête au préverbe. C’est cette valeur directive, semble-t-il, qui explique pour le VDI deerro le repérage du déplacement par l’adverbe déictique huc en (13), alors que la base est dépourvue d’orientation. Le bouc s’est éloigné du troupeau, et son errance l’a conduit près des bergers. (13) Huc mihi, dum teneras defendo a frigore myrtos, uir gregis ipse caper deerrauerat . . . (VERG. ecl. 7, 6–7) ‘En cet endroit, tandis que j’abritais du froid mes tendre myrtes, le mâle du troupeau, mon bouc, s’était égaré . . .’ 14 Voir Brachet (2000 : 62), Orlandini et Poccetti (2011 : 28–29) et les données statistiques de Cabrillana (1997a : 330).

118

Sophie Van Laer

2.4 Conclusion De cette deuxième section, nous retiendrons deux faits : – le complément exprimé ne relève pas nécessairement de L1 et ne prend pas nécessairement la forme d’un groupe prépositionnel introduit par une préposition identique au préverbe ; – seul decurro fait parfois intervenir un accusatif d’extension spatiale qui semble directement lié à la présence d’un préverbe. Cette remarque reflète la singularité de de- dans l’ensemble des préverbes étudiés : plutôt que de prendre appui sur L1 (exprimé ou implicite), il apporte à la base une valeur directive.

3 Les préverbes en lien avec le lieu final : ad-, in- et obDu fait de la fréquence d’occurro (voir tableau 3), ce deuxième ensemble de préverbés est moins nettement dominé par les VDD. La compatibilité avec les VDI reste très faible. Tab. 3: Nombre total d’occurrences du corpus. ‘SP’ : sens particulier (le préverbé ne dénote plus un déplacement dans l’espace, même figuré). VDD

VMD

VDI

eo

uenio

curro

salio

uolo

ad-











in-



SP







ob-













total

ambulo

total erro 



 











3.1 Le cas des VDD 3.1.1 Adeo et ineo À la différence de ce que nous avons vu pour ab- et ex-, les VDD préverbés en adet en in- sont systématiquement repérés par L2, lorsqu’ils ont un complément (voir tableau 4).

Les verbes de déplacement en latin : préverbation et arguments

119

Tab. 4: Catégorie de lieu des compléments exprimés. Légende : voir Section 2.1. L

L

L + L

L

ø

total

total 

adeo









ineo















total









Mais, comme le montre le tableau 5, le complément de lieu prend le plus souvent la forme d’un accusatif directif (non prépositionnel),15 alors même que la répartition des occurrences est à peu près égale entre poètes, historiens et autres prosateurs, ou entre déplacement en direction d’une personne et déplacement en direction d’un lieu pour adeo16 : Tab. 5: Forme syntaxique du complément de lieu (L2).

adeo

ad + ACC.

in + ACC.

ACC.

adv.

total



























ineo total



Comment interpréter la différence entre accusatif directif (sans préposition) et accusatif prépositionnel ? Sur le plan syntaxique, les deux constructions n’offrent pas les mêmes possibilités. Si L2 est non prépositionnel, il est traité comme un objet grammatical et peut devenir sujet en cas de transformation passive. Le groupe prépositionnel, quant à lui, se maintient à la voix passive et oblige à recourir au passif impersonnel. Ceci vaut lorsque L2 réfère à un inanimé (14a et 14b) ou à une personne (15a et 15b) : (14a) libri Sibyllini ex senatus consulto aditi sunt. (LIV. 5, 13, 5) ‘un sénatus-consulte renvoya aux livres Sibyllins.’

15 L’accusatif directif, qui est d’usage pour les noms de villes, est alors systématiquement utilisé pour le repère. 16 35 occurrences chez les poètes, 31 chez les historiens et 29 chez les autres prosateurs. 43 dénotent un déplacement vers un humain, 52 vers un lieu.

120

Sophie Van Laer

(14b) aditum est ad libros Sibyllinos. (CIC. Verr. II 4, 108) ‘on alla consulter les livres Sibyllins.’ (15a) Sol aditus. (OV. fast. 4, 583) ‘Le Soleil interrogé.’ (15b) aditum est ad Metellum. (CIC. Verr. II 3, 152) ‘on alla auprès de Métellus.’ Cette différence syntaxique a-t-elle une incidence sémantique ? Même Cicéron, qui semble privilégier l’emploi du groupe prépositionnel, présente des cas d’alternances de construction : (16a) totiens apud maiores nostros Siculi senatum adierunt. (CIC. Verr. II 2, 146) ‘tant de fois, du temps de nos ancêtres, les Siciliens se sont adressés au Sénat.’ (16b) Agit mecum et cum fratre meo, ut si nobis uideretur, adiremus ad eorum senatum. (CIC. Verr. II 4, 137) ‘Il propose à mon cousin et à moi d’aller devant leur sénat, si nous le voulions bien.’ Il semble difficile de préciser cette différence. Peut-être est-elle comparable à celle qui existe en français entre : (17a) Il habite Nantes. (où Nantes est traité comme un objet grammatical) (17b) Il habite à Nantes. (où Nantes est traité comme un lieu) (18a) Depuis ce jour, la peur l’habite.17 (18b) Depuis ce jour, la peur habite en lui.

17 Même si le complément prend la forme d’un objet grammatical, la transformation passive semble peu naturelle pour (17a) alors qu’elle est usuelle pour (18a) : Depuis ce jour, il est habité par la peur.

Les verbes de déplacement en latin : préverbation et arguments

121

3.1.2 Obeo Le répérage d’obeo est différent, ce qui s’explique par les deux valeurs que peut prendre le préverbe : – la valeur de “face à face” (enfrentamiento) définie par García-Hernández (1980 : 172). – la valeur de “recouvrement”, définie par Van Laer (1998 et 2010 : 80).18 C’est la valeur de recouvrement qui est à l’œuvre avec obeo. Lorsque le verbe dénote un déplacement dans l’espace, il se construit systématiquement avec un complément à l’accusatif d’extension spatiale, exprimant le lieu parcouru.19 (19a) Ego Siciliam totam quinquaginta diebus . . . obii. (CIC. Verr. I 6) ‘et moi, en cinquante jours, j’ai parcouru la Sicile tout entière.’ Même s’il présente une différence sémantique avec l’accusatif directif, le complément d’obeo est également traité comme un objet grammatical et peut devenir sujet en cas de transformation passive ou support nominal de l’adjectif verbal : (19b) propterea quod tum putant obeundam esse maxime prouinciam, cum in areis frumenta sunt. (CIC. Verr. II 5, 29) ‘pensant que la province soit être visitée surtout au moment où les blés sont sur les aires’

3.2 Le cas des VMD : accurro, incurro et occurro C’est également L2 qui repère les VMD formés sur la base -curro lorsqu’ils ont un complément. Dans le cas d’occurro, le préverbe a la valeur de “face à face”. La forme syntaxique du complément fait intervenir un nouveau cas, le datif. Ce datif ne se rencontre pas avec accurro dans notre corpus, où le repère est souvent effacé et prend une forme prépositionnelle quand il est exprimé. (20) Considius equo admisso ad eum accurrit. (CAES. Gall. 1, 22, 2) ‘Considius accourt vers lui (= César) à bride abattue.’ 18 Voir, en dernière analyse, García-Hernández (2016). 19 Obeo est alors paraphrasable par ‘couvrir de son déplacement’. Cet emploi est assez rare : 6 occurrences dans notre corpus.

122

Sophie Van Laer

Pour incurro, la distribution entre groupe prépositionnel et datif semble liée à des considérations génériques20 : les prosateurs privilégient le groupe prépositionnel, les poètes marquent une préférence pour le datif. Le choix de la préposition in contribue à générer la connotation hostile du déplacement.21 (21a) Catilina . . . in confertissumos hostis incurrit ibique pugnans confoditur. (SALL. Catil. 60, 7) ‘Catilina . . . se jette au plus fort de la mêlée, et tombe percé de coups en combattant.’ (21b) . . . Nec armentis incurrere fortibus ursi. (OV. met. 7, 546) ‘ . . . Les ours ont cessé d’attaquer les grands troupeaux.’ Occurro systématise le repère au datif (22a et 22b), sauf lorsque le déplacement s’oriente vers un lieu (23) : (22a) obuia cui (= Turno) Volscorum acie comitante Camilla / occurrit . . . (VERG. Aen. 11, 498–499) ‘Devant lui (= Turnus) paraît Camille, qui court à sa rencontre, accompagnée d’une troupe de Volsques . . .’ (22b) A. Postumius dictator, T. Aebutius magister equitum, . . . ad lacum Regillum in agro Tusculano agmini hostium occurrerunt. (LIV. 2, 19, 3) ‘Aulus Postumius, dictateur, Titus Aebutius, maître de cavalerie . . . rencontrèrent l’ennemi au lac Régille, sur le territoire de Tusculum.’ (23) ut quaeque pars castrorum nudata defensoribus premi uidebatur, eo occurrere et auxilium ferre. (CAES. Gall. 3, 4, 2) ‘chaque fois qu’un point du camp, dégarni de défenseurs, paraissait menacé, on accourrait à la rescousse.’ Cet emploi du datif est conforme à l’analyse qu’en fait Serbat (1996 : 184) qui y voit un “repère de visée”. Il semble directement lié à la présence du préverbe.

20 Nous nous appuyons ici sur un corpus élargi, prenant en compte toutes les œuvres des auteurs du corpus. 21 Elle s’explique par l’inadéquation entre la formalisation du repère (limite double) et le référent de ce repère (souvent un groupe d’individus). La même connotation est présente dans l’expression française familière foncer dans le tas ou rentrer dans le lard.

Les verbes de déplacement en latin : préverbation et arguments

123

3.3 Le cas des VDI : obambulo Dépourvus d’orientation, les VDI sont difficilement compatibles avec l’expression de L2, qui prédomine pour ces préverbes. D’où leur faible présence dans notre corpus. Nous n’envisagerons que le cas d’obambulo dont l’étude se fera à partir d’un corpus élargi. Le préverbe, doté de la valeur de face à face,22 donne une orientation indirecte au déplacement apparemment désordonné, et permet d’introduire un repère relatif autour duquel ce déplacement s’organise : (24) Non lupus insidias explorat ouilia circum nec gregibus nocturnus obambulat . . . (VERG. georg. 3, 537–538) ‘Le loup ne cherche plus à s’embusquer autour des bergeries et ne rôde plus la nuit près des troupeaux . . .’ L’objectif du loup est de rester suffisamment près des troupeaux pour pouvoir attaquer dès qu’une occasion favorable se présente. Mais le repère appelé par le préverbe peut ne pas être exprimé : (25) induitur faciem tauri mixtusque iuuencis mugit et in teneris formosus obambulat herbis. (OV. met. 2, 850–851) ‘II (= Jupiter) revêt l’apparence d’un taureau ; mêlé au troupeau, il mugit et promène ses belles formes sur le tendre gazon.’ Ici, le seul complément, qui est un circonstant,23 relève du lieu du procès. C’est le cadre dans lequel Jupiter, métamorphosé en taureau, se donne à voir. Le repère appelé par le préverbe (Europe et ses compagnes) reste implicite. Enfin, ce repère peut prendre la forme d’un groupe prépositionnel dénotant un lieu relatif, analysable comme un circonstant du procès : (26) milites inermes aut humi sedentes accubantesque aut obambulantes ante uallum portasque. (LIV. 25, 39, 8) ‘les soldats, désarmés, étaient assis ou couchés par terre, ou se promenaient devant le retranchement et les portes.’

22 Dans une occurrence hors corpus, ob- a une valeur de recouvrement (OV. met. 14, 188). 23 Le statut syntaxique exact du complément d’un VDI est complexe : le complément locatif ne se réalise pas nécessairement sous la forme d’un argument sur le plan syntaxique, mais il s’apparente à un argument logique ou sémantique du verbe, qui appelle une localisation.

124

Sophie Van Laer

Dans ce camp militaire carthaginois, les soldats semblent faire preuve d’indolence, mais évitent de s’éloigner trop de l’enceinte du camp. Le lieu relatif fait office de repère du préverbe, tout en définissant un espace pour le procès d’ambulo. La diversité des structures syntaxiques présentées montre que la forme prise par le repérage d’obambulo n’est pas stabilisée. Les différentes solutions retenues visent à rendre accessible un double repérage (le cadre du déplacement dénoté par -ambulo et le repère de visée appelé par ob-) à travers un seul syntagme

3.4 Conclusion Dans ce deuxième groupe de préverbés, généralement repérés par L2, nous voyons apparaître deux formes pour le complément qui semblent directement liées à la préverbation24 : – l’accusatif non prépositionnel, qu’il s’agisse d’un accusatif directif (adeo et ineo) ou d’un accusatif d’extension spatiale (obeo). – le datif, interprétable comme un ‘repère de visée’ (incurro et surtout occurro). Dans le cas d’obambulo, susceptible d’appeler un double repérage, la forme du repère n’est pas stabilisée mais les occurrences étudiées ne donnent jamais lieu à une double complémentation.

4 Les préverbes en lien avec le lieu médian : per-, trans- et praeterCe dernier ensemble, dominé par per- (voir tableau 6), présente une unité de fonctionnement qui apparaît en confrontant pour ce préverbe VDD et VMD. Nous étudierons donc successivement les différents préverbes.

24 Je rejoins ici les analyses de Serbat (2001 [1997] : 145–149) et de Palachi (2010 : 459 et 464–465).

Les verbes de déplacement en latin : préverbation et arguments

125

Tab. 6: Nombre total d’occurrences du corpus. ‘SP’ : sens particulier (le préverbé ne dénote plus un déplacement dans l’espace, même figuré). VDD

VMD

eo

uenio

curro

per-

SP





praeter-



trans-



total



salio

VDI

total

uolo

ambulo

erro



















  









4.1 Les préverbés en per- : peruenio et percurro Per- exprime un “parcours d’un bout à l’autre d’une limite double” (Pottier 1962 : 282). Le parcours dénoté peut connaître 2 repérages25 : 1) par un lieu envisagé dans son extension, exprimé sous la forme d’un accusatif d’extension spatiale. Le déplacement a lieu de sa limite initiale à sa limite finale. (27) Caesar . . . omnem agrum Picenum percurrit. (CAES. civ. 1, 15, 1) ‘César . . . parcourt tout le Picénum.’ 2) par le lieu final du déplacement, exprimé sous la forme d’un groupe prépositionnel. En ce cas, le déplacement est envisagé au moment où peratteint la limite finale du parcours. (28) neque prius fugere destiterunt quam ad flumen Rhenum . . . peruenerunt. (CAES. Gall. 1, 53, 1) ‘ils ne s’arrêtèrent qu’une fois parvenus au Rhin.’26 L’accusatif d’extension spatiale, non prépositionnel, peut être traité comme un objet grammatical, et devenir le support nominal d’un participe parfait passif : (29) orbe pererrato. (OV. fast. 4, 589) ‘à parcourir le monde.’

25 Ce qui permet de distinguer “verbes de parcours” (VP) et “verbes de parcours à visée finale” (VPF) : voir Van Laer (2010 : 90). 26 Traduction de la CUF légèrement modifiée.

126

Sophie Van Laer

4.2 Les préverbés en trans- : transeo Trans- exprime le “franchissement d’une limite double” (Pottier 1962 : 283). Son fonctionnement est assez comparable à celui de per-. – Le repérage peut se faire par le lieu final du déplacement, qui suppose le franchissement d’une limite correspondant à la limite finale du parcours. Le complément relève alors de L2. Il est de forme prépositionnelle.27 (30) ne propter bonitatem agrorum Germani, qui trans Rhenum incolunt, e suis finibus in Heluetiorum fines transirent. (CAES. Gall. 1, 28, 4) ‘car la bonne qualité des terres lui (= à César) faisait craindre que les Germains qui habitent sur l’autre rive du Rhin ne quittassent leur pays pour s’établir dans celui des Helvètes.’ – Lorsque le repérage se fait par l’espace ‘franchi’ (31) ou ‘longé’ (32, avec un effet de sens ‘dépassement’), il se réalise sous la forme d’un accusatif non prépositionnel d’extension spatiale. (31) si per uim nauibus flumen transire conentur. (CAES. Gall. 3, 11, 2) ‘s’ils (= les Germains) entreprennent de forcer avec leurs bateaux le passage du fleuve.’ (32) transit equom cursu frenisque aduersa prehensis. (VERG. Aen. 11, 719) ‘elle (= Camille) dépasse le cheval dans sa course, se retourne, saisit le mors.’ – Mais lorsque ce même espace est envisagé comme ‘traversé’,28 le complément peut prendre la forme usuelle du lieu médian, un syntagme prépositionnel en per : (33) Quare optumum factu uideri, per media eius castra palam transire. (SALL. Iug. 107, 5) ‘le mieux était donc, à son avis (= de Volux), de passer hardiment au milieu de son camp (= de Jugurtha).’ L’accusatif d’extension spatiale du lieu franchi peut devenir sujet en cas de transformation passive : (34) Rhodanus . . . nonnullis locis uado transitur. (CAES. Gall. 1, 6, 2) ‘Le Rhône est guéable en plusieurs endroits.’

27 24 occurrences relèvent de cet emploi dans notre corpus. 28 Pour les compléments référant à l’espace, le plus fréquent, dans notre corpus, est celui qui dénote l’espace franchi (65 occurrences, dont 49 chez César pour un cours d’eau). Les deux autres sont sensiblement plus rares (6 occurrences pour l’espace longé ou dépassé, 8 occurrences pour l’espace traversé).

Les verbes de déplacement en latin : préverbation et arguments

127

Mais, pour le lieu longé, d’emploi plus rare, Sénèque préfère recourir à un groupe prépositionnel, qui explicite le rôle sémantique du repère. Il emploie alors un passif impersonnel : (35) Praeter ipsum theatrum Neapolitanorum, ut scis, transeundum est Metronactis petenti domum. (SEN. epist. 76, 4) ‘Tu sais que, pour aller chez Métronax, il faut passer le long du théâtre de Naples.’

4.3 Les préverbés en praeter- : praetereo Praeter appelle un repère dénotant le lieu longé,29 qui se met à l’accusatif. Une seule occurrence fait intervenir un autre repérage, les adverbes huc et illuc, relevant de L2 (36) . . . Veniam peto pudentem ut . . . conserues puerum mihi pudice non dico a populo ; nihil ueremur istos qui in platea modo huc modo illuc in re praetereunt sua occupati. (CATVLL. 15, 2–3 ; 5–8) ‘. . . Je demande une faveur modeste : . . . préserve-moi cet enfant (=Juventius) de toute atteinte ; je ne parle pas de celles du public ; je ne crains pas les passants qui sur une place vont de côté et d’autre, tout occupés de leurs affaires.’ Huc et illuc ne réfèrent pas au lieu longé : ce sont les compléments de la base eo, l’ensemble du procès étant localisé par le circonstant in platea. Le repère appelé par le préverbe n’est pas exprimé, mais il est restituable contextuellement : il s’agit du puer mentionné au v. 5. Absorbés par leurs déplacements incessants, les occupati passent près de lui sans lui prêter attention.

29 Selon García-Hernández (1980 : 187), le préverbe signifie “por delante rebasando”.

128

Sophie Van Laer

4.4 Conclusion Les préverbes relevant du lieu médian sont fréquemment repérés par un accusatif d’extension spatiale qui peut devenir sujet en cas de transformation passive. Cet accusatif peut parfois être remplacé par un groupe prépositionnel introduit par per, qui constitue la forme usuelle du lieu médian. Sa valeur sémantico-logique peut aussi être explicitée par une autre préposition. Enfin, les préverbés en trans- et en per- admettent également un repérage par le lieu final, qui prend la forme d’un groupe prépositionnel. Nous avons relevé un exemple de praetereo faisant intervenir le lieu final, mais qui ne peut en aucun cas s’articuler au préverbe. Le repère appelé par ce dernier, contextuellement restituable, reste implicite.

5 Le cas du préverbe comCe préverbe, à la différence de ceux que nous avons étudiés précédemment, ne s’articule pas à un des lieux du déplacement. En association avec une base dénotant un déplacement, il présente 3 valeurs : ‘se déplacer ensemble vers une personne ou un lieu’ : le verbe est alors généralement repéré par le lieu final : (37) Quae fuit igitur causa cur cuncta ciuitas Lampsacenorum de contione . . . domum tuam concurreret ? (CIC. Verr. II 1, 80) ‘Quelle a donc été la cause qui a poussé les habitants de Lampsaque à se porter en masse . . . de l’assemblée où ils s’étaient réunis sur ta maison ?’ 2) ‘se déplacer ensemble’ d’où ‘se réunir, se regrouper’. En ce cas, le repère du déplacement est systématiquement effacé. (38) Nunc mihi cur cantent superest obscena puellae dicere ; nam coeunt certaque probra canunt. (OV. fast. 3, 675–676) ‘Il me reste à dire maintenant pourquoi les jeunes filles chantent des grivoiseries ; car elles se rassemblent pour chanter des obscénités d’un certain genre.’ 3) ‘aller trouver quelqu’un’/‘aller affronter quelqu’un’ : Ce dernier sens, moins prévisible, se rencontre essentiellement avec concurro, mais aussi avec conuenio. 1)

Les verbes de déplacement en latin : préverbation et arguments

129

(39) Venit Romam Ligus ; non dubitabat quin si ipse Verrem conuenisset, . . . commouere hominem posset. Domum ad eum uenit, rem demonstrat. (CIC. Verr. II 1, 126) ‘Ligus vient à Rome : s’il pouvait se rencontrer lui-même avec Verrès, cela ne faisait pour lui aucun doute, pourrait l’ébranler. Il va le trouver chez lui, il lui expose l’affaire.’ Il s’agit ici du déplacement d’un individu isolé, Ligus. Le principe d’explication par une arrivée conjointe ne fonctionne plus. D’autre part, le préverbé est encadré par deux emplois du verbe uenit. Il s’agit de rendre compte de la venue du personnage à Rome, dans la maison de Verrès. La présence d’ipse tout comme le caractère explicatif de la proposition où figure conuenisset indiquent que l’élément essentiel, c’est la mise en relation directe entre l’entité mobile, Ligus, et l’entité complément, Verrem. C’est le préverbe qui, nous semble-t-il, prédique cette mise en relation, Verrem apparaissant alors comme le repère. D’autres indices nous orientent vers cette lecture. – Les deux actants mis en présence peuvent occuper la fonction de sujet s’ils effectuent l’un et l’autre un déplacement pour se confronter : (40) non aliter Tros Aeneas et Daunius heros concurrunt clipeis . . . (VERG. Aen. 12, 723–724) ‘Ainsi le Troyen Enée et le héros fils de Danaus, d’un plein élan, heurtent leurs boucliers . . .’ – Dans le cas de concurro, si le repère est exprimé, il se met au datif et non à l’accusatif. Or, le datif tout comme l’accusatif non prépositionnel, sont les deux compléments dont la forme semble directement liée à la présence d’un préverbe.30 (41) Alta petunt ; pelago credas innare reuolsas Cycladas aut montis concurrere montibus altos. (VERG. Aen. 8, 691–692) ‘Ils gagnent le large. Sur la mer on croirait que flottent les Cyclades déracinées ou que de hautes montagnes contre des montagnes courent se jeter.’ Revenons au cas du repère à l’accusatif. Il présente une nouvelle fois les propriétés d’un objet grammatical :

30 Ceci n’est pas sans rappeler l’alternance entre adeo ou ineo, construits avec l’accusatif, tandis qu’incurro et occurro appellent un repère au datif. Voir aussi n. 24.

130

Sophie Van Laer

(42) Conlatinus cum L. Iunio Bruto uenit, cum quo forte Romam rediens ab nuntio uxoris erat conuentus. (LIV. 1, 58, 6) ‘Collatin avec Lucius Junius Brutus : ces derniers revenaient justement à Rome quand ils rencontrèrent le messager de Lucrèce.’ Le sujet d’erat conuentus est Conlatinus, l’agent du procès est ab nuntio uxoris. Le repère à l’accusatif devient donc sujet du verbe à la voix passive. À partir du moment où l’accusatif a les propriétés d’un objet grammatical, il semble saturer la structure argumentale du verbe sans laisser de place à l’expression du lieu où se produit la rencontre. Il peut néanmoins être utile de localiser cette rencontre. Cicéron, (39), rattache Romam au verbe uenit, Tite-Live (42) en fait le complément du participe apposé rediens. Une autre solution est de recourir au lieu du procès, qui n’est pas un complément du verbe mais un circonstant : (43) Te per eos dies nemo tuae domi conuenire, nemo in foro uidere potuit. (CIC. Verr. II 5, 137) ‘C’est toi qu’en ces jours là personne n’a pu trouver chez toi, personne voir sur le Forum.’

6 Conclusion Au terme de cette étude, il apparaît que le préverbe se satisfait généralement de la structure argumentale de la base, à savoir de l’expression d’un des lieux du déplacement. Même dans le cas de com-, le plus étranger aux lieux du déplacement, le préverbe n’occasionne pas la création d’un nouvel argument. Lorsque le verbe préverbé a un complément de forme prépositionnelle, le choix de la préposition ne semble pas être guidé par le seul souci de refléter le préverbe, en latin classique du moins. Nous avons relevé des possibilités de variations, l’emploi de in contribuant à générer un sens connotatif pour incurro ou l’explicitation par praeter du sens à donner au repère de transeo. Lorsque la préverbation met en place un autre repérage, elle ne génère pas une double complémentation. C’est ce que nous avons pu observer à propos d’obambulo, de praetereo ou de conuenio. L’effacement du repère est possible quand il est contextuellement restituable et présupposé par la présence du préverbe. L’autre solution étant l’effacement du complément requis par la base.

Les verbes de déplacement en latin : préverbation et arguments

131

Mais le préverbe peut modifier la forme syntaxique de l’argument complément. L’accusatif d’extension spatiale, l’accusatif directif et le datif semblent directement imputables au préverbe. Le datif peut commuter avec un groupe prépositionnel, tandis que l’accusatif, qu’il soit d’extension spatiale ou directif, tend à devenir un objet. Les deux réalisations syntaxiques visent à faire du repère un complément de verbe, qui ne s’articule plus directement à un des lieux du déplacement.

Abréviations ER ES L1 L2 L3 L4 VDD VDI VMD SP

= = = = = = = = = =

entité-repère entité située lieu initial lieu final lieu médian lieu du procès verbe de déplacement verbes de déplacement intrinsèque verbe de mode de déplacement sens particulier

Bibliographie CUF = Collection des Universités de France LLT-A = Library of Latin Texts – Series A (Brepols) Baudou, A. & Clément-Tarantino, S. (2015). À l’école de Virgile: Commentaire à l’Enéide, livre I. Villeneuve-d’Ascq: Presses universitaires du Septentrion. Boons, J.-P. (1987). La notion sémantique de déplacement dans une classification syntaxique des verbes de déplacement. Langue française, 76, 5–40. Brachet, J.-P. (2000). Recherches sur les préverbes de- et ex- du latin. Bruxelles: Latomus. Cabrillana, C. (1997a). Expresión casual de complementos de uenio y eo. Habis, 28, 325–336. Cabrillana, C. (1997b). Complementos direccionales en el marco predicativo de uenio. Minerva, 11, 117–137. García-Hernández, B. (1980). Semántica estructural y lexemática del verbo. Reus: Avesta. García-Hernández, B. (2016). El preverbio ob- en perspectiva diacrónica: Los significados adverbiales (‘detrás’, ‘encima’, ‘enfrente’). Dans F. Fuentes Moreno & M. del Castillo Herrera (Eds.), Quantus qualisque. Homenaje al profesor Jesús Luque Moreno (pp. 209–216). Granada: Universidad de Granada. Guillet, A & Leclère, C. (1992). La structure des phrases simples en français: Constructions transitives locatives. Genève: Droz.

132

Sophie Van Laer

Lehmann, C. (1983). Latin preverbs and cases. Dans H. Pinkster (Ed.), Latin linguistics and linguistic theory (pp. 145–165). Amsterdam/Philadelphia: John Benjamins. Langacker, R. W. (1987). Mouvement abstrait. Langue française, 76, 59–76. McGillivray B. (2013). Latin preverbs and verb argument structure: new insights from new methods. Dans E. van Gelderen, J. Barðdal & M. Cennamo (Eds.), Argument structure in flux: The Naples-Capri papers (pp. 119–134). Amsterdam/Philadelphia: John Benjamins. Orlandini, A. & Poccetti, P. (2011). La référence spatio-temporelle et métalinguistique des verbes de mouvement en latin et leurs évolutions romanes. Dans C. Moussy (Ed.), Espace et temps en latin (pp. 25–45). Paris: Presses de l’université Paris-Sorbonne. Palachi, C. (2010). Semántica léxica y sintaxis de los verbos latinos de movimiento: Relaciones entre verbos simples y sus derivados por medio de prefijos. Dans P. Anreiter & M. Kienpointner (Eds.), Latin Linguistics Today (pp. 461–473). Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck. Pottier, B. (1962). Systématique des éléments de relation. Paris: Klincksieck. Serbat, G. (1996). Préverbation et émergence d’un Datif post-verbal en Latin. Dans G. Gréciano & H. Schumacher (Eds.), Lucien Tesnière – Syntaxe structurale et opérations mentales (pp. 183–190). Tübingen: Max Niemeyer Verlag. Serbat, G. (2001) [1997]. Aperçu d’une analyse syntaxique des préverbes. Dans L. Nadjo (Ed.), Opera disiecta: Travaux de linguistique générale, de langue et littérature latines (pp. 143–152). Louvain/Paris: Peeters. Talmy, L. (1975). Figure and ground in complex sentences. Proceedings of the First Annual Meeting of the Berkeley Linguistics Society, 419–430. Vandeloise, C. (1986). L’espace en français. Paris: Éditions du Seuil. Van Laer, S. (1998). Le préverbe ob- en latin: Considérations sémantiques. Dans B. García-Hernández (Ed.), Estudios de linguística latina (pp. 1043–1059). Madrid: Ediciones Clásicas. Van Laer, S. (2010). La préverbation en latin: étude des préverbes ad-, in-, ob- et per- dans la poésie républicaine et augustéenne. Bruxelles: Latomus.

Words Word formation and etymology

Laurent Gavoille

Auctor adjectif verbal dans la périphrase auctor sum ‘conseiller’ Abstract: This chapter argues that in the archaic phrase auctor esse + quid or ut + subj. (‘to advise’), auctor should be considered as an adjective. Such a periphrasis could also throw light on the tricky relationship, from both morphological and semantical point of view, between augeō et auctor. The proposal is based on the following line of argumentation: (1) This archaic phrase must be distinguished from the current/classical construction auctor esse + gen. (‘to be the guarantor or the responsible of’), where auctor is a noun. Thus the agent noun auctor appears as secondary to the adjective. (2) The existence of the Vedic periphrastic future associating the suffix *-tér and the verb ‘to be’ (dātāsmi < *dā-tr̥-as-mi) invites us to consider this Latin expression too as a verbal periphrasis. In both languages the phrase has a directive and prospective sense. Such a verbal periphrasis is an isolated phenomenon in Latin, but it could be due to the fact that it consists in an old juridical expression. (3) From this Latin periphrasis the adjective and the noun became probably independent units, furnishing then a nominal basis which can explain the formation of augeō. The verb is not an IE causative but an essive or fientive (cf. its meaning ‘to increase’), secondarily built in Italic on a basis aug-. This basis, although often viewed as provided by augur, should be more efficiently derived from auc-tor sum: it could account for the semantic development of augeō, from ‘to grow’ to ‘to make grow’. Keywords: Latin linguistics, auctor (sum), augeō, verbal periphrasis

1 Introduction Cette contribution est consacrée à l’emploi de auctor comme adjectif verbal dans la périphrase auctor sum, qui signifie globalement ‘conseiller’ : sans revenir sur le sens de celle-ci, que j’ai déjà étudié ailleurs,1 il s’agira ici

1 Cf. Gavoille (2017) et Gavoille (2019). Laurent Gavoille, Université Bordeaux-Montaigne https://doi.org/10.1515/9783110647587-009

136

Laurent Gavoille

principalement d’analyser son rôle dans la constitution du verbe augeō, à considérer selon nous comme un dénominatif de auctor adjectif. Mais auparavant, il nous faut établir son caractère périphrastique, d’abord en montrant la nature adjectivale de auctor dans auctor sum, puis en comparant la formule latine avec les périphrases védiques correspondantes.

2 Auctor adjectif dans auctor sum Auctor sum ‘conseiller’ (+ quid ou complétive, cf. les exemples 4–5) est une formule peu représentée en latin (1 occurrence chez Ennius, 7 chez Plaute, 1 chez Térence et 6 dans la correspondance de Cicéron), qui constitue vraisemblablement un archaïsme et pour cette raison est parfois mal identifiée. Il convient de distinguer d’un côté la formule verbale auctor sum (souvent traduite par ‘conseiller, autoriser, permettre’), construite avec un accusatif (quid ?) ou une complétive au subjonctif et, d’un autre côté, la construction auctor sum ‘je suis garant, promoteur de’, où auctor, se comportant syntaxiquement comme un nom d’agent classique, peut être déterminé par un génitif ou un adjectif, et se trouver sur le même plan que d’autres noms d’agent (cf. 1–3) : (1)

Sed non erat nec recentior auctor nec huius rei quidem melior Dolabella. (CIC. Att. 9, 13, 1) ‘Mais il n’y a pas plus récent garant ni plus informé de la chose que Dolabella.’2

(2)

. . . atque huius deditionis ipse Postumius qui dedebatur, suasor et auctor fuit. (CIC. off. 3, 109) ‘. . . du reste, c’est Postumius en personne, lui qu’on livrait, qui fut le conseiller et le promoteur de cette remise à l’ennemi.’

(3) . . . ut apud posteros nostros non exstinctores coniurationis et sceleris, sed auctores et duces fuisse uideamur ? (CIC. dom. 101) ‘. . . pour que nos descendants se figurent que nous avons non pas étouffé, mais inspiré et dirigé une conjuration criminelle ?’

2 Sauf indication contraire, les traductions sont celles de la Collection des Universités de France aux Belles Lettres (Paris), parfois légèrement remaniées.

Auctor adjectif verbal dans la périphrase auctor sum ‘conseiller’

137

Ce n’est pas tant cette construction, courante en latin, qui retiendra notre attention, que la périphrase auctor sum dont le caractère verbal est clairement identifié par sa construction avec un adverbe : (4) . . .mihique ut absim uehementer auctor est . . . (CIC. Att. 15, 5, 2) ‘. . . il me conseille vivement d’être absent . . . ’ (trad. personnelle) (5) AN. Mi auctores ita sunt amici ut uos hinc abducam domum. PAN. At enim nos, quarum res agitur, aliter auctores sumus. (PLAVT. Stich. 128–129) ‘AN. Mes amis me conseillent de vous reprendre chez moi. PAN. Mais nous, que cela regarde, nous te conseillons le contraire.’ Si auctōres était un substantif, on aurait vraisemblablement †alii auctores sumus, et surtout, pour la forme, non pas auctōres mais auctrīces, puisque le sujet nos est féminin et que les noms en -trīx existent déjà chez Plaute (par exemple persuastrīx en Bacch. 1167). Dans un fragment de l’Évhémère d’Ennius, on trouve aussi auctor fuit à propos de Vénus, pourtant bien déterminée ensuite par un adjectif au féminin (sola) : (6) Artem meretriciam instituit auctorque mulieribus in Cypro fuit uti uulgo corpore quaestum faceret ; quod idcirco imperauit ne sola praeter alias mulieres impudica et uirorum adpetens uideretur. (ENN. frg. var. 142–145 Vahlen) ‘C’est elle qui la première institua le métier de courtisane et à Chypre recommanda formellement aux femmes de faire publiquement commerce de leur corps ; et elle les fit se conduire ainsi pour ne pas être la seule entre toutes les femmes à paraître impudique et coureuse d’hommes.’ (trad. personnelle) Dans les exemples (4), (5) et (6), la construction avec une complétive en ut ou ne suivi du subjonctif, ainsi que la synonymie avec instituit dans l’exemple d’Ennius, ne laisse pas de doute sur le caractère verbal de la formule. Auctor sum est une périphrase verbale où auctor se comporte comme un adjectif non sensible au genre grammatical, mais uniquement au nombre, et elle constitue un archaïsme en latin, dans la mesure où elle ne se rencontre que chez Ennius, Plaute, Térence et dans la correspondance de Cicéron (où elle représente un quart des attestations du syntagme auctor sum, alors que dans le reste de son œuvre, le taux tombe à 5 %). L’ancienneté de ces attestations de auctor adjectif tend à faire penser que le nom d’agent auctor est secondaire par rapport à cet adjectif. Quatre indices vont dans le même sens :

138

1)

Laurent Gavoille

L’apparition tardive de auctrīx (chez Tertullien) témoigne que le nom auctor ne s’est dégagé que progressivement de l’adjectif : il a d’abord pris une indépendance syntaxique, sans que soit ressenti le besoin de lui donner immédiatement une forme de féminin proprement nominale. C’est ainsi qu’Ovide juxtapose inuentrīx et le nom d’agent auctor, à propos du rapport de Minerve à la musique : (7) Sum tamen inuentrix auctorque ego carminis huius. (OV. fast. 6, 709) ‘C’est moi qui ai inventé cet art et en suis la patronne.’ (trad. personnelle)

L’allure figée de la formule : on ne trouve que l’affirmation à la 1ère personne auctor sum (qui peut évidemment être transformée en 3e personne en cas de style indirect) et, à la 2e personne, la question au style direct quid auctor es ?3 3) Son caractère juridique ou religieux : 2)

(8) sic augures dicunt : si mihi auctor es uerbenam manu asserere, dicito consortes. (VARRO ling. 6, 64–65) ‘ainsi les augures disent : si tu m’autorises à toucher le bouquet avec la main, nomme mes copartageants.’ (9) Sed, cum hoc iuris a maioribus proditum sit ut nemo ciuis Romanus aut sui potestatem aut ciuitatem possit amittere, nisi ipse auctor factus sit. (CIC. dom. 77) ‘Mais telle est la jurisprudence transmise par nos ancêtres qu’aucun citoyen romain ne peut perdre sa liberté ni sa citoyenneté sans son consentement officiel.’4 (trad. personnelle) 4) L’existence (dès Plaute) de auctōritās qui, comme tout abstrait en -tās, ne peut dériver que d’une base adjectivale.5 Cet argument morphologique oblige à poser un auctor adjectif. On peut ainsi ajouter auctor à la liste des vieux noms d’agent qui sont adjectifs avant d’être substantifs, comme dux (s’il est bien secondaire par rapport à redux, cf. Bader 1962 : 58) et princeps avec lesquels il d’ailleurs assez souvent associé dans les textes, comme dans cet exemple :

3 Cf. Gavoille (2017 : 56–58). 4 Cf. infra (12). 5 Par exemple liberālitās sur liberālis, suāuitās sur suāuis.

Auctor adjectif verbal dans la périphrase auctor sum ‘conseiller’

139

(10) . . . ut erga duces ipsos et principes atque auctores salutis meae satis gratus iudicarer . . . (CIC. Cato 24) ‘. . . que de montrer assez de gratitude à ceux mêmes qui ont conduit, mené et inspiré la défense de mes droits . . .’6 Auctor illustrerait donc ces propos de Bader : “Dans le système ancien, tout verbe est susceptible d’avoir à côté de lui deux sortes de nom, et deux sortes seulement : d’une part un substantif abstrait (‘nom d’action’), transposition nominale de la notion exprimée par le verbe . . . d’autre part un individualisant désignant une personne : un adjectif. Selon les rapports plus ou moins intimes qu’il entretient avec la conjugaison, cet individualisant peut rester un adjectif proprement dit (par exemple en latin l’adjectif en -nt- ou -to-), ou devenir un substantif, quand, au lieu d’être un déterminatif qui ‘définit une catégorie’, il indique, ce qui en est proche pour le sens, un ‘représentant d’une catégorie’ ; ce qu’on appelle ‘nom d’agent’ n’est rien d’autre qu’un adjectif (verbal) substantivé” (1962 : 190).7

3 Comparaison avec le futur périphrastique védique Une telle périphrase verbale avec une forme en -tor existe aussi en védique, où elle sert à former un autre futur que celui, plus courant, en -sya- doté d’un suffixe -s- désidératif.8 Le futur périphrastique comporte un dérivé en *-tér (comme dātár-) associé, aux première et deuxième personnes, à la copule as‘être’, d’où 1ère sg. dātāsmi < *dā-tr̥-as-mi, 2e sg. dātāsi, 3e sg. dātā.9 Pour la différence entre ces deux futurs, je renvoie aux travaux de Tichy, qui a bien montré que, du point de vue de la valeur, le futur en -sya- fonctionne comme un présent désidératif où l’accomplissement du procès est envisagé dans un futur indéterminé, alors que le futur périphrastique n’est employé que

6 Princeps est d’interprétation difficile : Festus le glose par primum captus, mais le sens actif ‘qui prend la première part’ existe (cf. Bader 1962 : 63–64) et c’est celui qu’on a dans les syntagmes avec auctor (voir aussi, entre autres, CIC. Balb. 61 ; prov. 25 ; Phil. 2, 71 ; Sull. 34). 7 Benveniste avait déjà noté cette particularité (1935 : 109) : “Les sens premier des mots en *-ter est simplement celui d’un adjectif marquant l’exercice d’une activité non transitive : le praetor (*prai-itor) est ‘celui qui marche à la tête (des armées)’ ; gr. βατήρ est glosé par βαίνων, comme un simple participe ; skr. gántar signifie de même ‘allant, venant’.” 8 Par ex. haniṣyántā, futur désidératif de hánti ‘tuer’, sur la racine *gṷhen-. 9 Cf. Mayrhofer (1964 : 88).

140

Laurent Gavoille

si les modalités du futur sont précisées par l’indication d’un moment déterminé, au moyen d’un adverbe de temps ou d’une conditionnelle. Dans le système du védique, ce futur périphrastique, qui a remplacé progressivement l’injonctif prospectif réservé à la proposition subordonnée, exprime un futur rendu incertain par les conditions qui lui permettront d’advenir.10 Il a ainsi le même rapport au temps que l’impératif futur, qui correspond à un ordre lié non pas au présent mais à des conditions précises, souvent énoncées dans une hypothétique. C’est pourquoi, morphologiquement cette fois, ce futur périphrastique n’est pas à mettre en relation avec le suffixe *-́tor de nom d’agent, mais avec le suffixe *-tér. Selon Tichy, *-́tor a une fonction généralisante, désignant un agent habituel ou même un agent indéfini, alors que *-tér renvoie à un agent lié à une situation unique (actuelle, passée, future, hypothétique).11 La signification de cette périphrase associant une forme *-tér et le verbe ‘être’ n’est évidemment pas la même en latin. Chaque système de langue est différent et l’explication fonctionnelle du futur périphrastique en védique, prenant la place de l’injonctif prospectif, ne saurait être transposée comme telle en latin. On peut en revanche dégager des traits communs (à partir desquels, par comparaison, il est éventuellement possible de remonter une valeur indo-européenne). Le premier point commun, morphologique et syntaxique, est l’association avec un impératif futur, qui exprime un ordre lié à des conditions précises, comme dans l’exemple de Varron – cf. (8) supra. Nous avons affaire, dans ce passage, à une formule juridico-religieuse, dont la structure rappelle celle des textes de lois : (11) Si pater filium ter uenumduit, filius a patre liber esto. (LEX XII tab. 4, 2 [Gaius inst. 1, 132 ; Vlp. reg. 10, 1]) Comme j’ai déjà pu le montrer, auctor sum est une vieille formule rituelle.12 Un texte de Cicéron, dans le discours De domo sua, permet de penser qu’elle appartient, fondamentalement, au domaine du droit, ou plus exactement à la jurisprudence de l’ancienne Rome :

10 Voir à ce sujet Tichy (1992a). 11 Cf. Tichy (1992b : 417–420). Cette valeur des noms d’agent en *-tér est liée à l’origine locative du suffixe : le type *dh3-tér-/gén. sg. *dh3-tr-és est issu du locatif singulier du nom d’action *dh3-tér(i) ‘celui qui est en train de donner (dans une situation donnée)’, cf. Tichy (1995 : 61), repris par Pinault (2003 : 160). 12 Cf. Gavoille (2017 : 60–61).

Auctor adjectif verbal dans la périphrase auctor sum ‘conseiller’

141

(12) Sed, cum hoc iuris a maioribus proditum sit ut nemo ciuis Romanus aut sui potestatem aut ciuitatem possit amittere, nisi ipse auctor factus sit – quod tu ipse potuisti in tua causa discere : credo enim, quamquam in illa adoptatione legitime factum est nihil, tamen te esse interrogatum auctorne esses ut in te P. Fonteius uitae necisque potestatem haberet ut in filio. (CIC. dom. 77) ‘Mais telle est la jurisprudence transmise par nos ancêtres qu’aucun citoyen romain ne peut perdre sa liberté ni sa citoyenneté sans son consentement officiel – ce que tu as pu apprendre toi-même dans ta propre cause : en effet, bien qu’on n’ait suivi dans cette adoption aucune forme légale, je pense qu’on t’a demandé si tu autorisais officiellement que P. Fonteius ait droit de vie et de mort sur toi, comme sur un fils.’ Ce texte a trait à la déchéance de la liberté et de la citoyenneté, qui ne peut être imposée par la force, mais exige la prononciation de certaines formules : l’un doit demander auctor es ut [ciuitatem amittas] ? (‘acceptes-tu solennellement de perdre ta citoyenneté ?’), l’autre répond auctor sum (‘j’accepte solennellement’). Le caractère juridique de la formule est confirmé par l’équivalence avec censeō, vieux verbe qui, faisant référence à une demande du roi au Sénat et à sa réponse, signifie anciennement, selon Ernout et Meillet (19594), ‘déclarer d’une façon formelle ou solennelle, exprimer un avis dans les formes prescrites’, conformément au sens de la racine *ƙeNs- ‘déclarer, annoncer’ ; (plus fréquemment) ‘estimer, évaluer, apprécier’, avec prise en compte de ce qui est jugé bon.13 Cette équivalence est illustrée par les deux exemples suivants de Cicéron. En (13), quid mi auctor es est repris par uide quid censeas ; en (14), à la demande petis quid sim tibi auctor répond censeo commorandum : (13) Etsi quid mi auctor es ? Aduolone an maneo ? Equidem et in libris haereo et illum hic excipere nolo . . . Sed ego ipse κεκέπφωμαι. Itaque posthac . . . Tu tamen uide quid de aduentu meo censeas . . . (CIC. Att. 13, 40, 2) ‘Cela dit, que me conseilles-tu ? J’accours et ou je reste ici ? Pour ma part, je suis plongé dans mes livres et ne veux pas recevoir ce garçon ici . . . Mais moimême je suis comme une girouette. Aussi, à l’avenir . . . À toi de voir ce que tu penses de ma venue éventuelle . . .’

13 Kümmel et Rix (2001 : 326). Sur le latin censeō et ses correspondants indo-iraniens, cf. García-Ramón (1993), plus particulièrement 115–119 pour le passage de ‘déclarer avec autorité’ à ‘définir, estimer’.

142

Laurent Gavoille

(14) Quoniam quid tibi permittatur cognosti, quid mihi placeat puto te scire oportere. Actis his rebus litterae a te mihi redditae sunt, quibus a me consilium petis quid sim tibi auctor, in Siciliane subsidas an ut ad reliquias Asiaticae negotiationis proficiscare . . . Sed ego, siue hoc siue illud est, in Sicilia censeo commorandum. (CIC. fam. 6, 8, 1–2 ; de Rome, décembre 46, à Cécina qui n’a pas le droit de rentrer en Italie mais qui a obtenu, grâce à l’entremise de Cicéron, de rester plus longtemps en Sicile) ‘Puisque te voilà informé de ce qui t’est permis, il faut, je pense, que tu saches ce qui me paraît bon. Une fois cette tractation terminée, on m’a remis une lettre de toi, où tu me demandes conseil : est-ce que je t’engage à demeurer en Sicile ou à partir régler ce qui reste de tes affaires en Asie ? . . . Pour ma part, dans un cas comme dans l’autre, je suis partisan du séjour en Sicile.’ Le sens juridique se retrouve chez Plaute, même s’il est détourné à des fins comiques. On trouve ainsi auctor sum dans la bouche de personnages qui, au lieu d’accepter un châtiment, le réclament d’eux-mêmes, ce que l’allocutaire a évidemment du mal à croire ; or la solennité de la formule auctor sum permet, malgré le caractère paradoxal du contenu, de rendre la déclaration digne de foi. Dans de tels contextes, la formule signifie ‘permettre, autoriser’, en synonymie avec sinō : (15) AG. Suspende, uinci, uerbera ; auctor sum, sino. MI. Si auctoritatem postea defugeris, ubi dissolutus sies, ego pendeam. (PLAVT. Poen. 146–148 : le jeune Agorastoclès est prêt à tout pour se faire pardonner de son esclave Milphion, qu’il a récemment battu, car il a besoin de ses services) ‘AG. Suspends-moi au gibet, enchaîne-moi, bats-moi ; je t’y autorise, je veux bien. MI. Si tu viens par la suite à retirer l’autorisation, une fois que tu seras détaché, c’est moi qui serai pendu.’ Le second point commun de auctor sum avec le futur périphrastique védique est l’aspect prospectif. Dans un contexte juridique, celui qui prononce la formule auctor sum ‘j’autorise’ ne se contente pas d’accepter mais, par là même, rend possible l’action à venir en lui donnant une garantie juridique. Dans l’occurrence du De domo sua, l’individu qui accepte de perdre sa liberté s’engage à ne pas revenir sur sa décision. Dans les emplois courants, celui qui dit auctor sum certifie, par l’énonciation de cette formule, que le conseil qu’il donne est bon et s’en porte garant. En termes pragmatiques, auctor sum est un acte de parole qui donne une force, un *augos (nom d’action abstrait, de genre neutre)

Auctor adjectif verbal dans la périphrase auctor sum ‘conseiller’

143

à un contenu propositionnel qui est à réaliser – exprimé le plus souvent dans une complétive au subjonctif, comme en (12). La dimension prospective est là, dans cette ouverture vers un procès dont la réalisation est rendue possible par l’acte de parole initial. Cet aspect est également à mettre en relation avec le suffixe véd. accentué -tár qui selon Renou, est plutôt ponctuel et décrit au premier chef « l’entrée dans une fonction ».14 Tichy, de son côté, précise que dātár- peut désigner un agent potentiel, comme le latin nemo est qui + subjonctif (‘il n’y a personne pour . . . ’).15 D’une manière générale, le type hystérodynamique16 a une valeur relativisante ou situative, ce qui s’accorde parfaitement avec les textes juridicoreligieux dans lesquels apparaît en latin la formule auctor sum. Les deux passages du De domo sua de Cicéron et du De lingua latina de Varron renvoient bien à des agents potentiels pris dans une situation précise. Ces deux points communs (valeur directive et prospective) avec le futur périphrastique védique se retrouvent dans la définition du sens fondamental de auctor sum qui est à la fois un conseil d’action et l’affirmation d’une garantie pour l’avenir.

4 Augeō dérivé de auctor adjectif L’identification de auctor comme adjectif dans la formule auctor sum permet de comprendre d’abord comment a pu se dégager, en latin, une base nominale aug-/auc-, et ensuite pourquoi augeō a pris une valeur factitive. Il est impossible de faire dériver auctor de augere, pour des raisons morphologiques et sémantiques. On pose souvent, en raison du sens de augeō ‘faire croître’, un thème causatif *h2oug-éi̯e-,17 mais le degré o ne peut

14 Cf. Renou (1938), cité par Tichy (1992b : 413 n. 8). 15 Cf. Tichy (1992b : 416). Pour la formule latine, voir par ex. CIC. Verr. I 46 Nulla est enim laus ibi esse integrum ubi nemo est qui aut possit aut conetur corrumpere, ‘Il ne mérite en rien d’être loué, celui qui reste intègre quand il n’y a personne qui ait le pouvoir ou qui essaie de corrompre.’ 16 C’est-à-dire, en indo-européen, le thème nominal athématique accentué sur le suffixe aux cas forts, sur la désinence aux cas faibles, dont relèvent les noms d’agent en *-tér : par ex., sur *deh3- ‘donner’, nominatif *dh3-tér-s / génitif *dh3-tr-és. Les noms d’agent en -*tōr relèvent, eux, du type acrostatique où la racine est accentuée dans tout le paradigme (cf. Schindler 1975 et Pinault 2003). 17 Cf. par ex. Kümmel et Rix (2001 : 274). Thème causatif posé également dans Ernout and Meillet (19594 : 58) et Pokorny (1959–1969 : 84–85).

144

Laurent Gavoille

pas expliquer aug- en latin. Aussi a-t-on recours à des subterfuges, comme une éventuelle analogie avec un thème aoriste sigmatique indo-européen qui expliquerait le parfait auxi (*h2ḗug/h2éug-s) – hypothèse fragile dans la mesure où le parfait sigmatique est surprenant pour un causatif. En fait, le véritable causatif indo-européen est fabriqué sur le doublet en -s (*h2u̯eg-s-) de cette racine, de sens peut-être inchoatif : ainsi véd. vakṣáyati sur *h2u̯ogs-é-i̯e- “faire s’accroître”.18 Dans ces conditions, on peut s’orienter vers une autre explication, qui est celle d’un fientif signifiant l’entrée dans un nouvel état ou d’un essif exprimant l’état, tous deux susceptibles d’expliquer le -ē- du verbe latin.19 Cette hypothèse est vraisemblable pour plusieurs raisons : 1) Ces verbes d’état peuvent être d’origine dénominative, comme rubeō sur *(h1)rudh- ‘rouge’, et souvent doublés d’un inchoatif en *-ske/o-, comme rubescō, albescō. C’est le cas justement de augeo qui connaît un inchoatif augescō. 2) Certains sont accompagnés d’un composé en -faciō pour former une périphrase factitive, comme arē-faciō en face de areō ; or justement existent auctificus et augificāre. 3) Augeō signifierait alors ‘entrer dans un état de force, devenir fort’, ce qui rend compte des attestations de augēre intransitif ‘croître’ : (16) Eodem conuenae complures ex agro accessitauere. Eo res auxit. (CATO orig. 20, ap. GELL. 18, 12, 7) ‘Venus de la campagne, de nombreux étrangers vinrent au même endroit sans discontinuer ; en conséquence, leur situation s’améliora.’ (trad. personnelle) (17) Ceterum illi . . . qui ignoscundo populi Romani magnitudinem auxisse, nusquam etiam tum Lepidum progressum aiebant . . . (SALL. or. Phil. 6) ‘Mais ceux qui . . . disaient que Rome devait sa grandeur à sa politique de pardon, ajoutant que du reste Lépidus n’avait pas encore bougé . . .’

18 Cf. Garnier (2010 : 448). 19 Le fientif a une structure *R(z)-éh1-/-h1-, c’est-à-dire racine au degré réduit, non accentuée, et suffixe alternant (pour l’accent et la quantité), type ἐμάνην ‘devenir fou’ (< *mn-éh1-) sur *men- ‘penser’. Le fientif est de structure R(z)-(e)h1-i̯é/ó-, c’est-à-dire racine au degré réduit, non accentuée, suffixe accentué, type lit. miniù ‘penser, se souvenir’. Sur cette question, cf. Kümmel et Rix (2001 : 25).

Auctor adjectif verbal dans la périphrase auctor sum ‘conseiller’

145

Toutefois cette hypothèse rencontre une difficulté morphologique, car les fientifs et les essifs sont bâtis sur le degré réduit de la racine qui ne peut expliquer aug- : *HuC- devrait évoluer en *uC-20 (donc †ug- en latin). La solution adoptée consiste à poser un éventuel *aug-ē-i̯e/o, dépourvu de laryngale car il s’agirait d’une formation italique, faite d’après un thème nominal aug- comme celui de aug-ur (*aug-os).21 La sémantique vient elle aussi ruiner toute dérivation possible de auctor à partir de augeo, car les sens du nom ne sont pas de tout parallèles à ceux du verbe. Augēre signifie le plus souvent ‘augmenter’, c’est-à-dire rendre quelque chose ou quelqu’un plus grand, avec un complément renvoyant à une quantité nombrable, comme numerum, ou bien ‘renforcer’, c’est-à-dire ajouter de la force à ce qui en a déjà (par exemple, animum, le courage) : il s’agit alors d’augmenter la qualité, qui peut avoir des degrés.22 Si auctor était le nom d’agent correspondant à augeō, on devrait le trouver complété par des compléments au génitif relevant des mêmes champs sémantiques : †auctor numeri, auctor animi par exemple, avec le même sens de ‘qui augmente, qui accroît’, mais ce n’est pas le cas. Chez Plaute, on ne trouve que des compléments au datif désignant le bénéficiaire de l’action ; à partir de l’époque classique apparaissent des génitifs, de noms de procès et de résultat où il n’est aucunement question de nombre, de quantité ou de degré.23 En aucune manière ces syntagmes ne pourraient être considérés comme la transposition de augēre (+ acc.). Le sens de auctor y est celui de ‘garant, conseiller, instigateur, promoteur’, et le complément renvoie à ce qui doit être garanti, promu : (18) . . . nisi ut erga duces ipsos et principes atque auctores salutis meae satis gratus iudicarer . . . (CIC. Cato 24) ‘. . . que de montrer assez de gratitude à ceux-mêmes qui ont conduit, mené, promu de la défense de mes droits . . .’ Ce sens de auctor est fondamental, car il est immotivé en contexte, alors que les acceptions de ‘vendeur’, ‘père’, ‘fondateur’ dépendent du complément (fundī, generis, ciuitātis).24 On peut raisonnablement supposer que cela s’explique par

20 Cf. Schrijver (1991 : 74–75). 21 Pour *aug-ē-i̯e/o, voir Vernet i Pons (2008 : 135) et Garnier (2010 : 448). 22 Pour une présentation et une analyse détaillées de ces exemples, cf. Gavoille (2019). 23 Pour un examen précis de ces syntagmes, cf. Gavoille (2019). 24 Cf. Gavoille (2019).

146

Laurent Gavoille

l’ancienneté et la primauté de la formule auctor sum ‘je me porte garant’ dans laquelle auctor est adjectif. Tout indique que le nom d’agent auctor est un dérivé interne de auctor adjectif, et non pas un nom d’agent dérivé de augeō. Mieux vaut donc partir, pour expliquer augeō, d’une base nominale aug- ‘fort’. Pour celle-ci, on mentionne souvent augur, mais je pense qu’il faut utiliser aussi auctor (dans auctor sum), qui peut d’ailleurs remonter à une racine au degré plein *h2éu̯g-tōr puisque le latin a mixé les deux anciens types de noms d’agent. L’avantage de auctor est double : 1) Il fournit un point de départ qui est, en latin, à la fois nominal et verbal (puisqu’il entre dans une périphrase verbale). Ainsi on comprend mieux qu’une base nominale aug- reposant sur *h2éu̯g- (comme véd. ójas- ‘force vitale’, av. aojah-, lat. augus-tus ‘solennel, consacré’) ait pu être réutilisée dans la constitution d’un verbe comme augeō.25 2) Auctor présente l’avantage d’être attesté (et non pas reconstitué comme l’ancien nom abstrait *augos masculinisé en augur) et d’expliquer, en vertu de sa valeur agentive, le sémantisme de augeō. Il faut en effet partir du sens de base auctor sum, qui est fientif : ‘je deviens fort’, c’est-à-dire ‘j’entre dans le statut de garant (par ma déclaration) relativement à un projet précis (désigné par quid ou par une complétive au subjonctif) ; l’auctor, quant à lui, est ‘celui qui est fort (garant, soutien, base) ou devient fort relativement à un projet précis’.26 À partir de là s’explique très bien aug-eō, sur radical auc- emprunté à auctor et avec suffixe fientif, signifiant d’abord ‘devenir fort, croître, se développer’ – voir l’exemple (16) de Caton. Mais auctor sum permet également d’expliquer le développement sémantique du verbe vers un sens factitif/causatif. D’un point de vue pragmatique, la périphrase latine constitue un acte de parole performatif déclaratif. Elle signifie ‘se déclarer formellement garant à propos d’une action à venir’, ou mieux ‘prendre la responsabilité de dire, à propos d’une action à venir, que . . . ’ ; mais le fait même de prendre la responsabilité de l’action à venir rend possible cette action. L’acte de parole déclaratif débouche secondairement (de manière

25 Augeō n’est pas isolé : il existe une petite série de verbes provenant de racines verbales indo-européennes, dont le sens est faible d’un point de vue verbal, qui n’ont pas conservé de thème radical athématique et pour lesquels on reconstruit une forme nominale de degré plein, qui permet d’éviter l’hypothèse d’un degré plein analogique. Par exemple aceō ‘être aigre’ sur *h2ek- ‘être acéré’ (essif normal *h2k-h1i̯é-) ; areō ‘être sec’ sur *h2eh1s- ‘se dessécher sous l’effet de la chaleur’ (essif normal *h2h1s-h1i̯é-). 26 Cf. Gavoille (2017 : 65–68).

Auctor adjectif verbal dans la périphrase auctor sum ‘conseiller’

147

perlocutoire : ce n’est pas signifié, c’est une conséquence) sur un acte directif (le conseil) si l’action à réaliser est à la charge de l’allocutaire : (19) Hirtius, quidem se afuturum . . . mihique ut absim uehementer auctor est, ille quidem periculi causa, quod sibi etiam fuisse dicit. (CIC. Att. 15, 5, 2) ‘Hirtius précise qu’il n’assistera pas à la séance . . . et me conseille énergiquement de n’y point assister ; il allègue le danger que je courrais et me dit que cette raison l’a déterminé, lui aussi.’ C’est ainsi que auctor sum peut être mis sur le même plan que imperāre, comme nous avons pu le voir dans l’exemple d’Ennius cité supra (6). Pour imperāre aussi, le sens ‘dire de faire’ est un développement secondaire (à partir de ‘faire faire’).27 Mais si imperāre suppose une annihilation de la volonté de l’autre, avec auctor sum l’un des deux protagonistes, loin d’être privé mécaniquement de volonté, délègue volontairement à l’autre, en lui demandant de prendre une responsabilité. Cette valeur directive secondaire de auctor sum explique que augeō, bâti sur le radical auc/g-, ait pu développer, à côté de son sens fientif, un sens factitif/causatif ‘faire croître’. On comprend mieux, ainsi, la transitivation observée dans des essifs comme augeō.28 À partir de là, une fois augeo constitué et intégré au système verbal latin, auctor sum, de formation isolée et non reconnaissable, tombe naturellement en désuétude.

5 Conclusion Auctor sum est une périphrase verbale ancienne dans laquelle auctor se comporte comme un adjectif. Ainsi auctor fait partie des vieux noms d’agent qui ont été adjectifs avant d’être substantifs. Certes isolée en latin, cette périphrase est à comparer au futur périphrastique védique associant un dérivé en *-tér et le verbe ‘être’, avec lequel la formule latine partage le fait d’être associée dans les textes à un impératif futur et de contenir un aspect prospectif. Auctor sum a joué un rôle important d’intermédiaire entre la racine *h2éu̯g- et la constitution du verbe augeō, dénominatif de l’adjectif auctor, aussi bien du point de vue morphologique que sémantique.

27 Cf. Gavoille (2014 : 215). 28 Cf. Garnier (2010 : 448), qui évoque aussi habeō.

148

Laurent Gavoille

Bibliographie Bader, F. (1962). La formation des composés nominaux du latin. Paris: Les Belles Lettres. Benveniste, É. (1935). Origines de la formation des noms en indo-européen. Paris: Adrien Maisonneuve. Ernout, A. & Meillet, A. (19594). Dictionnaire étymologique de la langue latine. Paris: Klincksieck. García Ramón, J. L. (1993). Lat. cēnsēre, got. hazjan und das idg. Präsens *ƙéns-e -ti (und *ƙn̥séi̯e-ti?) ‘verkündigt, schätzt’, Stativ *ƙn̥s-eh1- ‘verkündigt, gesachätzt sein / werden’. In G. Meiser (Ed.), Indogermanica et Italica. Festschrift für Helmut Rix zum 65. Geburtstag (pp. 106–130). Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Garnier, R. (2010). Sur le vocalisme du verbe latin: Étude synchronique et diachronique. Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck. Gavoille, L. (2014). Imperare: Du causatif au directif. In B. Bortolussi & P. Lecaudé (Eds.), La causativité en latin (pp. 199–220). Paris: L’Harmattan. Gavoille, L. (2017). La formule de conseil auctor sum dans la correspondance de Cicéron. Étude étymologique, pragmatique et sémantique. In É. Gavoille & F. Guillaumont (Eds.), Conseiller, diriger par lettre (pp. 53–70). Tours: Presses universitaires François-Rabelais. Gavoille, L. (2019). Auctor et les notions de ‘force’ et d’ ‘accroissement’: étude sémantique comparée du nom d’agent auctor et du verbe augēre. In É. Gavoille (Ed.), Qu’est-ce qu’un auctor? Auteur et autorité, du latin au français (pp. 25–40). Bordeaux: Ausonius Éditions. Kümmel, M. & Rix, H. (Eds.) (2001). Lexicon der indogermanischen Verben: Die Wurzeln und ihre Primärstammbildungen (2. Ed.). Wiesbaden: Reichert. Mayrhofer, M. (1964). Sanskrit-Grammatik mit sprachvergleichenden Erlaüterungen. Berlin/Boston: De Gruyter. Pinault, G.-J. (2003). Sur les thèmes indo-européens en *-u-: dérivation et étymologie. In E. Tichy, D. S. Wodtko & B. Irslinger (Eds.), Indogermanisches Nomen: Derivation, Flexion und Ablaut. Akten der Arbeitstagung der Indogermanischen Gesellschaft (Freiburg, 19. bis 22. September 2001) (pp. 153–188). Bremen: Hempen Verlag. Pokorny, J. (1959–1969). Indogermanisches etymologisches Wörterbuch. Bern. Renou, L. (1938). Le suffixe védique -tr̥- et les origines du futur périphrastique. Bulletin de la Société de linguistique de Paris, 39, 103–132. Schindler, J. (1975). L’apophonie des noms-racines indo-européens. Bulletin de la Société de linguistique de Paris, 67, 31–38. Schrijver, P. (1991). The reflexes of the Proto-Indo-European laryngeals in Latin. Amsterdam/ Atlanta: Rodopi. Tichy, E. (1992a). Wozu braucht das Altindische ein periphrastisches Futur? Zeitschrift der Deutschen Morgenländischen Gesellschaft, 142(2), 334–342. Tichy, E. (1992b). Zur Rekonstruktion der Nomina agentis auf *-tér- und *-tor-. In R. Beekes, A. Lubotsky & J. Weitenberg (Eds.), Rekonstruktion und relative Chronologie (pp. 411–420). Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Tichy, E. (1995). Die Nomina agentis auf -tar- im Vedischen. Heidelberg: Universitätsverlag Winter. Vernet i Pons, M. (2008). La segona conjugació verbal llatina: Estudi etimologic i comparatiu sobre l’origen protoindoeuropeu de la formació dels seus temes verbals. Barcelona: Promociones y Publicationes Universitarias.

Romain Garnier

Secondary forms reanalysed as primary ones: the crētus-class in Latin Abstract: This chapter will provide a brief overview of five Latin verbs commonly considered primary ones, as they seem at face value: (i) crēscō, crēuī, crētum ‘to grow’; (ii) discĕrnō, discrēuī, discrētum ‘to separate, part, divide’; (iii) contĕrō, contrīuī, contrītum ‘to crumble, to separate in small pieces’; (iv) requiēscō, -ēuī, -ētum ‘to rest’; and (v) uiēscō ‘to shrink up, shrivel, wither’. When we look more carefully, we find that those verbs could be accounted for as secondary verbal formations: in such a way, crēscō ‘to grow’ could be a back-formation of the preverbed accrēscō ‘to grow, increase’ (< *ad-car-ḗscō). The effect of syncope may have blunted the detail of the verbal pattern, making it unrecognisable. Such a synchronic reanalysis could explain why Lat. crēuī ‘I grew’ shows an intransitive value despite its active ending. The widely accepted assumption that crēuī would reflect a PIE root-aorist *k̑réh1-t is no longer sustainable. On a wider range, the internal evolution of Latin secondary verbs may affect the reconstruction of some PIE roots, including the most prominent ones: PIE *kreh1-(i)- ‘to sieve, to separate’, *ku̯i ̯eh1- ‘to rest, to relax’, *u̯i ̯eh1- ‘to bend, to plaid’. Lat. quatiō, quassī, quassum ‘to shake’ (usually explained by a PIE root *(s)kueh1t-) will be also accounted for. Keywords: Latin linguistics, primary verbs, secondary verbal derivation, syncope, reanalysis, PIE roots

1 Introduction The Latin verb nōscō, nōuī ‘to begin to know, become acquainted with’ is clearly a radical formation, indirectly reflecting a PIE *sk̑é/ó-formation on zero-grade, with ablaut leveling. We may assume a Pre-Lat. *gnā-sk-ō (< PIE *g̑n̥h3-sk̑-é/ó-), *gnō-w-ed (≪ PIE *g̑néh3-t), surfacing as (g)nōscō, perfectum (g)nōuit [3 sg.] ‘he recognised’ with a present stem made similar to the perfect (Garnier 2010: 159–189). The same holds for the privative adjective ignārus

Romain Garnier, Université de Limoges, Institut Universitaire de France https://doi.org/10.1515/9783110647587-010

150

Romain Garnier

‘ignorant’ (< PIE *ń̥-g̑n̥h3-ro-), the starting point of a denominative stem *ignārāre ‘to ignore’ whose vocalism was secondarily realigned to match with its antonym nōuī ‘I know’, producing ignōrāre. This sketch has been so far tentatively extended to other Latin scō-verbs, such as crēscō, crēuī, crētum ‘to grow’ and requiēscō, -ēuī, -ētum ‘to rest’, which are assumed to reflect PIE inchoative stems built on zero-grade, with the same ablaut leveling as has probably occurred in the prehistory of Lat. nōscō, nōuī, where the inherited *(g)nāscō was made similar to the perfect (g)nōuī. As a result, crēscō ‘to grow’ (perfect crēuī) is accounted for as the indirect reflex of *crāscō < PIE *kȓ̥ h1̑ sk-é/ó‘to grow’ (Kümmel LIV2017), which is compared to the Luw. (putative) pres̑ implied by the HLuw. middle partient stem *zarza- ‘to grow’ (< PIE *kȓ̥ h1-sk-é/ó-) ciple za+ra/i-za-mi-sa /zarzamis/ ‘heranwachsend’ as per Rieken (2003: 48–50). A very similar analysis led scholars to analyse Lat. requiēscō, -ēuī, -ētum ‘to rest’ as ̑ (LIV2: 393, sv. *kṷi̭eh1-‘ausruthe indirect reflex of a PIE inchoative *kṷih1-sk-é/óhen’, n. 7), whose root vocalism would have been leveled on the analogy of an old root aorist: PIE *kṷi̭éh1-t ‘he took his rest’, surfacing as Lat. quiēuī. Arm. ̑ would be echoing hangč‘im ‘I rest’ (< *han-gi-č‘im < Post-IE *°kṷih1-sk-[i̭]-é/ó-) ṷ ̑ requiēscō, which stands for *an-quī-scō (< PIE *án-k ih1-sk-e/o-). In the following sections, I aim to show that there is a more economical explanatory scheme, and that several canonical isoglosses should be given up for the sake of consistency.

2 Five ‘secondary’ Latin verbs 2.1 Lat. crēscō, crēuī, crētum ‘to grow’ The Latin verb crēscō, crēuī, crētum ‘to grow’ is accounted for as the reflex of a ̑ ̑ 1-s-m̥ ‘I made grow’, with a sigmatic aorist *krḗh PIE primary stem *kȓ̥ h1-sk-é/ó-, reflected by Gr. *κρῆσαι implied by the Hesychian heortonym κρησίπαιδα. The ̑ 1-t ‘to grow’ lacks any evidence. The ad hoc assumption of a root aorist PIE *kréh often adduced analysis of crēscō, crēuī, crētum relies on a Pre-Latin system *krāsk-ō, *krē-ṷ-ai̭, *krā-to- before a presumed ablaut leveling. Contrarily to what I assumed earlier (Garnier 2010: 166), such an explanation cannot be upheld: if ̑ 1-s-m̥ was the starting point of such an ablaut levelthe PIE sigmatic aorist *krḗh ̑ 1-), how could the donor form have lost ing (pace Kümmel LIV2017: 45, s.v. *kreh its causative meaning *‘to make grow’? It is more expedient to assume that accrēscō, ac-crēuī, ac-crētum ‘to increase, to grow’ (whose ante-class. antonym is ̮ dēcrēscō ‘to decrease’) reflect a secondary inchoative stem *ad-cĕr-ḗscō [–̀⚒̮ –́⚒],

Secondary forms reanalysed as primary ones: the crētus-class in Latin

151

̮ *ad-cĕr-ḗtus [–̀⚒̮ –́⚒]̮ with syncope. The rhythmic structure *ad-cĕr-ḗuī [–⚒ ̀ ̮ –́⚒], [–̀⚒̮ –́⚒]̮ was subject to syncope in Latin (Garnier 2012, with literature).1 Those present stems belong to the same pattern as ex-ŏl-ḗscō, ex-ŏl-ḗuī, ex-ŏl-ḗtum ‘to grow to an end, to pass away’ (< pre-Lat. *ex-ăl-ḗscō), built on the primary verb alō, -ere ‘to feed, to nourish’ (< PIE *h2él-e/o-).2 There is a striking parallel between accrēscō ‘to grow, increase’ and adolēscō ‘to grow up’ (< pre-Lat. *ad-ălḗscō).3 Kümmel (LIV2017) reconstructs two separate roots: PIE *k̑erh3- ‘to satiate, to feed’ and PIE *k̑reh1- ‘to grow up, to increase’. In my opinion, such an assumption is unnecessary, as such a situation is paralleled by Lat. alēscō (ante-class.) ‘to grow up, to increase’ vs. alō ‘to feed, to nourish’ (Cf. Gr. ἄναλτος ‘insatiable’). In Oscan, there is a subjunctive karanter ‘they feed on, they eat’ [3 pl. middle voice] which may reflect Proto-It. *kar-e/o- ‘to feed’ (< PIE *k̑r̥h1-é/ó-), whose middle voice forms meant something like ‘to feed on, to eat’. The PIE root was *k̑erh1- ‘to nourish, satiate, (make) grow’, as is evidenced by the theonym Cerēs (< *k̑erh1-ḗs), undoubtedly an inherited designation for the embodied growth.4 Turning back to crēscō, crēuī, crētum, I would suggest that the starting point was a zero-grade thematic root-present *car-ō, *car-ĕrĕ ‘to feed’ (< PIE *k̑r̥h1-é/ó-), cognate of Oscan karanter ‘uescantur’ and indirectly reflected by the gloss carensis: pistoribus ‘bakers’ (CGL 5.14.26).5 On this primary verb

1 According to Nishimura (2011: 32), the sonority (prominence) of vowels is a conditioning factor of syncope: he assumes that “/i/ and /u/ and [ǝ] were the most likely to delete in Latin”. We may analyse *accerēscō [ak.kǝ.rḗs.kō] as the reflex of *ad-car-ēscō with the regular Latin vowel weakening of [ar] to [ǝr], conditioning syncope, especially when the rhythmic structure was [–̀⚒̮ –́ ⚒̮]. 2 It is worthy to note that the inchoative stem abolēscō ‘to decay little by little, to vanish, cease’ (< *ab-ăl-ḗscō), may have had a doublet *dēlēscō (< *dē-ăl-ḗscō), with a perfect participle dēlētus (< *dē-ăl-ētus ‘abolished’), on which was back-formed a new causative verb dēleō ‘to abolish, destroy, annihilate’ (Garnier 2016a: 395). 3 According to Haverling (2000: 282), in these cases, ad underlines the direction of the development in question, like in ad-augesco ‘to grow, become bigger’. The couple accrēscō vs. dēcrēscō (both in Plautus and Cato) is well attested in the earlier periods (Keller 1992: 155). 4 The classical adjective crēber, -bra, -brum ‘thick, frequent’ is often referred to as the outcome of Proto-It. *krē-sro- (Hill 2006: 196) or *krē-ϑro- (De Vaan 2008: 144). I would rather assume a back-formation from Lat. crēbrēscō ‘to become frequent, to increase, grow strong’ (assimilation of *crēbĭlḗscō from *crēbĭlis). 5 Oscan *karensis [dat. pl.] ‘pistoribus’ (< present participle *karentsifs) rendered as carensis in the gloss. Note that Oscan praesentid ‘praesente’ is inflected as an i-stem. Assibilation in Oscan is attested for the PlN Bansae ‘Bantiae’ (locative), from *Bantyai. Strictly speaking, the (dialectal) assibilation may have been triggered by [y], not by [i]. If the present participle was

152

Romain Garnier

*car-ō, *car-ĕrĕ ‘to feed’ was built an inchoative *ad-car-ēscō ‘to grow, increase’ regularly reflected by accrēscō. The semantics are paralleled by adolēscō ‘to grow up’ (< pre-Lat. *ad-ăl-ḗscō) from alō ‘to nourish, to feed’. The vowel weakening (*accerēscō [ak.kǝ.rḗs.kō] < *ad-car-ēscō [ak.ka.rḗs.kō]) antedates the syncope (Lat. accrēscō [ak.krḗs.kō] < *accerēscō [ak.kǝ.rḗs.kō]). As a result, Lat. √crē- was reanalysed as a verbal root, producing the rare crēscō ‘to grow’6 which implies a synchronic reanalysis as a primary verb crēscō, perfect crē-uī similar to nō-scō, perfect nō-uī.

2.2 Lat. dis-cĕr-nō, dis-cr-ēuī, dis-cr-ētum ‘to separate, part, divide’ An analogical extension of this verbal pattern may account for Lat. dis-cĕr-nō (< *dis-kri-nō), dis-cr-ēuī, dis-cr-ētum ‘to separate, part, divide’ (< *dis-cĕr-ḗuī, -ḗtum)7 and spĕr-nō, spr-ēuī, spr-ētum ‘to despise, to scorn, spurn’ (< *°spĕrḗuī, -ḗtum), whose participles respectively stand for expected *dis-certus (< Proto-It. *dís-kri-to-) and *sprātus (< PIE *spr̥h2-tó-). Even if those verbs are not secondary, we may assume that an ablaut discernō [dis.kǝr.nō], *discertus [dis.kǝr.tus] (< *[dis.krǝ.tus] < Proto-It. *dis-kri-to-), was extended in -ētus, giving *dis-cĕr-ḗtus [dis.kǝ.rḗ.tus] ‘separated’ eventually syncopated in discrētus [dis.krḗ.tus]. Concerning spernō, we must assume that the starting point was maybe a eto-participle: Proto-It. *apo-spar-eto- (≪ PIE *ápo-spr̥h2-eto‘spurned’), which would have surfaced as Lat. *ab-sper-tus [as.pǝr.tus], extended in *asper-ētus [as.pǝ.rḗ.tus] > *asprētus [as.pǝ.rḗ.tus], which gave sprētus by depreverbation. The preverb ab- is needed to explain the semantics, because the PIE root *sperh2- means ‘to trample’ (De Vaan 2008: 580), as is evidenced by Hitt. išparra- ‘to trample’, YAv. spara- ‘to tread’, whereas Ved. √apa-sphari- rather means ‘to reject with the foot, to kick back’.

*karenti- ‘nourishing’, then the genitive plural was something like *karentíúm [karentyom], surfacing as (dialectal) *karensíúm [karentsyom]. The expected dative/ablative plural *karentis (< *karentifs) would then have been remade as *karensis. Note also Oscan caria ‘panis’. 6 As pointed out by an anonymous reviewer, crēscō is a very rare verb in the most ancient authors, scoring only nine attestations in Plautus, Terence and Cato. 7 The laryngealistic reconstruction of a root *kreh1-(i)- ‘to sieve, separate’ (LIV2: 366) is unnecessary.

Secondary forms reanalysed as primary ones: the crētus-class in Latin

153

2.3 Lat. con-tĕr-ō, con-tr-īuī, con-tr-ītum ‘to crumble, to separate in small pieces’ Lat. tĕrō, trīuī, trītum ‘to rub, rub to pieces’ is often accounted for in the literature by a root present *térh1-ti or *térh1-e/o- (Schrijver 1991: 396). According to the author, the perfectum trīṷī (< PIE *tréh1-i-) and the perfect participle trītus (< PIE *tríh1-to-)8 are built on an i-stem, which would be totally unparalleled in Latin. The ‘regular’ structure is to have an i-enlargement in the present stem, not in the preterite. To avoid this ‘anti-canonical’ structure, Kümmel accepts Meiser’s assumption of a doublet *√trei ̭(H)g̑- (LIV2: 632, n. 4 s.v. *terh1-)9 reflected by Lat. trīuī, trītum (which would point eventually to a case of suppletism). This is a far-fetched assumption, made in order to account for Gr. ́ ‘to rub’ (which ought to be left aside, being particularly obscure).10 In τρīϐω my opinion,11 the verbal system tĕrō, trīuī, trītum may be unified by assuming an athematic i-present: PIE *tórh1-i-, *tóh1-i- ‘to rub’, reflected by Lat. *tărĭō, -ĕrĕ surfacing as con-tĕr-ĕrĕ (< *con-tărĭō, -ĕrĕ) with the effect of vocalic reduć tion. The perfect participle con-trītus ‘rub to pieces’ may be the regular out12 ̮ ̮ ́ ́ The iō-verbs belonging to come of a *con-tĕrītus [–̀ ⚒–́ ⚒] (< *con-tărītus). the third conjugation in Latin were prone to hesitate between infinitives in -ĕrĕ and -īre (cf. Ennian ōua parīre ‘to lay eggs’). There is mention of an old perfectum teruī according to Charisius (GLK I: 248). This is a good argument to reconstruct an archaic pattern *ter(i)ō, -ĕrĕ, teruī, *terītus (trītus) ‘to rub’.

2.4 Lat. re-qui-ēscō, -ēuī, -ētum ‘to rest, be idle’ (Pl.+) 2.4.1 The classical analysis This scō-verb is commonly analysed as the reflex of a Proto-It. root *kʷiyē- (< PIE *kṷi̭eh1-).13 Latin shares an isogloss with Armenian for the PIE present morpheme 8 The root accent assumed here by Schrijver is only for the sake of Dybo’s law (according to which an oxytone accent *trih1-tó- would trigger a short -i- in Italo-Celtic). The very existence of this law in Latin is questionable, as I have already stated (Garnier 2016b). 9 Refering to Meiser’s unpublished Habilitation Vorgeschichte und Ausbildung des lateinischen Perfektsystems, Freiburg, §259. This odd assumption simply disappeared in the edited version (Meiser 2003: 228). 10 According to Beekes (2010: 1509), “no exact agreement exists outside Greek”. 11 For further discussion, see Garnier (2016a: 405–406). 12 The rhythmic structure [–̀⚒̮ –́⚒̮] was subject to syncope in Latin (Garnier: 2012). 13 See, for instance, De Vaan (2008: 508) and Lipp (LIV2: 393, s.v. *kṷi ̭eh1- ‘ausruhen’).

154

Romain Garnier

̑ *-sk-é/ó(which is quite rare, especially concerning the verbal system). Arm. hangč‘im [mid.] ‘I rest’ is from *hangi-č‘im ‘I rest’, regularly associated to an aorist hangeay ‘I rested’ (< *hangi-ay), the -i- being clearly preserved in the substantive Arm. hangist ‘rest’. The protoform leading to what we find in Armenian would be ̑ The preverb hansomething like Pre-Arm. *an-kʷī- ‘to rest’ (< *an-kṷih1-sk-[i̭]-é/ó-). is a variant of an- (= Gr. ἀν-, ἀνά-), as is evident from the doublets Arm. am-baṙnam ‘erheben’ (< PIE *án-bʰr̥gʰ̑ -) and am-berem ‘ertragen’ (< PIE *án-bʰer-e/o-)14 ́ ā˘ ‘to be mentioned by Pokorny (IEW: 39).15 The root is also reflected by OAv. √š(ii) calm, glad, rest’ (< Proto-Ir. *čyā-ya-ti), with a fossilised participle used as an adjeć tive: OAv. šiiāta‘glad, happy’, OP šiyāta- ‘peaceful, happy (on earth)’. Note OCS po-čijǫ ‘I rest’ vs. causative po-kojǫ ‘I calm (someone)’.16 The Latin perfectum quiēuit (if from Proto-It. *kʷiyē-w-ed) is considered as the reflex of an athematic root aorist PIE *kṷi̭éh1-t ‘he took his rest’.17

2.4.2 A new proposition In my opinion, the alleged equation between OAv. š ́iiāta-, OP šiyāta- and Lat. quiētus is a mistake. I suggest a segmentation *re-qui-ēscō, -ēuī, -ētum, on the same pattern as Lat. ex-ŏl-ḗscō, ex-ŏl-ḗuī, ex-ŏl-ḗtum ‘to grow to an end, to pass away’ (< *ex-ăl-ḗscō), built on the primary verb alō, -ere ‘to feed, to nourish’ (< PIE *h2él-e/o-), as stated above (Section 2.1). The ‘real’ primary root would be PIE *kṷei-̭ ‘to pile up, to heap’, reflected by Ved. √ci- (present stem cinóti ‘to heap’) and OAv. √caii-/š ́ii- ‘to heap up, gather’. Ved. kā́ya- m. ‘body’ is recognised as a metaphorical designation of the body as an ‘assemblage’ (Pāṇ. iii, 3, 41). This word points to a PIE action noun *kṷói-̭ o- m. ‘action of piling up’, possibly joined to a collective stem: PIE *kṷi-̭ éh2 ‘heap, pile of

14 The Armenian exhibits graphical doublets ham-baṙnam ‘erheben’ and ham-berem ‘ertragen’, which can be both accounted for by a confusion of the inherited preverb an-/am- with Arm. ham- ultimately borrowed from Parth. ham- ‘together’ (= Ved. sam- ‘id.’). In this respect, we may note here the etymologicaly related Xwar. hncy- (< Com. Ir. *ham-čyā-ya-) ‘to rest, relax’ vs. hnc’wy- ‘to let (it) rest’ (Cheung 2007: 38). 15 Contra: Klingenschmitt (1982: 70, n. 14) assumes that Arm. ham- is the regular reflex of PIE *sm̥- ‘together’, but this is far from convincing, since PIE *sm̥- ‘together’ as a verbal prefix is an Indo-Iranian innovation. 16 Imperfective OCS po-čivati ‘to lie’ (Ru. počivát’ [slavonism], present po-čije-: zdes’ počíet ‘here lies’). Compare also OPrus. etskī- ‘to rise (from the dead)’ < *at-si-kī- ‘to awake’ as per Vaillant (GCLS III: 279). 17 Lipp (LIV2: 393, sv. *kṷi ̭eh1-‘ausruhen’, n. 3), following Klingenschmitt (1982: 282).

Secondary forms reanalysed as primary ones: the crētus-class in Latin

155

leaves, *bed of leaves’, on which was eventually built a secondary derivative *kṷi-̭ eh2-ió̭ nt- ‘lying (in bed), resting’ and a (secondary) action noun *kṷi-̭ éh2ti- [f.] ‘action of lying’,18 which matches quite well with Proto-Ir. *čyā-ti- [f.] ‘action of lying (in peace), calm, peace’, which leads to OP š(i)yāti- [f.] ‘welfare, peace (on earth), happiness (after death)’. Once reanalysed as a ‘present’ participle, the adjectival stem PIE *kṷi-̭ eh2-ió̭ nt- ‘lying (in bed), resting’ was the source of a ‘stative’ present stem *kṷi-̭ eh2-iḙ́ /ó- ‘to rest in bed, be calm’, with a ‘new’ to-stem *kṷi-h2-tó- ‘quiet’. OAv. š ́iiāta- ‘happy’ and OP šiyāta‘peaceful, happy (on earth)’ may be secondary forms with full-grade, on the same pattern as Proto-Ir. *dāta- ‘put’ vs. Ved. hitá- ‘id.’ (< PIE *dʰh1-tó-, cf. Gr. θετός). Therefore, it may be unnecessary to assume a PIE etymon *kṷieh1-tóon the sole basis of OAv. š ́iiāta-, OP šiyāta- and Lat. quiētus. It is be possible to assume that *re-qui-ēscō, -ēuī, -ētum was in fact an inchoative stem built on an (unattested) participle *quītus ‘quiet’, the substantive requiēs, -em [f.] (variant: -ētem) being simply a back-formation from requiēscō (Garnier 2016a: 313), on the same pattern as the ‘athematic’ adjective mansuēs ‘tamed’ from mansuēscō ‘to tame’. See also prōlēs [f.] ‘offspring’ which may be back-formed on an inchoative stem *prōlēscō, -ēuī, -ētum ‘to grow forth, spring up, arise’ (< *prŏ-ăl-ēscō). The Latin adjective tranquillus ‘calm’ was brilliantly explained by Vine (2008) as a hypostatic adjective reflecting a Proto-It. locution *trans=kʷeslim ‘after the tempest’, with a name for ‘tempest’ ultimately built on the PIE root *kȗ ̭es- ‘to blow’.19 In my opinion, this cogent etymology needs a phonetical explanation for the (irregular) treatment of trans- (we would expect here Lat. **trās-quillus). As I have previously suggested (Garnier 2010: 172), we can assume here a contamination with an old preverbed stem *an-quillus ‘resting, quiet’ (< Proto-It. *an-kʷī-lo-), with the same preverb PIE *an- ‘above, upward’ possibly shared by Arm. hangč‘im ‘I rest’ and Gr. ἀνα-παύομαι ‘id.’ To sum up, Vine is perfectly right in assuming a Proto-It. locution *trans=kʷeslim ‘after the tempest’, but this analysis cannot account straightforward for Lat. tranquillus: the possibility of a contamination with another adjective *anquīlus/*anquillus may vindicate the unexpected treatment of tranquillus.

18 The oblique case of which (PIE *kṷih2-téi ̭-) is ultimately the source of Proto-Sl. *po-čìti. 19 For the semantics, cf. Ved. śúṣma- [adj.] ‘rushing’ (said of wind, fire, water) from PIE *k̑usmó-.

156

Romain Garnier

2.5 Lat. ui-ēscō ‘to shrink up, shrivel, wither’ The same analysis can be extended to Lat. ui-ēre ‘to bend, to plaid, to weave’ (pre-classical), which is joined to an inchoative stem ui-ēscō ‘to shrink up, shrivel, wither’, with a participle ui-ētus ‘bent, bent up’ used as an adjective: ‘shrivelled, wrinkled’ (Hor.+). The assumption of a PIE root *ṷi ̭eh1- ‘to bend, to plaid’ is not convincing (pace Kümmel LIV2: 695). In my opinion, the starting point could be the primary root PIE *ṷeh2- ‘to be bent’, reflected by Lat. uārus [adj.] ‘bent, grown inwards’ (< PIE *ṷéh2-ro-). We may assume then a present stem *ṷéh2-i- ‘to bend’, with a to-stem *ṷih2-tó- ‘bent, twisted’ (< *ṷǝh2-i-tó-)20 reflected by Ved. vītá- ‘turned to’ and by Lat. uĭtta f. ‘band (worn round the head)’, standing for *uīta (2) with the effect of the so-called lĭttera-rule and an early feminisation of a neuter, the form *uīta (2)/uĭtta ‘band’ standing for Stand. Lat. *uītum [n.] ‘what is twisted’. From this *uītus ‘bent’, the Latin language could create a causative verb ui-ēre ‘to bend, to plaid, to weave’, like ci-ēre ‘to put in quick motion’ vs. citus ‘quick, rapid’ and also an inchoative ui-ēscō ‘to shrink up, shrivel, wither’, *ui-ēuī, ui-ētus ‘bent, bent up, shrivelled, wrinkled’.

3 Appendix: Lat. quatiō, quassī, quassum ‘to shake’ As I have pointed out before (Garnier 2010: 97), it is unsatisfactory to assume a PIE Root aorist *kṷéh1t-s, *kṷéh1t-s, *kṷéh1t-st ‘to shake’ whose second-person singular would have resulted in Proto-Gr. *ékʷēs ‘you [sg.] shook’. According to Lipp (LIV2: 564, n. 2), such a form would have been the starting-point for a ‘new’ paradigm *ἔκʷην et *ἔκʷη following the pattern of Proto-Gr. *égʷān, *égʷās, *égʷā ‘I went’ (= Gr. ἔϐην, ἔϐης, ἔϐη).21 Paul the Deacon has conserved a verb intercutīre in the gloss Intercutitus uehementer cutītus, hoc est ualde stuprātus (P.-Fest. 100, 24 L.) ‘intercutītus means violently cutītus, that is strongly dishonored’. It may hardly be separated from per-cutiō, -ere ‘id.’ Such

20 The metathesised to-stem *ṷih2-tó- ‘bent, twisted’ was eventually the source of a secondary root *ṷei ̭h2-whose middle present *ṷéi ̭h2-o(ri) [3 sg.] is reflected by Hitt. *u̮eḫa(ri), surfacing as an enlarged form ṷeḫattari ‘he turns himself’ (Kloekhorst 2008: 995). 21 Contra: Lipp (LIV2: 564, n. 2). Schrijver (1991: 463) reconstructs a PIE present stem *(s)kṷóti- ‘to shake’ with delabialisation (but Gr. πάσσω would then remain unexplained).

Secondary forms reanalysed as primary ones: the crētus-class in Latin

157

a verb is not likely to have arisen from a locution ‘inter=cutem’ ‘under the skin’.22 We may rather assume Non-Stand.-Lat. *quā ́tit/*quắttit [3. sg.] (< *kʷā́́ (< *quăttīre) ́ 23 reflected by the preverbed particiti-i ̭e/o-) vs. infinitive *quătīre ́ ́ ple inter-cutītus /in.ter.kʷǝ.t.tus/ which may reflect *inter-quătītus with the regular Latin vowel weakening. The analogically shortened indicative quatiō (from an infinitive *quatīre) was reshaped into a hight-register standard Latin: quătĭō, -ĕrĕ ‘to shake’, on the same pattern as căpĭō, -ĕrĕ ‘to take’, răpĭō, -ĕrĕ ‘to seize and carry off, to snatch, carry away, tear’ (vs. Vulg. Lat. *rapīre surfacing as Fr. ravir ‘to seize, abduct’). We may then compare Proto-Celt. *kʷā-ti- ‘winnowing’, reflected by OIr. cáith [f.] ‘chaff, husk’, which is glossed by Lat. furfur/acus and has the secondary meaning: ‘scrap, rubbish’. MoIr. cáithim ‘to winnow’ is from Proto-Celt. *kʷāti-ye/o- ‘id.’ (which perfectly matches the putative Non-Stand.-Lat. *quā́tit assumed above). Matasović (2009: 175, s.v. *kʷā-ti- ‘chaff, husk’) is reluctant to account for the OIr. form by a PIE etymon *kṷoh1t-í- because ‘the reconstructed PIE root shape (PIE *(s)kṷeh1t-) is highly unusual’. Hom. ἐπι-πάσσω ‘to pour over, sprinkle’ (< *‘to shake’) may go back to Proto-Gr. *epi-kʷ-áō (which gave rise to first millennium Gr. *ἐπι-πάω/*-άζω, aor. °πάσσαι, participle ἐπί-παστος), alternating with *kuw-áō ‘to separate the grain from the chaff, to winnow’ (< PIE *(s)kuh1-eh2-iḙ́ /ó- ‘to shake’), with a regular glide à la Lindeman. Gr. κύᾰμος [m.] ‘bean’ could be the reflex of a secondary-stem *κυᾰ-μός ‘winnowed’, with the same pattern as Gr. χηρᾰ-μός [f.] ‘hole, cleft’ built on an unattested present stem *χηράω ‘to be empty’. Notwithstanding the Epidauric ἐπιπῆν ‘to pour, sprinkle’, Beekes (2010: 1155) assumes a PIE etymon *kṷéh2t-iḙ /o- in order to compare ἐπιπάσσω with Lat. quatiō, ‘ignoring the *ē to which the Epidauric form points’. The Doric form is not a root aorist: it belongs to a present stem *ἐπίπημι, infinitive ἐπιπῆν. Besides, in terms of Aktionsart, a root ‘to shake’ is not a good candidate for a telic root aorist. The expected Dor. *ἐπίπᾱμι was contaminated by an inherited action noun *πη-τύς (≪ Proto-Gr. *κʷη-τύς [f.] ‘winnowing’), reflected by the Laconian secondary derivative πήτεα· πίτυρα [n. pl.] ‘chaff’ (Hesychius.). This points to a PIE primary tu-stem *kṷéh1-tu- [m.] ‘shaking’ with full grade of the root.

22 Despite the awkward gloss Inter cutem flāgitātōs dīcēbant antīquī marēs, quī stuprum passī erant (P.-Fest. 98, 22 L.) ‘The Ancients called ‘dishonored under the skin’ the males who had undergone anal intercourse’. Note that we have to assume for the verb flāgitāre a meaning close to flāgitium ‘dishonor’. Adams (1982: 147) explains intercutiō by inter cutem. 23 This sound change is paralleled by Lat. ătrṓcem (< *ăt.trṓcem < *ātrṓcem) ‘dreadful, fierce’ vs. ā́ter ‘dark’.

158

Romain Garnier

The old to-stem of the primary root *(s)kṷeh1- ‘to shake’ (not †(s)kṷeh1t-) was PIE *(s)kuh1-tó- ‘shaken’. Cf. OE hūdenian ‘to shake’ (weak verb second class), which reflects West. *hūdanōjan ‘id.’ built on Proto-Germ. *χūđaz ‘shaken’ (< PIE *kuh1-tó-). PIE secondary present stem *(s)kuh1-teh2-i̭é/ó- ‘to shake, hurry’ (cf. OE scūdende, OCS skytati sę), with a doublet: PIE *(s)kuh1-eh2-i̭é/ó- ‘id.’. Remarkable is the ‘Italo-Celtic’ isogloss *kʷā-́ ti- [f.] ‘winnowing’ built on *kʷā-yé/ó- ‘to shake, to winnow’ (< PIE *(s)kuh1-eh2-i̭é/ó-), together with a very significant feature: the ‘Italo-Celtic’ (?) denominative present stem *kʷā-ti-yé/ó- ‘to shake, winnow’ (cf. MoIr. cáithim, Non-Stand.-Lat. *quāt́ it/*quăt́ tit).

4 Concluding remarks Should the views assumed here be accepted (at least partly), we may renounce for good unsatisfactory etymologies, such as Lat. crēscō ‘to grow’ (perfect crēuī) ̑ ‘to grow’ accounted for as the indirect reflex of *crāscō < PIE *kȓ̥ h1-sk-é/ó2017 (Kümmel LIV ), Lat. (dis-)crēuī ‘I sieved’ from Proto-It. *kreye- (Meiser 2003: 226), Lat. trīuī from PIE *tréh1-i- (Schrijver 1991: 396), Lat. re-qui-ēscō ‘to rest, be idle’ from a Proto-It. root *kʷiyē- (< PIE *kṷi̭eh1-) ‘to rest’ (De Vaan, 2008: 508; Lipp LIV2: 393, s.v. *kṷi̭eh1- ‘ausruhen’), Lat. quatiō ‘to shake’ from PIE *(s)kueh1t(Lipp LIV2: 564). On the other hand, we may now assume a whole set of new ré 1-i- ‘to constructions: PIE *kȓ̥ h1-é/ó- ‘to feed’ (Proto-It. *kar-e/o-), PIE *tórh1-i-, *tr̥h rub’, reflected by Lat. *tărĭō (preverbed con-terere), PIE action noun *kṷói̭-o- m. ‘action of piling up’, joined to a collective stem: PIE *kṷi̭-éh2 ‘heap, pile of leaves, *bed of leaves’, secondary derivative *kṷi̭-eh2-i̭ónt- ‘lying (in bed), resting’, ‘stative’ present stem *kṷi̭-eh2-i̭é/ó- ‘to rest in bed, be calm’, with a ‘new’ to-stem *kṷih2-tó- ‘quiet’. From the primary root *(s)kṷeh1- ‘to shake’, source of a secondary present stem: PIE *(s)kuh1-eh2-i̭é/ó- ‘to winnow’ (Italo-Celt. *kʷā-yé/ó-). This verb produced an action noun *kʷā-́ ti- [f.] ‘winnowing’, source of an Italo-Celtic denominative present stem *kʷā-ti-yé/ó- ‘to shake, winnow’ (cf. MoIr. cáithim, Noń Stand.-Lat. *quāt́ it/*quăt́ tit/ *quătīre).

Abbreviations Arm. Dor. Gr. Hitt. Hluw.

= = = = =

Armenian Doric Greek Hittite Hieroglyphic Luwian

Secondary forms reanalysed as primary ones: the crētus-class in Latin

MoIr. Non-Stand.-Lat. OAv. OCS OE OIr. OP OPrus. Parth. PIE Post-IE Pre-Arm. Pre-Lat. Proto-Celt. Proto-Germ. Proto-Gr. Proto-It. Proto-Ir. Proto-Sl. Ru. Skt. Stand. Lat. Ved. West. Xwar. YAv.

= = = = = = = = = = = = = = = = = = = = = = = = = =

159

Modern Irish Non-Standard-Latin Old Avestan Old Church Slavonic Old English Old Irish Old Persian Old Prussian Parthian Proto Indo-European Post-Indo-European Pre-Armenian Pre-Latin Proto-Celtic Proto-Germanic Proto-Greek Proto-Italic Proto-Iranian Proto-Slavic Russian Sanskrit Standard Latin Vedic Westic Xwarezmian Young Avestan

Bibliography CGL = Goetz, G. (1889–1923). Corpus glossariorum Latinorum (7 vols.). Leipzig. GCLS = Vaillant, A. (1950–1977). Grammaire comparée des langues slaves (5 vols.). Paris: Klincksieck. GLK = Grammatici Latini (Keil). Grammatici Latini ex recensione Henrici Keilii (7 vols.). Leipzig: Teubner. IEW = Pokorny, J. (1959). Indogermanisches etymologisches Wörterbuch (2 vols.). Bern/ Stuttgart: Francke Verlag. LIV2= Rix, H. et al. (20012). Lexikon der Indogermanischen Verben. Die Wurzeln und ihre Primärstammbildungen. Unter Leitung von Helmut Rix, bearbeitet von Martin Kümmel, Thomas Zehnder, Reiner Lipp, Brigitte Schirmer. Wiesbaden: Dr. Ludwig Reichert Verlag. Pāṇ. = Pāṇinī’s Aṣṭhādhyāyī (Pāṇinī’s Grammar in 8 vols.). Adams, J. N. (1982). The Latin sexual vocabulary. Baltimore: John Hopkins University Press. Beekes, R. S. P. (2010). Etymological dictionary of Greek, with the assistance of Lucien van Beek, 2 Volumes. Leiden/Boston: Brill.

160

Romain Garnier

Cheung, J. (2007). Etymological dictionary of the Iranian verb. Leiden Indo-European Etymological Dictionary Series. Edited by A. Lubotsky. vol. 2. Leiden/Boston: Brill. Garnier, R. (2010). Sur le vocalisme du verbe latin: Étude synchronique et diachronique. Innsbruck. Garnier, R. (2012). Allomorphisme et loi de limitation rythmique en latin. Bulletin de la Société de Linguistique de Paris, 107(1), 235–260. Garnier, R. (2016a). La dérivation inverse en latin. Innsbruck. Garnier, R. (2016b). Dérive postnéogrammairienne des laryngales. In J.-L. Léonard (Ed.), Actualité des Néogrammairiens, Mémoires de la Société de Linguistique de Paris, (vol. 23, pp. 159–182). Leuven: Peeters. Haverling, G. (2000). On Sco-verbs, Prefixes and Semantic Functions: A study in the development of prefixed and unprefixed verbs from Early to Late Latin (Studia Græca et Latina Gothoburgensia LXIV). Göteborg: Acta Universitatis Gothoburgensis. Hill, E. (2006). Lateinisch crēscō ‘wachsen’ etymologisch: urslavisch *krějǫ ‘genesen’, litauisch šeriù ‘futtern’ oder armenisch serem ‘erzeugen’? International Journal of Diachronic Linguistics and Linguistic reconstruction, 3, 187–209. Keller, M. (1992). Les verbes latins à infectum en -sc-, étude morphologique à partir des formations attestées dès l’époque classique (Collection Latomus vol. 216). Bruxelles: Latomus. Klingenschmitt, G. (1982). Das altarmenische Verbum. Wiesbaden: Dr. Ludwig Reichert Verlag. Kloekhorst, A. (2008). Etymological dictionary of the Hittite inherited lexicon (Leiden IndoEuropean Etymological Dictionary Series. Edited by Alexander Lubotsky, vol. 5). Leiden/ Boston: Brill. Kümmel, M. (LIV2017). Addenda et corrigenda zu LIV2www.oriindufa.uni-jena.de/iskvomedia/in dogermanistik/Kümmel_liv2_add.pdf. Matasović, R. (2009). Etymological dictionary of Proto-Celtic (Leiden Indo-European Etymological Dictionary Series. Edited by Alexander Lubotsky, vol. 9). Leiden/Boston: Brill. Meiser, G. (2003). Veni Vidi Vici, Die Vorgeschichte des lateinischen Perfektsystems. München: Beck. Nishimura, K. (2011). Syncope in Latin: A historical restatement of its conditions. International Journal of Diachronic Linguistics and Linguistic Reconstruction, 8(1), 1–41. Rieken, E. (2003). Hieroglyphen-luwisch zí+ra/i-la-mi-i („SCALPRUM.ARGENTUM“) su-ha-pana-ti: ein Kompositum und eine neue luwisch-lateinische Isoglosse. Historische Sprachforschung, 116(1), 33–53. Schrijver, P. (1991). The reflexes of the Proto-Indo-European laryngeals in Latin (Leiden Studies in Indo-European, vol. 2). Amsterdam/Atlanta: Rodopi. Vaan, M. de. (2008). Etymological dictionary of Latin and the other Italic Languages (Leiden Indo-European Etymological Dictionary Series. Edited by Alexander Lubotsky, vol. 7). Leiden/Boston: Brill. Vine, B. (2008). On the etymology of Latin tranquillus ‘calm’. International Journal of Diachronic Linguistics and Linguistic Reconstruction, 5(1), 1–24.

Benjamín García-Hernández

Origen y significado de dēsīderāre: nostalgia y deseo desde Plauto Abstract: In this chapter, we propose that desīderare is the intensive derived form of desīdere (“to go to the bottom”), as consīderare is of consīdere. This suffixal derivation is not different to that of tolerare and reciperare from tollere and recipere, respectively. The first meaning of desīderare would be “to be depressed, feel nostalgic” due to the absence of someone or loss of something. When this retrospectivity becomes prospective, the verb takes on the meaning “to desire” something in the future. This is the difference between defunctos desidero and defunctos uidere desidero. Keywords: Latin linguistics, nostalgia, desire, etymology, suffix -erāre

1 De consīderare a desīderare: Estado de la cuestión y nuevo origen El estudio del origen de desīderare ha estado siempre ligado al de consīderare. Ernout y Meillet (2001: s.v. sīdus) sostienen que se ha formado « sans doute sur consīdero, comme desero sur consero . . . 1: cesser de voir, constater [ou regretter] l’absence de; d’où “chercher, désirer” ».2 Lexicógrafos y gramáticos latinos no tuvieron dificultad en señalar el sustantivo sīdus como base derivativa de ambos verbos:

1 Cf. también s. v. sero: « desero: se détacher de, lâcher, déserter. D’abord terme de la langue militaire, formé sur conserere, dont il est l’opposé: deserere pugnam, d’après conserere pugnam ». 2 Las comillas dobles angulares indican «citas textuales», las dobles elevadas indican “significados” completos y las simples indican ‘semas’, esto es, rasgos significativos o bien las ‘traducciones’ de los textos latinos. Estas son del autor del trabajo. Agradecimiento: Hemos realizado este trabajo en el marco del proyecto de investigación dirigido por la Profª Rosario López Gregoris, con el título Drama y dramaturgia en Roma. Estudios filológicos y edición (Ref. FFI2016–74986–P). Benjamín García-Hernández, Universidad Autónoma de Madrid https://doi.org/10.1515/9783110647587-011

162

(1)

Benjamín García-Hernández

desiderare et considerare a sideribus dici certum est. (PAVL. FEST. p. 66, 7) ‘Es cierto que desiderare y considerare se dicen de las estrellas.’

En esa línea lexicógrafos y estudiosos modernos los han tomado por verbos denominativos parasintéticos, formados a la vez con prefijo y sufijo, como si sus primeros significados fueran “contemplar las estrellas” y “perder de vista las estrellas”, en una relación con la lengua augural análoga a la de contemplari.3 Esta explicación clásica no ha estado libre de la sospecha de ser una etimología popular. Las soluciones modernas del origen de consīderare y desīderare no son más satisfactorias. La del parentesco de sīdus con el lit. svidùs ‘brillante’ tiene cierta tradición (Pârvulescu 1980: 159–160). Otras más recientes ven en el sustantivo latino el étimo indoeuropeo *sīdhos. Este, con el valor de ‘dirección, meta’ (Rix 1985: 346) o de ‘marca’ (Nyman 1990: 51–68), habría pasado a denominar las estrellas como puntos de orientación.4 Según los datos recogidos en el artículo de consīderare en el TLL (vol. IV, p. 426, 1–432, 38), dividido en dos apartados fundamentales (I proprie: oculis. II translate: mente), no hay constancia de que este verbo se aplique a objetos astronómicos y estelares (caelum et stellas, p. 427, 6–11) hasta Valerio Máximo (1, 7 ext. 6), Séneca (benef. 4, 23, 4) y Aulo Gelio (2, 21, 2). Y según nuestro primer análisis semántico (García-Hernández 1976: 80–81), el verbo tiene el significado fundamental de “examinar de cerca” el valor de algo; lo que desdice de la supuesta acción de contemplar los astros. De hecho, es un verbo que Cicerón emplea varias veces para caracterizar la codicia de Verres que, tan pronto como contempla un objeto precioso, pasa a considerar su valor con la intención de apropiárselo: (2)

[Verres] accessit ad argentum, contemplari unum quidque otiose et considerare coepit. (CIC. Verr. II 4, 33) ‘[Verres] se acercó a los objetos de plata, comenzó a contemplar detenidamente uno por uno y a examinar su valor.’

La proximidad del objeto se manifiesta también en el empleo como término del examen médico (TLL s. v., I B. speciatim de opera medicorum morbum morborumue signa intuentium) y como término de la evaluación mercantil. En la acción visual (oculis), después de los usos absolutos (I 1), el primer ejemplo de empleo transitivo

3 Walde y Hofmann (1982: s.v. considero); Prellwitz (1903: 64–65, 72); Bader (1962: 134–135); Guiraud (1964: 29–30); Le Boeuffle (1977: 22–23); De Meo (1983: 135); Le Pennec-Henry (1987: 123). 4 De Vaan (2008: s.v. sīdus), y más en extenso García-Hernández (1991a: 92, 98).

Origen y significado de dēsīderāre

163

citado es el siguiente: « 2 aliquid: homines: SVET. Cal. 36 feminas ». Merece la pena aducir todo el texto, que hemos explicado con mayor detalle en García-Hernández (2000: 93–95): (3) quas [feminas] plerumque cum maritis ad cenam uocatas praeterque pedes suos transeuntis diligenter ac lente mercantium more considerabat. (SVET. Cal. 36, 2) ‘Y a estas [mujeres], invitadas a cenar por lo general con sus maridos, mientras desfilaban delante de sus pies, las examinaba diligente y lentamente al modo de los mercaderes.’ El sujeto de la acción es el propio emperador y no se trata de una pasarela para contemplar la belleza o modelos de moda. La actitud de Calígula parece más bien la de un mercader de esclavas que, en un examen visual detenido y próximo, evalúa el objeto de compra. En suma, considerare representa el examen visual, la acción visual evaluativa, ya se trate de la mirada codiciosa de Verres, de la mirada lasciva de Calígula o de una exploración médica. Por ello, es en principio una acción próxima al objeto; solo en una segunda fase debió de aplicarse a objetos distantes y lejanos, gracias al reclamo de la etimología popular: sidera considerare. Por cierto, este verbo latino bien podría representar la acción de los astrónomos modernos que, gracias a telescopios gigantes, acercan las estrellas a sus ojos. El análisis semántico y referencial nos llevó, en otro trabajo anterior (García-Hernández 1991a: 96–97), a abrir la posibilidad de que consīderare fuera un derivado intensivo de consīdere (“sentarse juntos”), de manera que su significado primario habría sido “sentarse juntos a examinar”, “ponerse a evaluar”. Hoy estamos convencidos, en efecto, de que consīderare no es sino la intensificación de la acción de consīdere; p. ej., de la que se dice de iis, penes quos arbitrium est, inprimis de iudicibus (TLL I 2 b, p. 433, 81–434, 21) o de iis, qui disputaturi considunt (I 2 e., p. 434, 22–37): (4) ut primum iudices consederunt, ualde diffidere boni coeperunt. (CIC. Att. 1, 16, 3) ‘Tan pronto como se sentaron los jurados, la gente de bien comenzó a desconfiar seriamente.’ En tal caso, la coincidencia de consīdere y sedēre en el contexto de consīderare constituiría una figura etimológica con mayor razón que la de sīdus:

164

Benjamín García-Hernández

(5) consedimus ante limen thalami . . . et quia considerantium haeserant uultus . . . (PETRON. 26, 4–5) ‘Nos sentamos ante el umbral de la cámara nupcial . . . y como, al coincidir mirando, se juntaban nuestras caras . . .’ (6) sedebamus ergo in puppi simul uniuersi et lucentia sidera considerabamus. (GELL. 2, 21, 2) ‘Nos sentamos, pues, todos juntos en popa y contemplábamos las brillantes estrellas.’ Pues bien, si la conexión de consīderare con sīdus se ha impuesto desde tiempos antiguos con la facilidad de una etimología popular, la referencia estelar de desīderare, inducida por la atribuida a aquel, ha resultado más forzada. Con razón se preguntaba Thurneysen (1906: 180): « Aber wie kommt desiderare zu seiner Bedeutung? Darauf fehlt bis heute eine befriedigende Antwort ». Así que, sentada la nueva hipótesis etimológica de consīderare, cabe proponer un origen análogo para desīderare, como derivado del ‘dinámico’ desīdĕre, no del ‘estático’ desĭdēre. Merced al valor “desde arriba” del preverbio de-” (García-Hernández 2002: 143–148), este verbo indica la acción de “irse, asentarse al fondo”, en claro contraste con tolli (“elevarse”): (7) si demiseris in aquam [ouum], . . . inane natat, plenum desidit. (VARRO rust. 3, 9, 11) ‘Si echas [un huevo] al agua, el huero nada, el lleno se va al fondo.’ (8) tollimur in caelum curuato gurgite, et idem subducta ad Manis imos desedimus unda. (VERG. Aen. 3, 564–365) ‘Nos elevamos al cielo sobre la ondulación de un remolino y, seguidamente, zafándose la ola, caímos en la profundidad de los Manes.’ El cambio semasiológico del movimiento físico (“irse al fondo”) al estado psíquico (“deprimirse, hundirse, abatirse”) es común en cualquier lengua dentro de un mismo verbo: lat. deprimere animum, deprimi (TLL vol. V, 1, p. 614, 52–73), esp. abatirse, deprimirse, hundirse (anímicamente), etc. La última acepción (1c) de desīdere en el OLD es “to be depressed” en sentido físico: desidentes sinus habent (uertebrae). No debería, pues, sorprender que desīderare, su derivado intensivo, haya asumido el sentido psíquico de “deprimirse”. Entre este significado hipotético (“deprimirse” por una pérdida) y el histórico “echar en falta” el ser u objeto perdido hay una línea de continuidad obvia: la que pasa del dolor del sujeto a la constatación de la ausencia

Origen y significado de dēsīderāre

165

del objeto. Ninguna dificultad vemos en ese cambio diatético que va del sufrimiento propio, en principio ‘intransitivo’, a su posterior proyección ‘transitiva’ sobre el objeto ausente. Es un desarrollo semasiológico similar al de considere “sentarse juntos” > considerare “sentarse juntos examinando” > “examinar de cerca”. Transitivaciones más o menos análogas a desīderare (< desīdere + -erare “deprimirse por una carencia” > “echar en falta”) se podrían aducir muchas. Dentro de la misma familia léxica: possīdere (< potis + sīdere “sentarse como dueño” > “tomar posesión de, ocupar”) y possidēre (< potis + sedēre “estar sentado como dueño” > “poseer”), que se transitivan tan pronto como sus contenidos se proyectan sobre un objeto. En un campo léxico próximo: los verbos gemere, flere, lacrimare, lugere, plorare, etc., pasan de la acción fisiológica intransitiva de “gemir, llorar” a la sentimental transitiva de “deplorar, lamentar”. Acompañando a desiderare en sentido prospectivo, puede verse una ilustrativa transitivación de somniare en (41): me desideres, me somnies. La evolución del contenido de desiderare consta, al menos, de tres importantes pasos semasiológicos: “*deprimirse por” > “echar en falta” > “desear”. El sufijo intensivo -erare, aplicado a la base de desidere (“irse al fondo”), lo sitúa en el ámbito de los sentimientos. El significado hipotético es intransitivo, pero se une al segundo (“echar en falta”) por el sentido negativo que emana del valor descendente del preverbio de-” (“desde arriba”, “cayendo”). Su transitivación no requiere más que la proyección de la acción sobre el objeto que causa la depresión. Una vez adquirido el significado transitivo “echar en falta”, se opera la inmediata transformación del valor negativo de la nostalgia en deseo positivo. Estos dos significados históricos, uno retrospectivo y el otro prospectivo, son las dos vertientes de un sentimiento doble, que no es como un monte que se sube y se baja, sino como un valle (de lágrimas) al que se desciende y del que se sube. Eso es así, porque desiderare representa primero el sentimiento negativo que resulta de una ausencia o una pérdida y a continuación la recuperación positiva del sentimiento. En esos dos significados se unen el final triste de un proceso y el deseo inicial de otro. Si ambos son como las dos caras de la misma moneda, la primera es la cruz. Los procesos formativos que defendemos aquí de consīderare y desīderare son paralelos; pero muy diferentes de las formaciones parasintéticas vigentes desde antiguo. No hay base nominal (sīdus), sino la base verbal sīdere “sentarse”, que recibe sendos preverbios: consīdere, desīdere. Luego estos se instituyen en bases del sufijo intensivo -erare. Sin necesidad de apelar al origen indoeuropeo, los dos verbos surgen dentro de la lengua latina y adquieren una morfología compleja; no por parasíntesis, sino por derivación intensiva en -erāre: consīdere > consīderare,

166

Benjamín García-Hernández

desīdere > desīderare. Ambos siguen el doble proceso compositivo y derivativo de reciperare: capere > recipere > reciperare sidere > considere > considerare sidere > desidere > desiderare.

2 Evolución polisémica de desīderare. Nostalgia y deseo desde Plauto Los usos de desiderare se distribuyen en el TLL (vol. V, 1, p. 701–710) en cinco acepciones: I i.q. amissum uel absentem requirere. II latiore sensu i.q. optare, postulare, cupere. III i.q. quaerere. IV i.q. necesse habere. V i.q. incertum esse. Las dos primeras son fundamentales y, en consecuencia, ocupan la mayor parte del espacio lexicográfico. La tercera, de la que se aportan solo dos testimonios, y la quinta, que contiene cuatro ejemplos de interrogativa indirecta, podían haberse integrado en la segunda. La cuarta, que comprende 51 líneas, conecta la primera acepción con la segunda, de manera que se puede establecer una triple gradación significativa: “echar en falta”, “necesitar”, “desear”. En esa posición intermedia, inmediatamente antes de la noción intencional de “desear”, como primer subgrado prospectivo, se sitúan los verbos desiderativos en -urire y otros análogos (García-Hernández 1980: 89–91) que expresan una necesidad fisiológica; así, se constata en esuriens (“tener hambre”) y sitiens (“tener sed”), respecto de cibum et potum desiderare: (9) si esuriens cibum aut potum sitiens desiderare, quaerere, uel etiam inuenisse uideatur . . . (MACR. somn. 1, 3, 4) ‘Si, estando hambriento o sediento, tuviera la impresión de desear, buscar o incluso haber encontrado comida o bebida . . .’ Los tres significados de desiderare trazan la línea evolutiva de la orientación retrospectiva a la prospectiva. Y los dos importantes, el primero (“echar en falta”) y el tercero (“desear”) expresan sentimientos contrapuestos que no dejan de influir en la variación de la construcción sintáctica. Una cosa será “añorar” a los difuntos (defunctos reges desiderant) y otra “desear” verlos, aunque sea en imagen: (10) dum defunctos eos desiderant in imaginibus uidere. (MIN. FEL. 20, 5) ‘Mientras desean verlos, una vez muertos, en efigies.’

Origen y significado de dēsīderāre

167

Con la noción de añoranza, la acción verbal afecta sobre todo a objetos ausentes de clase animada. En cambio, con la noción de deseo el objeto animado suele introducirse a través del infinitivo, según muestra el texto anterior (eos desiderant uidere), mientras el objeto inanimado es más directo: (11) cur pecuniam magnopere desideret . . . ? (CIC. Tusc. 5, 89) ‘¿Por qué va a desear dinero en gran cantidad . . . ?’ Desīderare se atestigua bien desde Plauto que aporta diez ejemplos: cuatro en uso absoluto, tres con acusativo y otros tantos con infinitivo (Lodge 1962: s.v.). En las dos primeras construcciones se pone de manifiesto la noción principal de nostalgia, atinente al sentimiento de ausencia o carencia del objeto (12, 13). Sin embargo, en los tres ejemplos con infinitivo la negación del verbo (non, haud, nihil desidero) viene a equiparar nociones retrospectiva y prospectiva: “no echo en falta” = “no deseo” (14): (12) misera amans desiderat. (PLAVT. Bacch. 208) ‘La pobre, enamorada, lo echa de menos.’ (13) . . . atque id se uolt experiri, suom abitum ut desiderem? (PLAVT. Amph. 662) ‘¿. . . y quiere comprobar cómo siento su ausencia?’ (14) ego bonum, quo malum accedit, mihi dari hau desidero. (PLAVT. Merc. 148) ‘Yo no deseo que se me regale un bien, al que se añade un mal.’ Los cuatro empleos de Terencio tienen una distribución análoga: dos en uso absoluto, uno con acusativo y otro precedido de negación con infinitivo implícito (nosse): (15) nostin qui sim? – non desidero. (TER. Ad. 177) ‘¿Sabes quién soy? – No deseo (saberlo).’ Hay ciertos géneros literarios y algunos autores especialmente sensibles a la noción nostálgica. Así, en el género epistolar en el que la ausencia del destinatario, amigo o familiar es tema recurrente; en la poesía elegíaca de Ovidio que, alejado de la patria, tanto añora seres queridos y lugares familiares5; en la

5 De ello se ha ocupado Fasce (1988: 74–81) en un buen estudio sobre el sustantivo desiderium.

168

Benjamín García-Hernández

historia de tema bélico, en el momento de señalar a los caídos en la guerra, a las bajas producidas en las batallas (TLL p. 703, 21–45): (16) in eo proelio non amplius ducentos milites desiderauit, sed centuriones, fortes uiros, circiter XXX amisit. (CAES. civ. 3, 99, 1) ‘Deploró la muerte de no más de doscientos soldados en el combate, pero perdió casi treinta centuriones, hombres aguerridos.’ La noción desiderativa, como valor innovado, crece especialmente en latín tardío. Mientras en Ovidio seis de nueve empleos mantienen el significado nostálgico de “echar en falta”, cuatro siglos más tarde, de los diecinueve usos de Macrobio trece son de “desear”. No obstante, la conciencia del significado nostálgico pervive hasta el final de la Antigüedad, particularmente en la memoria de gramáticos y comentaristas. Beda sitúa el significado “echar en falta” de desīderare como antecedente de la acción de concupiscere, cuyo significado no es otro que “desear”: (17) [homo] desiderando autem absentia concupiscit. (BEDA gramm. VII 270, 26–27) ‘Pero [el hombre] echando en falta lo ausente lo desea.’ El gramático viene a decir que la acción de “echar en falta” lleva implícita la de “desear” y lo entiende así porque desiderare contiene las dos. Es difícil establecer una división clara entre sentimientos de signo inverso, pues estos a menudo se superponen. El poeta Catulo expresó esa situación de confusión con la certeza que le sugería su propia vivencia: odi et amo. quare id faciam, fortasse requiris. / nescio, sed fieri sentio et excrucior (CATVLL. 85): ‘Odio y amo. Quizá preguntas por qué lo hago. No lo sé, pero siento que sucede y me atormento’. No obstante, nosotros clasificamos “echar en falta” y “desear” como términos ‘alternativos’, lo mismo que “odio” y “amo”, en la medida que, al prevalecer un significado, cede el otro, como si estuvieran en los platillos de una balanza. Se trata, evidentemente, de una clasificación significativa que es válida también para alternancias fisiológicas más o menos rigurosas: “aspirar” | “espirar”, “sorber” | “soplar”, “dormir | “despertar”. La modalidad alterna consta de un término ‘positivo’ y otro ‘negativo’. Determinar su orientación en un sentido u otro podrá depender de la perspectiva con que se asume la realidad. Dado el origen propuesto de desiderare, está claro que su significado negativo “echar en falta” antecede al positivo “desear”; pero aquel llega tras una situación positiva que se ha perdido (habere | carere), lo mismo que el odi de Catulo supone un amo previo. Cicerón lo expresa mediante una alternancia temporal en el texto (29): habuit, non habet [“caret”],

Origen y significado de dēsīderāre

169

desiderat. ¿Por qué es previo el significado negativo en desiderare? Sin duda, por el valor ‘ablativo, descendente’ que aporta el preverbio de- desde la base derivativa desidere (“irse al fondo”), valor que mantiene al crearse desid-erare con el hipotético significado “deprimirse”. La modalidad alterna que nos ha servido para dar cuenta del orden ‘negativo’ | ‘positivo’ de los dos significados históricos de desiderare (“echar en falta” | “desear”) es una relación intrasubjetiva, en cuanto que en el plano paradigmático se plantea respecto de un sujeto, como “dormir” | “despertar”. A ella se añaden las clases aspectuales, también intrasubjetivas, a diferencia de la diátesis, que es relación intersubjetiva. Aunque las llamamos relaciones intrasubjetivas e intersubjetivas en atención a la unidad y diversidad de la función de sujeto, no tienen carácter sintagmático, sino paradigmático. En la categoría nominal los términos mater .- filia coexisten en relación intersubjetiva, con independencia de cómo se presenten sintácticamente. Si alguien dice que es a la vez madre e hija, la identidad del sujeto sintáctico no afecta al plano paradigmático. En este se mantiene la relación intersubjetiva, con la particularidad de que entonces el proceso es doble y de expresión elíptica: ella será madre (de su hija) e hija (de su madre). En cambio, sponsa → uxor (“prometida” → “esposa”) constituye una relación intrasubjetiva de aspecto gradual. Al contrario del aspecto, la diátesis, sea gramatical o léxica, es siempre relación intersubjetiva; incluso si es reflexiva, gracias al desdoblamiento del sujeto en objeto. Se verá mejor en el siguiente paradigma verbal, en el que se incluyen términos en relación intrasubjetiva de aspecto gradual entre paréntesis. Léase primero sin estos, para comprobar la relación diatética: pater pecuniam filio (promittit →) dat .pecunia filio (datur → data est) est .pecuniam filius (accipit →) habet. La integración de los tres verbos (dat .- est .- habet) en el mismo sistema diatético permite entender su correspondencia en la expresión de la misma idea, según la norma de cada lengua: al. ‘es gibt’ dat, ingl. ‘there is’ est, esp. ‘hay’ y fr. ‘il y a’ habet.6 Cualquier unidad léxica se halla en el centro de una red de relaciones intrasubjetivas respecto de las que se sitúan en la misma perspectiva (promittere → dare) y de relaciones intersubjetivas respecto de las que se sitúan en perspectiva

6 Para mayor detalle sobre relaciones intrasubjetivas (modales y aspectuales) e intersubjetivas (diatéticas), véase García-Hernández (1991b: 131–147, 2014: 1–15).

170

Benjamín García-Hernández

diferente (dare .- accipere). Unas y otras ayudan a definir y configurar la delimitación de su significado. En el paradigma anterior habere se sitúa, como verbo de aspecto resultativo, al final de un proceso. Pero el fin de un proceso puede ser el principio de otro de signo contrario. Ahí surge ya una relación alterna, como la de pecuniam habere | perdere, filium habere | amittere. Estos dos verbos expresan aspecto ‘no resultativo’ respecto del ‘resultativo’ carere, que es el término alterno pleno de habere: perdere → carere, amittere → carere, habere | carere “non habere”. En esa situación de carencia, entendemos que se crea desiderare como ‘intensivo’ de sentimiento (“*deprimirse por”) sobre la base de desidere (“irse al fondo”). El nuevo verbo nace con una orientación retrospectiva (“*estar deprimido por”, “echar en falta”), de manera que puede ocupar la posición ‘resultativa’ de carere: filium amittere → desiderare: “perder un hijo” → “echarlo en falta”. A partir de ahí, se produce el cambio de orientación prospectiva, que da lugar a que el mismo verbo contenga dos significados de clase ‘alterna’: “echar en falta” | “desear”, como si fuera la alternativa de un movimiento respiratorio: “espirar” | “aspirar” (cf. esp. aspirar a “desear”). El significado prospectivo representa el inicio de un nuevo proceso que, como se verá, puede terminar otra vez en las acciones ‘resultativas’ de habere, uidere, inuenire, etc. (34–38).

2.1 La nostalgia, como resultado de una pérdida o efecto de una ausencia Ante todo, desiderare expresa un fuerte sentimiento de nostalgia por una pérdida (amissum), por una ausencia (absentem, suum abitum) o por lo que no se tiene (quod deest): (18) [pater] qui desideraret amissum [filium]. (SEN. contr. 1, 7, 13) ‘[El padre] que echaba en falta al [hijo] perdido.’ (19) parui nostra refert, qui desideramus amissos. (FRONTO p. 236, 24–25) ‘Poco importan nuestros asuntos, si echamos en falta a los seres que hemos perdido.’ (20) quanto diutius abest [filius] . . . magis desidero. (TER. Haut. 424–425) ‘Cuanto más dura la ausencia [del hijo] . . . más lo echo en falta.’ (21) . . . ut ea quae deerant non desiderarentur. (CIC. Brut. 234) ‘. . . de manera que no se echaría en falta lo que le faltaba.’

Origen y significado de dēsīderāre

171

Amittere se sitúa como grado aspectual ‘no resultativo’ de desiderare: pater (filium) amittit → desiderat. Ello quiere decir que desiderare con el significado de “sentir nostalgia, echar en falta” es el término ‘resultativo’ del proceso que lleva de la pérdida del ser querido a la pena que se siente por él. Esa relación secuencial da lugar a que la acción de amittere sea perfecta, cuando la de desiderare es presente: desideramus amissos (FRONTO p. 236, 24–25). La secuencia aspectual de “perder” → “añorar” (pater filium amittit → desiderat) es igualmente intrasubjetiva en pasiva: filius qui amissus est, desideratur; filius amissus et desideratus. En cambio, las acciones intransitivas de abire, abesse y deesse, que también anteceden a la de desiderare, mantienen una relación intersubjetiva con ella, de manera que esta no es su consecuencia aspectual, sino diatética: filius abit, abest, deest .- pater filium desiderat. Por ello, es suficiente poner en pasiva el último verbo, para que los dos términos de la relación intersubjetiva tengan, sintácticamente, el mismo sujeto: filius qui abit, abest, deest, desideratur. Así se explica que desiderare en pasiva sea sinónimo de tales antecedentes (TLL p. 704, 39: in passiuo fere i.q. deesse). La pérdida de un ser querido o apreciado y su ausencia rompen o interrumpen una relación familiar y lazos amorosos, amicales, patrióticos, etc. Estas son todas relaciones intersubjetivas: parentes .- liberi, uir .- uxor, inter fratres uel sorores, inter amantes, inter amicos, magister .- discipulus, patria .- ciuis, etc.: (22) quam tu filium tuom tam pater me meu’ desiderat. (PLAVT. Capt. 316) ‘Tanto como tú a tu hijo, me echa en falta mi padre a mí.’ (23) cum enim te desidero, fratrem solum desidero? (CIC. ad Q. fr. 1, 3, 3) ‘Pues cuando te echo en falta, ¿solo al hermano echo en falta?’ (24) eram cum amicis diu desideratis. (PLIN. epist. 5, 14, 8) ‘Estaba con amigos largamente deseados.’ (25) igitur uale, magister . . . et me, quantum amas, tantum desidera. (AVR. Fronto p. 31, 18–19) ‘Así pues, adiós, maestro . . . y échame en falta tanto cuanto me estimas.’ (26) hos ciues patria desideret? (LIV. 22, 60, 18) ‘¿A estos ciudadanos echaría de menos la patria?’ Por analogía la acción de desiderare puede aplicarse más allá a cualquier concepto que, como sujeto, mantiene una relación complementaria con otro: accusatio .- crimen (CIC. Cael. 6); ager .- aratorem (27); ager .- cultorem (CIC. Verr. II

172

Benjamín García-Hernández

3, 47); ara .- cruorem (CATVLL. 68, 79); castra .- imperatorem (CIC. Catil. 1, 10); palmes .- ulmum (28): (27) centum et unum aratores unus ager istius iniuria desiderat. (CIC. Verr. II 3, 120) ‘Un solo campo echa en falta a ciento un labradores por los agravios de ese.’ (28) stratus humi palmes uiduas desiderat ulmos. (IVV. 8, 78) ‘Tendido en el suelo, el sarmiento echa de menos los solitarios olmos.’ La secuencia aspectual pater (filium) amittit → desiderat y la diatética filius abest .- pater desiderat no deben hacernos olvidar la alternancia de partida habere | carere, mencionada al final del capítulo anterior, pues la acción nostálgica de desiderare es consecuencia de un haber tenido y no tener (29). En efecto, se echa en falta lo que está fuera del alcance de la mano o de la vista; lo contrario supondrá un despiste que dará lugar a una situación cómica, como la descrita en el texto plautino (30). Claro que el dominio y represión del sentimiento de carencia pueden causar la anulación estoica de la acción de desiderare, cuya negación (31) equipara los significados retrospectivo y prospectivo, según veíamos en Plauto (14): (29) habuit, non habet, desiderat, requirit, indiget. (CIC. Tusc. 1, 87) ‘Tuvo, no tiene, echa en falta, busca, siente necesidad.’ (30) fieri istuc solet, quod in manu teneas et oculis uideas, id desideres. (PLAVT. Trin. 913–914) ‘Suele suceder eso, que echas de menos lo que tienes en la mano o ves con tus ojos.’ (31) non caret is, qui non desiderat; ergo hoc non desiderare dico esse iucundius. (CIC. Cato 47) ‘No carece el que no desea. Por consiguiente, afirmo que este no desear es más agradable.’ Así pues, expresando el sentimiento de nostalgia, la acción de desiderare (“echar en falta, añorar”) se sitúa como término ‘resultativo’ de una carencia o pérdida (caret, amittit → desiderat) y como efecto ‘diatético’ de una ausencia (abest .- desiderat). Mientras estos términos clasemáticos delimitan el entorno significativo del verbo, el sufijo ‘intensivo’ -erare forma parte de su expresión.

Origen y significado de dēsīderāre

173

A reforzar el aspecto ‘tensivo’ de la acción nostálgica acuden los verbos del campo próximo de “flere” (32) y los de clase desiderativa para insistir en la idea de “desear” (cap. 2.2): (32) Hispaniae ipsos lugebant desiderabantque duces. (LIV. 25, 36, 16) ‘Las Españas lloraban y lamentaban la pérdida de los generales mismos.’

2.2 El deseo en el comienzo de nuevos procesos La orientación retrospectiva de la nostalgia se torna prospectiva, cuando el verbo adquiere el valor desiderativo y su acción se proyecta hacia adelante. De ahí que desiderare presente dos caras, como un Jano bifronte: la del aspecto ‘resultativo’ (“añorar”) que marca el final de un proceso de pérdida o ausencia; y la del ‘desiderativo’ que, como grado inicial (“desear”), abre nuevos procesos. Es este segundo valor el que señala el destino definitivo del verbo. En el texto siguiente aparece con otros desiderativos; entre expectare “esperar a ver”, que es ‘desiderativo’ por su significado, como sinónimo de uisere “ir a ver” (García-Hernández 1976: 91–94), y arcessere “hacer venir”, también con sufijo desiderativo: (33) qua re te expectamus, te desideramus, te iam etiam arcessimus . . . (CIC. Att. 1, 18, 1) ‘Por lo cual esperamos a verte, deseamos tu presencia, incluso ya te mandamos llamar.’ Mientras en la orientación retrospectiva desiderare expresa el contenido final en relación intrasubjetiva (pater filium amitit → desiderat) o intersubjetiva (filius abest .- pater filium desiderat), ahora con valor ‘desiderativo’ representa el grado inicial respecto de la acción ‘progresiva’ de quaerere (9, 38) y las ‘resultativas’ de inuenire (9, 34, 38), habere (35), uidere (10), audire (36), scire (37). La relación aspectual, que es paradigmática, no varía si estos verbos aparecen en forma finita o infinita: (34) imitetur ars igitur naturam, et quod ea desiderat inueniat. (RHET. Her. 3, 22, 36) ‘Imite, pues, el arte la naturaleza y encuentre lo que desea.’ (35) qui cum multa habeat, plura desiderat. (MIN. FEL. 36, 4) ‘Quien, aun teniendo mucho, desea más.’

174

Benjamín García-Hernández

(36) nunc quid est, quod quisquam ex uobis audire desideret . . . ? (CIC. S. Rosc. 104) ‘¿Qué hay ahora que alguien desee oíros . . . ?’ (37) concedo esse deos; doce me igitur unde sint, ubi sint . . .; haec enim scire desidero. (CIC. nat. deor. 1, 65) ‘Te concedo que existen los dioses; enséñame, pues, su origen, dónde están . . .; esto, en efecto, deseo saber.’ Cualquier proceso consta de un grado ‘no resultativo’ y otro ‘resultativo’, como si fueran un ‘infectum’ y ‘perfectum’ léxicos; así, hemos observado que la visión retrospectiva de la nostalgia se articula en “perder” → “echar en falta”. Ahora bien, a menudo el ‘no resultativo’ se divide en ‘ingresivo’ y ‘progresivo’, de manera que estos unidos al ‘resultativo’ componen un proceso trimembre: ‘ingresivo’ → ‘progresivo’ → ‘resultativo’. El aspecto ‘desiderativo’ es un subgrado del ‘ingresivo’ y, por tanto, desiderare lo representa como primer término en la triple articulación de la búsqueda de la verdad, descrita por un autor cristiano del s. V: (38) . . . eo flagrantius ametur ueritas quo uel diutius desideratur uel laboriosius quaeritur uel tardius inuenitur. (CLAVD. MAM. anim. p. 101, 7) ‘. . . tanto más ardientemente sea amada la verdad cuanto más tiempo se desea, más laboriosamente se busca e incluso más tarde se encuentra.’ desiderare tiene, pues, un doble aspecto gradual: es ‘resultativo’ en la orientación retrospectiva (“echar en falta”) e ‘ingresivo desiderativo’ en la prospectiva (“desear”). Además, contiene el aspecto ‘intensivo’, propio del sufijo -erare, que es común a las dos orientaciones. Esta superposición de dos clases aspectuales, la gradual y la tensiva, no deja de ser frecuente; así, spectare (“mirar atentamente”) es ‘no resultativo’ respecto del ‘resultativo’ uidere (“ver”) e ‘intensivo’ respecto de aspicere “mirar” (García-Hernández 2014: 3–4, 10–12). Pues bien, el aspecto ‘intensivo’ de desiderare puede explicitarse mediante los adverbios ualde (39), uehementer (40) o verse reforzado por sinónimos también intensivos, cuales exspectare (39, 41) y cogitare (41): (39) ualde te expecto, ualde te desidero. (CIC. Att. 2, 25, 2) ‘Con ansiedad te espero, con ansiedad deseo tu presencia.’ (40) amoque et laudo et uehementer desidero. (TER. Hec. 488) ‘La amo, la alabo y la deseo vehementemente.’

Origen y significado de dēsīderāre

175

(41) dies noctesque me ames, me desideres, me somnies, me exspectes, de me cogites. (TER. Eun. 193–194) ‘Que día y noche me ames, desees mi presencia, sueñes conmigo, me esperes, pienses en mí.’ La tensión de la añoranza y la intensidad del deseo tienen, pues, carácter predominante sobre el grado ‘resultativo’ que la acción de “añorar” representa respecto de una pérdida y sobre el grado ‘desiderativo’ que abre la perspectiva de un nuevo proceso.

3 Conclusión sobre el origen y evolución de desīderare Consīderare y desīderare no son formaciones parasintéticas sobre el sustantivo sīdus, según se ha creído tradicionalmente; ni siquiera son modificados preverbiales. Antes bien, son derivados de consīdere y desīdere con el sufijo intensivo -erare, como reciperare respecto de recipere o tolerare respecto de tollere. El movimiento físico indicado por desidere (“irse, asentarse al fondo”) es inverso al de tollere (“levantar”); pero la modificación intensiva que aporta el sufijo -erare en los derivados desiderare (“añorar, sentir nostalgia”) y tolerare (“soportar, tolerar”) es análoga. Partiendo, pues, del contenido ‘no intensivo’ de desidere, desiderare se carga de sentimiento: “irse al fondo” > “*deprimirse por una pérdida” > “echar en falta”. Luego, en su desarrollo semasiológico se produce un desplazamiento metonímico de la noción de nostalgia a la de deseo, como si fueran las dos caras, una negativa y la otra positiva, de la misma realidad. El punto de inflexión de la alternancia modal (‘negativa’ | ‘positiva’) reside en el cambio de la orientación retrospectiva a la prospectiva: “sentir una carencia” | “tener un deseo, un anhelo”. Sin salir de la esfera de los sentimientos, desiderare contiene en sí la semilla de una evolución que lo hace girar del final negativo de un proceso al comienzo positivo de otro. En ambos mantiene el aspecto intensivo que le aportó el sufijo -erare; pero en cada uno de ellos establece relaciones intersubjetivas o intrasubjetivas con verbos diferentes. En el proceso retrospectivo ocupa la posición final en la relación diatética abesse .- desiderare (“estar ausente” .“echar en falta” al que está ausente) y en la secuencia aspectual amittere → desiderare (“perder” → “echar en falta”). En cambio, en el proceso prospectivo ocupa el primer grado en las secuencias aspectuales desiderare → quaerere, inuenire, habere, uidere, audire, scire (“desear” → “buscar”, “encontrar”, “tener”, “ver”, “oír”, “saber”). En el primer proceso encuentra sinónimos en el

176

Benjamín García-Hernández

campo semántico de “flere”, en particular lugere; y en el segundo lo acompañan exspectare “esperar a ver”, uisere “ir a ver”, arcessere “hacer venir”, cupere “desear”, concupiscere. Unos y otros configuran su entorno significativo y ayudan a definir los significados de nuestro verbo. A diferencia de los paradigmas gramaticales, los del nivel léxico son abiertos. El léxico no deja de participar del sistema de la lengua y por ello es una fuente constante de gramaticalización. Desiderare es un buen ejemplo, pues coronó su evolución prospectiva en latín como auxiliar del grado ‘desiderativo’, al que dio nombre desde los gramáticos latinos: desideratiua uerba (PRISC. gramm. II 535, 9–11). En esa función vino a reemplazar los improductivos sufijos -(es)sere y -urire, particularmente en el latín vulgar. Pocos testimonios habrá tan ilustrativos como el de Petronio (101, 10), cuando escribe uisere desiderabit con el valor de “uidere desiderabit”, en vez de uiset. No es extraño, pues, que haya tenido una buena presencia en romance, en forma patrimonial o culta: fr. désirer (cf. ingl. to desire), it. desiderare, etc. (Meyer-Lübke 1972: § 2593).

Bibliografía OLD = Glare, P. G. W. (Ed.) (1968–1982). Oxford Latin Dictionary. Oxford: Oxford University Press. TLL = Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. Bader, F. (1962). La formation des composés nominaux du latin. Paris: Les Belles Lettres. De Meo, C. (1983). Lingue tecniche del latino. Bolonia: Pàtron. Ernout, A. & Meillet, A. (2001 [1932, 1985]). Dictionnaire étymologique de la langue latine. Paris: Klincksieck. Fasce, S. (1988). Nostalgia e rimpianto nel lessico psicológico latino. Sandalion, 10–11, 67–81. García–Hernández, B. (1976). El campo semántico de “ver” en la lengua latina: Estudio estructural. Salamanca: Universidad de Salamanca. García-Hernández, B. (1980). Semántica estructural y lexemática del verbo. Reus y Barcelona: Ediciones Avesta. García-Hernández, B. (1991a). Considero. Propuestas etimológicas y contenido semántico. Cuadernos de Filología Clásica. Estudios Latinos, 1, 87–98. García-Hernández, B. (1991b). The lexical system of intersubjective and intrasubjective relationships. En R. Coleman (Ed.), New studies in Latin linguistics: Selected papers from the 4th colloquium on Latin linguistics (pp. 129–149). Amsterdam: John Benjamins. García-Hernández, B. (2000). Considerare como lexema visual y término técnico mercantil: El valor heurístico de un texto de Suetonio. Voces, 10–11, 91–100. García-Hernández, B. (2002). El significado fundamental del prefijo latino de-. En L. Sawicki & D. Shalev (Eds.), Donum grammaticum: Studies in Latin and Celtic linguistics in honour of Hannah Rosén (pp. 141–150). Lovaina / París: Éditions Peeters.

Origen y significado de dēsīderāre

177

García-Hernández, B. (2014). Le système classématique des relations intersubjectives et intrasubjectives. En Dictionnaire historique et encyclopédie linguistique du latin. Université de Paris-Sorbonne, Centre Alfred Ernout. http://www.linglat.paris-sorbonne. fr/encyclopedie_linguistique:notions_linguistiques:semantique:systeme_classematique (acceso: 22 junio 2014). Guiraud, Ch. (1964). Les verbes signifiant “voir” en latin: Étude d’aspect. Paris: Klincksieck. Le Boeuffle, A. (1977). Les noms latins d’astres et de constellations. Paris: Les Belles Lettres. Le Pennec-Henry, M. (1987). Considerare, obserare, segregare, insinuare: hypostase ou parasynthèse? En Études de linguistique générale et de linguistique latine offertes en hommage à Guy Serbat (pp. 121–128). París: Société pour l’Information Grammaticale. Lodge, G. (1962). Lexicon Plautinum. Hildesheim: Olms. Meyer-Lübke, W. (1972). Romanisches etymologisches Wörterbuch. Heidelberg: Universitätsverlag Winter. Nyman, M. (1990). Hits and Misses: Lat. considerare and desiderare. Zeitschrift für Vergleichende Sprachforschung, 103, 51–68. Pârvulescu, A. (1980). Latin considerare et desiderare. Zeitschrift für vergleichende Sprachforschung, 94, 159–165. Prellwitz, W. (1903). Zur lateinischen Wortbildung. Considero, desidero und andere PräfixDenominativa. En Γέρας. Abhandlungen zur indogermanischen Sprachgeschichte: August Fick zum siebenzigsten Geburtstage gewidmet von Freunden und Schülern (pp. 63–76). Göttingen: Vandenhoeck und Ruprecht. Rix, H. (1985). Sūdor und sīdus. En Sprachwissenschaftliche Forschungen. Festschrift für J. Knobloch (pp. 339–350). Innsbruck: Universität Innsbruck. Thurneysen, R. (1906). Senium und desiderium. Archiv für lateinische Lexikographie, 14, 179–184. Vaan, M. de (2008). Etymological dictionary of Latin and the other Italic languages. Leiden: Brill. Walde, A. & Hofmann, J. B. (1982). Lateinisches etymologisches Wörterbuch. Heidelberg: Universitätsverlag Winter.

Luca Rigobianco

Latin dubenus ‘dominus’: an attempt at etymology Abstract: The aim of this chapter is to propose an etymology of the Latin word dubenus, attested only in Paulus’ epitome of Festus as an archaic word meaning ‘dominus’. After showing the genuineness of the tradition, I intend to account for such a gloss from a linguistic point of view. Abstractly, dubenus can be analysed as a noun or a substantive adjective derived with the suffix -ēnofrom a base *dubo-, nomen actionis in -bo- < *-bho- or in -mo- < *-mo- (with subsequent dissimilation of -m- to -b-) from the base du- < *deH3w- ‘to give’. As an alternative, dubenus can be a writing error for duhenus, i.e. duenos ‘bonus’ (< *deH3w-eno-) with h as a marker of hiatus and -us for -os as a normalized ending. In both cases, dubenus/duhenus would mean ‘dominus’ in relation to the notion of giving (*dubeno- ‘the one of the giving’, *dueno- ‘provided with gifts’), in accordance with a semantic evolution which can be suitably framed from a historical perspective. Specifically, it is possible to hypothesize that, in early Roman society, the dominus was the one who gave (and received) gifts in order to maintain both social relationships existing with his peers and hierarchies within society. More generally, dubenus/duhenus ‘dominus’ could be considered as a linguistic relic of an Indo-European ideological system in which, as shown by Benveniste, social status was determined by the exchange of gifts through a practice similar to the potlatch. Keywords: Latin linguistics, etymology, Paulus’ epitome, dubenus, duenos

1 Introduction The aim of this chapter is to propose an etymology of the Latin word dubenus. Dubenus is attested only in Paulus’ epitome of Festus as an archaic word meaning ‘dominus’ : Dubenus apud antiquos dicebatur, qui nunc dominus (PAVL. FEST. p. 67, 7). This gloss has been mostly considered as a corrupt gloss on philological and, secondarily, linguistic grounds. In what follows, I intend to show the genuineness of the tradition which preserve ‘dominus’ as the meaning of dubenus (see Section 2) and, consequently, to account for such a gloss from a linguistic point of view (see Section 3). Luca Rigobianco, Ca’ Foscari University of Venice https://doi.org/10.1515/9783110647587-012

Latin dubenus ‘dominus’: an attempt at etymology

179

2 The glossographic tradition Paulus’ gloss Dubenus apud antiquos dicebatur, qui nunc dominus is usually connected with two glosses reported in Pseudo-Philoxenus’ Latin-Greek glossary, which contains many glosses derived from Festus (Dammann 1894: 29–33; Laistner 1926: 130–131). The two glosses are Dubinus διϲϲοϲ (GLOSS. II 55, 62), i.e. ‘twofold, double’, also in the sense of ‘doubtful, ambiguous’ (GEL, s.v. δισσός : 1078), and Dubius δεϲποτηϲ (GLOSS. II 56, 6), i.e. ‘master, lord’ (GEL, s. v. δεσπότης : 961). The translation of dubenus as ‘dominus’ in Paulus has been ascribed by Scaliger (Scaliger 1576: 57–58) and Goetz (Goetz 1926: 46) to a misunderstanding, which would also be recognizable in the odd PseudoPhiloxenus’ gloss Dubius δεϲποτηϲ instead of an expected Dubius διϲϲοϲ. Therefore, Scaliger and Goetz have proposed to correct Paulus’ gloss, respectively, as follows: dubienus apud antiquos dicebatur, qui nunc dubius; dubenus apud antiquos dicebatur, qui nunc dubinus. In detail, Scaliger has restored an original form dubenus (based on the forms dubenus in Paulus and dubinus in Pseudo-Philoxenus), whose meaning would be ‘dubius’, whereas Goetz has considered dubenus and dubinus as two alternative forms meaning ‘dubius’. In particular, according to Goetz, the reason for such a misunderstanding (dominus instead of dubinus ‘dubius’) would have been the phonetic proximity of [b] to [m] (Goetz 1926: 46). However, there is no philological evidence to support the hypothesis of a corruption of Paulus’ gloss and the consequent correction of ‘dominus’ into ‘dubius’ or ‘dubinus’. The reading ‘dominus’ is found in all the manuscripts of Paulus’ epitome of Festus (except Monacensis 14734, where, according to Müller, the entry is omitted [Müller 1839: 67]), and, in addition, PseudoPhiloxenus’ gloss Dubius δεϲποτηϲ gives an indirect proof of the reliability of Paulus’ testimony. Indeed, the connection between the two glosses (Dubenus . . . dominus and Dubius δεϲποτηϲ) makes it clear that their source cannot but be Festus or Verrius himself (cf. Figure 1). Dubius instead of Dubenus in Pseudo-Philoxenus’ glossary could be one of the many mistakes which are found in this glossary (Goetz 1923: 31).1 In this regard, Lindsay, integrating the glosses transmitted by Pseudo-Philoxenus and by Paulus, has proposed to reconstruct the original text as . Dubenus apud antiquos dicebatur qui nunc dominus (Lindsay 1930: 177). In particular, the gloss Dubanus dubius has been restored starting from PAVL. FEST. p. 21, 14 Addubanum

1 See, for example, Dauus αφρων απειροϲ (GLOSS. II 37,12) and Daunum αφρονα (GLOSS. II 37, 42) instead of Dalivum . . . insanum (PAVL. FEST. p. 68, 1; in this regard, see Scaliger 1576: LV).

180

Luca Rigobianco

[Verrius (end of the 1st c. BC/beginning of the 1st c. AD)]

Festus (2nd c. AD) *Dubenus apud antiquos dicebatur, qui nunc dominus

Pseudo-Philoxenus (6th c. AD?) Dubius δεϲπоτηϲ

Paulus (end of the 8th c. AD) Dubenus apud antiquos dicebatur, qui nunc dominus

Figure 1: A possible reconstruction of the glossographic tradition.2

dubium, corrected into Ad dubanum dubium. In Paulus’ epitome the entry Dubanus would have been omitted, as is normal for epitomes,3 whereas in Pseudo-Philoxenus’ glossary the two glosses would have been confused and reordered as Dubinus διϲϲοϲ, instead of Dubanus διϲϲοϲ, and, after other entries, Dubius δεϲποτηϲ, instead of Dubenus δεϲποτηϲ.4 In any case, as stated above, there is philological evidence of a tradition, probably starting at least from Verrius or Festus, which preserve ‘dominus’ as the meaning of dubenus. On the basis of this evidence, it is appropriate to try to account for such a gloss from a linguistic point of view rather than – or at least before – correcting it.

3 An attempt at etymology 3.1 Introduction The explanation of dubenus ‘dominus’ within Latin and in comparison with other Indo-European varieties is somewhat difficult. In this regard, the 2 The dating of Pseudo-Philoxenus’ Glossary is questionable: see Rudorff (1866: 220–230), Loewe (1876: 182), Goetz (1910: 1439–1441), Laistner (1926: 125–136), and Lindsay (1930: 77–80). For the manuscript tradition of Pseudo-Philoxenus’ Glossary, see Goetz and Gundermann (1888: vii–xx). 3 ego prolixitate superflua quaeque et minus necessaria praetergrediens et quaedam abstrusa penitus stilo proprio enucleans, nonnulla ita, ut erant posita, relinquens, . . . conpendium optuli (PAVL. FEST. p. 1, 8–10). 4 According to Lindsay, in the version of Pseudo-Philoxenus’ glossary which has come down to us, the glosses Erant reordinatae . . . Erant divisae et decurtatae . . . Erant neglegentissime transcriptae. Denique plurimae erant aut consilio aut incuria praetermissae (Lindsay 1930: 80).

Latin dubenus ‘dominus’: an attempt at etymology

181

hypothesis of considering dubenus as a phonetic variant of dominus (Bréal 1899; Brugmann 1907: 636) is usually rejected because of the ad hoc phonetic processes assumed (LEW, s.v. dubenus: 375). At the same time, the comparison with what is known of varieties neighbouring Latin does not seem to support the hypothesis of dubenus being a loanword.5 A Latin word dubenus can be plausibly analysed as a noun or a substantive adjective derived with an n-suffix.6 More specifically, in the light of the knowledge of Latin derivational morphology, the most reasonable segmentation hypothesis is *dub-ēno-. The suffix -ēno- expresses relationship in nomina gentilicia (for example, Alfēnus < *alfo-) and in Picēnus/Picēnum ‘Picene’/‘Picenum’ < *piko- ‘woodpecker’.7 The isolation of a suffix -ēno- allows us to identify a base *dub- or, as an alternative, a thematic base *dubo-. Taking this into account, a possible starting point for an attempt at etymology could be the identification of the initial sequence du- with the Latin base du- < *deH3w-8 ‘to give’ (cf. duim ‘I may give’). Specifically, a base *dubo-, although without counterparts in Latin and in other Indo-European varieties, could be theoretically analysed as a nomen actionis in -bo- < *-bho- or in -mo- < *-mo- (with subsequent dissimilation of -m- to -b-) from the base du- ‘to give’ (for formal details, see Sections 3.3.1 and 3.3.2). Following this hypothesis, albeit speculative (at least apparently), dubenus would approximately mean ‘dominus’ as ‘the one of (-ēno-) the giving (*dubo-)’, in accordance with a semantic evolution which can be suitably framed from a historical perspective (see Section 3.2). However, in the light of the difficulties of accounting for such a base (*dubo-) from a linguistic point of view (see Sections 3.3.1 and 3.3.2), an alternative

5 Dubenus could be comparable to South Picenian qdufeniúí, if the analysis of qd(u)- as an outcome of *kĕd(u)- < ke- (preverb) + -d(u)- ‘to give’ is accepted (cf. Marinetti 1984: 41–42, 56–64 and Marinetti 1985: 109–112 for such an analysis of the verbal form kduíú in the South Picenian inscription CH 1). However, Rix (1994) has proposed to analyse South Picenian qd-/ kd- as an outcome of *kl- and, thus, qdufeniú- would correspond to Latin Clufennius. 6 On n-suffixes, see, in general, Leumann (1977 [1926–1928]: 320–329) and Weiss (2009: 287–290). 7 On Latin -ēno- (borrowed from Italic?) and its counterparts in other Indo-European languages, see Rix (1951: 242–245) and Rix (1972: 727–728). 8 I adopt the laryngeal theory as developed by Prosdocimi (since Prosdocimi 1985; on Latin dō-/da-/du- from a single root *deH3w-, see, specifically, Prosdocimi 1996: 285–288), despite the common assumption of *deH3w-, whence Latin du-, as a u-extension of a separate root *deH3-, whence Latin dō-/da- (LIV, s.v. *deh3-: 105–106; LIV, s.v.?*deh3u̯-: 107). In any case, a Latin base du- ‘to give’ is self-evident.

182

Luca Rigobianco

hypothesis can be taken into consideration. Specifically, dubenus could be a writing error for duhenus, i.e. duenos ‘bonus’ with h as a marker of hiatus and -us for -os as a normalized ending (see Section 3.3.3). If this is the case, the use of duenos ‘bonus’ for meaning ‘dominus’ could find an adequate explanation from a historical point of view assuming that, as already mentioned for the supposed base *dubo-, duenos is derived from the base du- < *deH3w- ‘to give’ – in this case, by means of the participial suffix -eno- distinct from -ēno- seen above9 – and, thus, is originally related to the ‘gift’ (approximately, ‘provided with gifts’; see Section 3.2).

3.2 Meaning According to the analysis of dubenus < *dub-ēno- outlined above (see Section 3.1), the meaning ‘dominus’ would have evolved from an original meaning related to ‘the giving’ (approximately, ‘the one of the giving’). In this regard, it is possible to hypothesize that, in early Roman society, dubenus ‘the one of the giving’ could correspond to ‘dominus’ (namely, one who has ius possessionis and iubendi potestas within a specific sphere [TLL, s.v. dominus : vol. V, 1, p. 1911, 48–49]) inasmuch as, at that time, the dominus was the one who gave gifts in order to maintain both social relationships existing with his peers and hierarchies within society. If the hypothesis is correct, Latin dubenus, as well as daps, damnum, etc., could be considered as a linguistic relic of an IndoEuropean ideological system in which, as shown by Benveniste (Benveniste 1948–1949 and Benveniste 1969: 65–79), social status was determined by the exchange of gifts through a practice similar to the potlatch (Malinowski 1922; Mauss 1923–1924). Specifically, in a society which works this way “[d]onner, c’est manifester sa supériorité, être plus, plus haut” (Mauss 1923–1924: 174). Such an ideological system has left many traces in the Indo-European languages and cultures. In particular, Watkins has demonstrated the presence of an exchange or reciprocity relation between the patron and the poet in the Celtic, Indic, Germanic, and Greek traditions (Watkins 1976; Watkins 1995: 68–84). Furthermore, dubenus would have a precise counterpart in Gothic Gibica, name of the Burgundian royal ancestor as attested in the Leges Burgundionum. Gibica is in all likelihood a transparent name meaning ‘dator, largitor’ (< Germ. *geba- ‘to give’ [Grimm 1841: 573]) and, thus, referring to the king as ‘the one who gives’.

9 For the different approaches to the etymology of Latin duenos, see n. 10.

Latin dubenus ‘dominus’: an attempt at etymology

183

If, alternatively, dubenus is a writing error for duhenus = duenos ‘bonus’ (see above, Section 3.1, and Section 3.3.3), the meaning ‘dominus’ could have arisen from an original meaning ‘provided with gifts’. The reconstruction of such an original meaning is based on an etymological analysis of duenos as a derivation from the base *deH3w- ‘to give’ with the participle suffix -eno-.10 The semantic evolution ‘provided with gifts’ > ‘dominus’ can be historically framed according to what has been said above: in a society in which potlatch is an effective practice, the dominus is at the same time both ‘the one who gives gifts’ and ‘the one who receives gifts’, that is to say a duenos (‘provided with gifts’; cf. CIC. Att. 8, 1, 3 bonorum, id est lautorum et locupletum).11 More generally, the semantic evolution could be theoretically comparable with that of optimus, which also has a social sense,12 if it is from *opi-timo‘richest’ < *opi- ‘wealth’,13 although the systemic conditions involved are probably different. In any case, the replacement of dubenus/duhenus, whatever its etymology, with dominus implies a social change, that is to say a change in the way of conceiving the authority and, consequently, of expressing it.

10 LEW, s.v. bonus: 111. For a different opinion, see DELL, s.v. bonus: 73 (“on est amené à rapprocher got. taujan «ποιεῖν, πράσσειν», tewa « ordre »”, gr. δύναμαι, et sans doute véd. dúvaḥ (gén. dúvasaḥ) « hommage », duvasyáti « il rend hommage») and EDL, s.v. bonus: 73–74 (“If OLat. due- reflects *duHe-, duenos could belong to Go. taujan, Gr. δύναμαι from PIE *deuh2- ‘to join, fit together’, or to Lat. duim from PIE *d(e)h3-u- ‘to give’ . . . The absence of vowel colouring is another problem: *duh2-eno- is expected to yield *duanos, whereas *duh3-eno- should yield *duonos (Schrijver 1991: 109f.)”). In this regard, I want to point out that duenos instead of *duonos is the expected outcome according Prosdocimi’s laryngeal theory (*dØH3w- > *du-), which makes the assumption of a metathesis (*deH3u > *deuH3; LIV, s.v. ?*deh3u̯-: 107) – with *H3 subsequently responsible for colouring adjacent e – superfluous (Prosdocimi 1996: 285–288). Neri’s hypothesis of a connection between duenos and the presumed Indo-European root *deH1w- ‘zusammenfügen, ordnen’ (Neri 2011: 144 n. 288; see also Kümmel 2015, s.v. Neu: *deh1u̯-) has been questioned by Neri himself (personal communication) after that the review of the etymology of Gothic tewa by Kroonen (2013, s.v. *tēwō-: 515) would have made the reconstruction of an Indo-European root *deH1w- unnecessary. 11 In this regard, cf. also Medium Irish den ‘firm, strong, powerful’, which is in a possible etymological connection with Latin duenos (see, most recently, Zair 2012: 137–138). 12 See TLL, s.v. bonus (in particular, vol. II, p. 2082, 30–78: “5. de viro bono . . . a. in vita publica”). 13 LEW, s.v. optimus: 216–217; DELL, s.v. *ops: 463–464. For a different opinion, see EDL, s.v. ob: 421.

184

Luca Rigobianco

3.3 Form 3.3.1 *dubo- < *d(e)H3w-bhó- (nomen actionis)? *dubo- could be analysed as a nomen actionis derived from the base *deH3w- ‘to give’ with the suffix *-bho- (*d(e)H3w-bhó- > *dubo- ‘the giving’). According to the traditional view, the Indo-European suffix *-bho- derives colour adjectives, animal names, and abstract nouns (including nomina actionis [Brugmann 1906 (1888): 386–390]).14 Nevertheless, in Latin the suffix -bo- < *-bho- is very rare, found only in very few colour adjectives (for example, albus ‘white’) and animal names (for example, columba ‘pigeon, dove’). The only possible instance of the use of -bo- in the derivation of abstract nouns in Latin could be *sabo- < *s(w)Ø-bhó- ‘selfhood’ from the base *s(w)e- ‘self’,15 whence Sabus ‘Sabus’ (name of a god, hypostasis of the ‘selfhood’ [Coarelli 1997: 204–207]), Sabīni ‘Sabines’ (literally, ‘those of the selfhood’ > ‘those who are themselves [in opposition to others]’), etc. (Rix 1957; De Simone 1992; Bader 1994; Prosdocimi 2011). However, *sabo- could also be analysed as a lexicalization by means of *-ó- (and not *-bhó-) from an enlarged base *s (w)ebh- ‘self’ (Bader 1982: 133–134; De Simone 1992: 232), which is found, for example, in the Greek reflexive pronouns σφῶν, σφίσι(ν), σφᾶς/ σφέα. Furthermore, the use of *-bho- in the derivation of abstract nouns has been questioned also for other Indo-European languages, in particular for Greek (Chantraine 1933: 262–264) and Balto-Slavic (Ambrazas 1993), casting doubt on its very existence. Consequently, the reconstruction of a base *dubo- < *d(e)H3w-bhó- ‘the giving’ is questionable and alternative hypotheses must be considered.

3.3.2 *dubo- < *d(e)H3w-mó- (nomen actionis)? *dubo- could be analysed as a nomen actionis or a verbal adjective derived from the base *deH3w- ‘to give’ with the suffix *-mo-16 (*d(e)H3w-mó- > *dumo- ‘the giving’, ‘given’; cf., for example, fūmus ‘smoke’ < *dhu(e)H2- ‘to breathe’),17

14 For a recent discussion on the different hypotheses about the origin of the suffix *-bho-, see Pinault (2016). 15 On să- < *s(w)Ø-, see Prosdocimi (2011: 260–262). 16 On the Indo-European suffix *-mo-, see Brugmann (1906: 245–251) and Hamp (1982–1983). According to Hamp, originally there was a distinction between nomina actionis in *-mo- with o-grade (beside feminization in *-mā with zero-grade) and verbal adjective in *-mo- with zerograde. 17 On -mo- in Latin, see Leumann (1977 [1926–1928]: 319) and Weiss (2009: 286).

Latin dubenus ‘dominus’: an attempt at etymology

185

with subsequent dissimilation of -m- to -b- in dubenus because of the -n- of the suffix (*dumo- → *dum-ēno- > *dubēno-). Such a formation (*deH3w-mó-) has no counterparts in other Indo-European languages18; however, it cannot be excluded that -mo- was a productive suffix in earlier Latin (Weiss 2009: 286). In any case, it should be noted that the assumption of a dissimilation of -m- to -b(*dum-ēno- > *dubēno-) is not firmly grounded. The only possible instance of a dissimilation of -m- to -b- in the middle of a word would be hībernus ‘of or belonging to winter (hiems)’, if the hypothesis of a derivation from *hiemernos (Leumann 1977 [1926–1928]: 231, 322; Sihler 1995: 211),19 instead of from *heimrinos (LEW, s.v. hiems : 645–646; EDL, s.v. hiems : 284–285; Weiss 2009: 164, 291),20 is tenable.21 A similar phenomenon, namely *m- > f- (probably via [β]) because of a following nasal, seems to be sporadically attested in Latin at the beginning of the word (Meillet 1918; Weiss 2009: 156): cf. forma < *morma (?) (LEW, s.v. forma : 530–531; DELL, s.v. fōrma : 247; EDL, s.v. fōrma : 233–234), formīca < *mormīka (LEW, s.v. formīca : 531–532; DELL, s.v. formīca : 247–248; EDL, s.v. formīca : 234), and formīdō < *mormīdō (LEW, s.v. formīdō : 532; DELL, s. v. formīdō : 248; EDL, s.v. formīdō : 234–235). In particular, the similarity of the two phenomena (-m- > -b-; m- > f-) lies in the correspondence with the Latin general outcomes of the Indo-European voiced aspirates respectively in the middle and at the beginning of the word (Leumann 1977 [1926–1928]: 163–175; Meiser 1998: 101–105; Weiss 2009: 72–79). Anyway, the participants in the dissimilation process are different (m–m vs. m–n) and the phonotactic context is different as well (word initial vs. word internal). Therefore, the reconstruction of a base *dubo- < *d(e)H3wmó- ‘the giving’ is also questionable. Additionally, Latin has dōnum ‘gift’ < *deH3wno-, which makes the existence of a formation from the same base with almost the same meaning (*dubo- < *d(e)H3w-bhó- /*d(e)H3w-mó- ‘the giving’) unlikely, although not impossible.

18 IEW, s.v. dō-: 223–226 (“Nominalbildungen: dō-no-m, dō-ro-m, dō-ti-s, də-ti-s ‘Gabe’”). 19 See also DELL, s.v. hiems: 293–294, where hībernus is derived from *heimrinos but -b- is considered as an outcome of dissimilation. 20 In this case, -b- would be due to development of the cluster -mr- and not to dissimilation: cf., for example, brevis < *mreĝhu̯i- (Leumann 1977 [1926–1928]: 190; Meiser 1998: 112; Weiss 2009: 163). 21 If this hypothesis is correct, dominus instead of *dobinus would have to be ascribed to the influence of domus.

186

Luca Rigobianco

3.3.3 Dubenus: a writing error for duhenus? In the light of the philological evidence of a tradition which preserve ‘dominus’ as the meaning of dubenus (see Section 2) and the difficulties of accounting for such a gloss from a linguistic point of view (see Sections 3.3.1 and 3.3.2), an alternative hypothesis can be taken into consideration. Specifically, dubenus could be a writing error for duhenus, i.e. duenos ‘bonus’ with h as a marker of hiatus (see below) and -us for -os as a normalized ending.22 The main reason for such an error could have been the recurrence of the cluster dub- in the previous entry (or entries, according to the reconstruction of Lindsay seen above): Dubat dubitat.

Duhenus apud antiquos dicebatur, qui nunc dominus. whence, by a mechanical error: Dubat dubitat.

Dubenus apud antiquos dicebatur, qui nunc dominus. (PAVL. FEST. p. 67, 6–7) Furthermore, dubenus instead of duhenus could have been facilitated by both the obscurity of such a word, by then no longer in use (apud antiquos dicebatur), and, possibly (but not necessarily), palaeographical reasons. In particular, the replacement of h (duhenus) with b (dubenus) could be easily explained, for example, by the use of half-uncial script, where the two letters are similar in shape (see, for example, the codex Casinensis 150).23 Anyway, I put any palaeographical consideration aside, since it is not possible to date the error (dub- instead of duh-) precisely. Indeed, it could have already occurred in Verrius and/or Festus. In any case, the agreement of the two sources – Pseudo-Philoxenus’ Glossary (Dubius δεϲποτηϲ) and Paulus’ epitome (Dubenus . . . dominus) – shows that the error (dub- instead of duh-) was in their common manuscript source/sources, which must precede Pseudo-Philoxenus’ Glossary, and that is the terminus ante quem.

22 A similar hypothesis has been advanced by Birt, according to whom dubenus would be an orthographic variant for *duvvenus/*dvenus, i.e. duenos (Birt 1897: 69–71). 23 For a general introduction to half-uncial script, in use from the fifth century AD, see Steffens (1910: vii–viii) and Bischoff (1986 [1979]: 99–108).

Latin dubenus ‘dominus’: an attempt at etymology

187

As an aside, I add that a replacement of h with b also occurs in the β family of Paulus’ manuscripts, where we read beluus instead of helvus (Helvacea genus ornamenti Lydii, dictum a colore boum, qui est inter rufum et album, appellaturque helvus, PAVL. FEST. p. 99, 12). The use of h for marking the hiatus (duhenus) is sporadically attested in Latin texts (Leumann 1977 [1926–1928]: 174). Nevertheless, the occurrence of such an orthographic convention in the senatus consultum de Bacchanalibus (dating from 186 BC; CIL I2, 581), where we read ahenam ‘made of bronze’ instead of aenam,24 can throw light on the origin of duhenus. In the middle of the third century BC, dwV- changed into bV-: cf., for example, Duellōna > Bellōna, *duene > bene, duidens > bidens, etc. (Meiser 1998: 111; Weiss 2009: 161). The later use of forms with dwV- instead of bV- as archaisms (cf., for example, duelonai instead of *belonai in the senatus consultum de Bacchanalibus) implies that (some) Latin speakers were aware of such a correspondence (duV-: bV-). In the light of this, the use of h in duhenus could have been introduced during the same period or, anyway, in a period in which there was awareness of such a correspondence (and duenos survived only in the glossographic tradition as an archaism meaning ‘dominus’ and no more related to duonos and bonus), in order to indicate a (non-etymological) reading [dŭĕ-] with hiatus instead of [dwĕ-] and the consequent lack of a corresponding form in bV- (**benus). Otherwise said, it is possible to reconstruct the following scenario. In the middle of the third century BC, the glossographic tradition preserved an archaic word duenos ‘dominus’. Such a word was no longer in use and Latin speakers who had access to the glossographic tradition were probably not aware of the etymological relationship with duonus ‘bonus’, because of the antiquity of the change -ue- > -uo- and different meanings. During the same period dwVchanged into bV- (for example, duonus > bonus). Obviously, such a phonetic change did not affect duenos ‘dominus’, because it existed only as a written word. Besides, after the change of dwV- into bV-, early forms as Duellōna, duellum, etc. survived for a long time as archaisms alongside Bellōna, bellum, etc.25 In contrast, duenos lacked a counterpart *benos. Consequently, a glossographer put an h between u and e in duenos for indicating a hiatus and, thus, a reading [dŭĕnos] instead of [dwĕnos]. Such a reading ([dŭĕnos]), although etymologically

24 What matters here is the possible use of this orthographic convention in Latin, regardless of its origin (see, for example, EDL, s.v. aes: 27–28 on the hypothesis of an Umbrian origin; in this regard, see, conversely, Prosdocimi (2015: 144–148) on early Romanization of Umbrian tradition). 25 A trisyllabic [dŭĕ]llum, used by later poets, is clearly secondary with respect to a disyllabic [dwĕ]llum, attested in Plautus and implied by bellum (see, for example, Clausen 1971).

188

Luca Rigobianco

unfounded – but this was, of course, unknown to the glossographer – was required in order to account for the lack of a counterpart *benos, contrary to couples as Duellōna: Bellōna. If the hypothesis is correct, Verrius and/or Festus (whence PseudoPhiloxenus and Paulus; see Section 2) preserve a form (duenos) which cannot be dated later than the end of the sixth century BC or the beginning of the fifth century BC, because of the retention of the diphthong -ue- (Meiser 1998: 82; Mancini 2004: 233; Weiss 2009: 139; Nussbaum: 2017). The preservation of such an ancient form can be explained in two ways. Specifically, one can assume that duenos in the meaning of ‘dominus’ did not change into duonos because of cultural conservatism or, more probably, that duenos ‘dominus’ was retained in its original form through a long antiquarian tradition – possibly, only in writing.26 In addition, a meaning ‘dominus’ for duenos can be suitable for its occurrences in the ‘duenos inscription’ (CIL I2 4). Specifically, as shown, among others, by Colonna (Colonna 1979), duenos med feced . . . duenoi can be interpreted as a dedicatory inscription ‘a dueno- had me made for a dueno-’, that is to say ‘a dueno- gave me to a dueno-’ – instead of an artisan’s signature (‘Dueno-/a dueno- made me for Dueno-/a dueno-’)27 – in which dueno- can have a social sense. In particular, on the basis of linguistic and cultural considerations, Colonna claims that in the seventh and sixth centuries BC duenos was the name by which the members of the aristocratic elite called themselves (Colonna 1979: 170). If this interpretation is correct,28 the ‘duenos inscription’ would be a further trace of an Indo-European ideological system in which social relationships among elite peers was maintained through the exchange of gifts (‘a master had me made for a master’).

4 Conclusion As seen above (see Section 2), there is no philological reason to question the reliability of the gloss dubenus ‘dominus’, transmitted by Paulus and, indirectly, by

26 In this regard, see, for example, Burroni-Brezigia (2017), who have convincingly shown the preservation in Paulus of an -osio genitive (amosio, amended into annosio), to be dated on phonetic grounds before the second half of the fourth century BC or even to the fifth century BC. See, however, Mancini (2016), on the accessibility to the oldest sources for the grammarians of the republican and imperial age. 27 For ‘to make something’ meaning ‘to have something made’, see also Prosdocimi (1981). 28 Many other interpretations of the ‘duenos inscription’ have been proposed. The literature on this topic is vast and cannot be cited here.

Latin dubenus ‘dominus’: an attempt at etymology

189

Pseudo-Philoxenus. However, such a form is difficult to justify from a linguistic point of view. In particular, an etymology from an alleged base *dubo- < *d(e) H3w-bhó- or *d(e)H3w-mó- ‘the giving’, though semantically plausible (*dub-ēno‘the one of (*-ēno-) the giving (*dubo-)’ > ‘dominus’; see Section 3.2), is formally problematic (see Sections 3.3.1 and 3.3.2). Therefore, it is preferable to consider dubenus as a writing error for duhenus, that is to say duenos, with dub- instead of duh- due to the recurrence of the cluster dub- in the previous entry. If it is the case, h would be a hiatus-marker which would have been put for indicating a(n etymologically unfounded) reading [dŭĕnos] instead of [dwĕnos] and, as a consequence, the lack of a counterpart *benos, in contrast to couples as Duellōna [dwĕ-]: Bellōna [bĕ-], etc. (see Section 3.3.3). Furthermore, a meaning ‘dominus’ for duenos is semantically plausible in the light of an etymology from the base *deH3w- ‘to give’ with the participle suffix -eno- (*deH3w-eno- ‘provided with gifts’). In particular, such a meaning (‘provided with gifts’ > ‘dominus’) fits well with an independently reconstructed Indo-European ideological system in which, as stated above, social status was determined by the exchange of gifts.

Bibliography CIL = Corpus Inscriptionum Latinarum (1862–). Berlin/Boston: De Gruyter. DELL = Ernout, A. & Meillet, A. (1959 [1932]). Dictionnaire étymologique de la langue latine. Paris: Klincksieck. EDL = Vaan, M. de. (2008). Etymological Dictionary of Latin and the other Italic Languages. Leiden/Boston: Brill. GEL = Liddell, H. G. & Scott, R. (1996 [1843]). A Greek-English Lexicon. Oxford: Clarendon. IEW = Pokorny, J. (2005 [1959]). Indogermanisches Etymologisches Wörterbuch. Tübingen/ Basel: Francke. LEW = Walde, A. & Hofmann, J. B. (1938–1954). Lateinisches etymologisches Wörterbuch. Heidelberg: Universitätsverlag Winter. LIV = Rix, H. (2001 [1998]). Lexikon der indogermanischen Verben. Wiesbaden: Reichert. TLL = Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. Ambrazas, S. (1993). Daiktavardžių darybos raida. Vilnius: Mokslo ir enciklopedijų leidykla. Bader, F. (1982). Autour du réfléchi anatolien: étymologies pronominales. Bulletin de la Société de Linguistique de Paris, 77, 83–155. Bader, F. (1994). Les noms des Aryens : ethniques et expansion. In F. Bader, Langues IndoEuropéennes (pp. 65–83). Paris: Centre national de la recherche scientifique. Benveniste, E. (1948–1949). Don et échange dans le vocabulaire indo-européen. L’Année sociologique, 3, 7–20. Benveniste, E. (1969). Le vocabulaire des institutions indo-européennes. Paris: Les Éditions de Minuit.

190

Luca Rigobianco

Birt, T. (1897). Beiträge zur lateinischen Grammatik. Sprach man avrum oder aurum? Frankfurt am Main: Suerländer. Bischoff, B. (1986 [1979]). Paläographie des römischen Altertums und des abendländischen Mittelalters. Berlin: Schmidt. Bréal, M. (1899). Variétés. Dubenus. Mémoires de la Société de Linguistique de Paris, 6, 261. Brugmann, K. (1906 [1888]). Grundriß der vergleichenden Grammatik der indogermanischen Sprachen (II, 1). Straßburg: Trübner. Brugmann, K. (1907). βάναυσος, μάναυεται, βανόν. Rheinisches Museum für Philologie, 62, 634–636. Burroni, F. & Brezigia, M. (2017). Lat. amosio: A previously unnoticed -osio genitive in Latin. Pallas, 103, 77–86. Chantraine, P. (1933). La formation des noms en grec ancien. Paris: Champion. Clausen, W. (1971). Dvellvm. Harvard Studies in Classical Philology, 75, 69–72. Coarelli, F. (1997). Il campo Marzio. I. Dalle origini alla fine della repubblica. Roma: Quasar. Colonna, G. (1979). Duenos. Studi Etruschi, 47, 163–172. Dammann, A. (1894). De Festo Pseudophiloxeni auctore. Commentationes philologae Ienenses, 5, 1–48. De Simone, C. (1992). Sudpiceno safīno-/lat. sabīno-: Il nome dei Sabīnī. AIΩN. Sezione linguistica, 14, 223–239. Goetz, G. (1910). Glossographie. In Paulys Real-Encyclopädie der classischen Altertumswissenschaft, 13 (coll. 1433–1467). Stuttgart: Metzler. Goetz, G. (1923). Corpus glossariorum latinorum: I, De glossariorum latinorum origine et fatis. Leipzig: Teubner. Goetz, G. (1926). Aus lateinischen Glossaren. In F. Kluge & O. Behaghel (Eds.), Festschrift Friedrich Kluge zum 70. Geburtstage am 21. Juni 1926 (pp. 39–51). Tübingen: Englisches Seminar. Goetz, G. & Gundermann, G. (Eds.) (1888). Glossae latinograecae et graecolatinae. Leipzig: Teubner. Grimm, J. (1841). Gibichenstein. Zeitschrift für deutsches Altertum, 1, 572–575. Hamp, E. P. (1982–1983). Indo-European substantives in *-mó- and *-mā́-. Zeitschrift für vergleichende Sprachforschung, 96, 171–177. Kroonen, G. (2013). Etymological Dictionary of Proto-Germanic. Leiden/Boston: Brill. Kümmel, M. (2015). Addenda und Corrigenda zu LIV2. www.martinkuemmel.de/liv2add.html (accessed: 13 October 2017). Laistner, M. (Ed.) (1926). Philoxeni glossarium. In W.-M. Lindsay, R.-G. Austin, M. Laistner & J.F. Mountford (Eds.), Glossaria latina, (II, pp. 123–291). Paris: Les Belles Lettres. Leumann, M. (1977 [1926–1928]). Lateinische Laut- und Formenlehre. München: Beck. Lindsay, W. M. (Ed.) (1930). Festus. In J. W. Pirie & W.-M. Lindsay (Eds.), Glossaria latina, (IV, pp. 71–467). Paris: Les Belles Lettres. Loewe, G. (1876). Prodromus corporis glossariorum latinorum. Leipzig: Teubner. Malinowski, B. (1922). Argonauts of the Western Pacific. London: Routledge & Kegan. Mancini, M. (2004). Latina Antiquissima II: Ancora sull’epigrafe del Garigliano. In V. Orioles (Ed.), Studi in memoria di Eugenio Coseriu (pp. 229–251). Udine: Forum. Mancini, M. (2016). I grammatici, lo standard e il latino arcaico. In M. Benedetti, C. Bruno, P. Dardano & L. Tronci (Eds.), Grammatiche e grammatici. Teorie, testi e contesti (pp. 85–140). Roma: Il Calamo.

Latin dubenus ‘dominus’: an attempt at etymology

191

Marinetti, A. (1984). Il verbo italico. Apporti dalle iscrizioni sudpicene. Linguistica Epigrafia Filologia Italica. Quaderni di lavoro, 2, 27–73. Marinetti, A. (1985). Le iscrizioni sudpicene. I. Testi. Firenze: Olschki. Mauss, M. (1923–1924). Essai sur le don. Forme et raison de l’échange dans les sociétés archaïques. L’Année sociologique, 1, 30–186. Meillet, A. (1918). À propos de latin formīca. Mémoires de la société de linguistique de Paris, 20, 115. Meiser, G. (1998). Historische Laut- und Formenlehre der lateinischen Sprache. Darmstadt: Wissenschaftliche Buchgesellschaft. Müller, K. O. (Ed.) (1839). Sexti Pompei Festi De verborum significatione quae supersunt cum Pauli epitome. Leipzig: Weidmann. Neri, S. (2011). Wetter. Etymologie und Lautgesetz. Jena: Friedrich-Schiller-Universität Jena. Nussbaum, A. J. (2017). The latin ‘bonus rule’ and benignus ‘generous, kind’. In B. S. S. Hansen et al. (Eds.), Usque ad radices: Indo-European Studies in honour of Birgit Anette Olsen (pp. 575–592). Copenhagen: Museum Tusculanum Press. Pinault, G.-J. (2016). Suffixes from roots: the case of PIE *-bho-. fachtagung-ig2016.univie.ac. at/uploads/media/Pinault.pdf (accessed: 13 October 2017). Prosdocimi, A. L. (1981). ‘fare’ e ‘fare per’. A proposito dell’iscrizione REI IX, 1. Studi Etruschi, 49, 357–362. Prosdocimi, A. L. (1985). Umbro furfa- ~ lat. forfex: -eH2 > -a vs. -eH2s > -eks. Archivio Glottologico Italiano, 70, 51–61. Prosdocimi, A. L. (1996). Appunti sul verbo latino (e) itálico, VII. Studi Etruschi, 61, 263–312. Prosdocimi, A. L. (2011). Ipseità e alterità tra etnonimia e poleonimia. Suessa, Suessula, Opikoi, Oinotroi e simili. In O. Paoletti & M. C. Bettini (Eds.), Gli Etruschi e la Campania settentrionale (pp. 251–281). Pisa/Roma: Fabrizio Serra. Prosdocimi, A. L. (2015). Le Tavole Iguvine: II, Preliminari all’interpretazione. La testualità: fatti e metodi. Firenze: Olschki. Rix, H. (1951). Picentes – Picenum. Beiträge zur Namenforschung, 2, 237–247. Rix, H. (1957). Sabini, Sabelli, Samnium. Ein Beitrag zur Lautgeschichte der Sprachen Altitaliens. Beiträge zur Namenforschung, 8, 127–143. Rix, H. (1972). Zum Ursprung des römisch-mittelitalischen Gentilnamensystems. In Aufstieg und Niedergang der römischen Welt (I, 2, pp. 700–758). Berlin/New York: De Gruyter. Rix, H. (1994). Südpikenisch kduíú. Historische Sprachforschung, 107, 105–122. Rudorff, A. F. (1866). Über die Glossare des Philoxenus und Cyrillus. Berlin: Königl. Akademie der Wissenschaften. Scaliger, J. (Ed.) (1576). Sex. Pompei Festi libros de verborum significatione Castigationes, Recognitae et auctae. Paris: Mamertum Patissonium. Sihler, A. L. (1995). New Comparative Grammar of Greek and Latin. New York/Oxford: Oxford University. Steffens, F. (1910). Paléographie latine. Trèves/Paris: Schaar & Dathe-Champion. Watkins, C. (1976). The etymology of Irish dúan. Celtica, 11, 270–277. Watkins, C. (1995). How to kill a Dragon: Aspects of Indo-European Poetics. New York/Oxford: Oxford University. Weiss, M. (2009). Outline of the Historical and Comparative Grammar of Latin. Ann Arbor/ New York: Beech Stave. Zair, N. (2012). The Reflexes of the Proto-Indo-European Laryngeals in Celtic. Leiden/Boston: Brill.

Kanehiro Nishimura

Fors and fortūna: linguistic and cultural aspects Abstract: The Latin words fors and fortūna are often used as if they are synonymous. However, some Roman authors make significant distinction in their use of these terms. This chapter attempts to explain their semantic differences by referring to the morphological structure of each form: fors and fortūna are formed with *-ti- and *-tu-, two different suffixes for creating abstract nouns. The examination of other sets in Latin of ti- and tu-stems from the same verbal roots leads us to differentiate these two derivations semantically. In particular, I argue that tu-stems refer to actions that are telic/bounded. Thus, while the underlying meaning of the ti-stem fors is ‘random chance (with the result undetermined or unpredicted)’, fortūna may have teleological implications (similar to English destiny). The collocation fors fortūna (often used as a combinatory theonym) is also considered. Keywords: Latin linguistics, word formation, *-ti-, *-tu-, lexical semantics

1 Fors and fortūna: synonymous? The words fors and fortūna are commonly used in Roman literature, and the semantic similarity or overlap (‘chance’, ‘luck’, etc.) between these forms has been widely acknowledged. Ernout and Meillet (1985: 249) suggest that the use of fors particularly in the nominative or ablative singular (see also Walde and Hofmann I: 534) resulted from the possibility that its other case forms (e.g. gen. sg. fortis) would have caused formal confusion with fortis ‘strong’ (cf. gen. sg. fortis); fortūna may thus have been utilized in order to fill the gap thus engendered. If so, one might expect the meaning of fors and that of fortūna to be similar to or coincide with each other. Acknowledgment: I owe a great debt of gratitude to Hiroaki Natsukawa, Tadashi Ogawa, Lucie Pultrová, Tamon Suzuki, Anthony Yates, and the two anonymous reviewers for their useful comments and suggestions. Special thanks go to Masato Konishi for his valuable advice on semantics. All remaining errors are of course my own. (The production of this chapter was supported by JSPS KAKENHI Grant Number JP16K02669.) Kanehiro Nishimura, Kobe City University of Foreign Studies https://doi.org/10.1515/9783110647587-013

Fors and fortūna: linguistic and cultural aspects

193

In fact, in the Loeb translation of Catullus by Cornish and Goold,1 for example, fors and fortūna are both simply translated as ‘fortune’ or ‘Fortune’2: saeva fors ‘fortune . . . full of spite’ (64, 169–170); fortuna mea ‘my fortune’ (64, 218); fortunae signa secundae ‘the tokens of prosperous fortune’ (64, 222); fors ‘Fortune’ (64, 366); fortuna casuque . . . acerbo ‘by fortune and bitter chance’ (68, 1); fortuna mihi tete abstulit ipsum ‘fortune has taken your own self away from me’ (101, 5). The use of fortūna in CIC. fat. 5–6 also draws our attention. In contrast to fātum ‘fate’, fortūna means something like ‘accident, chance’ – a translation commonly used for fors, too3: ne hercule Icadii quidem praedonis video fatum ullum; nihil enim scribit ei praedictum: quid mirum igitur ex spelunca saxum in crura eius incidisse? puto enim, etiamsi Icadius tum in spelunca non fuisset, saxum tamen illud casurum fuisse. Nam aut nihil omnino est fortuitum, aut hoc ipsum potuit evenire fortuna. (CIC. fat. 5–6) ‘Even in the case of the brigand Icadius I swear I can’t see any trace of destiny; for the story does not say that he had any warning, so that if a rock from the roof of a cave did fall on his legs, what is there surprising about it? for I suppose that even if Icadius had not been in the cave at the time, that rock would have fallen all the same, since either nothing at all is fortuitous or it was possible for this particular event to have happened by fortune.’ (trans. Rackham)

Further, in Paelignian, an Italic language spoken in central Italy (now extinct) and genetically related to Latin, there is a sequence forte faber (forte < *fortes, an equivalent of gen. sg. of Latin fors; cf. Zair 2017: 281), which has often been associated with fabrum esse suae quemque fortunae ‘every man is the fashioner of his own fortune’ (trans. Ramsey) of Appius Claudius Caecus (thought to be the first person to introduce this collocation) apud PS. SALL. rep. 1, 1, 2 (see Vetter 1953: 150 and Untermann 2000: 254). This interpretation suggests that Italic (*)fors and Latin fortūna are synonymous. We also have a collocation forte fortūnā ‘by chance’ and a combinatory theonym Fors Fortūna. The latter’s function as a goddess has long been a matter of debate, concisely summarized by Billington (1998). In the two forms being used together one might see a semantic congruity between fors and fortūna.

1 The translations of Latin passages cited in this work are based on the most recent editions of the Loeb Classical Library (the translators’ names will then be provided). 2 In this chapter, vowel length will be marked in linguistic discussion, but omitted in direct textual citations. 3 fortuītus (presumably based on fortu- like fortūna) in the same section also seems to mean ‘accidental’, although the logic of the sentence where the word appears is not necessarily clear. Cf. n. 4.

194

Kanehiro Nishimura

However, Manilius uses fors and fortūna in rather different contexts in his Astronomica (Book 1). It is thus worth considering whether his usage of fors and fortūna reflects the original one. With regard to fors in 1, 492–494, where the word is paraphrased as caeco . . . foedere, Manilius means that the universe is never controlled by such haphazardness: quis credat tantas operum sine numine moles ex minimis caecoque creatum foedere mundum? si fors ista dedit nobis, fors ipsa gubernet. (MANIL. 1, 492–494) ‘Who could believe that such massive structures [seas, lands, etc. – KN] have been created from tiny atoms without the operation of a divine will, and that the universe is the creature of a blind compact? If chance gave such a world to us, chance itself would govern it.’ (trans. Goold)

On the other hand, Manilius mentions the power of fortūna in 1, 508–512; his use of the word is meant to express something semantically close to ‘destiny’ or ‘fate’ (cf. Champeaux 1982: 430), as represented by fatis (abl. pl.) in the same passage: quot post excidium Troiae sunt eruta regna! quot capti populi! quotiens fortuna per orbem servitium imperiumque tulit varieque revertit! Troianos cineres in quantum oblita refovit imperium! fatis Asiae iam Graecia pressa est. (MANIL. 1, 508–512) ‘How many the realms that have collapsed since the sack of Troy! How many the peoples led into captivity! How oft has Fortune brought slavery and sovereignty in turn and returned in a different guise! To what high sovereignty has it rekindled the ashes of Troy, forgetful of the past! And Greece in its turn has been crushed by Asia’s destiny.’ (trans. Goold)

In the lines that follow, Manilius explains the regularity of the sun, moon, and stars as governed by not casus opus ‘the work of chance’ but magni numinis ordo ‘the order of a supreme deity’ (1, 531); the former can be regarded as almost equivalent to fors, but the latter to fortūna.4 Another passage from Manilius (1, 797–798) is also worth mentioning: et Cato fortunae victor, fictorque sub armis / miles Agrippa suae . . . ‘Here are Cato and Agrippa, who proved in arms the one the master, the other the maker of his destiny’ (trans. Goold). Although the interpretation of the passage is not immediately clear, fortūna here cannot be sheer ‘chance’. Even earlier, the following

4 Cf. MANIL. 1, 182: . . . fortuitos ortus surgentibus astris ‘(I cannot believe that) when stars appear to view their risings are the product of chance . . . ’ (trans. Goold); see Section 3 and n. 19.

Fors and fortūna: linguistic and cultural aspects

195

passage is found in Accius: Fortunane an forte repertus. ‘Was he through Fortune found or chance?’ (trans. Warmington). The distinction between Fortūna and fors is clear here, and it is also noted by Nonius: ‘Fors’ et ‘Fortuna’ hoc distant: fors est casus temporalis, fortuna dea est ipsa.’ ‘Fors’ and ‘Fortuna’ differ in this: ‘fors’ is a chance event of the moment: ‘Fortuna’ is the goddess herself’ (trans. Warmington). We have thus faced the question of whether fors and fortūna indeed differ from each other. Judging from several cases just cited, the answer seems to be positive. But how and why are they different from each other? In this chapter I will argue that the semantic distinction is the result of a morphological difference, that is, fors fortis as a ti-stem vs. fortū-na as a tu-stem derivative.

2 Derivations in *-ti- and *-tuThe two stem-building suffixes are generally thought to have existed in ProtoIndo-European (PIE) as *-ti- and *-tu-. Both of these have the function of deriving abstract nouns primarily from verbal roots,5 and the resulting derivatives are often used as infinitives in the daughter languages (see Fortson 2010: 107, 125). In the case of fors and fortūna the underlying verbal root is most likely to be PIE *bher- ‘carry’ (> Latin ferō, Greek φέρω; see Walde and Hofmann I: 534, Champeaux 1982: 430–432, and De Vaan 2008: 236). Derivatives from this root with the suffix *-ti- show a variety of meanings among Indo-European languages: e.g. Vedic bhr̥tí- ‘bearing, present’, Avestan paiti.bərəti- ‘offering’, Armenian bard ‘heap, bundle’, and Gothic ga-baurþs ‘birth’. This fact perhaps indicates relatively later derivation (see Lehmann 1986: 134) or reinterpretation of the derivative in each language/branch. The meaning ‘chance, luck’ of Latin fors seems to be yet more different from those of the above forms. For this reason Ernout and Meillet (1985: 249) dissociate Latin fors from the forms in other Indo-European languages (see also De Martino 2015: 248).6 However, the meaning ‘carry’ of the root and that of fors ‘chance, luck’ are not incompatible with each other from the typological standpoint: Chinese 運 (simplified as 运 in modern times; /yùn/ in Mandarin) means not only ‘carry, move’ as a verb but also ‘fortune, luck’ as a noun (borrowed into Japanese) – an almost exact

5 The denominal type like senectūs, ūtis ‘old age’ will be discussed in depth elsewhere. 6 De Martino’s connection (2015: 254–257) of Fortūna and the PIE root *u̯ert- ‘turn’, though semantically interesting, cannot be supported phonologically despite his phonetic analyses.

196

Kanehiro Nishimura

match with Latin. Therefore, the association of ‘carry’ and ‘chance, luck’ should not be excluded. Note also Old English ge-byrd (*bher- + *-ti-), which means ‘lot, fate’ (as well as ‘birth’) in some context, not terribly far from ‘chance, luck’ (see Bosworth and Toller 1898 s.v. and NIL 2008: 17).7 While it is certain that fors continues *bhr̥-ti- with the root in zero grade (see Untermann 2000: 304; Weiss 2011: 316; Lundquist 2015; and Zair 2017: 280–281), the ablaut in the root that underlies fortūna is not a simple matter. Leumann (1977: 353) states that abstract nouns in *-tu- showed full grade in the root in PIE, namely, *bhér-tu-, which is reflected in Vedic inf. bhártave, prábhartum. Greek φέρτρυς· ἆθλος, Θούριοι (Hesychius), if the second -ρ- is secondary (see NIL 2008: 24 n. 45), also appears to point to *bhér-tu-. We can find such ablaut in Latin likewise: e.g. versu- ‘circular movement’ < *u̯ért-tu(cf. vertō ‘turn’). According to Leumann, however, the root shape of tu-stems in Latin came to converge with that of past participles, the latter being based on the zero grade (e.g. *kwlh̥ 1-tó- > cultus ‘tilled’ → cultu- m. ‘tilling’); therefore, vorsu- exists as well (← vorsus ‘turned’ < *u̯r̥t-tó-). Alternatively, the root in zero grade might be original; cf. NIL’s (2008: 17) reconstruction *bhr̥-tu- for *for-tu- of fortūna. Given that fortūna may reflect a deinstrumental form (see Weiss 2011: 290), that is, something like *bhr̥-tu-h1-neh2- (Transponat) based on *bhr̥-tu-h1 or *bhr̥-teu̯-(e)h1 (?), the zero-grade root of a weak case is most likely in any case. The question would then arise as to the functional difference between *-tiand *-tu-. This issue has not always been treated explicitly in previous literature. With regard to Latin, Collin (1904: 458–459) suggests that although nouns in -ti- (often extended as -tiōn-; see Leumann 1977: 366; and also Weiss 2011: 312 with some reservations) and those in -tu- became synonymous in the course of time, the suffixes originally had ‘einen kleinen Bedeutungsunterschied’ – the former highlights an act (‘Handlung’) while the latter a result (‘Resultat’) of an act or a state (‘Zustand’) caused by it. Benveniste (1975: 96 n. 1) considers Collin’s definition ‘peu précise’ and suggests instead a semantic contrast between the two suffixes, *-ti- for ‘l’action objective’ vs. *-tu- ‘l’action subjective’ (1975: 112); for the latter he also uses the term ‘personelle’ (his emphasis) and ‘au point de vue du sujet’ (1975: 96).8 However, his definition of *-tu- is not incompatible with Collin’s. For Benveniste, the suffix denotes “l’action comme 7 The meaning ‘chance, luck’ could have originated from phrasal expressions like German Was der Tag bringt ‘what the day will bring’ and the like. 8 The starting point of Benveniste’s argument is sensory nouns and the like, that is, those designating personal reactions to various phenomena, such as gustus ‘tasting’ and spīritus ‘action of breathing’, the majority of which are derived with -tu-.

Fors and fortūna: linguistic and cultural aspects

197

subjective, émanant du sujet et l’accomplissant, en tant que predestination ou disposion interne . . . ” (1975: 112); this interpretation does not work without a (potential or expected) result of an act. Thus, as will become clear, Collin’s framework is not far-fetched at all, and Benveniste’s view overlaps with it. A survey of the Latin corpus turns up some other cases besides fors and fortūna in which a ti(ōn)-stem and a tu-stem derived from the same root are both used with (nearly) the same frequency. Despite their synonymous use in several cases, semantic distinction may be drawn between them: 1) occāsiō vs. occāsus. First, occāsus continues a prefixed form of *kadtu- (> cāssus [cf. QVINT. inst. 1, 7, 20] via Lachmann’s Law > cāsus ‘falling’). Among other things, it is used to mean ‘sinking (of the sun or stars)’ (cf. occidō ‘sink below the horizon’). This word, then, entails a goal-oriented movement with a more or less fixed route.9 The involvement of ‘goal’ is close to Collin’s notion, ‘Resultat’ of an act, but it is not exactly the same. More focus is laid on the orientation of an entity towards a target (cf. Benveniste’s ‘prédestination ou disposition interne’), which we can regard as a telic event (in the case of occāsus, probably an ‘accomplishment’ within the framework of Vendler 1957, i.e., [+telic] and [+duration]). On the other hand, occāsiō, based on *kad-tiōn- (but no †cāsiō), means ‘occasion, opportunity’, which is quite different from that of occāsus, as pointed out by Ernout and Meillet (1985: 82). They also mention the meaning that becomes evident in a later period, ‘cause, reason’, which is quite far from the notion of ‘goal’ (≈ ‘result’) assumed for occāsus. 2) āctiō vs. āctus. āctus ‘driving of cattle or carts’ keeps the basic meaning of the verb agō ‘drive’. As pointed out by Ernout and Meillet (1985: 16), āctus is also used as a land measure, necessarily a limited length or area, which is akin to Collin’s ‘Zustand’ of an act but with a nuance of reaching some limit (≈ endpoint). Although the base verb is essentially atelic, āctus as a measure may function as denoting that an event is bounded with an endpoint (cf. I walked today [atelic/unbounded] vs. I walked two miles today

9 The prefix ob- may entail a motion towards a goal (among other nuances; see Van Laer 2010: 79–81): ob-itus ‘approach’, for example, is far commoner than ob-itiō. Therefore, one may hypothesize that the prefix plays an important role in the selection of the tu-suffix for derivation. This conjecture is supported by the fact that, for example, ad-itus ‘access’, co-etus/co-itus ‘meeting’, con-ventus ‘action of coming together’, and sumptus ‘spending of money’ are usually (far) more frequent by token than their ti(ōn)-stem counterparts. However, the prefix having some sort of telic (or bounded) profile does not absolutely determine the selection of a suffix; otherwise, occāsiō could not exist.

198

3)

4)

5)

6)

7)

Kanehiro Nishimura

[bounded]). On the other hand, āctiō ‘action’ is differentiated from āctus in use (particularly in philosophy), as Ernout and Meillet state; their definition is ‘façon d’agir, action . . ., activité’, which is more focused on process. The ti-stem ars artis ‘art, craft’ also entails a similar nuance (‘façon d’être ou d’agir’ as defined by Ernout and Meillet 1985: 48) in contrast to the tustem artus (used in pl.) ‘joint in the body’, a point or endpoint where bones come together. cursus ‘action of running’, on the other hand, is akin to āctus. Besides the meaning as a simple action noun (like cursiō in Varro), it also means ‘path, course’, often with a predetermined goal and a fixed or controlled direction, as with occāsus discussed above. On the other hand, its corresponding ti-stem found in the accusative-adverb cursim ‘rapidly’ describes the manner of how things are proceeding. The same is true for the ti-stem passim ‘in a scattered manner’ (← pandō ‘spread out, extend’). The tu-stem counterpart passus ‘step, pace, stride’, however, represents a particular length, and is thus used as a measurement (see Ernout and Meillet 1985: 478). In this regard, it is similar to āctus and cursus.10 The ti-stem sēnsim ‘slowly’, likewise, is an adverb that describes a process of an act. On the other hand, its base verb sentiō may function as both a verb of perception (‘perceive [at a certain moment]’, an ‘achievement’ per Vendler 1957: 146–147) and a state verb (‘feel that . . . ’ or the like). For the former, telicity is implied, and the meaning of the latter is a result of the former’s lexical aspect or Aktionsart (see Vendler 1957: 154). This semantic feature may be the reason for the u-stem sēnsus ‘capacity to perceive by the senses’ (immanent to a person, even when s/he is asleep; see Vendler 1957: 152–153, 156).11 In fact, some of the sensory and intellectual verbs (almost) exclusively derive tu-stems: e.g. gustus ‘tasting’, intellectus ‘action of understanding’ (vs. rare intellectiō ‘synecdoche’), questus ‘regret’ (vs. rare questiō ‘action of complaining’), and tumultus ‘mental disturbance’. The ti-stem statim ‘immediately’ also depicts how things happen. However, it is difficult to find a semantic difference between another ti(ōn)-stem, that is, statiō ‘state of standing still’ and the tu-stem status ‘condition of being in a standing position’ in terms of their primary meanings. In this respect, the meaning of the verb statuō ‘place so as to remain upright’ provides

10 Also mēnsūra (< *-tū-) ‘action of measuring; length, area (etc.)’ vis-à-vis mēnsiō (< *-tiōn-) ‘action of measuring’ (rare), both based on mētior ‘measure’. 11 con-sēnsus ‘concord’ is also more frequent than con-sēnsiō ‘concord’ (see Ernout and Meillet 1985: 614).

Fors and fortūna: linguistic and cultural aspects

199

a hint of the semantic profile of its derivational base status: the action of the verb entails the completion of erecting an entity.12 This telic aspect of the verb (typically clear from the meaning; see Pustejovsky 1995: 16) seems to be associated with the suffix -tu-. 8) Of some relevance is the pair pars ‘portion, share’ / partim ‘in part’ vs. partus ‘action of giving birth’ (pariō ‘give birth’; Ernout and Meillet 1985: 483 *‘procurer’ > ‘procurer un enfant au mari’). Although the ti-stems are difficult to interpret in terms of aspect, the tu-stem clearly highlights the ‘climax’ (cf. Vendler 1957: 145–146) of a woman’s delivery. 9) The tu-stem adverb nātū (← abl. sg.) ‘by birth’ has much in common with partus. nātū is often used for denoting a person’s age or the age difference between persons (e.g. maior nātū ‘older’); it thus refers implicitly to the time of a person’s birth, that is, to the completion of her/his mother’s delivery (cf. Ved. jā́tu adv. ‘from birth, ever’). nātiō ‘people’ only rarely implies this nuance. The deified version of nātiō in CIC. nat. deor. 3, 47, though related to childbirth, seems to preside over the process of it (not only its final stage)13 as a protecting goddess: Quodsi tales dei sunt, ut rebus humanis intersint, Natio quoque dea putanda est, cui cum fana circumimus in agro Ardeati rem divinam facere solemus; quae quia partus matronarum tueatur, a nascentibus Natio nominata est. (CIC. nat. deor. 3, 47) ‘And if it is the nature of the gods to intervene in man’s affairs, the Birth-Spirit also must be deemed divine, to whom it is our custom to offer sacrifice when we make the round of the shrines in the Territory of Ardea: she is named Natio from the word from being born (nasci), because she is believed to watch over married women in travail.’ (trans. Rackham)

Here the theonym Nātiō is hermeneutically connected to nāscēns ‘being in the process of birth’, in which the lives of mothers and unborn babies are at stake. 10) How about birth’s antonym, death? The underlying verb ‘die’ typically represents a telic event (‘achievement’ as per Vendler 1957: 146 n. 6). Although its conspicuous endpoint would lead one to expect the derivation via *-tufor ‘death’, it is mors (mor-ti-) that is commonly used for that meaning. Note, however, that the word may refer to a wider time span than the punctual point of one’s expiring: e.g. PLAVT. Men. 411 tertium Liparo, qui in morte regnum Hieroni tradidit ‘thirdly Liparo, who passed the kingdom on to Hiero when he died’ (trans. De Melo), where Hiero seems to have

12 See Nishimura (2006) for Umbrian 3. sg. impv. II restatu ‘restore’ (via haplology) < *re-statutu < *re-statu-i̯e-tōd. 13 CIL I2 60 NATIONV CRATIA ‘nationis gratiā’ perhaps also has the same nuance.

200

Kanehiro Nishimura

declared (or reconfirmed) his will shortly before his death. So we can consider the ti-stem mors capable of showing the process of dying (as with a Biblical expression, Hebrews 11:21 ‘ . . . when he was dying . . . ’) or its abstract notion. It is also possible that there is a relic of a tu-stem derived from the same verbal root in the Latin vocabulary; mortuus ‘deprived of life, dead’ is a candidate. De Vaan (2008: 390) argues that “[t]he formation of mrtu̯o- ‘dead’ for PIE *mr̥to(as in Slavic) may be due to a contamination with *mr̥u̯o- as attested in Celtic”. Weiss (2011: 442) also posits a sort of contamination of mortus ‘dead’ and the antonym vīvus ‘alive’. However, for phonological and morphological reasons, Livingston (2004: 9–10) suggests that mortuus was derived from the tu-stem by means of -o- (i.e. *mortu-o-).14 Her idea is semantically not far-fetched at all based on our discussion so far. Indeed, tu-stems for ‘death’ are seen in derivatives of eō ‘go’ with some prefixes, such as ob-itus, inter-itus, and ex-itus, as well as cāsus, a euphemism for ‘death’ (cf. n. 15). The adjective mortuus clearly presupposes one’s last breath, a completion of a living life. It is certainly true that there are several cases where it is difficult to retrieve clear-cut semantic features of ti- and tu-stems at first glance because their meanings are too similar or too far from each other: e.g. cantiō ‘singing, song’ vs. cantus ‘id.’; pōtiō ‘act of drinking’ vs. pōtus ‘id.’; sitis ‘thirst’ vs. situs ‘neglect, disuse; physical deterioration’. Nevertheless, from the examples (1–10) discussed above, it follows that tu-stems tend to be accompanied with a nuance of orientation or direction of an act with a fixed goal/endpoint. Collin’s description (as well as Benveniste’s) is thus on the right track, and a similar type of explanation also appears in Pultrová (2011: 87–88). As far as we know from our data, it does not matter whether the lexical aspect of an underlying verb is telic or not: if the verb is telic, -tu- simply highlights telicity of an act; even if the verb is atelic, the suffix seems to make an event bounded in conspiracy with some contextually conceivable endpoint.15 On the other hand, ti-stems tend to lack this nuance.

14 *mor-teu̯-o- (with the tu-suffix in the full grade) > *mortou̯o- > *mortu(u)o- is phonologically also possible. 15 cāsus ‘falling down’, although the act of falling may continue forever theoretically, is usually bounded by a specific endpoint or (pragmatically) the ground. In fact, the word also functions as a euphemism of ‘death’, usually a non-durative telic event. The fact that lāpsus ‘action of falling’ (← lābor ‘slide’) is far more common than lapsiō ‘act of falling (into error)’ may be explained likewise. In the case of saliō ‘leap’, the verb’s telic aspect is a decisive factor for the derivation of saltus ‘jump, leap’ (no †saltiō).

Fors and fortūna: linguistic and cultural aspects

201

3 fors vs. fortūna: difference and confusion What we have seen so far suggests that there is an underlying semantic motivation for the selection of *-ti- or *-tu- in deriving abstract nouns; in particular, the latter is semantically specific. Fortson (2010: 108) says that between forms in *-ti- and those in *-tu-, the latter are usually used for so-called supines that indicate purpose (≈ goal/object) ‘in order (to)’: e.g. Latin questum ‘to complain’, Vedic ánu prá-voḍhum ‘to advance’, and Old Church Slavonic sŭpatŭ ‘to sleep’. In Latin, the tu-suffix is also included in future participles in -tūrus, which imply that one is aiming to do an act. Note also the Sanskrit gerundive suffix -tavyà- ‘to be . . . ’ (e.g. jani-tavíyà- ‘to be begotten’; see Gotō 2013: 141). Although this is most likely a later creation (no occurrence in Rigveda; see Whitney 1889: 346), *-tu- (~ *-teu̯- > Sanskrit -tav-) is at least there as an underlying element expressing obligation or necessity.16 It can thus be said that *-tufunctions as a marker of goal/purpose while *-ti- does not. It is then reasonable to assume that the same distinction underlies the ti-stem fors and the tu-stem derivative fortūna. With regard to the passage of Accius, cited above, Billington (1998: 130) says that this playwright “differentiates between a goddess Fortuna and an abstract concept of chance”, “ . . . allowing for chance [= Fors] to be either good or bad”. Her statement coincides with what we have found from the survey of several ti-stems (vis-à-vis tu-stems), namely their tendency as abstract nouns not to denote a final result; fors provides an either good or bad outcome. Returning to Manilius’ text, on the other hand, fortūna seems to be used to explain a goal-oriented (or result-conscious) algorithm or system of the universe.17 This is due to the goal-oriented property of -tu- according to our theory. The inclusion of this suffix indicates that the term fortūna is infused with a teleological notion (recall the gloss of Greek φέρτρυς in Hesychius, that is, ἆθλος ‘prize of contest’, which also implies a goal of an act); linguistic analyses thus support the semantic differentiation of fors and fortūna. The aforementioned phrase in the ablative forte fortūnā ‘by chance’ and the twofold theonym Fors Fortūna, although fors and fortūna have been judged to differ from each other, can also be explained in our framework. Walde and Hofmann (I: 534) regard fortūna as originally an adjective modifying fors. This understanding is reasonable in view of numerous adjectives in -no- (e.g. Rōmānus ‘Roman’,

16 The connection with Greek -τέος (< *-teu̯o-?; cf. Benveniste 1975: 104 and Rix 1992: 237) has been disputed; see Hutton (1993: 127). 17 Cf. Champeaux (1982: 430): “ . . . Fortū-na, exprimant, comme le parfait, l’action achevée, serait . . . la déesse qui a effectivement accompli l’« action de porter » . . . ”

202

Kanehiro Nishimura

aēnus ‘of bronze’, etc.; see Leumann 1977: 322–329). They thus tentatively gloss this noun phrase as ‘glücklicher Zufall’ (see also NIL 2008: 23 n. 44). As stated above, ti-stems are abstract nouns that tend to lack the nuance of goal or direction. In the case of fors, we can suppose the ‘detachment’ of one who observes things to happen or happening, that is, s/he does not necessarily detect intentional causes (by gods) or teleological mechanisms behind phenomena around her/him. Thus, Walde and Hofmann’s translation of fors as ‘Zufall’, that is, ‘random chance’, is well supported. On the other hand, some others may judge the things happening around them to be destined or promised in advance. Such intuition cannot be represented by fors; the adjective fortūna with its implicit teleological nuance due to *-tu- is therefore added. The simple fors (with the default value) is thus juxtaposed to fors fortūna ‘glücklicher Zufall’ (< ‘random chance that is decreed to bring about [good] luck’), to whom (Fors Fortūna as a deity) one can pray for good luck before things develop.18 In fact, the idiom forte fortūnā means not always ‘by chance’ but also ‘by good luck’ (Fors Fortūna, likewise), as Latte (1960: 179–180 n. 7) suggests by adducing a few cases from Terence (see also Billington 1998: 129). Note that Fors Fortūna remains a fickle goddess in a way (OV. fast. 6, 784 dubiae . . . deae; see Billington 1998: 136, 138); this is because the head of this noun phrase is Fors, an unpredictable deity of chance/luck (which is used alone for Fors Fortūna in OV. fast. 6, 775 as deam . . . Fortem [cf. Fortunae Fortis in 773] to meet metrical conditions). We are thus led to the idea that Fors Fortūna differs from simple Fortūna because the latter is not fickle but rather ‘(well-)destined’ from the etymological standpoint. The same is true for fortūna as an abstract noun (see the passages of Accius and Manilius in Section 1). Recall that we started this chapter with the confusion of fors and fortūna in use. The almost synonymous use of these two words is noticeable in Catullus. The passage of Cicero we discussed also implies semantic overlap of these words. The case of Appius Claudius Caecus is likewise perplexing. The question would thus arise as to why such confusion took place. Note that even if fortūna as a predetermined mechanism governs everything and/or the goddess Fortūna knows everything in advance, humans remain ignorant of their future and accordingly vulnerable to unpredictable events (see also Champeaux 1982: 433–434). Here we can see a sort of ambivalence of fortūna, and a semantic convergence point of fors and fortūna is thus conceivable. An ancient annotation to TER. Phorm. 841: O Fortuna, O Fors Fortuna ‘O Fortune, O Lucky Fortune!’ (trans. Barsby) is also suggestive in this regard. Brown (2006: 332) quotes

18 Fors Fortūna perhaps started to look like an appositional syntagm later (cf. Iūpiter ‘[orig.] Father Sky’).

Fors and fortūna: linguistic and cultural aspects

203

Donatus’ comment on this line that “Fortuna was the goddess of uncertain events, Fors Fortuna of happy outcomes.” Although Fortuna and Fors Fortuna are treated as two different divinities, Donatus seems to have interpreted Fortuna as if it was synonymous with the simple fors. Likewise, OV. fast. 6, 576 caecaque in hoc uno non fuit illa (sc. Fortuna) viro (sc. Servius Tullius, a devotee of the goddess) ‘ . . . and he was the only man for whom she was not blind’ (trans. Frazer and Goold) implies as much chronic volatility of Fortūna as fors.19 However, it seems that such a merger between fors and fortūna was not brought to completion. Despite his indiscriminate use of these two terms elsewhere (see Section 1), Cicero hints at a difference between them by saying as follows: Fortunaque sit vel Huiusce diei (nam valet in omnis dies) vel Respiciens ad opem ferendam vel Fors, in quo incerti casus significantur magis, vel Primigenia, a gignendo comes. (CIC. leg. 2, 28) ‘We may also have as gods Fortune, or the Fortune of This Day, for that applies to every day, or Fortune the Provident, that she may help us, or Chance Fortune, which refers particularly to the uncertainty of future events, or First-born Fortune, our companion from birth.’ (trans. Keyes)

In this passage Cicero presents several facets of the deity Fortuna. One of these is identified with Fors, which has to do with incerti casus ‘the uncertainty of future events’. However, the other ones have rather different overtones: Fortuna Huiusce diei appears to make reference to a specific day (as with daily horoscopes in newspapers or the Internet in our time); further, Fortuna Respiciens (‘mindful’) looks like a guardian deity who already knows what happens in the long-term future. From this observation it follows that authors who use fors and fortūna without clear distinction may still be conscious of some difference between these two terms.

4 Conclusion We have examined how different fors and fortūna are from each other based on the linguistic analyses of the Indo-European suffixes *-ti- and *-tu-; the latter has a specific semantic function denoting telicity or boundedness while in the former such nuance is normally absent. It is true that fors and fortūna are confounded with each other by not a few authors; fortūna seems to fluctuate between fors and fātum depending on authors and/or contexts. However, our

19 See also the meaning of fortuītus as mentioned in n. 3 and n. 4.

204

Kanehiro Nishimura

linguistic investigation has revealed differences in usage in certain other texts, which support a semantic distinction of the two suffixes. Between fors and fortūna, the latter is more conspicuous as a theonym. Indeed, the tu-suffix is included in some other divine names like Mātūta, Neptūnus, Portūnus, and probably Fātuus (see De Vaan 2008: 205). What we have found in this chapter may shed new light on linguistic and cultural analyses of such theonyms in the future.

Bibliography CIL = Corpus Inscriptionum Latinarum. (1862–). Berlin: Berlin-Brandenburg Academy of Sciences and Humanities. NIL = Wodtko, D. S., Irslinger, B. & Schneider, C. (2008). Nomina im Indogermanischen Lexikon. Heidelberg: Universitätsverlag Winter. Benveniste, É. (1975). Noms d’agent et noms d’action en indo-européen. Paris: Librairie d’Amérique et d’Orient. Billington, S. (1998). Fors Fortuna in ancient Rome. In S. Billington & M. Green (Eds.), The concept of the goddess (pp. 129–140). London/New York: Routledge. Bosworth, J. & Toller, T. N. (1898). An Anglo-Saxon dictionary: Based on the manuscript collections of the late Joseph Bosworth. Oxford: Clarendon Press. Brown, P., trans. (2006). Terence: The comedies. Oxford: Oxford University Press. Champeaux, J. (1982). Fortuna: Recherches sur le culte de la Fortune à Rome et dans le monde romain des origins à la mort de César. I: Fortuna dans la religion archaïque. Rome: École Française de Rome. Collin, C. (1904). Zur Geschichte der Nomina actionis im Romanischen. Archiv für lateinische Lexikographie und Grammatik mit Einschluss des älteren Mittellateins, 13, 453–473. De Martino, M. (2015). Arcana verba: Fortuna e Iuppiter nel loro background indoeuropeo. II: Il « motivo della Sorte esteso ». Bari: Edipuglia. Ernout, A. & Meillet, A. (1985). Dictionnaire étymologique de la langue latine: Histoire des mots (4th rev. and enl. ed.). Paris: Librairie C. Klincksieck. Fortson IV, B. W. (2010). Indo-European language and culture: An introduction (2nd ed.). Malden/Oxford/Chichester: Wiley-Blackwell. Gotō, T. (2013). Old Indo-Aryan morphology and its Indo-Iranian background. Vienna: Verlag der Österreichischen Akademie der Wissenschaften. Hutton, W. (1993).The meaning of qe-te-o in Linear B. Minos, 25–26, 105–131. Latte, K. (1960). Römische Religionsgeschichte. Munich: Beck. Lehmann, W. P. (1986). A Gothic etymological dictionary. Leiden: E. J. Brill. Leumann, M. (1977). Lateinische Laut- und Formenlehre. Munich: Beck. Livingston, I. (2004). A linguistic commentary on Livius Andronicus. New York/London: Routledge. Lundquist, J. (2015). On the accentuation of Vedic -ti- abstracts: Evidence for accentual change. Indo-European Linguistics, 3, 42–72.

Fors and fortūna: linguistic and cultural aspects

205

Nishimura, K. (2006). Umbrian restatu: A restitution. Indogermanische Forschungen, 111, 182–191. Pultrová, L. (2011). The Latin deverbative nouns and adjectives. Prague: Carles University in Prague/Karolinum Press. Pustejovsky, J. (1995). The generative lexicon. Cambridge, MA: The MIT Press. Rix, H. (1992). Historische Grammatik des Griechischen: Laut- und Formenlehre (2nd corrected ed.). Darmstadt: Wissenschaftliche Buchgesellschaft. Untermann, J. (2000). Wörterbuch des Oskisch-Umbrischen. Heidelberg: Universitätsverlag Winter. Vaan, M. de. (2008). Etymological dictionary of Latin and the other Italic Languages. Leiden/ Boston: Brill. Van Laer, S. (2010). La préverbation en latin: Étude des préverbes ad-, in-, ob- et per- dans la poésie républicaine et augustéenne. Brussels: Éditions Latomus. Vendler, Z. (1957). Verbs and times. The Philosophical Review, 66, 143–160. Vetter, E. (1953). Handbuch der italischen Dialekte, I. Band: Texte mit Erklärung, Glossen, Wörterverzeichnis. Heidelberg: Universitätsverlag Winter. Walde, A., & Hofmann, J. B. (1938–54). Lateinisches etymologisches Wörterbuch. Heidelberg: Universitätsverlag Winter. Weiss, M. (2011). Outline of the historical and comparative grammar of Latin (2nd corrected printing). Ann Arbor/New York: Beech Stave Press. Whitney, W. D. (1889). Sanskrit grammar (2nd ed.). Cambridge, MA/London: Harvard University Press. Zair, N. (2017). The origins of -urC- for expected -orC- in Latin. Glotta, 93, 255–289

Simona Georgescu

The world as a yawning gap New insights into the etymology of Lat. mundus ‘world’ Abstract: It is traditionally supposed that Lat. mundus ‘world’ had the original meaning of ‘ornament’, and gained the meaning of ‘world’ simply by calquing Gr. κόσμος, while mundus ‘hole, gap’ (considered to be of Etruscan origin) is just an homonym of mundus ‘world’ (cf. TLL). I shall argue that it is precisely the meaning ‘hole, gap’ that represents the key towards understanding the origin of mundus ‘world’ – the two meanings being strongly related to each other. By analysing the data provided by our Latin corpus, we find that the primordial meaning of mundus,-i must have been that of ‘cavity’, evolving towards the image of a yawning gap, just like Gr. χάος, afterwards designating ‘a gap leading to the underworld’, hence the ‘underworld’ itself; at the same time, still having as a starting point the shape of a cavity, yet turned upside down, it could evolve towards the meaning of ‘vault’ of the sky (as it appears in some of its most ancient occurrences), and then, by extension, it came to designate the whole universe. Keywords: Latin linguistics, mundus, etymology

1 Introduction The Thesaurus Linguae Latinae provides four different entries for the form mundus: – mundus 1 adj. ‘clean’ – mundus 2 s. ‘ornament’ – mundus 3 s. ‘world’ – mundus 4 s. ‘pit’ – as a path to the underworld.1

1 I will use these numbers (1–4) in the rest of the article to identify the different words or uses. Acknowledgment: I need to express my gratitude to my former student Remus Iosifescu, currently studying for a master’s degree at the University of Leiden, who not only provided me a significant number of studies that are absent from the Romanian libraries, but also showed me the weak points of my demonstration and added essential information for the present research, especially concerning Indo-European phonetics. Simona Georgescu, University of Bucharest https://doi.org/10.1515/9783110647587-014

The world as a yawning gap

207

The large and discontinuous semantic area covered by these words – if we regard them as homonyms – has, since antiquity, given birth to a great variety of explanations, either based on different etymological approaches, or promoting different relations between these meanings. Most of the existing hypotheses seem to overlook mundus 4 (considered to be of non-Indo-European origin) and to establish various possible chronological combinations between the adjective and the first two nouns. In this chapter, we shall argue that mundus 2 ‘ornament’, mundus 3 ‘world’, and mundus 4 ‘pit’ are, in fact, one and the same word: we only have to start from mundus 4 in order to understand its (Indo-European) origin and the semantic evolution leading to the other two meanings. To approach this issue, we start from a semantic overview of what seems to be the oldest meaning of mundus – that of ‘pit’ –, to then arrive at possible Indo-European cognates and to extract the most plausible root. Afterwards, in order to find the relation between the meanings ‘pit’ and ‘world’, we use as basic theoretical background the principles provided by cognitive etymology: thus, an insight into some cognitive patterns which determine the manner of perceiving the world in various cultures offers a different perspective on the relation between mundus 3 and mundus 4.

2 An outlook on the etymological propositions Among these four entries, mundus 3 (‘world’) seems to represent the most longstanding etymological crux.

2.1 Latin etymologists The traditional view is that mundus 3 had the original meaning of ‘clean’ or ‘ornament’ and later on acquired the meaning of ‘world’ simply by coining the Greek word κόσμος. This theory was actually put out by Latin authors: Appellatur a caelatura caelum, graece ab ornatu κόσμος, latine a puritia mundus. (VARRO Men. 420)2 A perfecta absolutaque elegantia. (PLIN. nat. 2, 8) Cum ornamenti nomine sit penes Graecos mundus. (TERT. adv. Hermog. 40, 2)

2 The quotes are taken from TLL, vol. VIII, p. 1634–1640. Translations, when present, are my own.

208

Simona Georgescu

Cui et apud Graecos ornamenti et cultus, non sordium nomen est. (TERT. adv. Marc. 1, 13, 3) Vtraque lingua κόσμον ab ornatu, mundum de lumine dixit. (PAVL. NOL. carm. 32, 197)

2.2 Modern etymologists Most modern linguists appeared to have simply adopted this explanation: – Buck (1949): “Lat. mundus [‘world’] is the result of semantic borrowing, starting as a literary imitation of Gr. κόσμος. It is the same word as mundus used of a woman’s ‘ornaments, dress’, this being related to the adj. mundus ‘clean, elegant’.” – Pokorny (1959) derives mundus ‘clean’ from *meu-1 ‘washed’, supposed to have also provided the noun mundus ‘world’ (calque on Gr. κόσμος) – DELL (2001 [19321]): mundus ‘world’ “semble bien être le même mot que mundus ‘parure’, qui a été choisi pour désigner le ‘monde’, sans doute à l’imitation du gr. κόσμος” The latest Latin etymological dictionary, De Vaan (2008), does not help us either: it still considers mundus ‘world’ to be of unknown origin “maybe related to mundus ‘ornament’”. The only difference is that De Vaan inverts the relation between those meanings: ‘ornament’ is the secondary meaning, coming from ‘world’, via ‘equipment’, and not vice-versa. There are many etymological propositions outside the world of dictionaries. Vendryes (1914) takes as a starting point the meaning of ‘hemispherical cavity’ dug in the ground, through which the living could communicate with the dead (cf. CATO ap. FEST. p. 154). Thus, he sees in mundus a word related to fundus, identical to Celt. dubno- ‘bottom, ground’. DELL, who considers this meaning as “un employ special”, admits at best “une contamination du groupe-trouble de fundus et du mot mundus, indépendant, pour désigner une entrée du monde infernal”. But, by considering that for the Latin speakers mundus in its ordinary acceptation has never designated anything else but the celestial vault (mentioning ENN. frg. var. 9, mundus caeli), DELL concludes that this meaning is incompatible with that of ‘bottom’, ‘fundament’, and that it is very possible that the infernal mundus had nothing to do with the celestial mundus, as the first one must be of Etruscan origin.3

3 The supposed Etruscan origin in based on a hapax munθ, of unknown meaning (cf. Puhvel 1976: 164 n. 47), and on what seems to be a female divine name munθu, munθχ, munθuχ on some mirrors. Anyhow, “explanation per obscurius must be discouraged”, concludes Puhvel 1976: 164, n. 47.

The world as a yawning gap

209

Kretschmer (1927: 196)4 thought about an etymological connection between mundus and Germ. Mund ‘mouth’ (Go. munþs from *munþaz), by virtue of mundus’ meaning of ‘entrance to the underworld’. He suggests an etymon *muntos as an equivalent to Germ. *munþaz. In an attempt to explain its phonetic evolution, Kretschmer assumes that it was by analogy with fundus that the original group -nt- turned to Lat. -nd-. Unfortunately, this proposition was not much taken into account: Evangelisti (1969: 358) considers it to be part of “il regno della fantasia”, while others do not even mention it. Pisani (1962: 10–11) takes as a starting point the phrase in mundo habere or in mundo esse, which he considers equivalent to the Italian phrase avere in pugno, nella mano (Germ. auf der Hand liegen, cf. Lat. mani-festus, in manu esse). He proposes an ancient meaning of ‘hand’ for mundus in this particular context – “che lo rende confrontabile col germ. *mund m.f. ‘mano’ (OHG munt, ASax. mund, etc.)”. As manus has always been connected by lexicographers to Germ. *mund (cf. DELL, s.v. manus), Pisani reanalyses Lat. mundus and Germ. *mund as derived from *mn̥dh- (with an n̥ “che questa volta ha dato in lat. un per la vicinanza della labiale”), which coexisted with the form manus on normal grade; cf. Germ. Mandel ‘handful’, Gr. μανδάκης.5 Not knowing how to connect the meaning of ‘hand’ with that of ‘world’ (therefore analysing them as different words), Pisani offers an original etymological proposition: he relates mundus to ASax. mud- (mut-)spelli-, originating OHG muspilli ‘world destruction’; ASax. mūd- can go back to a form *munth-, and this one can meet with mundus in a reconstructed *muntho-. He also cites ISID. orig. 20, 3, 4: nam mus terra, unde et humus. Although a hapax, mus is considered by Pisani as a phonetically plausible evolution of *munth-s (“nome radicale accanto al tematico *muntho-”), just like the alternation pos / pons in VARRO ling. 5, 4. In regard to the adj. mundus, he maintains the opinion in DELL: “il semble qu’il y ait eu deux (ou trois) mots différents, un adjectif mundus et un substantif

4 As part of a short note concerning Ernst Schwenter’s book, Die primären Interjektionen in den idg. Sprachen mit besonderer Berücksichtigung des Griech., Lat. und Germanischen (Heidelberg, 1924): when mentioning mugeo, musso etc, he points out that Germ. Mund should also be considered as part of this series; this assertion leads him to assume that Lat. mundus must be related to Germ. Mund. 5 He compares mundus with unda, supposedly *untha, -ia, which accounts for ASax. yđ, OSax. udia, OHG undea etc. See also Meillet (1920: 253) for a phono-morphological approach to Lat. unda.

210

Simona Georgescu

mundus ‘parure’ et ‘monde’”. Pisani insists on the primacy of the meaning ‘world’ and considers mundus muliebris a calque on Gr. κόσμος γυναικεῖος. Puhvel (1976: 165–167) considers that the original form was the adjectival one, mundus ‘washed’, hence ‘neat, prepared, proper’, and its neuter mundum (cf. LVCIL. 519) meant ‘neat state, readiness’ in an abstract sense, but also ‘neatly arranged stuff, gear, trimmings’ in a concrete manner. By coming in semantic contact with Gr. κόσμος, it assimilated the ornamental and philosophical meaning of the latter (‘decoration’, and ‘firmament, universe’). According to Puhvel (1976: 167), “Lat. mundus describes in origin that which has been so cleansed.” Evangelisti (1969) connects mundus with Skr. mandala- ‘circle, sphere’ and ‘magical circle’, a theory that was happily embraced by Canevascini (1995: 341) and Milani (2009: 349),6 but considered as ‘unconvincing’ by Puhvel (1976: 165). What we can keep is Evangelisti’s suggested evolution ‘ditch’ > ‘sky’ > ‘ornament’ (the semantic filiation should be preserved, unlike its proposed origin which should be reinterpreted). To draw a first conclusion concerning this variety of etymological propositions, we shall quote Dumézil (2000 [19741]: 356): “la notion de mundus est l’une des plus controversées de la religión romaine”. After describing the rituals concerning this subterranean mundus, Dumézil concludes: “il n’est pas possible de reconstruire ce que nos informateurs eux-mêmes ne comprenaient plus et dont ils ont, somme tout, très peu parlé. L’étymologie n’aide pas: le mot mundus ne s’éclaire pas par l’indo-européen” (2000 [19741]: 358).

3 A semantic approach 3.1 Mundus’ original meaning It does seem that the oldest meaning is that of ‘hole in the ground’, so old that even the Romans themselves no longer understood its relationship to mundus ‘world’ and did not perceive the roots of the ancient ritual. Except for Vendryes, all etymologists either exclude or do not exploit sufficiently a relation between mundus 3 ‘world’ and mundus 4 ‘pit’, ‘underworld’.

6 On the other hand, Milani acknowledges that the etymological problem of mundus has not yet been solved: “Nonostante validi approfondimenti, resta ancora aperto il problema della sua origine e degli sviluppi semantici. Ci si chiede: il qualificante mundus e il sostantivo mundus hanno la stessa origine? Sono due sfaccettature dello stesso lessema? Si tratta solo di omofoni? I vari significati del sostantivo mundus “ornamento”, “volta celeste”, “universo, mondo”, “pozzetto infero” sono sviluppi semantici o indicano lessemi diversi?” (2009: 347).

The world as a yawning gap

211

I believe that this relation is precisely the key to the origin of mundus ‘world’. I shall argue that mundus 3 and mundus 4 are one and the same word. Let us first take a look at mundus 4. TLL presents it as part of the formula mundus patet: Mundus cum patet, deorum tristium atque inferum quasi ianua patet. (VARRO frg. Macr. Sat. 1, 16, 18) ‘when mundus opens, it is like an open door of the sad and the underworld gods.’

3.2 Umbilicus urbis In the Roman Forum there still lies a brick structure (cf. Figure 1): a type of cave inside of which there is a pit traditionally considered as leading to the underworld, the world of the dead. Three times per year this pit was open and the dead could rise through it. The Romans called it, according to Festus, Cereris mundus.7 Dumézil (2000 [19741]: 356) believes that this denomination corresponds to a localization: “le mundus était in sacro Cereris c’est à dire sans doute dans une dépendance du temple de la déesse. Normalement fermé, il ne s’ouvrait que 3 jours par an.” The derivative adjectival form mundalis appears in association with the same goddess in an inscription from Capua, sacerdos Cerialis mundalis (CIL X 3926, ap. TLL vol. VIII, p. 1621, 34–35), which attests the strong relation between mundus and Ceres.8 A very important piece of evidence confirming that the mundus was a construction (not only in Rome, but in other towns as well) is offered by the following inscriptions: T. Muttius . . . Celer . . . theatrum, mundum, gradus faciendos cura(vit). (from Corfinium, probably late republican, discussed by Devijver and Van Wontergehm 1983)9 sanctum mundum Attinis p(ro) r(editu) a fundament(is) Helviae Abascante et Capitolina f(ilia) . . . p(ecunia) s(ua) f(ecerunt). (from Lucania, INSCR. Année Épigr. 1979 n. 195)

7 Cereris qui mundus appellatur, qui ter in anno solet patere (FEST. p. 142). 8 See also APVL. apol. 13 maius piaculum decernis speculum philosopho quam Cereris mundum profane videre, interpreted in TLL as “apparatus quidam mysticus”, under mundus 2 (TLL vol. VIII, p. 1634, 47–49), while it should probably appear under mundus 4 (cf. Devijver and Van Wontergehm 1983: 500). 9 CIL IX 3173 erroneously prints alvendum; other restitutions contain fundum or eundum instead of mundum, which is the correct reading according to Devijver and Van Wontergehm (1983: 491).

212

Simona Georgescu

Figure 1: Umbilicus urbis (also described as Mundus), the Roman Forum; restoration dating from third century AD.10

The second inscription attests that the mundus 4 could also be associated with other gods, in this case, with Attis. It seems to be “une sorte de crypte, aménagée a fundamentis, ‘pour le retour’ du dieu (c’est-à-dire sa résurrection célébrée à chaque printeps)” (INSCR. Année Épigr. 1979 n. 195). In a commentary of Festus on Cato, we find an interesting explanation for the usage of mundus with the meaning ‘underworld’ as a result of the transfer

10 The photo was taken by myself in the Roman Forum, Rome, 2015.

The world as a yawning gap

213

of the name mundus ‘world’ to this reality, transfer which is based on the formal similarity of the two mundi: qui quid ita dicatur sic refert Cato in commentariis iuris civilis: ‘mundo nomen impositum est ab eo mundo, qui supra nos est: forma enim eius est adsimilis illae’; eius inferiorem partem veluti consecratam dis Manibus clausam omni tempore nisi his diebus, qui supra scripti sunt . . . (FEST. p. 154) ‘the name of the ‘mundus’ was given from that mundus that is above us: because its form is similar to that one [from above]; its inferior part, just as it is consecrated to the Manes gods, is closed all the time, except for those days mentioned before.’

The essential aspect for understanding the significance of this mundus is exactly its shape, on which even our ancient sources insist: forma < . . . > eius est < . . . > adsimilis illae (FEST. p. 154). Thus, the shape of the subterranean mundus is similar to the sky vault, the upper vault, eo mundo, qui supra nos est (as Festus puts it). As even for Cato the meaning was difficult to explain, probably because it had already lost its frequency, we argue that this must be the oldest meaning, and thus the one that should be put in first place. There is yet another significance of mundus: Plutarch (The life of Romulus 11) describes a ritual accomplished by Romulus, in which he carves a hole in the ground, where he throws “the first offerings that are good according to tradition and necessary according to nature”, and, moreover, a handful of dust from each country that his companions came from. Βόθρος γὰρ ὠρύγη περὶ τὸ νῦν Κομίτιον κυκλοτερής, ἀπαρχαί τε πάντων, ὅσοις νόμῳ μὲν ὡς καλοῖς ἐχρῶντο, φύσει δ' ὡς ἀναγκαίοις, ἀπετέθησαν ἐνταῦθα. καὶ τέλος ἐξ ἧς ἀφῖκτο γῆς ἕκαστος ὀλίγην κομίζων μοῖραν ἔβαλλον εἰς ταὐτὸ καὶ συνεμείγνυον. (PLVT. Rom. 11) ‘A circular trench was dug around what is now the Comitium, and in this were deposited first-fruits of all things the use of which was sanctioned by custom as good and by nature as necessary. And finally, every man brought a small portion of the soil of his native land, and these were cast in among the first-fruits and mingled with them.’11

And then he writes down an essential phrase: Καλοῦσι δὲ τὸν βόθρον12 τοῦτον ᾧ καὶ τὸν ὄλυμπον ὀνόματι μοῦνδον. (PLVT. Rom. 11) ‘Τhey call this trench, as they do the sky, by the name of “mundus.”’13

11 Translation by Perrin (1914). 12 The same word is used in the Odyssey 11, 25 to designate the hole in the ground made by Odysseus when communicating with the dead. 13 See also the testimony offered by SERV. auct. Aen. 3, 134 quidam aras superorum deorum volunt esse, medioximorum id est marinorum focos, inferorum vero mundos.

214

Simona Georgescu

Then Plutarch adds another significant note: εἶθ' ὥσπερ κύκλον κέντρῳ περιέγραψαν τὴν πόλιν. ‘Then, taking this as a centre [i.e. the mundus that was dug first], they marked out the city in a circle round it.’

We can thus understand that mundus was, moreover, the omphalos of the city – as the omphalos in Delphoi – and the brick structure in the Roman forum preserved its original function. TLL considers this denomination to be ex confusione orta vel ficticia (‘born out of confusion or fictive’). But if we take it for granted, we can see the essential part: mundus was on the one hand a round space marked by a trench where Romulus reconstructs the world in miniature, on the other hand a gap leading to the underworld, and the underworld itself. Its shape is that of a receptacle, a pit, a yawning gap.

4 Possible Indo-European cognates The exact meaning of ‘opening’, ‘gate’ seems actually to have been the original meaning of the Germanic words designating the ‘mouth’, reconstructed in Proto-Germanic as *munþa (cf. Kroonen 2013, s.v.): Go. munþs, ON munnr, Far. munnur, muđur, OE muþ, E mouth, OFri. mūth, OHG mund, Germ. Mund, etc. These forms indicate a PIE root *mnt-o, which, according to the same lexicographer, must have given Lat. mentum ‘chin’, MW mant ‘mandible, gums’. This root may also be identified in Gr. μασάομαι ‘to chew, bite’ and – a very important piece of evidence – in Lat. mando ‘to chew’. The testimony of this verb is meaningful not just because of the attestation of this root in Latin, but also because it facilitates our understanding of the phonological processes which led to the Latin form mundus.

5 Phonetics This etymology, as we have shown above, was advanced by Kretschmer in 1927, but only as a short note without further explanations. Perhaps that is why he was not taken seriously by other linguists. The problem of his proposition was the lack of a strong semantic explanation and the poor phonetic evidence.

The world as a yawning gap

215

If we start from the above-mentioned Etymological Dictionary of ProtoGermanic of Kroonen (2013), we must presuppose a PIE root *mnt-o-, designating the ‘mouth’. The stem was differently reconstructed by Pokorny as *menth-, a form that is difficult to accept because of the voiceless aspirate. What we should emphasize is that Pokorny also derives Lat. mando from the same root as the Germanic words designating the mouth. Kroonen considers as the base for mando a PIE form *mnt-néh2-, which went through the following phonetical processes: (1) Dissimilation of the first n; (2) Schwa secundum in the resulting form *mt-nā; and (3) Prenasalization under Thurneysen’s law. De Vaan reconstructs a form *mt-n(e)-h2- for mando, with a parallel evolution as *pt-n(e)-h2- > pando. Taking as a starting point the equivalences Lat. unda – Skr. udán, udnás and Lat. fundus – Skr. budhnás, to which he added Lat. pando – Gr. πετάννυμι,14 Lat. panderent – Osc. patensíns, Thurneysen (1883) drew the conclusion that a PIE -tn- / -dn-/ -dhn- yielded Lat. -nd-.15 His theory was strongly supported by Sturtevant (1944), who demonstrated the identity between the suffix -nd- of Latin gerundive and the Hittite suffix of verbal action nouns -tar < *-tr, with its genitive -nnas going back to *-tnos. He adds to Thurneysen’s list the example of Lat. praehendo, Gr. χανδάνω, who must go back, according to his demonstration, to IE *ghəd-nā (Sturtevant 1944: 210). Szemereny (1950) reopens the problem and adds, among other examples (prandium < *prətnos, etc.), that of mundus ‘clean’. Following the theory of PIE -tn-/ -dn- > Lat. -nd-, he proposes a form *mūtnos (a thematicized adjectival form of the noun *mūter / mūtnes), on a root for which he establishes the meaning of ‘washed’.16 A form *mudnos had also been proposed as the basis of mundus by Schulze (1934: 471), and accepted by Leumann (1977: 201), who adds the possibility of a *mūtnos. Going back to our starting point, in either of the two possible forms proposed for mando, De Vaan’s *mt-n(e)-h2- or Kroonen’s *mnt-néh2-, we should accept the change -tn- > -nd-. If we apply the same PIE root, *mnt- or *mt- to

14 All the three examples were regarded as indisputable by Stolz and Schmalz (1928) and Leumann (1977). 15 He supports this theory with further examples from the Romance languages, where this phonetic tendency leads to similar results (1883: 303): in Mio Cid, Spanish 2nd pl. imperatives dad, tened, cortad, followed by a pronoun in Dative or Accusative nos, thus dadnos, tenednos, cortadnos are many time written dandos, tenendos, cortandos. 16 He regards the noun mundus ‘world’ as a homophone coming from *movendos. Since we are talking about the phonological aspect, we will only take into account the phonetical explanation he gives for the form mundus, which could sustain an original *mtn-.

216

Simona Georgescu

mundus, we could easily reconstruct a form with o-grade *mot-n- / *mont-n- (in this case, with dissimilation of the first -n-), corresponding to Pokorny’s root on e-grade *ment- and alternating with the zero-grade stem present in ProtoGermanic.17 It is acknowledged that a PIE -o- before one of the sonants -l-, -r-, -m-, -n- followed by consonant closes to -u-. Here are some examples provided by Leumann (1977: 48): -or- + cons > -ur-: furnus vs. fornax; curvus vs. corvus; furmica vs. formica; -ol-+cons. > -ul-: stultus vs. stolidus, multa vs. the ancient form molta; omb- > umb-: umbo, umbilicus, rumpia, Gr. ῥομφαία. There are also cases of -om- turning to -um- before vowel: Numidae < Gr. Νομάδες, umerus < *omesos, hum-us vs. Gr. χθον-. The same closure of -o- to -u- takes place before a guttural nasal, as in onc-, ongu- > unc-, ungu-; uncus vs. Gr. ὄγκος, hunc vs. Old Latin honce, cūncti vs. OLat conctos, unguis vs. Gr. onux, homunculus, nuncupare (unc < ŏnc), for *nōmicupare. And, most relevant for us, there are examples of -und-, -unt- going back to -ond-, -ont- (including the Future Passive Participle *o-n-do- > -undo-): frundes (ENN. ann. 261), frunte CIL, frō(n)s, frūs, Populea fruns (ENN. ann. 577). We find more examples by simply consulting De Vaan (2008): mulgeo < PIt. *molgeje, mulctra < *m(e/o)lg-tro-, uls < PIt. *ol-tero, ol-tamo (ultimus), ulmus < PIt. *e/olmo among others. We could thus consider a proto-form *mo(n)t-n-o- standing for mundus, from a PIE *m(e/o)(n)t-o- probably designating the ‘mouth’. The other Latin descendant, mando, ‘to chew’, that all consulted dictionaries establish as derived from this root, provides important evidence that this stem has been inherited in Latin, with a meaning very closely related to that of ‘mouth’. The transparent formal relation between mando and mundus demands a semantic justification. Therefore, we shall try to find out what the mouth and the world have in common.

6 Semantics 6.1 ‘Gap’ – ‘mouth’ The case of Lat. pando is helpful once again: De Vaan (2008) derives, from PIE *pt- (whose nasalized form provided pando < pt-n(e)-h2) with secondary full

17 Pisani (1962), in order to explain the sequence -un- in mundus equivalent to Germ. -un- in Mund (here an expected evolution of a vocalic -n-, which should have provided in Latin -en-), presumed an influence of the preceding labial m- on the vowel; Evangelisti (1969) proposed the same influence of the preceding m- over the vowel (this time -o-, as he reconstructs the form *mondo-).

The world as a yawning gap

217

grade, a form pat-, to be found in pateo, passus < * patio-, dispennite < *dis-pat -n-, and Osc. patensins. Similarly from a phonetic point of view, we find Latin words containing a root mat- whose etymology is described as unknown. This goes for matula ‘a vessel for liquids, jar’, matella ‘a vessel, pot’ (both evident derivatives from a root *mat-) and *matia ‘intestines’ (a glossary word, mentioned by Ernout/Meillet, which could also be reconstructed on the base of the derivative matiarius ‘sausage-makers’ and of the Romance words meaning ‘intestines’ leading to this etymon). If we start from the same root *mt-, which provided mando (parallel to pt- > pando and pateo), we could find here an older meaning than that of ‘mouth’: that is ‘cavity’, ‘recipient’, ‘pit’ (supposed as well by the English etymological dictionaries for mouth). In fact, according to Hilmer’s theory (1918), the concept of ‘cavity’ appears to be one of the primary concepts which can evolve either towards the meaning of ‘cavity in the body’ (therefore designating the ‘abdomen’, ‘intestines’, ‘vulva’, ‘mouth’ – a semantic evolution to which we can subsume Lat. *matia ‘intestines’ and at least PG *muntha ‘mouth’), or to different types of cavities – ground holes or various recipients (see Lat. matula and matella). If we regard mundus as based on the same root, we can now understand that its original meaning must have been that of ‘cavity’, ‘hole’, ‘pit’. Therefore, between mundus 3 and mundus 4 according to the TLL order, we must give priority to mundus 4 ‘pit’ as the oldest in Latin.

6.2 ‘Gap’/‘pit’ – ‘world’ In order to explain how this word could have evolved semantically from ‘pit’ to ‘world’, we shall adduce similar examples from different languages. As we have previously mentioned, Vendryes (1914) thinks that mundus is related to Celt. *dubno- ‘bottom, ground’. Although this etymology cannot be retained, what we should focus on is that the Celt. root *dubno- ‘bottom, ground’, mostly recurring in words meaning ‘profound’ (cf. W dwfn, f. dofn, Bret. don) and related to Indo-European words with the same meaning (Go. diuths, OHG tiof, Lit. dubus ‘profound, hollow’), has evolved, in Ir. domun, to the meaning ‘world’ (cf. Dumnorix ‘king of the world’). “D’après l’étymologie, le nom du monde en celtique aurait été tiré d’un mot signifiant le ‘fond’” (Vendryes 1914: 307). If we take a look at the Greek language, we find that the equivalent word for mundus, meaning ‘underworld’, is no other than χάος. The semantic parallel with this word can give us a new perspective on the evolution of mundus. Gr. χάος originally meant ‘abyss’, the primary abyss that constituted the universe before any element came into being. Ancient Greeks believed it was a sort of

218

Simona Georgescu

cavity – above and below. In the Orphic tradition, it is described as μέγα χάσμα πελώριον ἔνθα καὶ ἔνθα, “a deep gap, in one direction and the other”, meaning above and below (PROCL. In Remp. II 138; fr. 66 Kern, ap. Hâncu 2018: 91). Its etymology is illuminating: it is related to χάσκω ‘to yawn, gape, open the mouth wide’, χαίνω ‘id.’, χάσμα s. ‘yawning’, and, outside the Greek language, with the Germ. Gaumen ‘the palate, the roof of the mouth’. Thus, the most appropriate translation for χάος would be the one proposed by Cornford (ap. Hâncu 2017: 111), ‘yawning gap’. In various Greek epic poems describing the original state of the world, the terms χάος and χάσμα replace each other, confirming that χάος did not have a different significance from χάσμα ‘yawning mouth’ or ‘the act of yawning’.18 This original yawning gap appears in many different mythologies as the primary state of things, before the world came into being. It appears in the Babylonian genesis, personified as Apsu (probably from api ‘pit’, cf. Vieyra 1961, and related to Gr. ἄβυσσος).19 The Orphic myth of the world genesis is built around the idea of an original dragon mouth, the open mouth of Cronos who had three different heads, in the center being that of a dragon (δράκω – literally ‘serpent’). Scandinavian mythology, too, speaks of an abyss that was literally yawning everywhere. Even Chinese mythology refers to an original yawning gap. Those are just a few examples for what we can describe as a mythological constant: an original abyss that can be seen as an open mouth of a certain deity. Moreover, we cannot ignore the importance of the act of ingurgitating in mythology: deities that swallow other deities or primary elements to give birth to different ones (e.g. Cronos eating his children). In a different perspective, the particular form of an open mouth, the vault supposed by it – should it be up or down – can be reproduced by the form of the egg that appears to be, in certain mythologies, the primary element; or – perfectly understandable – the original uterus that gives birth to the world, and that is central in certain mythologies (cf. Kernbach 1996 [19781]). Going back to the Greek χάος, we find it significant that later on, in the Greek literature, the referential area of this word was reduced to the ‘underworld’, to a certain gap that relates the world of living to the world of the dead.20

18 “Χάος < *χάϝος beside χαῦνος, like ἔρεβος to ἐρεμνός < *ἐρεβνός. As a basic meaning ‘hole, empty space, yawning opening’ is quite thinkable for χάος. Therefore χάος and χαῦνος were since long connected with χάσκω, χάσμα, χανεῖν, χήμη, etc.” (Beekes 2010, s.v. χάος; cf. DELG). 19 Later on, this Apsu was to become a synonym of ‘world of the dead’. 20 See the example of Hesiod (736–740), who places “the terrible houses of dark Night” (Νυκτὸς ἐρεμνῆς οἰκία δεινά) in “a great chasm” (χάσμα μέγ[α]).

The world as a yawning gap

219

And that is exactly what Lat. mundus represents in an archaic period, in the most ancient tradition.21 Thus, if we do take as a fundament the meaning of ‘mouth’, we find a possible parallel with Gr. χάος: just as χάος designated the original yawning gap, then the primordial world (as it appears in Hesiodos), mundus might have followed the same trajectory. The meaning of χάος, after evolving to ‘infinite concave space’, has been restrained once again to the tenebrae infernales and to ‘abyss’ – which continues to be its meaning in modern Greek. If we take a look to mundus’ early occurrences we shall find it referring, in a very concrete manner, to the vault of the sky: Mundus caeli vastus constitit silentio et neptunus . . . undis . . . pausam dedit . . . (ENN. frg. var. 9) Magni per caerula mundi. (LVCR. 5, 772) Cohum enim apud ueteres mundum significat. (DIOM. gramm. I 365, 18)

An even more obvious equivalence between mundus and caelum is to be found in the following examples: Caelum, quod vere mundus vocatur. (MACR. somn. 1, 20, 8) Mundus proprie caelum vocatur. (MACR. somn. 2, 11, 12)

In another context, we find mundus in an obvious upward pointing position (understood as ‘sky vault’) in its relation with the Earth. De mundi sideribus. (ACC. trag. 678) Quis sub hoc mundo mihi dabitur maritus? (SEN. Herc. O. 402) Sidera quis mundumque velit spectare cadentem? (LVCAN. 2, 289)

Of great importance in understanding its semantic evolution we find the following acceptance: terra . . . ut in media mundi regione quiescat (LVCR. 5, 534), an image that places the Earth between two vaults, one above, another one below. We can thus suppose that Lat. mundus underwent the same type of semantic change as Gr. χάος, that is restriction of meaning: from the original ‘yawning gap’ that was everywhere, its usage became restricted to the upper vault or the

21 The same semantic evolution is followed by Apsu, see n. 19.

220

Simona Georgescu

subterranean vault, in each case maintaining the concave shape. According to our testimonies, it came to be more frequently used for the ‘vault of the sky’, from where it enlarged its meaning to ‘sky as cosmic universe’; in fact, even the conceptualization of the ‘universe’ in our modern way creates the image of cosmic space, where volens-nolens we include the Earth and ourselves. Thus mundus came to designate the same concept: the world as a whole.

7 Mundus 1 ‘clean’ and mundus 2 ‘ornament’ The investigation is not over until we explain the situation of the two other mundi in TLL: the adj. ‘clean’ and the noun ‘ornament’ – supposed to have been either the fundament for ‘world’ (as most linguists have argued, in a sort of tradition going back to Varro; cf. Puhvel 1976, Canevascini 1995, etc.), or derived from ‘world’ (cf. De Vaan 2008). We reaffirm the statement of DELL, also backed up by Pisani (1962: 12): “il semble qu’il y ait eu deux (ou trois) mots différents, un adjectif mundus et un substantif mundus ‘parure’ et ‘monde’”. Thus, we follow Schulze (1934), Szemerenyi (1950), Puhvel (1976), Leumann (1977) and TLL in the proto-form they propose for the adjective mundus: *mudnos < *mewd- which can be detected in OCS myti ‘wash’, Lit. máudyti ´bathe´, Gr. μυδάω ‘be wet’, etc. Although deriving it from a different root, *mud-no- ‘happy’ (relating it to Skr. múd ‘joy, delight’, módate ‘to be happy’ etc), De Vaan (2008) also separates the adjective from the noun. We consider that, originally, we have two Latin homonyms: an adjective mundus ‘clean’ < *mew-d- and a noun mundus ‘cavity’, ‘pit’ < PIE *m(e/o)(n) t-n-, which evolved to ‘world’. The noun mundus ‘ornament’ could either be a semantical derivation of the adjective ‘clean’ or represent a simple calque on Gr. κόσμος ‘ornament’ and ‘world’.22 In this case, the meaning ‘ornament’ would be a third one in Latin, after mundus had already been provided with the meaning ‘world’.23 Actually, this meaning is not certainly attested until Cicero, thus in a period where Greek

22 Unfortunately, the limits of this article do not allow us to continue the research on mundus 2. We can only present here a hypothesis that will have to be developed on another occasion. 23 This idea had already been formulated by Pisani (1962: 11), who started from the Greek syntagm κόσμος γυναικεῖος to explain mundus muliebris as a calque.

The world as a yawning gap

221

cultural influence was at its strongest, manifested either by borrowings or by calques.24 The position of this marginal meaning could have been centralized by an association with the adjective ‘clean, elegant’, cognitively remitting to ‘adornment, jewel’. It was the folk etymology deriving ‘world’ from ‘clean’ or ‘ornament’ that led to the etymological propositions formulated by the Classical and Late Latin writers (see Section 2.1). At the same time, the parallel with Gr. κόσμος ‘ornament’, ‘world’ reinforced the relation between the three meanings of the two original homonyms: ‘clean’ on the one hand and ‘world’ and ‘ornament’ on the other hand.

8 Conclusions After having analysed what seems to be the oldest meaning of Lat. mundus, that of ‘pit’, ‘hole in the ground’, we proposed an etymological relation with the Germanic words designating the ‘mouth’ (as an opening, a gap). It appeared thus that the meaning of ‘world’ was acquired only later on, as a semantic evolution based on the analogy between the shape of a cavity and the vault of the sky – extending afterwards its semantic area to the whole universe. If we go back to our starting point, we should reconsider the order of the four forms established by TLL: we assume that there were two original homonyms, (1) mundus adjective and (2) mundus noun. The latter’s initial meaning must have been that of ‘pit’, evolving to ‘world’, and then, by calque on Greek, to ‘ornament’.

Abbreviations ASax. Bret. Celt. E Far. Germ.

Anglo-Saxon Breton Celtic English Faroese German

24 Although TLL cites as a first documentation of mundus ‘ornament’ a passage of Lucilius legavit quidam uxori mundum omne penumque (LVCIL. 519), the meaning of the apparent neuter mundum is not at all clear; another obscure passage from ACC. trag. 654, quoted by Festus, cum virginali mundo clam patre could be an isolated use previous to its lexicalization (if the meaning is really that of ‘ornament’).

222

Gk. Go. IE Ir. Lat. Lit. MW OCS OE OFri. OHG OLat ON Osc. PG PIE PIt. Skr. W

Simona Georgescu

Greek Gothic Indo-European Irish Latin Lithuanian Middle Welsh Old Church Slavic Old English Old Frisian Old High German Old Latin Old Norse Oscian Proto-Germanic Proto-Indo-European Proto-Italic Sanskrit Welsh

Bibliography CIL = Corpus Inscriptionum Latinarum (1862–). Berlin/Boston: De Gruyter. DELG = Chantraine, P. (1968). Dictionnaire étymologique de la langue grecque. Histoire des mots. Paris: Klincksieck. DELL = Ernout, A. & Meillet, A. (2001 [19321]). Dictionnaire étymologique de la langue latine. Paris: Klincksieck. TLL = Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. Beekes, R. (2010). Etymological dictionary of Greek. Leiden/Boston: Brill. Buck, C. D. (1949). A dictionnary of selected synonyms in the principal Indo-European languages: A contribution to the history of ideas. Chicago/Illinois: The University of Chicago Press. Canevascini, G. (1995). On Latin mundus and Sanskrit muṇḍa. Bulletin of the School of Oriental and African Studies, 58(2), 340–345. Devijver, H. & Wontergehm, F. van (1983). Un mundus (Cereris) a Corfinium: Nuova lettura e interpretazione dell’iscrizione CIL IX 3173 = ILS 5642. Historia, 32, 484–507. Dumézil, G. (2000 [19741]). La religion romaine archaïque. Paris: Payot. Evangelisti, E. (1969). Una congruenza lessicale latino-indiana, a proposito del mundus sotteraneo. In V. Pisani (Ed.), Studi linguistici in onore di Vittore Pisani (vol. 1, pp. 347–366). Brescia: Paideia. Hâncu, M. (2017). Understanding chaos in ancient Greek literature. In M.-L. Dumitru Oancea (Ed.), Inventrix et erudita magistra. Studia in honorem Anae-Cristinae Halichias (pp. 109–126). București: Editura Universității din București.

The world as a yawning gap

223

Hâncu, M. (2018). Lupta cu haosul. Terminologia cosmogoniilor presocratice. Nova Studia Classica. Nova Series IV. București: Editura Universității din București. Hilmer, H. (1918). The origin and growth of language. The Journal of English and Germanic Philology, 17(1), 21–60. Kernbach,V. (1996 [19781]). Miturile esențiale. București: Univers Enciclopedic. Kretschmer P. (1927). Review of Ernst Schwenter, Die primären Interjektionen in den idg. Sprachen mit besonderer Berücksichtung des Griech., Lat. und Germanischen. Glotta, 15, 195–196. Kroonen, G. (2013). Etymological dictionary of Proto-Germanic. Leiden/Boston: Brill. Leumann, M. (1977). Lateinische Laut- und Formenlehre. München: Beck. Meillet, A. (1920). Les noms du « feu » et de l’« eau » et la question du genre. Mémoires de la Société de linguistique de Paris, 21, 249–256. Milani, C. (2009). La nozione di ‘mondo’ in alcune lingue indoeuropee. In C. Milani, R. B. Finazzi & P. Tornaghi (Eds.), Varia inguistica (pp. 337–372). Milano: EDUCatt. Perrin, B. (1914) Plutarch. Lives, Volume I: Theseus and Romulus. Lycurgus and Numa. Solon and Publicola (Loeb Classical Library 46). Cambridge, MA: Harvard University Press. Pisani, V. (1962). Obiter scripta III. Paideia, 17, 1–15. Pokorny, J. (1959) Indogermanisches etymologisches Wörterbuch. Bern: Francke. Puhvel, J. (1976). The origin of Greek kosmos and Latin mundus. The American Journal of Philology, 97(2), 154–167. Schulze, W. (1934). Kleine Schriften. Göttingen: Vandenhoeck & Ruprecht. Stolz, F. & Schmalz, J. H. (1928). Lateinische Grammatik (5th edition, reedited by M. Leumann and J. B. Hofmann). Munich: Beck. Sturtevant, E. H. (1944). Hittite verbal nouns in -tar and the Latin gerund. Language, 20, 206–211. Szemerenyi, O. (1950). The Latin gerundive and other -nd- formations. Transactions of the Philological Society, 16, 169–179. Thurneysen, R. (1883). Urspr. dn, tn, cn im Lateinischen. Zeitschrift für vergleichende Sprachforschung, 26, 301–314. Vaan, M. de (2008). Etymological dictionary of Latin and the other Italic languages. Leiden/ Boston: Brill. Vendryes, J. (1914). La famille du latin mundus « monde ». Mémoires de la Société de Linguistique de Paris, 18, 305–310. Vieyra, M. (1961). Les noms du ‘mundus’ en hittite et en assyrien et pythonisse d’Endor. Revue hittite et asianique, 19, 47–55.

Eleonora Litta & Marco Passarotti

(When) inflection needs derivation: a word formation lexicon for Latin Abstract: The Word Formation Latin (WFL) project has been awarded a Marie Curie Individual Fellowship to create a language resource consisting of a derivational morphological lexicon of the Latin language, which connects lexical elements on the basis of word formation rules. In WFL, lexemes are segmented and analysed into their derivational morphological components, in order to establish relationships between them on the basis of derivational or compounding processes. This chapter illustrates the methodology lying behind WFL, its lexical basis and the choices taken in order to maintain consistency in the resource, when encountering a number of cases raising theoretical issues. The online graphical query system used to access WFL is described, as well as a few examples of linguistic investigations that are made easier through the use of such a resource. Keywords: Latin linguistics, derivational morphology, word formation, Classical Latin, language resources

1 Introduction The investigation of lexical data of classical languages through the use of language resources and Natural Language Processing (NLP) tools has witnessed a surge of interest in the past two decades. The employment of digital resources has changed the way classicists approach their investigation of texts. As far as Latin is concerned, today several textual and lexical resources, as well as NLP tools, are being used in lexicographic research. These resources are annotated corpora, treebanks,1 and digitised lexica.2 Beside

1 Ancient Greek and Latin Dependency Treebank (perseusdl.github.io/treebank_data, accessed: 20 October 2017), PROIEL Treebank (proiel.github.io, accessed: 20 October 2017), Index Thomisticus Treebank (itreebank.marginalia.it, accessed: 20 October 2017). 2 A Latin Dictionary by Lewis and Short (www.perseus.tufts.edu, accessed: 20 October 2017), Thesaurus Linguae Latinae (www.degruyter.com, accessed: 20 October 2017). Eleonora Litta, Marco Passarotti, Università Cattolica del Sacro Cuore https://doi.org/10.1515/9783110647587-015

(When) inflection needs derivation: a word formation lexicon for Latin

225

these language resources, there are NLP tools, such as morphological analysers,3 part-of-speech taggers4 and syntactic parsers.5 In this context, word formation processes and relations are rarely treated by computational lexical resources or by morphological analysers, but notions of word formation are of the highest importance for the construction of computational linguistics applications (Ten Hacken 2002). In the derivation process, the added morpheme almost always possesses a well-defined meaning and all lemmas characterised by the addition of the same morpheme acquire a set of common semantic properties (for example, all derived lemmas containing the suffix -(t)or are characterised, in Latin, by an agentive or instrumental nuance). Enriching textual data with derivational morphology tagging promises to provide strong outcomes. It can organise the lexicon at higher level than words, by building word formation based sets of lemmas sharing a common ancestor. Moreover, information on word formation can act as a bridge between morphology and semantics, since core semantic properties are shared at different levels by words built by a common word formation process. A number of lexical resources for derivational morphology of modern languages have been made available over the past few years.6 The first steps towards constructing a lexicon based on word formation for Latin were made by Passarotti and Mambrini (2012), when they proposed a way to semi-automatically extract word formation rules from a list of lemmas, and pair each lemma to their morphologically simpler (i.e. non-derived) counterpart. A few years later, the Word Formation Latin project (WFL) received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No. 658332WFL, to expand on those efforts and create a derivational lexicon for Classical Latin. This lexicon was ultimately included in the automatic

3 See, for instance, Lemlat (Passarotti et al. 2017): www.lemlat3.eu (accessed: 20 October 2017). 4 TreeTagger, a stochastic tool that has been trained also on Latin (currently two parameter files for Latin can be found at www.cis.uni-muenchen.de/~schmid/tools/TreeTagger, accessed: 20 October 2017). 5 See the recent efforts described in Ponti and Passarotti (2016) to improve reliability of automatic syntactic parsing of the Index Thomisticus Treebank. 6 Noteworthy is Word Manager, a system for morphological dictionaries available for German, English and Italian (Domenig and Ten Hacken 1992). More recent times have seen the production of the lexical network DeriNet for Czech (Ševčíková and Žabokrtskỳ 2014), the derivational lexicon DErivBASE for German (Zeller et al. 2013), derIvaTario for Italian (Talamo et al. 2016) and the derivational and morpho-semantic resource Démonette for French (Hathout and Namer 2014).

226

Eleonora Litta & Marco Passarotti

lemmatiser for Latin Lemlat creating a more complete resource for the study of Latin morphology (Litta et al. 2016). The WFL lexicon can be browsed in an easy-to-use web application at wfl. marginalia.it, and it can be downloaded as part of the Lemlat 3.0 database from github.com/CIRCSE/LEMLAT3. This chapter presents the data contained in the WFL lexicon, the methodology employed to build it, and describes a few problems that can arise from trying to maintain consistency throughout the whole resource and the solutions that we have employed in order to overcome them. Plans for potential future enhancements to the resource are reported in the chapter’s conclusions.

2 Methodology From a theoretical point of view, Word Formation Rules (WFRs) in WFL were conceived according to the so called Item-and-Arrangement (IA) model, which considers word forms either as simple (non-derived) morphemes or as a sequence of morphemes (base and affixes) having both form and meaning (Hockett 1954). IA was chosen as the theoretical model supporting WFL for two main reasons: first, it emphasises the semantic significance of affixal elements as they are found in the lexicon; secondly, IA was the model adopted by other derivational lexica like Word Manager, after which WFL was designed. For what the construction of the word formation relationships is concerned, WFL uses a step-by-step morphotactic approach. Each word formation process is treated individually, and the lemma resulting from a WFR is usually richer (containing more morphemes) than the input, with the exception of conversion, which only involves a change of part of speech (PoS). Each output lemma can only have one source, except in the case of compounds, where it is possible to have two (or three) input lemmas for one output lemma. The lexical basis for WFL is the same as that of the morphological analyser and lemmatiser for Latin Lemlat which has been collated from three dictionaries covering Classical and Late Latin: Oxford Latin Dictionary (Glare 1982); Ausführliches lateinisch-deutsches Handwörterbuch (Georges 1913–1918); Laterculi vocum latinarum (Gradenwitz 1904). It contains 40,014 lexical entries and 43,432 lemmas (as more than one lemma can be part of the same

(When) inflection needs derivation: a word formation lexicon for Latin

227

lexical entry).7 Additionally, the lexical basis of Lemlat has recently been enriched with the integration of most of the Onomasticon (26,250 out of 28,178 lexical entries) contained in the Forcellini lexicon (Budassi and Passarotti 2016; Forcellini 1864–1926). Lemlat contains every string of characters required in the inflectional paradigm of each lemma, like the uninflected parts of irregular supines (duc-, ductfor duco ‘to lead’), or the stem of the genitive of imparisyllaba nouns and adjectives (crimen, crimin- ‘accusation’), fundamental for the automatic processing of WFRs, as well as including graphical variants, like obf-/off- in offero ‘to put oneself forward, cause to be encountered’ (Passarotti and Mambrini 2012). These strings of characters are used by Lemlat while morphologically analysing and lemmatising input word forms, which are automatically segmented into formative elements. Among these, the lexical element is called LES (for ‘LExical Segment’). This is the invariable part of the inflected forms, i.e. the sequence – or one of the sequences – of characters that remains the same in the inflectional paradigm of a lemma (hence, the LES does not necessarily match the word stem). For example, poet is the LES for the lemma poeta ‘poet’, as it is the sequence of characters that does not change in the different forms of the lemma poeta: poet-a, poet-ae, poet-am, poet-ae, poet-arum, poet-as, poet-is. Lemlat includes a LES archive, in which each LES is assigned a number of inflectional features. Among these, there is a tag for the gender of the lemma (for nouns) and a code (CODLES) for its inflectional category. For instance, the CODLES for the LES poet is N1E (first declension irregular nouns) and its gender is M (masculine). In the case of irregular nouns, as for poeta, there is also a field in the archive (LEM) containing information on how to recognise an irregular ending (poeta can sometimes appear with nom. sg. poetes) during the lemmatisation. WFL makes use of the LES archive together with a list of 69,682 lemmas automatically extracted from the Lemlat dataset. Both lists were added as tables to the relational database used while building WFL. In WFL, each morphologically derived lemma is assigned a WFR and is paired with its source lemma. All those lemmas that share a common (not derived) ancestor belong to the same ‘word formation family’. For instance, lemmas damno ‘to condemn’, damnosus ‘that causes loss’ and damnatio ‘condemnation’ all belong to the word formation family whose ancestor is the lemma damnum ‘loss’. WFRs are modelled as directed one-to-many input–output relations between lemmas.

7 These lexical resources, even if not as philologically accurate as the Thesaurus Linguae Latinae (TLL 1900–), have offered the opportunity to include lists of words for letters N and Q-Z, currently not yet available in TLL.

228

Eleonora Litta & Marco Passarotti

The WFL database was built in three steps. First, WFRs were found. Then, they were applied automatically to lexical data (which allowed to find additional WFRs). Lastly, all results of the automatic procedure were thoroughly checked manually in order to remove duplications, lacunas and inconsistencies. About half of affixal WFRs were found in a semi-automatic fashion by extracting from the list of lemmas the most frequent sequences of characters occurring on the left (prefixes) and on the right (suffixes) side of lemmas. All the remaining rules were found consulting existing literature on Latin derivational morphology (Fruyt 2011; Jenks 1911; Oniga 2007), or through empirical evidence from the data.8 The rules were recorded in a database table, where each rule is classified by type (derivation – prefixation, suffixation, conversion – and compounding), input and output PoS (V-to-A, N-to-N, A + V = V, etc.) and is assigned the required inflectional category and gender for its input and output lemmas. Compounding and conversion WFRs were listed by considering all the possible combinations of main PoS (verbs, nouns, adjectives). For instance, there are four possible types of conversion WFRs involving verbs: V-To-N (claudo > clausa; ‘to close’ > ‘cell’), V-To-A (eligo > elegans; ‘to pick out’ > ‘accustomed to select, tasteful’), N-To-V (magister > magistro; ‘master’ > ‘to rule’), A-To-V (celer > celero; ‘quick’ > ‘to quicken’). The classification of WFRs employed in WFL is based on Martinet’s traditional classification where, unlike in some more recent studies on Latin word formation (Benedetti 1988; Fruyt 2002 and Fruyt 2011), prepositions (e.g. cum ‘with’ or in ‘in’) are not considered compounding constituents, but prefixes (Martinet 1979: 235–243). In case of unclear affixations, we consider Oxford Latin Dictionary and Georges as primary reference manuals. The use of these two dictionaries insures consistency: because they are the main source for the lemmas included in WFL, all, even scarcely used, words in the lexical basis can be found in Oxford Latin Dictionary and Georges. Oxford Latin Dictionary especially contains guidance on the distinction between affixes and actual lemmas, and lemmas’ formative elements are clearly specified. This means that words including what Oxford Latin Dictionary considers a prefix, such as quadriennium ‘period of four years’ (quadri- ‘consisting of four of the things following’ + annus ‘year’, and not quatuor ‘four’ + annus) are included among prefixal derivations, while other similar lemmas formed by numerals, like sexennium ‘period of six years’, on the other hand, are 8 For example, when treating a fairly productive rule such as deverbal adjectives with suffix -bund-, we realised that a few of them were derived from nouns. We then added the rule N-To-V -bund- to the WFR list. In WFL, V stands for verbs, N for nouns, A for adjectives, PR for pronouns, and I for invariable lemmas (e.g. adverbs).

(When) inflection needs derivation: a word formation lexicon for Latin

229

labelled as N + N = N compounds, because Oxford Latin Dictionary categorises sex ‘six’ as a noun. Together with the list of WFRs, the main tables of the WFL database are the LES archive of Lemlat and the list of its lemmas (each assigned its PoS, inflectional category and, for nouns only, gender). Lemmas and WFRs are paired automatically using Structured Query Language (SQL) queries that match candidate input–output lemmas for each WFR. However, the automatic procedure is not sufficient for building the word formation families. The automation serves only to provide roughly filtered data that must be checked manually. Manual checking allows us to solve problems of overproduction and underproduction of candidate pairs. These duplicate results need to be analysed and rectified. There can be only one simple input lemma for each output lemma, and only one WFR associated with a derivative lemma. The most common reason behind overproduction of candidate pairs is homography. Homography can occur at different levels when we search for pairings. Let us consider, for example, the case of sero. During the application of prefixal rules to the list of lemmas, an SQL query returns two columns, respectively providing a list of candidate input lemmas and one of candidate output lemmas, selected by their PoS and sequence of characters. The candidate output is the same as the input with the addition of a string of characters, the prefix, preceding the same string of characters as it is displayed in the input. There are two 3rd conjugation verbs sero in Lemlat, with different inflections and meanings (1 ‘to plant’, 2 ‘to entwine’), and two verbs subsero (1 ‘to plant as a replacement’ and 2 ‘to insert below’). The simple pairing SQL query currently used does not discern which subsero comes from which sero, so it returns a list of all the possible combinations of input and output candidate lemmas. The distinction is performed manually on the basis of inflection, the details of which are deductible from the basis for perfectum recorded in Lemlat’s LES archive: sero (1), sevi, satus > subsero (1), subsevi, subsatus /sero (2), serui, sertus > subsero (2), subserui, subsertus. Apophony and assimilation, or obscure morphotactics, on the other hand, are the most common causes of underproduction.

3 WFL on the web In order to enable users to access WFL, we have developed a specific web application, where the relationships between lemmas of the same word formation

230

Eleonora Litta & Marco Passarotti

family are represented as a tree graph. In this graph, a node is a lemma, and an edge is the WFR used to derive the output lemma from the input one (or two/ three, in the case of compounds), along with any affix used. The entire database is thus like a big graph represented as a collection of edges, and the set of word formation families is simply the set of connected subgraphs. The website has been designed keeping in mind the kinds of queries and results that a linguist would be interested in. There are four distinct perspectives to query WFL: 1) By WFR: the primary interest is the behaviour of a specific WFR. For example, it is possible to view and download a list of all verbs that derive from a noun with a conversive derivation process (e.g. radix ‘root’ > radicor ‘to grow roots’); 2) By affix: it acts similarly as above, but works more specifically on affixal behaviour. For example, this perspective enables to retrieve all masculine nouns featuring the suffix -tor and to verify how many of them correspond to a female equivalent ending in -trix; 3) By PoS: the primary interest is the PoS of input and output lemmas. This view is useful for studies on macro-categories of morphological transformation, like nominalisation and verbalisation; 4) By lemma: it focuses on both derived and non-derived lemmas. It supports studies on the productivity of one specific morphological family or a set of morphological families. The results of these browsing options are of three types: a) lists of lemmas resulting from a query, that can be downloaded in a .txt file; b) derivational graphs: this type of graph represents the derivational chain (or cluster) for a specific lemma, which includes all the lemmas derived from the lemma selected, as well as all those it is derived from; c) a summary of the application of a given WFR to different PoS and the resulting lemmas. An important design aspect of the WFL web application is the fact that it limits queries that produce no results. Queries could produce no results if they search either for unattested WFRs, or for WFRs not yet included in WFL. Providing users with all the possible combinations of PoS, WFRs and affixes would result in quite long lists, thus requiring users to run single queries to find manually which of them have no occurrences in WFL. Instead, this can be easily inferred from the interface, as it is expected that one possible combination that is unavailable in the interface does not correspond to any word in WFL. For instance, the suffix -ace- is available only for denominal

(When) inflection needs derivation: a word formation lexicon for Latin

231

adjectives (N-To-A, argilla ‘clay’ > argillaceus ‘made of clay’), which means that it is not at work for all the other possible combinations of input–output PoS. In the web application, the four perspectives on queries mentioned above are implemented as four different screens, accessed via a top-level menu. For WFRs and affixes, the basic type (e.g. ‘Prefixation’ for WFRs, or ‘Prefixes’ for Affixes) is chosen via tab buttons, and for all perspectives the finer grained choices are specified via drop-down menus. The difference between querying the database by WFRs and by affixes is reflected in the priority of drop-down menus. For WFRs, first a WFR type (or types) is chosen (e.g. V-to-V for deverbal verbs), and then any desired affixes. The choice of the WFR type updates the second drop-down menu to restrict the affixes to just those that occur with the selected WFR type. A similar interaction holds for affixes. The PoS-based query option does not have an intermediate level of selection, but the choice is made via a series of drop-down menus. For each possible item involved in a WFR (one or two base input lemmas – the latter for compounds – and the output), there is the choice of PoS, and then refinements of that PoS: these are inflectional categories for all PoS (declension for nouns, classes for adjectives and conjugation for verbs), as well as gender for nouns. The options for the inflectional categories are limited to those appropriate for the PoS chosen. Querying WFL by lemma is performed by radio buttons, which allow for the selection of ‘all lemmas’, ‘only roots’ of derivational clusters (not derived lemmas) or ‘only derived lemmas’. The list of lemmas with their PoS (and gender, for nouns) is shown in a list, which can be filtered with the employment of common regular expression queries. The three types of query results are visualised in distinct ways in separate windows, interacting across the result types. Clicking on a lemma in the list opens its derivational graph in a separate window. In the graph of Figure 1, nodes are filled with lemmas and edges are labelled with affixes or input–output PoS (in the case of compounding and conversions). The selected lemma is shown inside a box. Clicking on any lemma in the graph replaces the current derivational graph with the one for the clicked lemma, moving the focus of the derivational trail. Clicking on an edge label in the derivational graph opens a new window (Figure 2) which provides a visualisation summarising the application of the corresponding WFR by PoS, a left-rooted tree, with the name of the affix as the root (first level of the tree), and all the combinations of the input and output PoS with their refinements (N1, N2, N3: first, second, third declension noun; A1: first class adjective)

232

Eleonora Litta & Marco Passarotti

abigo -(t)io(n) abactio

-ace

-(t)or abactor

V-To-N abactus

N1 ⇒ A1

12

N2 ⇒ A1

11

N3 ⇒ A1

9

Figure 1: Derivational graph of abigo ‘to drive away’.

Figure 2: Derivational Graph for WFR N-to-A -ace.

as second level branches, giving the number of lemmas for each input–output combination. The graph is collapsable so the user can focus on certain subsets only. As the subsets change, the list of lemmas is updated to reflect just the subsets that are selected. An additional feature of querying the lexicon by WFRs and by affixes is to search across the full derivational path of lemmas, thus providing results that go beyond the ‘outermost’ WFR. By selecting the ‘include as intermediate’ option, one can search not only for all the lemmas derived by a specific WFR but also for those that include at least one lemma produced by that WFR along their derivational path. For instance, with this option selected, among the results of a query that searches for deverbal adjectives formed with suffix -bil- is not only the adjective affabilis ‘that can be easily spoken’, but also the noun affabilitas ‘courtesy’ which has a deverbal adjective formed with suffix -bil- along its derivational path as it is derived from affabilis. All results can be downloaded: the list of lemmas as a tab-delimited text file, while the derivational graphs and WFR trees as images. Moreover, in the case of compounds, the user can choose whether to visualise or not both the roots of compounds, thus resulting in a multi-tree graph rather than a simple tree. It has to be remembered that, while searching by WFRs or PoS, a few peculiarities of the Lemlat lexical basis can result in a rather unconventional classification of the rules, which impacts especially (but not solely) on searches performed on compounds. For instance, participial adjectives are not included in the Lemlat lexical basis, because they are considered part of the verbal paradigm. This means that certain compounds that would be expected to have an adjective (A) as one of

(When) inflection needs derivation: a word formation lexicon for Latin

233

their constituents have a verb (V) instead, e.g. adjective altisonus (altus + sono) ‘that sounds high up, sublime’ can be found among V + V = A compounds rather than among A + V = A. Also, adjectival adverbs are considered in Lemlat adverbial cases of the adjectival inflection, hence a word such as dulciloquus (dulce + loquor) ‘sweet talking’ is to be found among A + V = A, rather than I + V = A.

4 Theoretical issues As mentioned above, WFRs are conceived according to the IA model. This means that the construction of the lexicon favours a morpheme-based approach to morphology. Moreover, this model is put in practice through the use of input–output relationships between lemmas, in a morphotactic approach that has been prioritised over philological considerations. The use of directed edges in our derivational trees implies that one word formation process has happened before the other, and that one given lemma specifically derives from another. The main issue in the development of a derivational morphology resource for a language like Latin – even though WFL focusses on Classical Latin – is that of the diachronic distribution of the lexicon, together with the fact that there are no native speakers to help with considerations about the transparency or opacity of a dubious derivation. As a result, we are forced to commit to a clear-cut work policy in order to ensure consistency throughout the lexicon. Three major factors are considered when in doubt: a) theoretical statements (Unitary Base Hypothesis,9 Item-and-Arrangement) and previous research on word formation; b) dictionaries (Oxford Latin Dictionary and Georges): we consider whether or not they support our analysis; c) semantic motivations. Overall coherence in the design of the lexicon is what generally drives decisions, but case-by-case analysis is necessary in order to establish a motivated derivational graph.

9 According to the Unitary Base Hypothesis, WFRs can only operate over a single type of syntactically or semantically defined base (Aronoff 1976; Scalise 1983).

234

Eleonora Litta & Marco Passarotti

4.1 Compounding or suffixation? While treating certain lemmas, we had to take a decision on whether they are the result of compounding or whether certain second constituents need to be considered as suffixoids. For example, -fex, -fico and -ficus (from base form fec- from facio ‘to make’) are considered suffixes denoting ‘making’ in Oxford Latin Dictionary. Yet, against Oxford Latin Dictionary’s position, we have chosen to consider -fex and -fico as second constituents for compounds, following authoritative bibliography on the subject (Brucale 2012; Jenks 1911; Oniga 1988 and Oniga 1992). For what -ficus (again from facio, forming adjectives) is concerned, on the other hand, we have taken a non-uniform approach: it was decided to treat certain lemmas ending in -ficus that had a corresponding verb ending in -fico, as V-to-N conversions instead. The reasoning behind this decision originated from the fact that, in general, linguistic research on Latin word formation considers compounding – comparatively to what happens in other Indo-European languages – as a poorly productive phenomenon (Fruyt 2002). For this reason, one can imagine a different formation process to have happened instead. In the case of -fico and -ficus, for example, we hypothesise the verbal compound to have formed before the adjective, as the main meaning of such compounds is almost always the result of a performed action (damnificus ‘that causes loss’ < damnifico ‘to cause loss’ < damnum + facio). This also means that rather than having another separated compounding rule for damnificus, which would have connected the two lemmas at a higher, more distant, level (from damnum and facio, but not with each other), the two lemmas are directly connected in the same word formation family. However, out of 92 adjectives ending in -ficus contained in the lexical basis, 65 do not have a -fico verb counterpart to be connected to through conversion, hence they are considered the result of compounding. This has ultimately created an inconsistency in the resource, that needs to be either rectified or justified theoretically.

4.2 Prefixation or conversion? Or ‘the chicken or the egg’ dilemma Occasionally, we had to decide whether to prioritise one derivation process over another. In certain cases, it is difficult to ascertain whether prefixation happened before or after conversion – or suffixation – diachronically. Let us consider, for example, the choice between the following two options: 1) sono ‘to make a noise’ > resono ‘to produce a prolonged sound’ > resonus ‘reverberating/echoing’, or 2)

(When) inflection needs derivation: a word formation lexicon for Latin

235

sono > sonus ‘sounding’ > resonus. In this case the derivation trail 1) is preferred over 2), as it absolves all three of the deciding factors outlined above. Generally speaking, conversion and prefixation could have happened at any point of the derivation process, but when trying to establish whether the prefixation process happened before or after certain conversive processes, we tend to give precedence to verbal prefixation, as it appears more productive (Fruyt 2002; Oniga 1992).10 Georges avails this hypothesis by stating that resonus comes from resono, and semantically the meaning of ‘reverberating/echoing’ seems to be more likely an inheritance from the verb meaning ‘to produce a prolonged sound’, rather than merely from the adjective ‘sounding’. In a similar way, supported by all or a combination of the three factors above, the derivation properus ‘quick’ > propero ‘to hurry’ > depropero ‘to hasten’ > deproperus ‘hastening’ is chosen over properus > deproperus. On the other hand, notice how, for semantic reasons, sonus > absonus ‘of unpleasant sound’ > absono ‘to have an unpleasant sound’, as the semantic emphasis for the verbal counterpart here seems to derive from the adjective. However, sometimes, when looking into prioritising prefixation over suffixation, relying on what the dictionary tells us makes things uneven within the same word formation family. See for example the case of horreo ‘to become stiff/ tremble (with fear)’ > abhorreo ‘to recoil from’ > abhorresco ‘to become disgusted’, but horreo > horresco ‘to stand up stiffly / become agitated’ > inhorresco ‘to become stiff (with cold)’ described in Budassi and Litta (2017).

4.3 Fictional entries Another phenomenon that required an unconventional solution was the presence of parasynthetic derivations formed through the simultaneous addition of a suffix and a prefix, or a prefix and conversion. See for example indolesco ‘to feel painful’, which, according to the Oxford Latin Dictionary, is formed by in + doleo + sco, or bicolor ‘of two colours’, an adjective resulting from prefixation of noun color ‘colour’ with prefix bi- indicating something ‘consisting of two things’. Bicolor has been inserted into WFL simply as the result of a prefixal WFR involving a change of PoS (N-To-A), while indolesco has created more complicated issues. The project’s morphotactic approach forced us to create a number of fictional lemmas in order to fill a gap in the derivation tree. These are not reconstructed lemmas, and are not expected to have ever existed, but only serve as

10 In WFL there are 4,829 prefixed V-To-V vs. 1,253 V-To-N and 176 V-To-A conversions.

236

Eleonora Litta & Marco Passarotti

a ‘mechanical’ connection between two lemmas. For instance, *indoleo only acts as a trait d’union between doleo ‘to feel pain’, and indolesco ‘to feel painful’, hence creating two formative processes instead of one.

4.4 Backformation Backformation is a derivation process that involves the removal of an affix (or a supposed one), so that a new word is created by analogy with similar looking existing ones, as for example galeo ‘to equip with a helmet’ < galeatus (adj.) ‘wearing a helmet’, grego ‘to collect into a flock’ < aggrego/congrego ‘to bring people together’, irascor ‘to be angry’ < iratus ‘angy’, (g)nascor ‘to be born’ < natus ‘born’. At the time of writing, these and other backformation processes are not marked in WFL, but are portrayed as follows: galeo < galea ‘helmet’, galeatus < galea, aggrego/congrego < grego, but iratus > irascor (A-To-V -sc-), (g)nascor = root verb (i.e. not derived). In the future, we are planning on adding backformation as a de facto WFR, and mark backformations to the tree graphs so that the edge can point in the opposite direction, or be recognisable by a different colour.

5 Conclusions and future work In this chapter, we have introduced and described Word Formation Latin, a new language resource for Latin derivational morphology. WFL contributes to fill a gap in the world of computational lexical resources for Latin. Not only it offers a practical interface for deeper investigations into Classical Latin derivational morphology, but by being integrated in the morphological analyser for Latin Lemlat, it turns it into an even more powerful tool for the automatic processing of Latin texts. The large-scale exploration of Latin word formation provided by WFL enables scholars to support their investigations and conclusions with comprehensive and replicable empirical evidence (see, for example, the study on productivity of affixal rules described in Litta et al. 2017). On the other hand, the resource has potential for improvement under a few aspects. First, the addition of medieval lexical data, which is currently underway, from the Medieval Latin lexicon of Du Cange (1883–1887), would allow to broaden the lexical coverage of the resource across different chronological periods.

(When) inflection needs derivation: a word formation lexicon for Latin

237

Another very interesting development that is being planned for WFL is its inclusion into the software for textual reuse detection TRACER (Büchler et al. 2014),11 to test the potentials of derivational morphology information in the analysis of the use of cognate words as echoing lexical choices, stretching WFL’s use to the domain of philological research.

Abbreviations A (WFL) CODLES (WFL) I (WFL) IA LES (WFL) NLP PoS PR (WFL) SQL WFL WFR V (WFL)

adjective code for the inflectional category of a lemma in Lemlat invariable lemma Item-and-Arrangement Lexical Segment Natural Language Processing part of speech pronouns Structured Query Language Word Formation Latin Word Formation Rule verb

Bibliography A Latin Dictionary by Lewis and Short: www.perseus.tufts.edu (accessed: 20 October 2017). Ancient Greek and Latin Dependency Treebank: perseusdl.github.io/treebank_data (accessed: 20 October 2017). Index Thomisticus Treebank: itreebank.marginalia.it (accessed: 20 October 2017). PROIEL Treebank: proiel.github.io (accessed: 20 October 2017). Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. https:// www.degruyter.com/view/db/tll (accessed: 4 June 2018). TreeTagger: www.cis.uni-muenchen.de/~schmid/tools/TreeTagger (accessed: 20 October 2017). Aronoff, M. (1976). Word formation in generative grammar. Cambridge, MA: Massachusetts Institute of Technology Press. Benedetti, M. (1988). I composti radicali latini: Esame storico e comparativo. Pisa: Giardini. Brucale, L. (2012). Latin compounds. Probus, 24, 93–117. Büchler, M., Burns, P. R., Müller, M., Franzini, E. & Franzini, G. (2014). Towards a historical text re-use detection. In C. Biemann & A. Mehler (Eds.), Text mining, theory and applications of natural language processing (pp. 221–238). Cham: Springer.

11 www.etrap.eu/research/tracer/ (accessed: 20 October 2017).

238

Eleonora Litta & Marco Passarotti

Budassi, M. & Passarotti, M. (2016). Nomen omen: Enhancing the Latin morphological analyser Lemlat with an onomasticon. In Proceedings of the 10th SIGHUM Workshop on Language Technology for Cultural Heritage, Social Sciences, and Humanities(LaTeCH 2016) (pp. 90–94). Berlin: The Association for Computational Linguistics. Budassi, M. & Litta, E. (2017). In trouble with the rules: Theoretical issues raised by the insertion of -sc- verbs into Word Formation Latin. In E. Litta & M. Passarotti (Eds.), Proceedings of the workshop on resources and tools for Derivational Morphology (DeriMo) (pp. 15–26). Milano: Educatt. Domenig, M., & Hacken, P. ten (1992). Word Manager: A system for morphological dictionaries (vol. 1). Hildesheim: Georg Olms Verlag. Du Cange, C. (1883–1887). Glossarium Mediae et Infimae Latinitatis. Niort: L. Favre. Forcellini, E. (1864–1926). Lexicon Totius Latinitatis. 4th rev. ed. F. Corradini & J. Perin. 4 vols. Padova: Typis Seminarii. Fruyt, M. (2002). Constraints and productivity in Latin Nominal Compounding. Transactions of the Philological Society, 100(3), 259–287. Fruyt, M. (2011). Word-Formation in Classical Latin. In J. Clackson (Ed.), A Companion to the Latin Language (pp. 157–175). Chichester: Wiley-Blackwell. Georges, K. E. (1913–1918). Ausführliches lateinisch-deutsches Handwörterbuch aus den Quellen zusammengetragen und mit besonderer Bezugnahme auf Synonymik und Antiquitäten unter Berücksichtigung der besten Hilfsmittel ausgearbeitet (8. verb. und verm. Aufl. von H. Georges,2 Bde.). Hannover: Hahn. Glare, P. G. W. (1982). Oxford Latin Dictionary. Oxford: Oxford University Press. Gradenwitz, O. (1904). Laterali vocum Latinarum. Leipzig: Hirzel. Hacken, P. ten (2002). Word formation and the validation of lexical resources. In M. González Rodríguez & C. Paz Suárez Araujo (Eds.), Proceedings of the Third International Conference on Language Resources and Evaluation (LREC 2002) (pp. 935–942). Las Palmas, Canary Islands, Spain: European Language Resources Association. Hathout, N. & Namer, F. (2014). Démonette, a French derivational morpho-semantic network. Linguistic Issues in Language Technology, 11(5), 125–168. Hockett, C. F. (1954). Two models of grammatical description. Morphology: Critical Concepts in Linguistics, 1, 110–138. Jenks, P. R. (1911). A manual of Latin word formation for secondary schools. Boston/NewYork/ Chicago: DC Heath & Company. Litta, E., Passarotti, M. & Culy, C. (2016). Formatio formosa est: Building a word formation lexicon for Latin. In Proceedings of the Third Italian Conference on Computational Linguistics(CLiC–it 2016) (pp. 185–189). Napoli: Accademia University Press. Litta, E., Passarotti, M. & Ruffolo, P. (2017). Node formation: Using networks to inspect productivity in affixal derivation in Classical Latin. In Proceedings of the 2nd International Conference on Digital Access to Textual Cultural Heritage, DATeCH2017 (pp. 103–108). New York, NY: Association for Computing Machinery. Martinet, A. (1979). Grammaire fonctionelle du français. Paris: Didier. Oniga, R. (1988). I composti nominali latini: Una morfologia generativa. Bologna: Pàtron. Oniga, R. (1992). Compounding in Latin. Rivista Di Linguistica, 4, 97–116. Oniga, R. (2007). Il latino: Breve introduzione linguistica. Milano: Franco Angeli. Passarotti, M. & Mambrini, F. (2012). First steps towards the semi-automatic development of a word formation-based lexicon of Latin. In Proceedings of the Eighth International

(When) inflection needs derivation: a word formation lexicon for Latin

239

Conference on Language Resources and Evaluation (LREC 2012) (pp. 852–859). Istanbul: European Language Resources Association. Passarotti, M., Budassi, M., Litta, E. & Ruffolo, P. (2017). The Lemlat 3.0 package for morphological analysis of Latin. In G. Bouma & Y. Adesam (Eds.), Proceedings of the NoDaLiDa 2017 Workshop on Processing Historical Language (pp. 24–31). Gothenburg: NEALT Proceedings Series 32. Ponti, E. M. & Passarotti, M. (2016). Differentia compositionem facit: A slower-paced and reliable parser for Latin. In Proceedings of the Tenth International Conference on LanguageResources and Evaluation (LREC 2016) (pp. 683–688). Portorož, Slovenia: European Language Resources Association. Scalise, S. (1983). Morfologia lessicale. Padova: Clesp. Ševčíková, M. & Žabokrtskỳ, Z. (2014). Word-formation network for Czech. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14) (pp. 1087–1093). Reykjavik: European Language Resources Association. Talamo, L., Celata, C. & Bertinetto, P. M. (2016). DerIvaTario: An annotated lexicon of Italian derivatives. Word Structure, 9(1), 72–102. Zeller, B. D., Snajder, J. & Padó, S. (2013). DErivBase: Inducing and evaluating a derivational morphology resource for German. In Proceedings of the Annual Meeting of the Association for Computational Linguistics (pp. 1201–1211). Sofia: Association for Computational Linguistics.

Marco Budassi, Eleonora Litta & Marco Passarotti

What’s beyond ‘inchoatives’? Derivation types on the basis of -sc- verbs Abstract: This chapter presents an investigation on the productivity of words derived from -sc- verbs attested in Classical Latin, as a way of adding to the rich debate on their semantic significance. The work is performed through the use of empirical data provided by the Word Formation Latin lexicon, a derivational morphology resource for Classical and Late Latin. After a brief introduction on the issues regarding -sc- verbs, the derivations with -sc- verbs in input are taken into consideration and analysed by word formation type. The results of this examination are discussed with special emphasis on quantitative as well as diachronic and semantic data. Keywords: Latin linguistics, -sc- verbs, word formation, derivational morphology

1 Introduction The class of -sc- verbs (also called inchoative verbs) is remarkably broad in Latin. Traditionally, such verbs were considered dynamic/intransitive counterparts derived from stative/transitive base verbs (e.g. augesco ‘I grow’, intransitive < augeo ‘I increase’, transitive, Da Tos 2013), or descriptions of the beginning of a situation (e.g. calesco = calere incipio ‘to become warm’, Viti 2015). In fact, Latin -sc- verbs are much less homogeneous. Sometimes, -sc- is part of archaic verbs roots (e.g. pasco ‘I feed’, which describes a non-terminative process of eating < PIE *peh2-); other -sc- verbs are back-formed from participles (e.g. nascor ‘I am born’ < natus ‘born’). Furthermore, -sc- verbs are not only deverbal: quite a number of -sc- verbs are derived from adjectives (e.g. albesco ‘I grow white’ < albus ‘white’) as well as from nouns (e.g. puellasco ‘I become like a girl’ < puella ‘girl’). For all these reasons, the label ‘inchoative’ usually employed to identify -sc- verbs has been in some dispute. Haverling (2000) provides a rich treatise on the semantic discrepancies existing among different phases of the Latin language as far as -sc- verbs are concerned, and highlights the differences occurring between unprefixed and prefixed -sc- verbs. The issue is thus quite tangled, but treated at large. However, an Marco Budassi, Università di Pavia Eleonora Litta and Marco Passarotti, Università Cattolica del Sacro Cuore https://doi.org/10.1515/9783110647587-016

What’s beyond ‘inchoatives’?

241

analysis of all lemmas morphologically derived from -sc- verbs could perhaps offer a suitable bedrock to go further in this matter. Such investigation is made possible through the use of the Word Formation Latin (WFL) lexicon (Litta et al. 2016; Litta and Passarotti, this volume), which records all word formation processes active in Classical and Late Latin, connecting lemmas on the basis of word formation rules (WFRs). Using the exhaustive empirical data provided by WFL, we can attempt to make an evaluation of the productivity of -sc- verbs, as well as of their impact on the whole Classical and Late Latin lexicon. The chapter is organised as follows: Section 2 provides an outline of the methodology used in this study. In Section 3 derivations with -sc- verbs in input are detailed and categorised. In Section 4, the results of our analysis are discussed with special emphasis on quantitative as well as diachronic and semantic data.

2 Methodology In order to perform our analysis, we have used data from the word formation–based lexicon of Latin WFL, a derivational morphology resource for Classical and Late Latin. The contents of WFL are lemmas connected by relationships established on the basis of word formation rules (WFRs).1 The lexical basis used for building WFL matches the one provided by the morphological analyser for Latin Lemlat 3.0 (Passarotti et al. 2017), originally counting 43,432 lemmas from three authoritative Classical Latin dictionaries (Oxford Latin Dictionary, Glare 1982; Ausführliches lateinisch-deutsches Handwörterbuch, Georges 1913–1918; Laterculi vocum latinarum, Gradenwitz 1904) and recently enlarged by the addition of 26,250 onomastic lemmas from Forcellini’s Onomasticon (Budassi and Passarotti 2016). These dictionaries are usually considered containing only Classical Latin. However, Georges also includes a significant number of lemmas from later periods. Moreover, as shown in Passarotti et al. (2017), Lemlat’s coverage of wordforms throughout Latinity has been proven to be fairly high.2 For this

1 WFL has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 658332-WFL. 2 In order to evaluate Lemlat’s coverage potentials, a list of all wordforms from all periods of Latin extracted from the Thesaurus Formarum Totius Latinitatis – CETEDOC Index of Latin Forms (TF-CILF) database (Tombeur 1998) in CD-ROM format, were extracted and lemmatised with Lemlat 3.0. In this process, 72.25% of all wordforms from the CD-ROM database were lemmatised. However, a look at the forms that were not analysed (hence not included in the lemmatiser’s lexical basis), revealed that among the 153,447 forms not analysed by Lemlat, there are mainly

242

Marco Budassi, Eleonora Litta & Marco Passarotti

reason, we believe the lexical basis used to build WFL to be sufficiently representative of both Classical and Late Latin word formation processes. In WFL, all those lemmas that share a common (non-derived) ancestor belong to the same word formation family. Each relationship in the lexicon corresponds to a different type of WFR. Each word formation family is represented in the WFL web application3 as a tree-graph, where a node is a lemma, and an edge is the WFR used to derive the output lemma from the input one (or two, in the case of compounds), along with any affix used. By using the data contained in WFL, we have built some network visualisations that exemplify how derivation works around -sc- verbs (Figure 1). Each node in the network corresponds to a specific set of lemmas that undergo the same word formation process (e.g. the node labelled ‘V-to-V sc’ contains all those -sc- verbs that are derived from other verbs). In order to perform a quantitative survey of the frequency distribution of lemmas derived from -sc- verbs, we have compared derivation data with data extracted from Brepols’ Library of Latin Texts – Series A (LLT). LLT is a database collecting more than 3,200 Latin texts ranging from antiquity to the literature of the patristic period, from a vast body of medieval material and from collections of Neo-Latin works. LLT enables to search through texts also by wordform, which proves particularly useful when performing studies on the productivity of certain word formation processes. In this study, we have limited our searches to the periods of ‘Antiquitas’, ‘Aetas Patrum I’ and ‘Aetas Patrum II’, which roughly correspond to the period of Latinity covered by the Thesaurus Linguae Latinae.4 For what -sc- verbs are concerned, we have used Haverling’s classification of such verbs in four categories (Haverling 2000): 1) category 1: archaic verbs formed directly from stems or participles with the help of the -sc- suffix (e.g. cresco ‘I grow’), as well as verbs derived by adding the enlarged suffix -(n)isco(r) (e.g. fruniscor ‘I make use of something’), and backformations from participles (e.g. gnascor < natus) or from verbs no longer in use (e.g. apiscor ‘I reach’);

sequences of letters (e.g. aaa), numbers (e.g. CCC), and extremely rare word forms (e.g. aaliza, 1 occurrence in the Middle Ages). Hence, in terms of textual occurrences, which is a much more reliable and relevant unit of analysis, Lemlat managed to analyse 98.345% of the occurrences in the texts of the corpus the TF-CILF is based on. 3 The WFL lexicon can be queried online at wfl.marginalia.it (last accessed: 02 May 2018), and its source database can be downloaded from github.com/CIRCSE/WFL. 4 Although the TLL represents the best and most accurate lexical resource for Latin, for the purposes of this study we had to draw our conclusions by relying on other resources, due to the unavailability of a significant portion of data for letters Q-Z in the TLL.

What’s beyond ‘inchoatives’?

243

2)

category 2: intransitive verbs derived from transitive verbs (e.g. inveterasco ‘I get old’ < invetero ‘I allow to become old’); 3) category 3: dynamic verbs derived from stative verbs (e.g. acesco ‘I get sour’ < aceo ‘I am sour’); 4) category 4: -sc- verbs derived from adjectives (e.g. siccesco ‘I get dry’ < siccus ‘dry’) or nouns (e.g. puellasco ‘I become like a girl’ < puella ‘girl’).

3 Derivations from -sc- verbs Figure 1 shows all the derivation processes that have -sc- verbs as input in a network-like representation. This visualisation represented a starting point from which further considerations were drawn.

V–To–A iu V–To–A ul

A–To–V sc V–To–V i

V–To–A bil

V–To–V it V–To–V sc V–To–V uri

N–To–V sc

V–To–N ul

V–To–N (t)or

V–To–N (t)io/(t)ion V–To–N i

Figure 1: Latin derivational patterns from -sc- verbs.

The three greyed-out nodes represent deverbal (‘V-to-V sc’), denominal (‘N-to -V sc’) and deadjectival (‘A-to-V sc’) -sc- verbs contained in the WFL lexical basis. The derivational patterns adding a suffix to one of these three -sc- verbs categories are represented as edges pointing to non-highlighted nodes. Prefixed -scverbs are excluded from this analysis due to the blurred lines around when and how these were formed (see Haverling 2000; Crocco Galèas and Iacobini 1993; Budassi and Litta 2017). In addition, -sc- verbs like pasco are not represented in Figure 1, as their -sc-suffixation dates back to pre-attested phases of Latin or

244

Marco Budassi, Eleonora Litta & Marco Passarotti

even to Proto-Indo-European (see Section 1). Thus, they are not considered derived -sc- verbs in WFL. -sc- verbs were the base for the production of verbs, nouns and adjectives, although the distinction between the latter two is often unclear and hard to demonstrate when reading the sources.

3.1 Verbs derived from -sc- verbs ‘V-to-V -sc-’ verbs are the input of three derivational processes producing verbs in output: ‘V-to-V ’ (-ent-i-o/or), ‘V-to-V (i)t’ (-it-o) and ‘V-to-V ur’ (-ur-io). In WFL, only one verb is derived from a -sc- verb by adding the suffix -i-: adulescentior ‘to behave in a youthful manner’. Adulescentior derives from the present participle of adulesco ‘to grow, to develop’, adulescens. In the Lemlat lexical basis (hence in WFL) adulescens is not included as a lemma, because it is considered part of the inflectional paradigm of adulesco. This is due to a methodological choice adopted by the developers of Lemlat: participles are characterised by a ‘double’ nature, which collocates them halfway through between verbs and nouns or verbs and adjectives. For this reason, adulescens and similar cases are treated in the lexical basis of Lemlat as verbs, and are lemmatised under the verbal lemma (in this case, adulesco). This means that, in WFL, adulescens > adulescentior is considered among V-to-V derivations rather than a N-to-V (or A-to-V). A major problem raising with adulescentior consists in the fact that this lemma can either be a form of the above-mentioned verb adulescentior, or a comparative form of participle adulescens. Therefore, in analysing data, a context-by-context disambiguation must be performed. On the basis of LLT, only one instance of the verb adulescentior, as second indicative present adulescentiaris, can be found in Varro (1). (1)

quidem ut facias censeo, quoniam tu quoque adhuc adulescentiaris (VARRO Men. 550)

All other instances of adulescentior or adolescentior are the comparative form of participle adulescens. The ‘V-to-V (i)t’ derivation process was more productive, see e.g. sciscitor/o ‘to try to get to know’ < scisco ‘to get to know’. Four other cases belong to this class: discito ‘to acquire knowledge’ < disco ‘to try to learn’, noscito ‘to investigate’ < nosco ‘to find out’, pascito ‘to graze’ < pasco ‘to feed’, and vescitor ‘to have, to enjoy’ < vescor ‘to make use of, to enjoy’. Of these, the last two have

What’s beyond ‘inchoatives’?

245

no occurrences in LLT for what antiquity and both patristic eras are concerned. Interestingly, in these four cases, all the verbs in input to the ‘V-to-V (i)t’ derivation belong to Haverling’s Category 1 verbs. As seen in Section 2, verbs belonging to this category are rather particular and old (they date back earlier than 100 BC), as their derivation from a non -sc- verb is archaic and unattested.5 Only one verb is derived through a ‘V-to-V ur’ derivation process, adulescenturio ‘to want to behave in a youthful manner’.6 Table 1 summarises the number and types of V-to-V derivations with a -scverb in input, and specifies their frequency of occurrence (‘Tokens’) extracted from LLT. Table 1: V-to-V derivations with a -sc- verb in input. Input WFR

Output WFR

Lemma

V-to-V -sc- verb

V-to-V i

adulescentior

V-to-V -sc- verb

V-to-V (i)t

sciscitor/o (discito) noscito pascito vescitor

V-to-V -sc- verb

V-to-V ur

adulescenturio

Tokens

WFR frequency





    







The ‘V-to-V (i)t’ process is by far the most productive way of deriving other verbs from -sc- verbs. This can be claimed both on the basis of the number of attested verbs which are derived according to that pattern (5), and on the basis of the frequency of occurrence of such forms in LLT (360). Table 1 displays also two verbs, pascito and discito, which are not attested in antiquity and the patristic eras if looking through LLT, but they appear in Georges and TLL with quotations respectively from Varro (VARRO rust. 3, 16, 19 feras apes dico, quae in silvestribus loci pascitant) and Cyprianus Gallus (CYPR. GALL. exod. 698 dum causas discitat omnes). Although these quotations appear to be disputed

5 Some of them (e.g. nosco) have even direct relatives with the -sc- suffix in other IndoEuropean languages (e.g. Ancient Greek γιγνώσκω ~ Latin nosco). 6 Once again, this verb derives from adulescens, so it could also be considered as the outcome of a N-to-V derivational process.

246

Marco Budassi, Eleonora Litta & Marco Passarotti

philologically, both lemmas appear in Georges, hence they are part of our lexical basis. In TLL, pascito is understood to be a spelling variant of pastito, iterative form of pascere. Discito, conversely, is probably a false transmission (attested in cod. A; cod. C has dehoscitat, corrected by Peiper in his edition in noscitat). Two different strategies must hence be adopted in the respect of these two lemmas : pascito can be included in the analysis, while discito must be rejected. This is the reason why discito is in brackets in Table 1 above. Note that in Table 1 pascito is attested by 0 tokens. This, although misleading, is due to the fact that values in Column ‘Tokens’ represent the number of attestations of lemmas in LLT, regardless of further occurrences of such lemmas in other resources. Given that pascito belongs to our lexical basis, however, it is listed as a ‘V-to-V (i)t’ derived type.

3.2 Nouns (and participles) derived from -sc- verbs Productive nominalisations from -sc- verbs involve the suffixes -(t)or, -(t)io/(t) ion- and -i- (Table 2 and 3). The only -(t)or noun derived from a -sc- verb is aeruscator ‘beggar’ < aerusco ‘to beg’. In Haverling’s (2000) view, aerusco has a non-terminative meaning (‘to go begging for money’) and dates back to the period which goes from 100 to 200 AD.7 The only ‘V-to-N (t)io’ is coruscatio ‘lightening’ < corusco ‘to make a rapid movement, to flash’, a rather archaic verb, which precedes the first century BC (Haverling 2000).

Table 2: Summary of V-to-N derivations with a -sc- verb in input (‘V-to-N i’ lemmas excluded; see Table 3). Input WFR

Output WFR

Lemma

Tokens

WFR frequency

V-to-V -sc- verb

V-to-N (t)or

aeruscator





V-to-V -sc- verb

V-to-N (t)io/(t)ion

coruscatio





7 Nonetheless, the origin of this verb is rather obscure.

What’s beyond ‘inchoatives’?

247

From WFL data it turns out that the most productive V-to-N derivation attested in Latin is by far the ‘V-to-N i’ type. A portion of ‘V-to-N i’ derived nouns is formed from Haverling’s Category 1 -sc- verbs: i.e. agnoscentia ‘awareness’ < agnosco ‘to recognise’ < nosco ‘try to get to know’ < PIE *gneh3. A second subgroup of ‘V-to-N’ derived nouns is based on -sc- verbs that are either denominal or deadjectival. Two nouns are part of this subgroup: irascentia ‘fury’ < irascor/o ‘to get angry’ < iratus ‘anger’ through a process of backformation (Haverling 2000), and quiescentia ‘repose’ < quiesco ‘to repose in sleep’ < quies ‘the repose of sleep’. Nevertheless, most of ‘V-to-N i’ nouns are derived on the basis of deverbal -scverbs, e.g. concrescentia ‘coagulation’ < concresco (< con- + creo + -sc-). However, while looking at the occurrences of these nouns in the database, their nominative, vocative and ablative singular can often be confused with present participle neuter plural direct cases of their original verb. For instance, compare the role of the occurrences of decrescentia to be found in antiquity, in examples (2) and (3) below: (2)

inter solem vero et lunam cum distet totum mundi spatium et lunae orienti sol trans contra sit ad occidentem, eam, quo longius absit, a radiis remissam xiiii die plena rota totius orbis mittere splendorem, reliquos que dies decrescentia cotidiana ad perfectionem lunaris mensis versationibus et cursu a sole revocationibus subire sub rotam radios que eius, et ita menstruas dierum efficere rationes. (VITR. 9, 2, 2)

(3) diffugere nives, redeunt iam gramina campis arboribus que comae; mutat terra vices et decrescentia ripas flumina praetereunt. (HOR. carm. 4, 7, 3) In the case of other lemmas, there is no ambiguity, as shown by examples (4) and (5): both instances of erubescentia, from the first patristic era, can be recognised as being forms of the derived noun in -i-: (4) illi qui in partibus uerecundioribus corporis contracta uexatione conscientiam medentium uitant et ita cum erubescentia sua pereunt. (TERT. paenit. 10, 1) (5) nam est quaedam confusio temporalis utilis, perturbatio animi respicientis peccata sua, respectione horrentis, horrore erubescentis, erubescentia corrigentis. (AVG. in psalm. 30, 2 serm. 1, 5 p. 232) For this reason, the frequency values in Table 3 are split between occurrences that are surely forms of the noun (Column ‘Non-ambiguous occurrences’; e.g.

248

Marco Budassi, Eleonora Litta & Marco Passarotti

Table 3: V-to-N -i- derivations with a -sc- verb in input. Output WFR

Lemma

V-to-N i

adulescentia adolescentia agnoscentia cognoscentia concrescentia concupiscentia convalescentia crescentia decrescentia dinoscentia discentia erubescentia excandescentia excrescentia expavescentia ignoscentia irascentia nascentia noscentia quiescentia refrigescentia reminiscentia resipiscentia

Non-ambiguous occurrences

Ambiguous occurrences

                      

                      

adulescentiam), and occurrences that could be either a nominative, vocative or ablative singular noun, or a neuter plural direct case of present participle (Column ‘Ambiguous occurrences’; e.g. adulescentia). Once again, where we have value 0 on the table, it means that the word is included in the source dictionaries, but not attested in LLT. See, for example, agnoscentia in (6): (6) ut in te uno (deo) et invisibilitatis plenitudo, quod pater filio, et visibilitas agnoscentiae, quod filius patri in operatione sancti spiritus deberet, ageretur. (PRISCILL. tract. 11, 142)

What’s beyond ‘inchoatives’?

249

3.3 Adjectives derived from -sc- verbs Three types of adjectives are derived from -sc- verbs. These are: -iv- (‘V-to-A iv’),8 -ul- (‘V-to-A ul’) and -bil- (‘V-to-A bil’) (Table 4). There are two lemmas for adulescentulus in WFL, one is the masculine noun adulescentulus, -i ‘a young man’, the other is the adjective adulescentulus, -a, -um ‘very young’. As for the nouns and participles ending in -ia above, there are a certain number of occurrences that are undoubtedly adjectival (these are 494 for adulescentulus and 113 for adolescentulus), while a series of occurrences cannot be distinguished unless a careful reading of all sources containing them was performed (and even then, ellipsis might make the distinction impossible). The only productive way of deriving adjectives from -sc- verbs in Latin adds suffix -bil- to the base verb. There are three subgroups of derivations in this class. The first one is straightforward, e.g. concupiscibilis ‘desirable’ < concupisco ‘to desire ardently’. The second sub-class of ‘V-to-A bil’ adjectives comes from Haverling’s Category 1 -sc- verbs: agnoscibilis ‘recognisable’ < agnosco ‘to recognise’ < nosco ‘try to get to know’ < PIE *gneh3. There is a third subgroup of ‘V-to-A bil’ adjectives made of only one deverbal adjective, the base verb of which is not deverbal but deadjectival through backformation. This adjective is irascibilis ‘hot-tempered, irascible’, which comes from irascor/o ‘to get angry’ < iratus ‘angry’. Table 4: Summary of V-to-A derivations with a -sc- verb in input. Input WFR

Output WFR Lemma

V-to-V -sc- verb V-to-V -sc- verb, N-to-V -sc- verb

V-to-A iv V-to-A bil

concupiscentivus agnoscibilis cognoscibilis concupiscibilis dinoscibilis erubescibilis ignoscibilis immarcescibilis irascibilis nascibilis noscibilis putrescibilis

8 v = u in WFL, hence the label ‘V-to-A iu’ in Figure 1.

Tokens

WFR frequency

           

 

250

Marco Budassi, Eleonora Litta & Marco Passarotti

4 Discussion The data described above show how only three derivation processes were productive9 with -sc- verbs as input: ‘V-to-V (i)t’, ‘V-to-N i’ and ‘V-to-A bil’. These are the derivation types we will focus on in this Section. The ‘V-to-V it’ derivation-type counts 4 derivations. The number is rather low if compared to the total amount of ‘V-to-V (i)t’ derivations in WFL and amounts to 1.497% of total (334 verbs, e.g. audio ‘to hear’ + -it- > audito ‘to hear frequently’). What these four verbs have in common is that they all belong to Haverling’s (2000) ‘Period I’ (namely, early Latin). Furthermore, 3 out of the 4 verbs (nosco, pasco and vescor) are part of that group of archaic verbs that do not show an attested clear-cut pattern of derivation. On the contrary, the -scderivation of scisco is clear and attested (scio ‘to know’ > scisco ‘to try to get to know’), albeit archaic. Perhaps one of the main reasons why only -it- verbs are formed only from this small group could thus lie in their chronology.10 The use of suffix -it- was very specific in early Latin. According to Viti (2015), suffix -it- was indeed employed to express imperfective aspect, either continuous or habitual (e.g. huius sermo haud cinerem quaeritat, PLAVT. Mil. 1000), and aspectual distinctions deteriorate in the history of the Latin verbal system (Viti 2015). As a consequence, frequentative verbs might have been an attempt to restore an aspectual opposition that would otherwise have gone lost. For Haverling, Category 1 verbs displaying a -sc- derivation dating back to Proto-Indo-European “have a non-terminative semantic function and describe Activities rather than Accomplishments and Achievements or States” (Haverling 2000: 148). This function is very similar to the one that Viti recognises for suffix -it in early Latin. We may hypothesise that archaic -sc- verbs were originally featured by an imperfective meaning. In early Latin, however, due to the increasing loss of aspectual distinction within the verbal paradigm, suffix -it- was added to maintain (or stress) the imperfective meaning of verbs such as nosco, pasco and vescor. This way, the oppositions nosco ~ noscito, pasco ~ pascito and vescor ~ vescitor were created. A first proof of this fact comes from the earliest attestations of such verbs featured by both -sc- and -it-. They are attested (at least) since the first century BC, which is precisely when we would expect them to be created, that is when the aspectual system of Latin verbs was getting blurred.

9 With productive we mean: with a number of derivations > 1. 10 If we consider the class of ‘V-to-V (i)t’ derivations in broader terms, it comes immediately to light how input verbs are featured by old attestations (e.g. alo ‘to feed’ < alito ‘to feed proficiently’, where alo belongs to early Latin).

What’s beyond ‘inchoatives’?

251

Second, the following passages are worth noting: (7) miris modis di ludos faciunt hominibus / mirisque exemplis somnia in somnis danunt : / ne dormientis quidem sinunt quiescere. (PLAVT. Rud. 593–595) (from Haverling 2000: 143) (8)

deinc ut quiescet priu’quam id, quod petit, perfecerit. (PLAVT. Mil. 214) (from Haverling 2000: 143)

In example (7), the meaning of quiesco is ‘to be resting’, thus non-terminative. Conversely, in example (8), quiesco means ‘to calm down’. According to Haverling (2000: 149), “such phenomena are left-overs from an earlier linguistic stage or the result of reinterpretations of single verbs”. Examples (7) and (8) show an early loss of the aspectual distinctions conveyed by -sc- in the earliest phases of the language. At that point, perhaps, -it- suffixed verbs were sometimes created to re-store (or re-stress) a specific imperfective value of verbs. The same kind of evidence is offered by sciscitor. Apparently, sciscitor would contradict the above-described reconstruction: we have an underived verb attested in Latin (scio ‘to know’, Haverling’s P111) (9) and the suffixed verb (scisco ‘to try to get to know’, P1), where -sc- adds an imperfective/conative value to the simple verb (10). A suffixation with -it- would have been redundant. Nevertheless, from Haverling’s P2 henceforth, scisco acquires the meaning of ‘to vote, to approve’, which is once more perfective (11). It is at this point that the -it- suffixation occurs, re-creating a verb (sciscitor) with an imperfective meaning (‘to try to get to know by enquiry’) (12). The chronology of first attestations of sciscitor seems confirmed by a search in LLT, fitting within Haverling’s P1. (9) profecto de auro nil scio nisi nescio. (PLAVT. Bacch. 324) (from Haverling 2000: 174) (10) ibi me inclamat Alcumena : iam ea res me horror adficit / erilis praevortit metus : accurro, ut sciscam quid velit. (PLAVT. Amph. 1068–1069) (from Haverling 2000: 174) (11) quae scisceret plebes aut quae populus iuberet . . . iuberi vetarique voluerunt. (CIC. Flacc. 15) (from Haverling 2000: 174) (12) Epicuri ex Velleio sciscitabar sententiam. (CIC. nat. deor. 1, 17)

11 P1 stands for ‘period I’ (to 100 BC), P2 stands for ‘period II’ (100 BC–0), P3 stands for ‘period III’ (0–AD 100).

252

Marco Budassi, Eleonora Litta & Marco Passarotti

In the end, the fact that the TLL attested pascito is thought to be a spelling variant of pastito (i.e. a form showing the iterative -it- suffix),12 strongly supports the hypothesis formulated above. Pasco behaves like a non-derived lemma, and suffix -it- happens to emphasise its imperfective meaning, operating an actual ‘re-derivation’ of the word. This is coherent with what is described for scisco above. Table 5: Input -sc- verbs of ‘V-to-N i’ derivations and period of attestation. P

P

P

adulesco

concresco

expavesco*

cresco

concupisco

decresco

convalesco

excandesco

erubesco

excresco

refrigesco

resipisco

dinosco

agnosco cognosco disco ignosco nascor/o nosco reminiscor irascor quiesco

As for deverbal nouns, the most productive ‘V-to-N’ derivation type adds suffix -i- to the base -sc- verb. Table 5 summarises the period of attestation of each of the 22 -sc- verbs that generate a nominalisation in -i, according to Haverling.

12 TLL reads: pāscito (vel pāstito), -āre. a pascere frequentative (aut a stirpe praes. aut a p. p. p.). i. q. (frequenter vel more solito) pascere : VARRO rust. 3, 16, 19 apes ferae, quae in silvestribus locis -ant (pasc- ed. princ.et unum apographon cod. Marc., past- cetera apogr.).

What’s beyond ‘inchoatives’?

253

Although the period of attestation of all the -sc- verbs into account is fairly archaic (with the exception of expavesco, the unprefixed form of which however dates back to P2), the chronological evidence is not as clear-cut as in the case of ‘V-to-V (i)t’ derivations. Also positing a semantic constraint as dynamicity + non-terminativity at the basis of the selection of input verbs of ‘V-to-N i’ derivations appears rather unconvincing, at least synchronically. Apparently, from a semantic perspective both input verbs and output nouns behave homogeneously, given that verbs in input describe quite uniformly dynamic nonterminative events (e.g. disco ‘to try to learn’)13 and output -i- nouns from these verbs describe events that are the result of such dynamic non-terminative process (e.g. excandescentia ‘eruption of anger’ < excandesco ‘to become hot’; refrigescentia ‘refreshment’ < refrigesco ‘to cool down’). Nevertheless, by broadening the analysis to verbs that can be input of a ‘V-to-N i’ derivation but are not -sc- verbs, it comes to light how even stative verbs can produce nouns by attaching suffix -i(careo ‘to be devoid of’ + -i- > carentia ‘lack’). As for deverbal adjectives, the most productive kind of V-to-A derivation adds the suffix -bil- to a base -sc- verb. In Table 6 the 11 verbs in input to ‘V-to-A bil’ derivations are associated with their period of attestation in Haverling’s terms. The period of attestation of all the -sc- verbs into account is fairly archaic (none of them goes beyond P2). Nonetheless, the higher degree of productivity of the ‘V-to-A bil’ WFR leads us to conclude that it is unlikely that a chronological constraint determined such a small group of verbs. It is however crucial to stress how all -bil- adjectives from -sc- verbs show late attestations in Latin: their first occurrence is in the Aetas Patrum (third century to 735 AD). Clearly, hence, the ‘V-to-A bil’ derivational pattern involves -sc- verbs only at a later stage. Other -bil- adjectives are widely attested much earlier (e.g. CIC. fam. 5, 12, 5 si vero exitu notabili concluduntur).

13 Certain prefixed verbs belonging to the class of input -sc- verbs describe various change of state that could hardly be considered as non-terminative processes (e.g. dinosco ‘to distinguish’). According to Haverling, on the basis of verbal families in which the unprefixed counterpart of early-attested prefixed verbs is not present, it is possible to maintain that the prefix was needed to express a “terminative or momentaneous sense in the earlier periods of Latin” (Haverling 2000: 149). The derivational process of prefixation seems thus in these cases secondary (in diachronic terms) with respect to the suffixation of -sc-. As a consequence, such cases in our analysis can be left out and we can focus on their unprefixed non-terminative counterparts only without major concerns. This way, problematic cases such as dinosco ‘to distinguish’ (terminative) do not invalidate the semantic homogeneity of the whole class of -scverbs input forms of ‘V-to-N i’ derivations, as they can be led back to their unprefixed nonterminative corresponding form (nosco ‘to get to know’).

254

Marco Budassi, Eleonora Litta & Marco Passarotti

Table 6: Input -sc- verbs of ‘V-to-N bil’ derivations and period of attestation. P

P

agnosco

concupisco

cognosco

erubesco

ignosco

immarcesco*

nascor/o

putresco

nosco

dinosco

irascor

5 Conclusions and future work In this chapter, we proposed a large-scale exploration of derivations with -scverbs in input. Our study highlighted that only three derivational patterns with -sc- verbs in input (‘V-to-V (i)t’; ‘V-to-N i’; ‘V-to-A bil’) behave as actual productive derivations in Classical Latin. It thus remains to be analysed why -sc- verbs were not allowed to participate in other productive derivations (e.g. ‘V-to-N (t)or’). This will be matter of investigation in future work. The same verbs – which are also coherently arranged according to both a diachronic parameter and a semantic parameter (see Section 4) – are constantly productive among the three derivation types, as shown by Table 7. In certain cases (e.g. nosco), a verb occurs at least once per derivation type; in others (e.g. irascor/o), a verb is used as input of two derivation processes. This trend is even clearer if we take into account the morphological families of verbs rather than different forms considered in autonomy: if we consider different verbs as part of one morphological family only (e.g. concresco, cresco, decresco and excresco ∈ cresco), we can distinctly see how verbs that are input forms of the three major derivational patterns involving -sc- verbs are always the same. This analysis paves the way to a glimpse of the semantics of such verbs. Basically, the verbs listed in Table 7 describe events of growing (e.g. adulesco, cresco, excresco), decaying (e.g. decresco, immarcesco, putresco), knowing (e.g. agnosco, nosco, reminiscor) or feeling (e.g. erubesco, expavesco, irascor). Similar verb classes are featured by a general dynamic non-terminative meaning. This distribution seems thus relevant as it shows that not every semantic sub-type of -sc- verbs (sketched in Section 1) can be input of a deverbal derivation, but only

What’s beyond ‘inchoatives’?

Table 7: Input -sc- verb of ‘V-to-V (i)t’, ‘V-to-N i’ and ‘V-to-A bil’ derivation types. Lemma

V-to-V it

V-to-N i

V-to-A bil

adulesco

X

agnosco

X

X

cognosco

X

X

concresco

X

concupisco

X

convalesco

X

cresco

X

decresco

X

dinosco

X

disco

X

erubesco

X

excandesco

X

excresco

X

expavesco

X

ignosco

X

X

X

X

X X

immarcesco irascor

X

X

nascor

X

X

X

X

nosco

X

pasco

X X

putresco quiesco

X

refrigesco

X

reminiscor

X

resipisco

X

scisco

X

vescor

X

255

256

Marco Budassi, Eleonora Litta & Marco Passarotti

a certain class of -sc- verbs participates in derivational patterns. Noticeably, the dynamic non-terminative meaning constitutes the bulk of the semantic value conveyed by suffix -sc- in earliest phases of Latin. This fact, in addition to the generally archaic attestation of verbs considered, seems to suggest that only a subclass of early commonly used -sc- verbs were successfully derivable. In this perspective, most of -sc- verbs involved in further derivational patterns emerged to be part of Haverling’s well-defined Category 1 of -sc- verbs. Our future work plans consist mainly in extending the lexical basis underlying WFL (i.e. the Lemlat’s one) with all the lexical entries of Du Cange’s (1883–1887) Glossarium mediae et infimae latinitatis. This way, also derivational patterns attested in later phases of Latin will be analysed. Finally, we aim to perform a more detailed consideration of the chronology of lemmas derived from -sc- verbs. Also, a deeper survey of the semantic features of both -sc- verbs and non -sc- verbs that are input of these derivation processes will follow this step. This operation will allow us to understand whether the low productivity of derivations with -sc- verbs in input is due to chronological, socio-linguistic or rather semantic constraints.

Abbreviations LLT = Library of Latin Texts, Series A. Brepolis Databases. Brepols clt.brepolis.net/llta/ Default.aspx (accessed: 4 June 2018) WFL= Word Formation Latin lexicon wfl.marginalia.it (last accessed: 02 May 2018) WFR= Word formation rule

Bibliography Budassi, M. & Passarotti, M. (2016). Nomen omen. Enhancing the Latin morphological analyser Lemlat with an onomasticon. In Proceedings of the 10th SIGHUM workshop on language technology for cultural heritage, social sciences, and humanities (LaTeCH 2016) (pp. 90–94). Berlin: The Association for Computational Linguistics. Budassi, M. & Litta, E. (2017). In trouble with the rules: Theoretical issues raised by the insertion of -sc- verbs into word formation Latin. In E. Litta & M. Passarotti (Eds.), Proceedings of the workshop on resources and tools for derivational morphology (DeriMo) (pp. 15–26). Milano: Educatt. Crocco Galèas, G. & Iacobini, C. (1993). Lo sviluppo del tipo verbale parasintetico in latino: I prefissi ad-, in-, ex-. Quaderni Patavini di Linguistica, 31–68. Da Tos, M. (2013). The Italian finire-type verbs: A case of morphemic attraction. In S. Cruschina, M. Maiden & J. C. Smith (Eds.), The boundaries of pure morphology (pp. 45–67). Oxford: Oxford University Press.

What’s beyond ‘inchoatives’?

257

Du Cange, C. (1883–1887). Glossarium mediae et infimae latinitatis. Niort: L. Favre. Georges, K. E. (1913–1918). Ausführliches lateinisch-deutsches Handwörterbuch aus den Quellen zusammengetragen und mit besonderer Bezugnahme auf Synonymik und Antiquitäten unter Berücksichtigung der besten Hilfsmittel ausgearbeitet (8. verb.und verm. Aufl. von H. Georges, 2 Bde.). Hannover: Hahn. Glare, P. G. W. (1982). Oxford Latin Dictionary. Oxford: Oxford University Press. Gradenwitz, O. (1904). Laterali vocum latinarum. Leipzig: Hirzel. Haverling, G. (2000). On Sco-verbs, prefixes and semantic functions: A study in the development of prefixed and unprefixed verbs from early to late Latin (Studia Græca et Latina Gothoburgensia LXIV). Göteborg: Acta Universitatis Gothoburgensis. Litta, E., Passarotti, M. & Culy, C. (2016). Formatio formosa est. Building a Word Formation Lexicon for Latin. In Proceedings of the Third Italian Conference on Computational Linguistics (CLiC–it 2016) (pp. 185–189). Napoli: Accademia University Press. Passarotti, M., Budassi, M., Litta, E. & Ruffolo, P. (2017). The Lemlat 3.0 package for morphological analysis of Latin. In G. Bouma & Y. Adesam (Eds.), Proceedings of the NoDaLiDa 2017 Workshop on processing historical language (pp. 24–31). Gothenburg: NEALT Proceedings Series 32. Tombeur, P. (Ed.) (1998). Thesaurus formarum totius latinitatis a Plauto usque ad saeculum XXum. Turnhout: Brepols. Viti, C. (2015). The use of frequentative verbs in Early Latin. In G. Haverling (Ed.), Latin linguistics in the early 21st century: Acts of the 16th international colloquium on Latin linguistics (pp. 170–182). Uppsala: Uppsala Universitet.

Davide Bertocci & Francesco Pinzin

Towards a morpho-syntactic analysis of -ī-scō and -ā-scō verbs Abstract: The goal of this chapter is to identify the morpho-syntactic rules which govern the formation of -sco verbs in Latin in a Constructivist framework. We assume a coherent syntactic structure for all productive -sco verbs, in which a stative sub-component and an inchoative one are both active. Then we concentrate on the vocalic element which is required between the root and the suffix -sco itself. In particular, we claim that the choice of either -ē- or -ā- or -īcan be predicted according to the status of the underlying lexical element. A list of roots which can only compare in stative configuration typically select -ē- (‘Caland’ type, e.g. rubesco), which is also related to stativity for diachronic reasons. Non-stative roots, nouns, adjectives instead select -ā- and -ī- either on the model of corresponding -āre and -īre verbs or as proper thematic elements, which are inserted in order to allow the insertion of -sco. Keywords: Latin linguistics, -sco suffix, thematic vowels, stative verbs, morpho-syntax

1 Introduction The target of this contribution is to propose a set of formal rules that is able to give reason of the distribution of the vocalic elements in the Latin inchoative verbs. The inchoative verbs are characterized by the presence of the suffix -sco (e.g., vetera-sco ‘I become old’). The suffix -sco is always preceded by a verbal theme, meaning that the base of the verb needs a vocalic element bridging it with the -sco suffix. If the -sco suffix is morpho-phonologically stable, the thematic element is not; it alternates between -ē-, -ī- and -ā-. We claim that a simple set of three rules is able to predict the distribution of the vocalic element. The three rules apply to three different sets of bases; the first one requires the insertion of -ē- and is lexically restricted to a set of roots which show a consistent paradigmatic coherence (‘Caland’ roots, see Section 5); the second rule also requires -ē-, probably influenced by the first one, and targets the denominal and deadjectival derivations, having, thus, a broader synchronic distribution; the Davide Bertocci, Francesco Pinzin, Università degli studi di Padova Goethe Universität Frankfurt am Main https://doi.org/10.1515/9783110647587-017

Towards a morpho-syntactic analysis of -ī-scō and -ā-scō verbs

259

third rule, finally, applies to the deverbal inchoatives and requires the maintenance of the thematic vowel of the verb from which the inchoative is derived. This set of rules predicts a greater variability in the realization of the vocalic element in the deverbal set, a prediction that is borne out. There is, finally, a set of denominal and deadjectival inchoatives which do not present the -ē- vowel, in contrast with the second rule. These exceptions will be treated in a specific section, showing that, in many cases, there is an idiosyncratic explanation for their deviant behaviour.1 Our analysis has the advantage of identifying a coherent relationship between the element -sco and the vocalic elements preceding it. We show that, even if both the vocalic elements and the -sco suffix are inherited, their relationship is ruled by internally coherent synchronic constraints. The contribution is organized as follows. In Section 2, we propose a descriptive categorization of the data. In Section 3, we introduce a few notions about the framework that we adopt for the analysis. In Section 4, we propose a formal account for the syntactic and semantic contribution of the -sco suffix to the verbal derivation. In Section 5 and Section 6, we present the set of rules regulating the insertion of the vocalic element bridging the base and the -sco suffix. The contribution is closed by the concluding remarks (Section 7).

2 A descriptive categorization The verbs presenting the -sco suffix in their morphophonological form are about 250.2 A first major differentiation is in order: there is a group of -sco verbs from an existing Latin base, and a second one whose members seem opaque and underivable. An example of a verb belonging to the first category is crassesco ‘I become fat, thick’ from crassus ‘fat, thick’. This verb is morphophonologically derivable adjoining to the adjective in its undeclined form, crass-, the suffix -sco. The final form involves a vowel -ē- bridging the two elements. The

1 It is relevant to note that there are verbs for which it is unavoidable to propose a nonsynchronic analysis (inherited schemes); e.g., posco ‘I demand, I request’. These verbs will not be analysed. 2 To collect the data, we used the Oxford Latin Dictionary and, when possible, the Thesaurus Linguae Latinae. Additionally, we searched for the occurrences of the relevant verbs with the Library of Latin texts (A Series). We restricted our research to the appropriate chronological limits. Our list does not include the prefixed verbs, meaning we listed all the prefixed variants of a single verb under the same lemma, e.g., the verb acesco ‘I become sour’ has four prefixed variants with the prefixes co-, ex-, in-, per-. We listed all these variants under the same lemma.

260

Davide Bertocci & Francesco Pinzin

characteristics of this vowel will be analysed later (Section 5). An example of a verb belonging to the second category is posco ‘I demand, request’. In this case the verb cannot be decomposed in a base synchronically attested outside the verbal paradigm itself + -sco; the -sco suffix is still recognizable, but there is no adjective/noun/verb from which this verb may be derived. The target of this contribution is to propose an analysis of the verbs in which the -sco morpheme is consistent with synchronic morphological and semantic restrictions, looking at the realization of the vocalic element bridging the -sco morpheme with the underlying lexical item. This point is crucial since, whereas in verbs like (g)nosco or posco, -sco- there is no synchronic mechanism linking an existing lexical item to the -sco element (the complex root+-sco is inherited), in the productive formations, -sco requires a stem, that is a thematized root.3 In the second group, moreover, the meaning contribution of the -sco element is not structurally coherent, that is, in this class -sco covers a complex set of meanings, ranging from the inchoative one of (g)nosco to the iterative one of posco. Given that, we focus on the productive derivations involving the -sco morpheme, setting apart the second category of verbs. From the point of view of the lexical morpheme, it is possible to subdivide the first category of inchoatives among denominals, deadjectivals and deverbals. An example of a denominal inchoative is gemmesco ‘I become like a gemstone’ from gemma ‘gemstone’. An example of a deadjectival derivation is duresco ‘I become hard’ from durus ‘hard’. An example of a deverbal derivation is (e)dormisco ‘I sleep off something’, from dormio ‘I sleep’. It is possible to also categorize as deverbal a relevant group of about 50 inchoative verbs which enter in a specific paradigmatic relation with a set of related nominal/ adjectival formations in -or/-idus (further details in Section 5). An example of such a verb is rubesco ‘I become red’, which is related to the stative verb rubeo ‘I am red’, to the noun rubor ‘redness, blush’ and to the adjective rubidus ‘red’. From now on, for the sake of simplicity, we will refer to this kind of verbs as ‘rubesco type’. From the point of view of the morphophonological composition, the verbs belonging to the first category can be further subdivided looking at the vowel bridging the -sco ending with the base:

3 For posco this is shown by the fact that the root, taking away the -sco element, is hardly transparent due to phonological changes which affected it in diachrony. Moreover, the iterative/intensive meaning of the suffix can be explained only in a comparative (cf. NHG fragen ‘to question’ vs. forschen ‘to research’ < *‘to question over and over again’. Many thanks to one of the anonymous reviewers for suggesting us this example) and reconstructive perspective (Leumann 1977: 535–539).

Towards a morpho-syntactic analysis of -ī-scō and -ā-scō verbs

261

– verbs in -ē-sco: this type represents the vast majority of the -sco verbs. E.g., acerbesco ‘I become sour’, calvesco ‘I become bald’, ditesco ‘I become rich’, pinguesco ‘I become fat’, rubesco ‘I become red’, arboresco ‘I grow in a tree’ and lapidesco ‘I become stonelike’; – verbs in -ī-sco: this type represents a small subset of the -sco formations. E.g., (e)dormisco ‘I sleep off’, (per)prurisco ‘I itch’, laetisco ‘I rejoice’ and longisco ‘I grow long’; – verbs in -ā-sco: this type is even more restricted than the -ī-sco one. E.g., gemmasco ‘I grow sprouts’ (re)puerasco ‘I become a child (again)’, generasco ‘I am born’, integrasco ‘I become whole’ and amasco ‘I fall in love’. The relationship between the lexical and the morphophonological categorization is complex. The most transparent lexical category is the one containing the deverbal inchoatives which enter in the -esco/-eo/-or/-idus paradigm, exemplified above with rubesco ‘I become red’. In this case the bridging vowel is always -ē-.4 The deadjectival verbs usually present the -ē- vowel too. Among these about 80, only seven do not present this vowel only. Among these seven verbs, three have -ī-: unisco ‘I get unified with someone/something’ from unus ‘one’, longisco ‘I become long’ from longus ‘long’ and laetisco ‘I rejoice’ from laetus ‘happy’.5 Two alternate among -ē- and -ī-: vil(e/i)sco ‘I become worthless’ from vilis ‘worthless’ and viv(e/i)sco ‘I come to life’ from vivus ‘alive’.6 Two, finally, alternate among ē and -ā-: veter(a/e)sco ‘I become old’ from vetus ‘old’, tener (a/e)sco ‘I become tender’ from tener ‘tender’. The denominal verbs, about 60, show greater variation. Among them, 17 present the possibility of having a different bridging vowel. Thirteen present the vowel -ā-: roborasco ‘I become strong’ from robor ‘strength’, pullulasco ‘I sprout’ from pullulus ‘sprout’, granasco ‘I flourish’ from granum ‘grain’, gemmasco ‘I sprout’ from gemma ‘sprout’, germinasco ‘I sprout’ from germen ‘sprout’, gelasco ‘I freeze’ from gelus ‘frost’, corporasco ‘I incarnate’ from corpus ‘body’, (ef)ferasco ‘I become wild’ from ferus ‘wild’, gallulasco ‘I change voice during

4 There is only one exception: putesco ‘I become rotten’. This verb is attested as putisco in a single occurrence from Cicero, mainly present in recent codices (cf. TLL s.v. p. 2752, 36–37; we thank one of the anonymous reviewers for pointing it out): animum illi pecudi datum pro sale, ne putisceret. (CIC. fin. 5, 38). 5 Many of this unexpected -i verbs can be considered hapax or are philologically contested. 6 Alternation between -i and -e is frequent in the manuscripts (we thank an anonymous reviewer for pointing out this issue). Given this fact, these exceptions look less relevant.

262

Davide Bertocci & Francesco Pinzin

boyhood’ from gallulus ‘small rooster’, (il)luculasco ‘I dawn’ from lux ‘light’ or luculentus ‘full of light’, puellasco ‘I become a child’ from puella ‘female child’, puerasco ‘I become a child’ from puer ‘child’ and vesperascit7 ‘it becomes dark’ from vesper ‘evening’. Two present both the -ā- and the -ē- vowel: purpur(a/e) sco ‘I become red as purple’ from purpura ‘purple’ and tenebr(a/e)sco ‘I become dark’ from tenebra ‘darkness’. Two, finally, present both the -ī- and the -ēvowel: luc(e/i)sco ‘I become bright’ from lux ‘light’ and call(e/i)sco ‘I become hard’ from callus ‘callus’.8 It is of note that, also in this case, the most frequent template involves the -ē- element. The deverbal inchoatives that do not belong to the category mentioned previously (-esco/-eo/-or/-idus paradigm) show the greatest variation. The remaining deverbals are 24: 8 present the -ē- vowel (e.g., apparesco ‘to begin to appear’, augesco ‘to begin to grow, to increase’); 5 the -ā- vowel (e.g., amasco ‘to begin to love’, generasco ‘to be generated’, integrasco ‘to begin anew’); 9 the -ī- vowel (e.g., cupisco ‘to wish’, dormisco ‘to fall asleep’); and, finally, 2 alternate between -ē- and one of the other two (ingem(e/i)sco ‘to complain, to moan’, (con/de/ad)sud(a/e)sco ‘to sweat much’). From this short overview, it is possible to conclude that the deverbal inchoatives, if not belonging to the ‘rubesco type’, present greater variation with respect to the denominal and deadjectival verbs. The denominals, then, with respect to the deadjectivals, present many more cases in which the -ā- vowel is present.

3 Framing the analysis The analysis of the -sco verbs relies on the decomposition of their meaning. We want to identify the specific contribution of the three elements that constitute the inchoative verbal forms: the verbal root, the vowel and the -sco- morpheme. The target, as already mentioned in the introductory remarks, is to propose an account capable of predicting the stem-vowel variation in the attested verbs. The -sco morpheme, as will be clarified below, is related to the Aktionsart of the verbal event. The analysis of the structuring of the verbal event is taken from a constructivist perspective (Hale and Keyser 2002; Marantz 2005, 2013; Mateu and Acedo-Matellán 2012; Acedo-Matellán and Mateu 2013; Cuervo 2014, 2015).

7 vesperascit ‘it becomes dark’, for obvious reasons, is presented using the third person and not the first one. 8 Due to the existence of adjective callidus and lucidus both callesco and lucesco might be regarded as part of the Caland system (we thank an anonymous reviewer for this note).

Towards a morpho-syntactic analysis of -ī-scō and -ā-scō verbs

263

This framework conceives verbal events as the by-product of the subsequent merging of different syntactic heads. The speaker is endowed with a set of universal syntactic heads with a specific semantic denotation. Each head is related to the introduction of a specific argument. – v-be°: identification stative event, related to the introduction of a holder. – v-go°: uncontrolled dynamic event, related to the introduction of an undergoer. – v-do°: controlled dynamic event, related to the introduction of a doer. The composition of these heads in different phrases creates the relevant verbal structure, as it is possible to observe in the following examples: (1) a. [v-goP [v-beP]] → Change of state event (e.g., the chair broke.) b. [v-doP [v-beP]]→ Causative event (e.g., John broke the chair.) c. [v-doP] → Activity (e.g., John walks.) The lexicon (formed by a list of roots) does not project autonomous phrases. When a root is merged in a syntactic derivation, it is directly adjoined to a syntactic head.9 The Latin sentence in (2), for example, represents a causative event, in which there are two phrases: a controlled dynamic phrase (v-doP) and an identification stative phrase (v-beP). These two phrases are related to the presence of two arguments, a holder and a doer. The doer is in control of the dynamic event that brings the state, of which the holder is the subject, into existence. The root is firstly merged into v-be° and then moves to v-do° (Fig. 1). (2)

alii verum. . .hordeum . . . molunt. (PLIN. nat. 18, 28) others.NOM instead barley.ACC mince.3.PL.PRS.ACT10 ‘Others, instead, mince the barley.’

4 The role of -sco Before proposing an analysis of the vocalic element bridging the -sco morpheme and the verbal base, we must address the issue of the role of -sco

9 In support of this view, see Alexiadou (2014). 10 For the abbreviations we follow the Leipzig Glossing Rules (see: www.eva.mpg.de/lingua/ resources/glossing-rules.php).

264

Davide Bertocci & Francesco Pinzin

v-doP DOER alii

v-goP v-go° √mol

v-beP HOLDER

v-beP

hordeum

v-be° (√mol)

Figure 1: Syntactic structure for example (2).

in Classical Latin. The -sco morpheme in Classical Latin has been frequently described, in its productive uses, as a morpheme that adds dynamicity to a stative verbal derivation (Haverling 2000, 2010). (3) ager aret. (COLVM. 2, 8, 5) field.NOM be.dry.3.SG.PRS.ACT ‘The field is dry.’ (4) dum arescunt mea (vestimenta). (PLAVT. Rud. 574) while become.dry.3.PL.PRS.ACT my.NOM.PL (clothes.NOM) ‘While my clothes dry.’ Mateu (2017), following the description put forward by Haverling, proposes a formal account in which the -sco morpheme is treated as the Latin morphological output11 of the v-go° (see Mateu 2017: ex. 27). The inchoative derivations describe change of states, meaning that the derivation of these verbs involves two argumental phrases, a stative phrase and an uncontrolled dynamic phrase (v-goP) that leads to the state. He proposes to mark the lower stative phrase as

11 The term morphological output belongs to the Distributed Morphology framework (DM, Halle and Marantz 1993). In this framework, morphemes are not conceived as carrying syntactic features and projecting syntactic phrases, they are post-syntactically matched with the derived structure, meaning that they are late-inserted. Morphology is the output of the syntactic derivation. Syntax, consequently, is derived merging a set of functional heads, not a set of morphemes.

Towards a morpho-syntactic analysis of -ī-scō and -ā-scō verbs

265

a locative Central Coincidence Relationship (Hale and Keyser 1993, Hale and Keyser 2002), a v-beP in our terms.12 v-goP UNDERGOER mea vestimenta

v-goP v-beP

v-go° -sc

HOLDER

v-beP

(mea vestimenta)

v-be°

√ar+e Figure 2: Syntactic structure for example (4).

The same derivation adequately describes the deadjectival/denominal cases, the only difference being the merging position of the formative of the verb. In Figure 2 the formative of the verb, √ar- (a defective root), is directly merged in the stative head (v-be°), realized as -ē-. In a deadjectival/denominal derivation (ex. 5) on the other side, the formative of the verb comes from a previously derived phrase, an a(djectival)P or a n(oun)P. Since a phrase cannot be merged in a head position, it follows that an inchoative denominal/deadjectival derivation will have the following structure (Fig. 3): (5) novella castanieta calvescunt. (COLVM. 4, 33, 3) new.NOM.PL chestnut.forests.NOM become.bald.3.PL.PRS.ACT ‘The new chestnut forests become bald.’ The formative of the verb is merged in the complement-of-the-state position and is then incorporated into the verbal event. The rest of the derivation mirrors the derivation in Figure 2.

12 Mateu (2017) focuses on the contribution of the prefixes to the final meaning of the inchoative verbs. The unprefixed variant just described is non-telic, meaning that the event is conceived as durative and not entailing an endpoint. If, on the other side, the verb is prefixed (exaresco), it describes a telic event, meaning that there is a final endpoint. We will not focus on this alternation, for further analyses specifically on the inchoative verbs see Haverling (2000) and Mateu (2017). For a more general analysis of the transition events and of the contribution of the verbal prefixes in a comparative view see Acedo-Matellán (2016).

266

Davide Bertocci & Francesco Pinzin

A first note on the positions of the vocalic element. In Figure 2, the -ēvowel is directly adjoined to the root (√ar+ē-) and not part of the root itself. The vocalic element cannot be conceived as part of the root, since it disappears when the element is not verbal but adjectival (ar-idus ‘dry’).13 As we will clarify in the next section, this vocalic element belongs to the stative phrase, meaning that it appears in specific verbal contexts involving a v-be°. In Figure 3, consequently, it is merged in the v-be°. v-goP

UNDERGOER

v-goP

novella castanieta v-go°

-sc

v-beP

HOLDER

v-beP

(novella castanieta) v-be°

-e

aP

calv-

Figure 3: Syntactic structure for example (5).

Up to this point, we proposed a specific morphosyntactic rule for the insertion of the -sco morpheme: we described it as the Classical Latin morphological output of the v-go° in a change of state derivation. (6) [v-go°] ↔ sc / [v-goP [v-beP]] This generalization is clearly too strong; it would predict that in Classical Latin the -sco morpheme should be present in each change of state derivation. This is clearly not the case, since there are change of state verbs, like morior ‘I die’, that do not show a -sco morpheme.14 This rule only holds true if we restrict its scope to the productive denominal/deadjectival/verbal derivation and to the

13 Moreover, -ē- is not a proper part of the verb stem, since it does not appear in the perfect. The synchronic connection of -ē- with the stative phrase is coherent with its origin from the PIE stative morpheme *-eh1- (see further). 14 For an analysis of this kind of change of state verbs see Gianollo (2010, 2014) and Pinzin (2017).

Towards a morpho-syntactic analysis of -ī-scō and -ā-scō verbs

267

aresco category. A general description of the morphological means used in Latin to derive the change of state verbs is beyond the scope of this contribution. For our target, that is the analysis of the distribution of the vocalic elements in the inchoative derivations, this simple rule will do the job.

5 The strong pattern, the -ē- morpheme In the general description proposed in Section 2, we noted the consistent presence of the -ē- vowel in the ‘rubesco type’. There is a single exception, signalled in fn. 4. This single exception, anyway, does not invalid the general pattern, that is strong and evident. Why is this vowel present in this kind of verbs? A first possible answer could be to take the -ē- vowel as part of the root. However, this proposal, as briefly explained above, is not adequate; the adjective and the noun derived from the same root do not show an -ē- vowel: al-gidus ‘cold’, algor ‘coldness’.15 It is of note that in the nominal derivations -ē- is also absent; -ē- only appears when the root is in a verbal derivation, may the verbal derivation be stative or inchoative (see ex. (3) and (4)). We proposed to relate the -sco morpheme to the higher part of the inchoative derivation, v-goP (see Figures 2 and 3). The -ē- vowel, instead, is not only present when the derivation is inchoative, it is also present when the derivation is stative. A straightforward proposal could be to relate it to the projection of v-be°. This proposal would put on the same morphosyntactic level both the -sco morpheme and the -ē-, meaning that both elements would be the morphological realization of a specific syntactic head. There is, however, a difference between these two elements: the -sco morpheme is more productive, it is conceivable as an autonomous morpheme actively used by Latin speakers, that is, it has a broader distribution; -ē-, on the other side, looks more constrained: no new stative verbs marked by the -ē- vowel are derived during the classical

15 For a morphosyntactic analysis of the adjectives of the -idus kind, that are beyond the scope of this contribution, see Mateu (2017). In a diachronic perspective, Bozzone (2016) showed that the Caland series like rubesco/rubor/rubidus are the remnant of an ancient class of PIE ‘verbal’ adjectives. Crucially, it is assumed that the IE stative formation with *-ē-i̯e/o was a secondary output of those adjectives: this can explain why in our rubesco type a canonical derivation is non-transparent. On this topic, see also Rau (2009) and, as regards to the Latin adjective in -idus, see Nussbaum (1999). Furthermore, as we will recall further, the archaic nature of -ē- can shed light on its decay as a productive deverbal suffix (cf. Section 6).

268

Davide Bertocci & Francesco Pinzin

period, the entire set of stative verbs in which the -ē- vowel is present is inherited.16 A way of signalling the limited productivity of -ē- is to propose a lexically restricted rule of insertion: (7) [v-be°] ↔ -ē- / {√ar-, √alg-, . . .} This rule of insertion restricts the presence of -ē- to the derivations involving v-be° and the specified roots. Combining the rule in (6) and the rule in (7), we predict the presence of the -ē- vowel both in the stative and inchoative derivation (both involve v-be°) and the presence of -sco only in the inchoative derivation (in the stative derivation v-go° is absent). There is an evident issue with respect to this proposal, anyway. The -ē- vowel is also consistently present in the denominal/deadjectival inchoatives, as noted in Section 2. The rule in (7) is not enough. In the denominal/deadjectival inchoatives the -ē- vowel is present only in the inchoative derivations. The correspondent statives do not show -ē-, the v-be° is lexicalized by means of the projection of the verb esse ‘to be’. An example is the verb sterilesco ‘I become infertile’ from sterilis ‘infertile’. The only way to derive a stative predication involving sterilis ‘infertile’ or a morphophonologically related root is by means of esse ‘to be’. (8)

quae sterilis est vacca, taura which.NOM.SG infertile.NOM.SG is cow.NOM taura.NOM appellata. (VARRO. rust. 2, 5, 6) called.NOM.SG ‘An infertile cow is called taura.’

This means that -ē- vowel, in these denominal/deadjectival cases, is always related to the presence of the -sco morpheme and, consequently, to the presence of v-go°. We can derive this distribution by means of a specific rule: (9) [v-be°] ↔ -ē- / [v-goP [v-beP [n/aP]]]

16 A close relationship between -ē- and stativity, despite its incoherent synchronic distribution, is clear on diachronic and comparative grounds (Jasanoff 2003): Italic data confirm that, even without a consistent stative paradigm, -ē- can refer to the result of a change-of-state process, as forms like Umbrian maletu ‘crumbled’ or vaseto ‘empty’ > ‘lacking’ (both associated to -ā- verbs) show (cf. Bertocci 2014).

Towards a morpho-syntactic analysis of -ī-scō and -ā-scō verbs

269

This means that, whenever a deficient,17 and thus incorporated (see Figure 3), nominal/adjectival element is merged in a change of state structure, -ē- will appear to mark the v-be°. This rule does not contrast with the rule in (7), since they scope on two different sets, the denominal/deadjectival inchoatives and the ‘rubesco type’ verbs. It is possible, however, to hypothesize a connection between the two rules. The rule in (9) is an extension of the rule in (7). Every Latin verb – except for some irregular verbs (‘unregelmäßige’ verbs, Leumann 1977: 521) – needs a thematic vowel.18 The choice of the -ē- vowel in the denominal/deadjectival inchoatives depends on the existence of the ‘rubesco type’; it is an extension of the ‘rubesco type’ to the new inchoative derivations, meaning that, even if -ē- is not productive anymore in Classical Latin, it is still maintained as a marker in the new denominal/deadjectival inchoative derivations, on the basis of the ‘rubesco type’ pattern. As we will propose in the next Section, the ‘rubesco type’ is strongly attractive, some inchoative verbs that usually have their own thematic vowel, coming from a previous use of the root in, e.g., a causative or activity derivation, are attracted by the ‘rubesco type’ and start presenting the -ē- vowel. Up to this point, we proposed two morphosyntactic rules governing the presence of the -ē- vowel in the inchoative verbs. One rule predicts that with a specific set of roots -ē- will always mark the presence of the v-be° (‘rubesco type’); the second one predicts that every new inchoative denominal/deadjectival derivation will present the -ē- vowel bridging the nominal/adjectival element with the -sco morpheme. We proposed that the choice of the -ē- vowel as a bridging element is due to the influence of the ‘rubesco type’, that is present as a strong paradigmatic element in the grammatical competence of the Latin speakers. As already noted in Section 2, however, the data do not perfectly fit this proposal. First of all, there is a set of deverbal inchoatives, as amasco ‘I fall in love’ from amo ‘I love’, that we still have not taken into consideration. In these cases, the choice of the bridging vowel is highly variable. We will treat these verbs in the next Section. Secondarily, there is a group of denominal/deadjectival verbs that do not present the -ē- vowel as a bridging element, e.g., longisco ‘I become long’ from longus ‘long’, puellasco ‘I become a child’ from puella ‘female child’ and puerasco ‘I become a child’ from puer ‘child’. We will treat these cases in the next Section.

17 Deficient means that it is not an autonomous noun or adjective. In order to be an autonomous noun/adjective an n/aP needs more structure, as the D(eterminer) P(hrase). 18 As for thematic vowels, see Embick and Halle (2005) and Bertocci (2017).

270

Davide Bertocci & Francesco Pinzin

6 -sco with -ā- and -ī- stems As we said above, -ē- is selected as the most reliable bridging element for -sco verbs for both syntactic and morphological reasons; in particular, it looks structurally related to v-be modification not only at a synchronic level, but especially considering its history and its comparative distribution. Nevertheless, some -sco verbs do select for -ā- and -ī- stems. The first point to consider is the syntactic structure of -ā-sco and -ī-sco verbs, and the lexical elements they involve. As far as the syntactic structure is considered, the -asco and isco verbs conform to the template we proposed before, in which there are two event phrases, a v-goP introducing a v-beP: (10) caput deponit, head.ACC.SG. down.put. 3.SG.PRES.ACT. PVB. condormiscit. (PLAVT. Curc. 360) sleep.become.3.SG.PRES.ACT. '(he) nods off, begins to sleep' Moreover, in some cases -sco has a valency change effect. In roborasco for instance it converts the causative roborare in a change of state predicate: -sco licenses v-goP and allows an identification between the subject and the condition of being strong, exactly as in the -ē-sco verbs. A second point to consider is the lexical item associated with -sco. As anticipated above, whereas the -ē-sco type only admits a specific set of roots (the ‘Caland’ ones), or nouns and adjectives involving identification, -ā-sco and -ī-sco can derive from verbs as well (amasco, cupisco); conversely, some nouns, from which one could expect an -ē-sco verb, derive an -ā-sco or -ī-sco one (ve-terasco, puellasco). A first hypothesis considers the possibility that the existence of an underlying verb of the first or of the fourth class acts as a model. That is, amasco seems directly derived by amāre, with its thematic categorization. Under this view, -sco would be able to head a v-goP, irrespectively of the presence of -ē-, provided that the underlying verb: 1) allows a stative interpretation (e.g., I fall in the state of be in love); 2) is thematized (in this case, -sco can apply to the -ā- or the -ī- stem). Consequently, we should state a more general rule according to which, when a v-be° is not overtly realized, -sco can license v-go only if a thematic stem is involved. The result of this rule would be that only -ā- or -ī- verbs can host -sco:

Towards a morpho-syntactic analysis of -ī-scō and -ā-scō verbs

271

(11) -sco ↔ [v-goP [v-beP]] / √-ā-,-īAccording to rule (11), we can also explain the late denominal verbs like puerasco, which are problematic with respect to the rule proposed in (9). Here again we can assume, firstly, that despite the absence of -ē-, the morpho-syntactic structure is compatible with the -ē-sco verbs treated above. In puerasco, corporasco, gallulasco, etc. the noun (e.g., puer) means the entity with which the subject identifies itself, that is, a v-be phrase is involved. The availability of an identification reading in these nouns allows for a v-go/v-be structure, triggering a derivation by -sco. When this happens, the selection of -ā-sco is possible for two reasons: on the one hand, if a -ā-re verb is present (e.g., corporare triggering corporasco), it acts as a model for a bare derivation process headed by -sco. On the other hand, as a default thematic element, -ā- can be conceived as the preferred element for repairing the lack of an -ē- stem. There is, finally, a small set of -ī-sco verbs like longisco, unisco, laetisco (see above) in which the selection of -ī- as a thematic element instead of -ā- does not seem triggered by an attested -ī-re verb. However, at least for unisco, the ‘pronominal’ forms unius and unī could have enhanced the verbal stem unī-. This claim has welcome outcomes, since it explains why, in the later stages of Latin, the restrictions embedded in -sco weakened, enabling it to have a larger distribution than in the archaic phase. In turn, we can also imagine that gradually the iconicity of -sco became less transparent, leading to the striking morphological optionality of the termination -iscere/-escere of ancient and modern Italian ferisco, favorisco, punisco, etc., cf. Rohlfs (1969: 242). In order to motivate the selection of -ā- and -ī-, finally, it is to recall that, keeping apart the most ancient nucleus of -sco verbs ((g)nosco, posco, etc.), which are probably inherited from PIE, -sco cannot apply directly to a root, but it requires to be merged with a stem, that is, with a thematized root. As far as -āand -ī- as ‘thematic’ element are concerned, we want to stress that -ā- and -īshare a number of properties: both look related to the derivation of verbs from nouns/adjectives (multare, curare); they are the only possible output for parasynthetic formations (impedīre, dērīvāre), and more generally -ā- appears with secondary formations like intensives and causatives or with metaplasms due to prefixation (experīre < parĕre, dĭcāre < dīcĕre, placāre: placēre, edŭcāre < dūcĕre, cantāre < canĕre, gestīre: gestāre: gerĕr); both -ā- and -ī- are controversial as regards to their etymology, this means that many verbs of the first and of the fourth class have not cognate forms in other IE languages, and that their diffusion in Latin is an innovation (Bertocci 2017). Once we state that -sco requires a stem, then, the selection of -ā- or -ī- as bridging element for -sco verbs can be expected,

272

Davide Bertocci & Francesco Pinzin

since they are productive, represent an innovation, and are the typical thematic elements of Latin verb.19 However, the rule (11) itself fails to prevent some nouns to produce both -ēsco verbs and -ā-sco or -ī-sco ones, e.g., sudesco (VARRO. ling. 5, 109) /sudasco (PLAVT. Cas. 360). Yet, we do not consider this as a structural objection, since, at the one hand, most of these data are very sporadic formations (gallulasco, corporasco are late hapax); on the other hand, we believe that -ē-sco type should be considered as a default option, always available with the adequate syntactic environment. Our claim is that these tokens do not modify the general frame, in which -ē-sco is the default strategy, and only under specific conditions -ā-sco or -ī-sco can arise.

7 Conclusion We have analysed the distribution of the suffix -sco paying attention to the morpho-syntactic conditions which allow its insertion as an inchoative morph. In particular, we have shown that -sco license the v-go head (Section 4) when a vbe head is also active in the lexical structure of a verb. This condition is typically represented by the -ē-sco verbs from nouns and adjectives (sterilesco) and from a set of roots we labelled as ‘Caland’, as in the diachronic terminology (algesco). In both cases, we assumed that -ē- is transparently related to the stative component of the event. The second point we addressed is the selection of the lexical item on which -sco applies: keeping apart a small set of verbs like posco, -sco always requires a stem, that is, a thematized root. Unexpectedly, not only the stative related -ē-, but -ā- and -ī- are also involved. Our hypothesis to explain -ā-sco, -ī-sco verbs is that -ā- and -ī- can be selected as bridging elements at two conditions: (1) when an underlying -ā-re or -ī-re verb (or a backformed stem, see longisco < longitūdō) is available; (2) when a stative component is semantically conceivable, but it is not overtly realized through -ē-. The selection of the bridging vowel associated to -sco, then, depends on the morpho-syntactic properties of the events that -sco verbs describe. The fate

19 We do not take a position on the choice between -ā- and -ī-; However, beside the verbs which can be thought as derived by -ā-re or -ī-re (finisco, amasco), we observe that -ā- and -īare deeply related in the late history of Latin. At least in Italo-romance varieties, -are and -ire terminations are often alternating giving rise to frequent metaplasms, so that some scholars (Prosdocimi 1991) traced them back to the Latin itself.

Towards a morpho-syntactic analysis of -ī-scō and -ā-scō verbs

273

of -ē-, which is not productive in Latin outside stative verbs like timeo, together with the availability of -ā- and -ī- as thematic vowels, makes -ā-sco and -ī-sco possible as well, even if less frequent throughout the Latinity.

Abbreviations √ aP DM nP vP

= = = = =

root adjectivalizing phrase Distributed Morphology nominalizing phrase verbalizing phrase

Bibliography TLL: Thesaurus Linguae Latinae (1900–). Berlin/Boston: De Gruyter (accessed: 19 February 2019, from www.degruyter.com/view/db/tll). Acedo-Matellán, V. & Mateu, J. (2013). Satellite framed Latin vs. verb framed Romance: A syntactic approach. Probus, 25, 227–265. Acedo-Matellán, V. (2016). The morphosyntax of transitions. A case study in Latin and other languages. Oxford: Oxford University Press. Alexiadou, A. (2014). Roots don’t take complements. Theoretical Linguistics, 40(3/4), 287–297. Bertocci, D. (2014). Per una grammatica dell’umbro delle Tavole di Gubbio: tra testo e grammatica. Riflessioni di metodo e un caso di studio nella morfologia del verbo. In O. Paoletti & M. C. Bettini (Eds.), Gli Umbri in età preromana: Atti del XXVII Convegno di Studi etruschi ed italici (pp. 553–564). Pisa/Rome: Fabrizio Serra. Bertocci, D. (2017). Latin 1st class -ā- verbs as thematic formations: On the deficiency of IE roots. Pallas. Revue d’études antiques, 103, 45–52. Bozzone, C. (2016). The origin of the Caland system and the typology of adjectives. IndoEuropean Linguistics, 4, 15–52. Cuervo, M. C. (2014). Arguments for a root. Theoretical Linguistics, 40(3/4), 375–387. Cuervo, M. C. (2015). Causation without a cause. Syntax, 18(4), 388–424. Embick, D. & Halle, M. (2005). On the status of stems in morphological theory. In T. Geerts, I. van Ginneken & H. Jacobs (Eds.), Romance languages and linguistic theory 2003 (pp. 59–88). Amsterdam: John Benjamins. Gianollo, C. (2010). I verbi deponenti latini e l’unità della flessione in -r. Incontri triestini di filologia classica, 8, 23–49. Gianollo, C. (2014). Labile verbs in Late Latin. Linguistics, 52(4), 945–1002. Glare, P. (Ed.) (2012 = 1982). Oxford Latin Dictionary. Oxford: Oxford University Press. Hale, K. & Keyser, S. J. (1993). On argument structure and lexical expression of syntactic relations. In K. Hale & S. J. Keyser (Eds.), The View from Building 20. Essays in Linguistics in Honor of Sylvain Bromberger (pp. 53–110). Cambridge, MA: The MIT Press.

274

Davide Bertocci & Francesco Pinzin

Hale, K. & Keyser, S. J. (2002). Prolegomenon to a theory of argument structure. Cambridge, MA: The MIT Press. Halle, M. & Marantz, A. (1993). Distributed morphology and the pieces of inflection. In K. Hale & S. J. Keyser (Eds.), The View from Building 20. Essays in Linguistics in Honor of Sylvain Bromberger (pp. 111–176). Cambridge, MA: The MIT Press. Haverling, G. (2000). On Sco-verbs, Prefixes and Semantic Functions: A study in the development of prefixed and unprefixed verbs from Early to Late Latin (Studia Græca et Latina Gothoburgensia LXIV). Göteborg: Acta Universitatis Gothoburgensis. Haverling, G. (2010). Actionality, tense and viewpoint. In P. Baldi & P. Cuzzolin (Eds.), New perspectives on Latin syntax (vol. 2, pp. 277–523). Berlin/Boston: De Gruyter. Jasanoff, J. H. (2003). Stative *-ē- revisited. Die Sprache, 43(2), 127–170. Leumann, M. (1977). Handbuch der Altertumswissenschaft Lateinische Grammatik. Tl.1: Lateinische Laut- und Formenlehre (5th ed.). Munich: Beck. Marantz, A. (2005). Objects out of the lexicon! Argument-structure in the syntax, manuscript. MIT, web.mit.edu/marantz/Public/UConn/UConnHOApr05.pdf (accessed: 11 January 2016). Marantz, A. (2013). Verbal argument structure: Events and participants. Lingua, 130, 152–168. Mateu, J. & Acedo-Matellán, V. (2012). The manner/result complementarity revisited: A syntactic approach. In M. C. Cuervo & Y. Roberge (Eds.), The end of argument structure? (pp. 209–238). Bingley: Emerald. Mateu, J. (2017). State and change of state in Latin: A view from the lexicon-syntax interface. In E. Castroviejo, O. Fernández-Soriano & I. Pérez Jiménez (Eds.), Boundaries, phases, and interfaces. Case studies in honor of Violeta Demonte (pp. 344–366). Amsterdam/ Philadelphia: John Benjamins. Nussbaum, A. J. (1999). *IOCIDUS. An account of the Latin adjectives in -idus. In E. Eichner & H.C. Luschützky (Eds.), Compositiones indogermanicae in memoriam J. Schindler (pp. 377–420). Praha: Enigma Corporation. Pinzin, F. (2017). Deposing deponency: Latin non-denominal deponents are not grammatically idiosyncratic verbs. Journal of Latin Linguistics, 16(1), 11–42. Prosdocimi, A. L. (1991). Tra romanzo e indoeuropeo: Il latino sommerso. In L. Vanelli & A. Zamboni (Eds.), Per Giovan Battista Pellegrini. Scritti degli allievi padovani (vol. 2, pp. 517–643). Padova: Unipress. Rau, J. (2009). Indo-European nominal morphology: The decads and the Caland system. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft. Rohlfs, G. (1969). Grammatica storica dell’Italiano e dei suoi dialetti (vol. 2). Torino: Einaudi.

Words Linguistic systems in contact

Theodor Georgescu

Les mots latins d’origine grecque avec diffusion panromane : comment et pourquoi ? Abstract: We focus on ten of the Greek words borrowed in Latin that have become Panromance lexemes (according to Fischer, 1985: aer, angelus, blastimare, bracchium, cannabis, castanea, chorda, coma, corona, marmor, masticare, pascha, petra, sabbatum, saccus, spatha, stuppa). We aim at analyzing the reasons of their adoption in Latin, their possible rivalry with old Latin words, as well as the parameters of their Romance survival. We also try to see what words have penetrated through orality – thus testifying to the daily needs of the Latin language in its dizzying growth – and which others have been borrowed through cultural exchanges, only later being integrated into the common language. Keywords: Latin linguistics, Greek borrowings in Latin, Panromance words

1 Introduction L’ouvrage est axé sur les grécismes du latin qui sont devenus des lexèmes panromans, en visant tant les caractéristiques de leur trajet jusqu’à l’intégration en latin, que les paramètres de leur survivance romaine. Le REW contient environ 900 grécismes qui sont devenus romains, à priori à travers le latin. Selon la liste proposée par Fischer (1985 : 130–132), il y en a 17 qui peuvent être considérés comme panromans : aer, angelus, blastimare, brachium, cannabis, castanea, chorda, coma, corona, marmor, masticare, pascha, petra, sabbatum, saccus, spatha, stuppa.1 La seule raison pour laquelle on peut les analyser comme un conjoint est leur destin : à la différence des autres grécismes latinisés, conservés dans des territoires discontinus, ce groupe de dix-sept mots d’origine grecque semble avoir fait partie du noyau du latin oral et s’être diffusé à travers une aire continue, en occupant tout le territoire de l’ancien Empire Romain.

1 La même liste proposée par Fischer (1985), fondée sur DELL, est utilisée comme base provisionnelle pour le Dictionnaire Étymologique Roman (DÉRom), cf. Buchi et Schweickard (2014). Theodor Georgescu, University of Bucharest https://doi.org/10.1515/9783110647587-018

278

Theodor Georgescu

Les raisons de leur pénétration en latin peuvent être différentes : il y a des lexèmes qui ont probablement pénétré à la fois avec l’objet désigné (cannabis, stuppa, etc.) et d’autres qui, en désignant des réalités déjà existantes, ont été adoptés à côté des vieux mots latins (e.g. brachium, coma, masticare, etc.). Dans ce cas-là, les emprunts pouvaient soit nuancer un contenu, soit remplacer un signifiant rendu insuffisant par rapport aux nécessités d’expression. Il y a aussi des mots propres au christianisme (angelus, blastimare, pascha, sabbatum) devenus des concepts culturellement indispensables. Les limites de cet ouvrage ne permettent pas de s’occuper de manière approfondie de ces derniers mots du domaine de la religion. On se concentrera donc sur les autres noms, dont on choisira dix pour l’analyse. Les rapports linguistiques entre les Grecs et les Romains, qui constituent le thème d’un chapitre dans la plupart des traités d’histoire du latin, ont été analysés de manière approfondie particulièrement par Frédérique Biville, qui a consacré plusieurs articles à ce sujet (cf., entre autres, Biville 1989 ; 1992 ; 1995 ; 1998). On se propose, dans cet ouvrage, de déterminer la place que se réservent les dix emprunts choisis dans le lexique latin, par rapport soit avec les mots autochtones qui désignaient la même réalité (si c’est le cas), soit avec les autres lexèmes du même champ sémantique, de même que révéler les raisons de leur conservation dans les langues romanes, aux dépens des lexèmes originairement latins (encore une fois, au cas où ceux-ci existaient réellement). Pour ce qui est du corpus, nous partons du TLL et de la base des dates PHI pour extraire les contextes les plus édifiants à cet égard, à partir desquels on peut distinguer les étapes de pénétration et d’adaptation de ces hellénismes en latin.

2 Les types d’emprunts Il y a deux niveaux d’hellénisation du latin : l’un est populaire – la voie orale –, qui conduit à la diffusion des termes du grec contemporain dans le basilècte, tandis que l’autre est représenté par la voie culte – écrite –, spécifique à l’acrolècte. D’un point de vue normatif, les emprunts se partagent, selon Terlingen (1943), entre emprunts « par nécessité » et emprunts « pour ornement », une terminologie que nous adopterons pour l’article présent.

Les mots latins d’origine grecque avec diffusion panromane

279

Il faut noter que la distinction populaire/littéraire ne coïncide pas – et ne doit pas se confondre – avec les deux catégories normatives (en d’autres mots, on ne peut pas appliquer l’étiquette d’ « emprunts nécessaires » aux mots pénétrés par la voie populaire et d’ « emprunts ornementaux » aux lexèmes entrés par la voie littéraire). La distinction n’est pas valide non plus d’un point de vue diachronique, contrairement à l’idée suggérée par Fischer (1965 : 51), qui affirmait que dans l’étape la plus ancienne avaient pénétré les emprunts de première nécessité, et que c’est uniquement par la suite que sont apparus les emprunts scientifiques, moins nécessaires d’un point de vue pratique. Des emprunts « pratiques » sont également les termes pénétrés avec un nouveau concept, concret ou abstrait, en se situant dans la même catégorie que les emprunts nécessaires. Par conséquent, ce ne sont pas seulement cot(h)urnus, un type de chaussure, ou palanga ‘rouleaux de bois pour le déplacement des vaisseaux’ qui sont des emprunts pratiques, mais aussi comoedia, aer ou phantasia (cf. CIC. ac. 1, 40). Les emprunts d’ordre psychosociologique (« pour ornement ») peuvent se rencontrer dans toutes les époques (par exemple, les interjections babae, bombax etc). Quintilien mentionne la raison pratique de l’emprunt de mots grecs (cf. Devoto 1969 : 224) : le grec est la réserve vers laquelle on se dirige chaque fois qu’on est confronté à un vide lexical en latin. confessis quoque Graecis utimur verbis ubi nostra desunt, sicut illi a nobis nonnumquam mutuantur. (QVINT. inst. 1, 5, 58)

Le degré d’adaptation à la langue latine est un indice concernant la voie et la période de pénétration. Il n’est pas exclu qu’on ait des doublets phonétiques et morphologiques, voire même sémantiques parfois : tumum (populaire) vs thymum (littéraire) ‘thym [plante]’ < gr. θύμον2 cicinus (populaire) vs cycnus / cygnus (littéraire) ‘cygne [oiseau]’ < gr. κύκνος citera vs cithara3 ‘cithare’ < gr. κιθάρα scema, -ae vs schema, -tis ‘attitude, figure [géométrique]’ < gr. σχῆμα, -τος elephantus, -i et *olifantus vs elephas, -ntis ‘éléphant [animal]’ < gr. ἐλέφας, -ντος.

2 La différence orthographique n’implique pas nécessairement une différence de prononciation entre les romains éduqués et les autres, c’est plutôt un change orthographique intervenu dans la période classique. 3 Cf. cithara non citera (PROB. app. gramm. IV 197, 26).

280

Theodor Georgescu

Il est évident que les variantes ayant subi des changements phonétiques et morphologiques remarquables, qui les ont permis de mieux s’intégrer dans le système latin, font partie de la couche populaire, tandis que la tentative de conserver la graphie, la prononciation et la catégorie flexionnelle indique un emprunt par la voie culte. En même temps, on peut établir l’époque de leur pénétration dans la langue populaire,4 avec la mention qu’on atteste parfois la pénétration d’un même mot dans des étapes différentes, dans le même registre : voir l’exemple de *olifantus, plus ancien que elefantus, hérité dans les langues romaines dans la première forme : v.it. lio(n)fante, v.fr. olifant, prov. olifan.5 Parfois, le terme emprunté à une époque archaïque (par la voie populaire), déjà intégré dans la langue (et dans le système phonétique latin), perd, du point de vue du parleur, la relation avec la langue-source et entre en opposition avec sa variante empruntée postérieurement (vu comme terme autochtone vs terme étranger) : apuam nostri, aphyen Graeci vocant. (PLIN. nat. 31, 95 [ap. Biville 1991 : 54])

La différence d’ordre diastratique se reflète de même dans la diffusion des emprunts dans le territoire du latin ainsi que dans leur transmission dans les langues romaines. Les emprunts plus anciens et usuels, hérités dans les langues romaines, sont généralement ceux populaires. Nous essayerons donc d’établir une catégorisation de ces douze grécismes qui ont eu la chance d’une diffusion panromane, selon les critères suivants : la raison de l’adoption, la voie de pénétration (dans la mesure où l’on peut la déterminer), le degré d’adaptation à la langue latine, le type d’emprunt à partir de la distinction « par besoin » / « pour ornement ». Selon un classement en grandes lignes, on aura : des noms de produits nouveaux, des noms désignant des éléments (concepts) déjà existants, mais avec une spécialisation sémantique (souvent faisant partie du lexique spécialisé), et des noms appartenant à des langages spécialisés (religion), dont on ne parlera pas dans cet article.

4 La pénétration dans la langue culte est plus facile à établir, en suivant leur attestation dans les textes. 5 Cf. REW 2841 ; quand même, REW ne reconstruit pas étymon *olifantus, en considérant tous ces termes comme descendants de elephas. Le changement de e- en o- devant -l- exilis n’est pas régulier (e- change en o- devant -l- pinguis : elaiFa> olaiua, *welo> uolo), contre DELL (s.v. elephantus) qui considère que la voyelle o « est conforme aux exigences de la phonétique latine », en montrant seulement sa surprise face à l’absence d’attestations de cette variante.

Les mots latins d’origine grecque avec diffusion panromane

281

3 Des noms de produits nouveaux – des emprunts pénétrés en même temps que le concept désigné 3.1 Noms de produits naturels, probablement importés D’un point de vue de leur adaptation, ces mots ont toujours eu une résonance étrange pour les parleurs ; ils sont, quand même, nécessaires, parce qu’ils expriment des réalités empruntées au monde grec, donc sans dénomination latine. Il est aussi possible que ces emprunts aient nuancé le contenu d’une réalité existante : il s’agit, dans ce cas-là, d’un nom plus restreint, plus précis, en désignant un certain type d’une catégorie d’objets qui avait une dénomination générique en latin, mais pas de noms spécifiques. Le grécisme est préférable dans cette situation, étant plus facile à utiliser qu’une paraphrase en latin. Le type d’emprunt est par besoin, avec l’importation du produit même.

3.1.1 Cannăbis Cannăbis ‘chanvre’, emprunté au gr. κάνναβις, est lui-même importé d’une langue orientale ; le mot, probablement d’origine scythe ou traque, était ressenti comme étranger dans le bassin méditerranéen.6 Il apparaît en latin avec des formes parallèles cannabus (um ?), -i / cannaba, ae, ce qui atteste la tendance d’intégrer le mot dans les types flexionnelles plus communs. En latin le mot est attesté depuis Varron : fiunt de cannabi, lino, iunco, palma. (VARRO rust. 1, 22, 1) cannabis grana. (COLVM. 11, 2, 75) e cannabi funes. (PLIN. nat. 19, 29)

Dans ses premières attestations, le concept est associé au monde grec et, en plus, relié avec autre mot d’origine grecque qui devait partager le destin de celui-ci dans les langues romaines,

6 Isidore associait le cannabis avec canna ‘canne, jonc mince plus petit que le roseau’, luimême emprunté au gr. κάννα : cannabum a similitudine cannae vocatum (ISID. orig. 19, 27, 3).

282

Theodor Georgescu

stuppa : Graeci magis cannabo et stuppa ceterisque satiuis rebus (naues suebant). (VARRO frg. Gell. 17, 3, 4)

Cette attestation est très importante car elle indique le domaine technique où le matériel était utilisé, à savoir la construction navale. C’est, peut-être, à travers le vocabulaire spécifique du domaine naval que le mot a pénétré en latin, tout comme d’autres mots appartenant à la même aire sémantique. Auparavant, les Latins utilisaient le lin, linum, donc l’importation de la nouvelle plante venue de l’Est se fait avec le nom spécifique. L’intégration en latin est démontrée par ses dérivées : cannabetum ‘chènevière’, cannabinus ‘de chanvre’. À l’époque basse apparaissent les formes cannabus, cannaba, cannabum, can(n)ape, canapa. Les formes romaines remontent à deux types : cannabis, conservé dans les dialectes suditaliques et sardes,7 et canapis/can(n)apus, le plus fréquent, qui se conserve dans les principales langues romaines.8

3.1.2 Stǔppa Stǔppa ‘étoupe’, sous-produit fibreux non tissé issu essentiellement du travail du chanvre ou du lin, est emprunté au gr. στύπ(π)η ‘filasse, étoupe’, attesté seulement dans le Lexicon de Suidas, qui cite Flavius Josephus : Ἰώσηπος· πληρώσαντες τὴν στοὰν στύππης, πρὸς δὲ ἀσφάλτου τε καὶ πίσσης (SVID. 1259). La forme classique était στυ(π)πεῖον ‘paquet d’étoupe, étoupe’.9 Les grammairiens latins étaient conscients non seulement de l’origine grecque, mais aussi de la provenance dialectale de ce mot : stuppam linum impolitum appellant Graeci Dorii. (FEST. p. 317)

La tournure latine, linum impolitum, qui expliquait le nom grec, est beaucoup moins pratique d’un point de vue de son utilisation dans la langue. Tout comme cannapis, stuppa est un terme technique emprunté probablement par l’intermédiaire du vocabulaire des navigateurs, car l’étoupe était utilisée par les Grecs pour coudre des éléments des navires, comme nous avons vu dans l’exemple de Varron (frg. Gell. 17, 3, 4).

7 Neap. kanneve, kannele, abruzz. kaneve, log. kánau, campid. kanniu. 8 Roum. cânepă, it. canapa, veron., venez. kánevo, fr. chanvre, prov. canebe, esp. cáñamo, port. cannamo. 9 HERODOT. 8, 52, 1. XEN. Cyr. 7, 5, 23. MEN. frg. 684.

Les mots latins d’origine grecque avec diffusion panromane

283

conpletas onerarias naves taeda et pice et stuppa reliquisque rebus. (CAES. civ. 3, 101, 2)

Les dérivés et les formes romaines en prouvent l’intégration : stuppeus ‘d’étoupe’, stupparius ‘servant à faire l’étoupe’, stuppator ‘calfat’.10

3.1.3 Castanea Castanea ‘châtaigne’/‘châtaignier’, mot emprunté au grec pluriel (κάρυα) καστάνεια ou κασταναῖα ‘noix de châtaigne’, est un adjectif dérivé de κάστανον, lui-même d’origine étrangère. Initialement, il était utilisé aussi comme adjectif spécificatif accompagnant le nom nux, en délimitant une seule espèce du contenu trop large de nux, qui désignait tout fruit à écale et à amande. L’importation du grec est donc faite par calque linguistique κάρυα καστάνεια (interprété comme singulier ?) / nux castanea. Par ellipse à partir de ce syntagme, castanea devient nom et se conserve dans cette forme dans les langues romaines. L’origine grecque était connue par les grammairiens latins : castaneam Latini a Graeco appellant vocabulo. Hanc enim castaniam vocant propter quod fructus eius gemini in modum testiculorum infra folliculum reconditi sunt, qui, dum eiciuntur, quasi castrantur. (ISID. orig. 17, 7, 25)

Attesté depuis Virgile, le mot désignait tant l’arbre, que le fruit : castaneae hirsutae (l’arbre) (VERG. ecl. 7, 53) et castaneae molles (le fruit) (VERG. ecl. 1, 81). Pour ce qui est de la voie et la période de pénétration, il doit être un emprunt populaire dont l’héritage romain relève de deux types phonétiques qui semblent avoir coexisté dans l’oralité : *castănia et *castinia. DELL considère que le vocalisme avec ă intérieur maintenu semble montrer que l’emprunt ne remonte pas à la période la plus ancienne. Meillet (2004 [19281] : 90), Sganzini (1937, ap. DÉRom s.v. */kasˈtani-a/) et FEW (2, 466b, 467a) postulent l’antériorité du protoromain */kasˈtɪni-a/ par rapport à */kasˈtani-a/. Quand même, en analysant les données romaines, le DÉRom estime que l’évolution phonétique régulière ne constitue pas une preuve de sa pénétration tardive face à la forme *castinia : la variante Elle est ancienne, mais postérieure au type I castania, conclut le DÉRom.11

10 Les formes romanes : roum. stupă, it. stoppa, log. istuppa, engad. stoppa, friul. stope, fr. étouppe, esp., port. estopa. 11 Cf. aussi Ernout (1954 : 32–34) (« doublet[s] dialectal tardif[s] ») et Biville (1995 : 110–111, 136–137).

284

Theodor Georgescu

3.2 Des noms pénétrés en même temps que l’objet-spécifique de la culture-source Saccus ‘sac, besace’ provient du gr. σάκκος (emprunté au sémitique saq ‘étoffe grossière’) ‘cilice (tunique), tapis, couverture’ et aussi ‘sac’. C’est dans ce sens secondaire que le mot est passé en latin ; toutefois, dans le latin biblique, il a aussi le sens du classique cilicium ‘pièce d’étoffe en poil de chèvre, vêtement de crin grossier’ (cf. DELL s.v.). Depuis les premières attestations le mot fait partie d’expressions, ce qui atteste une intégration à une époque ancienne : ad saccum ire. (PLAVT. Capt. 90) ‘aller prendre la besace, aller mendier’

L’usage du diminutif saccǔlus ‘petit sac, bourse’ et du verbe saccare ‘filtrer’, confirme le sentiment que le mot est complètement latinisé nummos in publico de sacculo effundentem. (PETRON. 71, 9) alii tritum in aqua triplici linteo saccant. (PLIN. nat. 33, 104)

On ne sait pas si le mot a été emprunté à la fois avec l’objet, ou s’il a été adopté pour désigner un certain type de sac, fait d’un certain matériel. En tout cas, il n’était pas ressenti comme grec, à la différence des autres mots discutés, par conséquent on peut parler d’une intégration totale à une époque ancienne.12

4 Des noms désignant des éléments (concepts) déjà existants, mais avec une spécialisation sémantique C’est le cas des concepts ou des produits qui existaient déjà, mais qui n’avaient pas une dénomination latine spécifique, c’est-à-dire étaient désignés avec le

12 Les formes romanes : roum. sac, it. sacco, friul. sak, fr., prov., cat. sac, esp., port. saco.

Les mots latins d’origine grecque avec diffusion panromane

285

nom générique dont ils étaient en effet des subordonnés taxonomiques (des produits spécifiques d’une classe d’objets, occupant une aire sémantique restreinte).

4.1 Marmor Marmor ‘marbre’, mot emprunté au gr. μάρμορος ‘morceau de pierre’,13 plus précisément ‘pierre blanche, marbre’, doit être relié à μαρμαίρω ‘briller’ (cf. DELG).14 Tout comme dans le cas de castanea, ce nom offre la possibilité d’une restriction sémantique du nom générique lapis, qui était largement utilisé y compris pour ce type spécifique de pierre. Donc, marmor est adopté comme un hyponyme nécessaire, en nuançant le contenu trop étendu du lapis. Son usage est ancien, depuis Ennius et Caton, même si dans ses premières attestations le mot est utilisé avec le sens figuré de ‘surface unie de la mer, la mer’,15 premièrement avec le sens d’‘écume de la mer’ par association avec la couleur : verrunt extemplo placidum mare : marmore flavo / caeruleum spumat sale conferta rate pulsum. (ENN. ann. 384).16 ut mare, cum magni commorunt aequora venti, / vertitur in canos candenti marmore fluctus. (LVCR. 2, 767).17 Cf. marmor sermo Graecus est a viriditate. (ISID. orig. 16, 5, 1)

Les syntagmes latins utilisés pour désigner le type spécifique de pierre étaient normalement composés du nom lapis et d’un adjectif exprimant la source ou la couleur : Parius lapis (‘de Paros’) (VERG. Aen. 1, 592), albus lapis (HOR. sat. 1, 6, 116), candidus lapis (CATO agr. 38, 2). La spécificité se manifeste aussi dans une autre direction, le pl. marmora étant utilisé, selon Isidore, pour nommer les pierres plus grandes qui brillent par frottement : marmora dicuntur eximii lapides, qui maculis et coloribus commendantur. (ISID. orig. 16, 5, 1).

13 Cf. μαρμάρῳ ὀκριόεντι ‘une pierre rugueuse’ (HOM. Il. 12, 380, Od. 9, 499). 14 Cf. μάρμαρον ἢ λίθον λευκήν (HIPPOCR. mul. 2, 185). 15 LVCR. 2, 767. VERG. Aen. 7, 28. CATVLL. 63, 88. 16 Cf. spumas virentis maris “flauom marmor” appellavit (GELL. 2, 26, 23). 17 Plus tard, par métonymie il pouvait dénommer la mer même : infidum remis impellere marmor (VERG. georg. 1, 254).

286

Theodor Georgescu

Marmor va donc occuper une place spécifique comme hyponyme de lapis et se conservera comme tel dans les langues romaines,18 tandis que lapis sera transmis aussi avec une restriction sémantique, comme ‘pierre tombale’. L’hyperonyme de marmor conservé dans toutes les langues romaines sera petra. Les multiples dérivés attestent l’intégration totale : marmorarius, marmoro, marmorosus, marmusculum.

4.2 Chorda/corda Chorda/corda ‘boyau, corde’ est emprunté au gr. χορδή pl. ‘boyaux, tripes’, sg. ‘saucisse, boudin’, et ‘corde’ d’instrument de musique, faite de boyau. Le sens de ‘corde d’un instrument’ est ancien19 et les composés grecs contenant cet élément révèlent surtout ce sens-là (χορδοπώλης ‘marchand de cordes’, χορδότονος ‘aux cordes tendues’, χορδοτόνον ‘chevalet’, χορδολογέω ‘accorder [un instrument]’). Ce mot a été emprunté aux côtés d’autres termes préexistants qui désignaient la ‘corde’ en général. La différence réside, d’une part, dans le matériel dont la corde est faite (des intestins), et de l’autre part dans son appartenance au vocabulaire technique de la musique, où le mot était déjà inclus en grec. Il s’agit donc d’un emprunt par nécessité, qui permet d’exprimer une réalité exacte et est limité au domaine spécifique de la musique : en latin chorda est utilisé d’abord dans le sens technique de ‘corde d’un instrument de musique’, avec ses premières attestations à Cicéron, Varron et Lucrèce. L’emprunt a été facilité par l’étymologie populaire, qui l’associait à cor : hinc etiam appellatam aestimamus chordam, quod facile corda moveat. (CASSIOD. var. 2, 40, 12) chordas autem dictas a corde, quia sicut pulsus est cordis in pectore, ita pulsus chordae in cithara. (ISID. orig. 3, 22, 6)

Le sens propre, ‘tripe, boyau’ est attesté secondairement dans le vocabulaire technique de la cuisine, par influence directe du grec : cocleas singulas et cordae frusta et hepatia. (PETRON. 66, 7)

18 Les formes romanes : roum. marmură, it. marmo(re), engad. marmel, friul. marmul, fr. marbre, prov. marbe, cat. marbre, esp. mármol, port. marmore. 19 HOM. Od. 21, 407. HYMN. HOM. Merc. 51. PIND. Pyth. 2, 128.

Les mots latins d’origine grecque avec diffusion panromane

287

Cette utilisation était connue par les grammairiens.20 Toutefois, le sens classique est de ‘corde d’un instrument de musique’, ce qui indique que sa pénétration en latin a eu lieu sous l’empire de ce sémantisme – donc, on peut l’attribuer au lexique spécifique des musiciens grecs si présents à Rome.21 D’un point de vue de l’adaptation au système phonétique latin, on atteste l’hésitation dans les manuscrits entre ch / c, ce qui prouve une certaine conscience de sa provenance étrangère. À basse époque, chorda a commencé à être employé comme synonyme de fūnis22 et c’est avec ce sens-là qu’il est demeuré dans les langues romaines.23 Comme dérivé, on peut mentionner chordula, auquel il faudrait ajouter le composé chordifex.

4.3 Spatha Spatha ‘battoir’, ‘spatule’, ‘épée large et longue’, est emprunté au gr. σπάθη. Les attestations en latin sont plus tardives que dans les autres cas discutés (depuis Varron). Le latin avait le nom gladius, désignant une épée, un glaive court, typiquement romaine : la spatha était un certain type d’épée, plus longue, importée des Grecs. L’emprunt est de nouveau fait par nécessité. Mais ce n’est pas seulement ce sens-là qui a joué un rôle important dans son futur trajet romain. Le sémantisme du grec, beaucoup plus large, a fourni des usages spéciaux en latin : il pouvait désigner, tant en grec qu’en latin, une ‘spatule’,24 la ‘partie plate de la rame’ ou plus précis la ‘tige de la feuille et de la fleur mâle du palmier’,25 ainsi qu’une ‘baguette des tisserands pour serrer le tissu’26 – ce qui représente son plus ancien sens en grec. En plus, le sens qui a fait une vraie carrière dans les langues romaines a été celui utilisé comme terme technique en médicine, par métaphore basé sur la similitude, ‘omoplate’. En latin le diminutif spat(h)ula est attesté pour ‘épaule, omoplate’ (d’animal) chez Apicius.27 20 Nam veteres Graeci intestina chordas vocaverunt (CAEL. AVR. acut. 3, 17, 144). 21 Nam voces ut chordae sunt intentae, quae ad quemque tactum respondeant (CIC. de orat. 3, 216) ; chordarumque sonos (LVCR. 4, 584) ; e septem cordis citharae (VARRO ling. 10, 46). 22 Qui semble designer une corde plus grosse que restis – avec lequel il entre en concurrence. 23 REW 1881 : chorda ‘Saite’, ‘Strick’, ‘Seil’. 2. ‘Kaldaunen’ ; roum. coardă, it., log., engad. corda, friul. kuarde, fr. corde, prov., cat. corda, esp. cuerda, port. corda. 24 Permisceto spatha lignea (COLVM. 12, 22, 1). 25 Palma sola . . . in spathis habet fructum (PLIN. nat. 16, 112). 26 Quemadmodum subtemen . . . spatha coire cogatur et iungi (SEN. epist. 90, 20). 27 Cf. spatula porcina (APIC. 4, 3, 4).

288

Theodor Georgescu

Pour ce sens-là, le latin avait humerus, qui désignait proprement l’‘épaule’, tergum ‘dos’ et dorsum ‘dos, croupe’. L’usage de spatha était initialement limité aux animaux, mais quelques langues romaines ont élargi son domaine d’application à l’homme et à toute la partie arrière du corps (ex. roum. spate ‘dos’). Le sens ‘d’épée’ est hérité dans la plupart des langues romaines.28 En effet, il semble que le sens ‘d’épée’ soit plus tardif en latin que celui anatomique. Il a probablement pénétré comme terme technique dans plusieurs domaines où l’on avait un vide lexical mais où l’objet ou le concept était connu : le vocabulaire des tisserands, celui de la médicine,29 le lexique de l’anatomie et de la botanique ; mais, en même temps, il a pénétré aux côtés d’un nouveau objet importé des Grecs, l’‘épée longue’.

4.4 Parties du corps ou actions qui concernent le corps D’un certain point de vue, on aurait pu intégrer spatha dans cette catégorie, mais étant donné que son sens est plus large et qu’il fait partie des divers vocabulaires spécialisés, on préfère l’intégrer dans la catégorie précédente.

4.4.1 Bra(c)chium Bra(c)chium ‘bras (depuis la main jusqu’au coude)’, ‘membre de devant’, ‘bras’, emprunté au gr. βραχίων, représente, de nouveau, le cas d’un mot qui venait à occuper une place vide dans le vocabulaire latin : manus était restreint à la ‘main’ (vue comme réceptacle) depuis l’IE, puis il y avait le ‘coude’ cubitus et humerus, comme parties séparées, mais pas le tout. De même, lacertus est adopté pour la partie supérieure du bras, par extension ‘muscles de l’épaule’ et ‘muscles’ en général, ‘force musculaire’, par une métaphore similaire à musculus < mus ‘souris’. bracchia . . . validis ex apta lacertis. (LVCR. 4, 829) subiecta lacertis / bracchia sunt. (OV. met. 14, 304–305)

28 REW 8128 : aroum. spată, roum. spată ‘Schulter’ spate ‘Rücken’, it. spada, fr. épée, cat. espasa, esp., port. espada. 29 En chirurgie, il désigne une forme d’attelle.

Les mots latins d’origine grecque avec diffusion panromane

289

Dans ces contextes, bracchium désigne la partie inférieure du bras, par opposition avec lacertus. En effet, il n’y avait pas un nom IE pour ‘bras’. Selon DELG, le ‘bras’ en grec s’appelle πῆχυς ‘coude’, ἄγκων ‘courboure du bras, coude’, βραχίων ‘bras’, χείρ ‘main’. Le sens propre de βραχίων était originellement la partie humérale du bras ; Chantraine (DELG) donne crédit à l’étymologie proposée par Pollux, un comparatif de βραχύς : βραχίων ὅτι ἐστὶ τοῦ πήχεως βραχύτερος (POLL. 2, 138). Donc le gr. βραχίων est emprunté par besoin, mais pour le bras entier – étendu sémantiquement pour couvrir un vide lexical, et attesté en latin depuis les plus anciens auteurs (ENN., PLAVT.). Les grammairiens latins expliquent l’emprunt au grec : brachium nos, Graeci dicunt βραχίων, quod deducitur a βραχύ, id est breve, eo quod ab umeris ad manus breviores sint quam a coxis plantae. (PAVL. FEST. p. 31)30

Le latin avait aussi armus ‘haut de bras (avec l’épaule)’ et spécialement ‘épaule’ (semble-t-il mot d’origine indo-européenne) ; selon les Latins, armus est réservé aux animaux, umerus aux hommes. Toutefois, un témoignage de Festus nous indique un usage qui couvre l’aire du ‘bras’ et de ‘l’épaule’ : armillas . . . dictas esse existimant quod antiqui umeros cum brachiis armos vocabant (PAVL. FEST. p. 23). La langue de l’Eglise semble avoir contribué à la conservation du mot bracchium, car ici il est aussi symbole de la puissance – voir le surnom du Christ, bracchium domini.

4.4.2 Coma Coma ‘chevelure (de l’homme)’, emprunté au gr. κόμη ‘chevelure’, est un mot de caractère surtout poétique. Cette fois-ci, il semble que c’était le besoin stylistique qui ait mené à son adoption en latin. Le mot est attesté depuis Ennius (il est absent chez Plaute, Térence et Caton), mais les grammairiens étaient toujours conscients de l’origine grecque du mot : coma est Graecus sermo. nam comas Graeci κόμας a secando nominant, unde et κείρειν tondere. (ISID. orig. 11, 1, 30)

Il est aussi attesté avec la forme de la déclinaison grecque comē, ēs, acc. comēn ‘barbe de bouc (plante)’ (PLIN. nat. 27, 142).

30 Pour ce qui est de l’orthographe, on trouve aussi la forme bracio (LEX repetund. [CIL I2 583] 52), bracia (bracias) TAB. devot. Audollent 135 A 2. B 2. 190, 10.

290

Theodor Georgescu

Le latin avait un terme générique, crinis, qui signifiait, en général, ‘poils’, comme partie du corps par opposition aux autres parties, ou capillus – ‘le poil de la tête’, sans idée accessoire. « Coma et caesaries contiennent une idée accessoire esthétique, une belle chevelure, avec la différence que dans coma on considère cette beauté du côté de la grâce et de la mollesse et, dans caesaries, du côté de la force » (Barrault 1853 : 458). De là, coma désigne plus particulièrement la chevelure des femmes et des enfants ; caesaries, celle des hommes : a caedendo dicta caesaries ; ergo tantum virorum est (SERV. Aen. 1, 590). Caesaries se dit quelquefois à propos des femmes, mais avec une idée de force, en parlant des Furies. De plus, caesaries est tout à fait poétique. Le mot a été latinisé assez vite pour fournir des dérivés proprement latins : comans ‘chevelu, pourvu d’une chevelure ou d’une crinière’ (attesté à partir de Virgile), comatus ‘chevelu’, comula ‘petite chevelure’. Dans son héritage romain, coma a perdu sa nuance poétique et, régionalement, il est passé à designer le poil long des animaux : roum. coamă, it. chioma, prov., esp., port. coma ‘crinière’.

4.5 Aer Aer ‘air’, provient du gr. ἀήρ. Pour ce concept, le latin avait expérimenté différentes possibilités d’expression31 : ventus,32 animus/anima – normalement pour l’air qui se trouve dans les organes, dans le corps humain – ou caelum.33 Aucun de ces lexèmes n’exprimait exactement le concept de ‘fluide gazeux’, chaque mot ayant une aire référentielle déjà définie et différente du concept dont les gens commençaient à être conscients (il va sans dire que l’aire n’est pas visible, et est donc moins perceptible que les autres éléments naturels). Avec cette conceptualisation, le besoin d’expression devient impérieux, d’autant plus que le grec a utilisé un nom spécifique pour cette notion. Aer est dèjà présent chez Plaute : Iubeas una opera me piscari in aere, / venari autem rete iaculo in medio mari. (PLAVT. Asin. 99–100).

Les auteurs latins étaient cependant conscients de l’origine grecque du mot :

31 Cf. Georgescu (2012 : 124–125). 32 Et densis aquila pennis obnixa volabat vento, / quem perhibent Graium genus aera lingua (ENN. ann. 149). 33 Namque et hoc « caelum » appellavere maiores, quod alio nomine « aera » (PLIN. nat. 2, 38).

Les mots latins d’origine grecque avec diffusion panromane

291

animali spirabilique natura cui nomen est aer – Graecum illud quidem sed perceptum iam tamen usu a nostris ; tritum est enim pro Latino. (CIC. nat. deor. 2, 91).

La perception qu’avaient les écrivains cultivés de son origine grecque se manifeste aussi dans la conservation de la déclinaison grecque (acc. sg. aera est le plus usité par Cicéron, Sénèque). Il est remarquable que cette forme d’accusatif grec, aera, se conserve comme telle dans certains idiomes romans (e.g. it. aria).34

5 Conclusions Nous avons analysé la manière et la raison de la pénétration en latin des mots grecs devenus panromans. Comme nous avons essayé de démontrer, tous ces grécismes ont été empruntés par besoin. Il s’agit soit de mots qui ont pénétré en même temps que le produit qu’ils désignaient, soit de lexèmes qui apportaient une spécialisation sémantique pour des concepts déjà existants. La première catégorie continue d’avoir en latin une résonance exotique, en revanche la seconde est complètement intégrée dans le vocabulaire. Le succès de ces mots, attesté par leur héritage panroman, s’explique par ce qu’ils expriment des réalités immédiates toujours présentes dans la vie, dont la dénomination, une fois intégrée dans la langue, devient indispensable.

Abréviations abruzz. aroum. campid. cat. engad. fr. friul. gr. IE it. log. neap. prov. port. roum.

abruzzais aroumain campidanais catalan engadinois français frioulan grec indo-européen italien logoudorien napolitain provençal portugais roumain

34 Dans le reste, roum. aier/ aer, friul. ayer, fr. air, prov., cat., esp., aire, port. ar.

292

esp. venez. véron. v.fr. v.it.

Theodor Georgescu

espagnol vénitien véronais ancien français ancien italien

Bibliographie DELL = Ernout, A. & Meillet, A. (19594 [19321]). Dictionnaire étymologique de la langue latine: Histoire des mots. Paris: Klincksieck. DELG = Chantraine, P. (2009 [1968]). Dictionnaire étymologique de la langue grecque: Histoire des mots. Paris: Klincksieck. DERom = Buchi, E. & Schweickard, W. (Eds.) (2014). Dictionnaire étymologique roman: Genèse, méthodes et résultats (Beihefte zur Zeitschrift für romanische Philologie 381). Berlin/Boston: De Gruyter. FEW = Wartburg, Walther von et al. (1922–2002). Französisches etymologisches Wörterbuch: Eine Darstellung des galloromanischen Sprachschatzes (25 vols.) Bonn/Heidelberg/ Leipzig/Berlin/Bâle:Klopp/UniversitätsverlagWinter/Teubner/Zbinden. PHI = Packard Humanities Institute – Classical Latin Texts. latin.packhum.org/index (dernier accès: 31 janvier 2019). REW = Meyer-Lübke, W. (19353). Romanisches etymologisches Wörterbuch. Heidelberg: Universitätsverlag Winter. TLL = Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. Barrault, E. (1853). Traité des synonymes de la langue latine. Paris: Hachette. Biville, F. (1989). Grec et latin: Contacts linguistiques et création lexicale. Pour une typologie des hellenismes lexicaux du latin. Dans M. Lavency & D. Longrée (Eds.), Cahiers de l’Institut de Linguistique de Louvain, Actes du Ve Colloque de Linguistique latine (pp. 29–40). Louvain-la-Neuve: Borzée. Biville, F. (1991). L’emprunt lexical, un révélateur des structures vivantes des deux langues en contact (le cas du grec et du latin). Revue de Philologie, de littérature et d’histoire anciennes, 65, 45–58. Biville, F. (1992). Le grec parlé en latin vulgaire: Domaines lexicaux, structures linguistiques d’accueil. Dans M. Iliescu & W. Marxgut (Eds.), Latin vulgaire – latin tardif III, Actes du IIIeme Colloque international sur le latin vulgaire et tardif (pp. 26–39). Tübingen: Max Niemeyer Verlag. Biville, F. (1995), Les emprunts du latin au grec: Approche phonétique, II. Vocalisme et conclusions. Louvain: Peeters. Biville, F. (1998). Compétence bilingue latino-grecque et manipulations interlinguistiques. Dans C. Brixhe (Ed.), La koiné grecque antique. 3, Les contacts (pp. 145–160). Nancy: A.D.R.A. Devoto, G. (1969). Storia della lingua di Roma. Bologna: Cappelli. Ernout, A. (1954). Aspects du vocabulaire latin. Paris: Klincksieck. Fischer, I. (1965). Istoria limbii române: vol. I, Limba latină. București: Editura Academiei. Fischer, I. (1985). Latina dunăreană. București: Editura Științifică și Enciclopedică.

Les mots latins d’origine grecque avec diffusion panromane

293

Georgescu, S. (2012). On the integration of Greek words into Latin: The case of aer. Analele Universității din București, 61(1), 119–129. Haugen, E. (1950). The analysis of linguistic borrowing. Language, 26(2), 210–231. Meillet, A. (2004 [19281]), Esquisse d’une histoire de la langue latine. Paris: Klincksieck. Rey-Debove, J. (1973). La sémiotique de l’emprunt lexical. Travaux de linguistique et de littérature, 11(1), 109–123. Sganzini, S. (1937). La castagna nell’alta Italia e nella Svizzera italiana. Vox Romanica, 2, 77–103. Terlingen, J. (1943). Los italianismos en español desde la formación del idioma hasta principios del siglo XVII. Amsterdam: N.V. Noord-Hollandsche uitgevers maatschappij.

Chiara Fedriani, Maria Napoli & Nadia Rosso

Forms and functions of Greek words in Late Latin literary texts: a corpus-based approach Abstract: This chapter investigates forms and functions of Greek words in Late Latin literary texts which are part of a corpus specifically designed to study textual bilingualism. Firstly, we will offer a quantitative account of the distribution of Greek words across different genres, along with a semantic analysis which considers the most represented lexical fields. Secondly, some illustrative cases will be discussed in order to show the main functions of using Greek in a selection of Late Latin texts, namely Pelagonius’ Ars ueterinaria, Macrobius’ Saturnalia, and Servius’ In Uergilii Aeneidos Libros. Finally, we shall focus on a specific word category, that of compounds created with Greek lexical material, providing a morphological classification and a quantitative survey of their textual distribution. This chapter provides evidence for the advantages of examining ancient bilingualism, combining quantitative and qualitative perspectives thanks to a corpus-based approach. Keywords: Latin linguistics, textual bilingualism, single-word switches, compounds, Grecisms, textual genres

1 Introduction This chapter provides a corpus-based study of the use of Greek words in Late Latin literary texts belonging to different genres and covering a long historical period, from the third to the seventh centuries CE.

Acknowledgement: The research which led to this chapter is related to the Textual Bilingualism in Latin Project, granted by the Compagnia di San Paolo Foundation, to which we are grateful. Many thanks also go to the University of Eastern Piedmont for its financial support. We would like to thank Luigi Battezzato and two anonymous reviewers for their useful comments. This chapter is the result of joint work of the authors. However, Maria Napoli wrote Sections 1 and 4, Chiara Fedriani wrote Section 2, Nadia Rosso wrote Section 3. Chiara Fedriani, Università degli Studi di Genova Maria Napoli, Nadia Rosso, Università del Piemonte Orientale https://doi.org/10.1515/9783110647587-019

Forms and functions of Greek words in Late Latin literary texts

295

This corpus was built as part of the Textual Bilingualism in Latin (TBL) project and was designed to enable queries on specific phenomena from GreekLatin contact. Such phenomena were manually identified and then annotated using a tagset which was worked out by our research team as specific to textual bilingualism.1 Within this tagset, a fundamental distinction is made between two major types of phenomena, namely what we have labelled code-switching and single-word switch: in our terminology, code-switching refers to cases of switching from Latin to Greek in the form of a sentence, whereas single-word switches are insertions made up of a single word in Greek. This chapter deals with this latter type of phenomenon, with the aim of providing both a qualitative and a quantitative account of single-word switches in our corpus.2 Illustrative examples of our object of investigation are provided under (1) and (2); note that we have considered all words of Greek origin, irrespective of whether they were written in Greek (1) or in Latin script (2): (1)

quarum (sc. partium animae) prima est ratio quam λογιστικόν appellant, secunda animositas quam θυμικόν uocant, tertia cupiditas quae ἐπιθυμητικόν nuncupatur. (MACR. somn. 1, 6, 42)

(2)

. . . et id aperte noluisse dicere, sed ostendere per periphrasin. (SERV. Aen. 7, 740)

An important methodological point needs to be made from the outset, namely that, for the creation of our annotated corpus, we have decided to consider all Greek words used in isolation (written in Greek or in Latin script) as singleword switches, without distinguishing between loanwords and non-loanwords. In order to exclude very common integrated Greek forms from the annotation process, since they do not provide relevant information as to the degree of bilingualism of a text, we have annotated those Latinized words of Greek origin which are not lemmatized in a reference dictionary,3 also when attested in 1 The TBL corpus is available online for free at tbl.uniupo.it. The website also provides the project description, information on the XML tags we have used to annotate the texts, and an updated bibliography on bilingualism, with a special focus on language contact between Latin and Greek. 2 Analyses of forms and functions of code-switching in the TBL corpus, in turn, can be found in Fedriani and Napoli (2018) and Fedriani, Napoli and Rosso (forthcoming). 3 Specifically, Lewis and Short (1879). A Latin Dictionary. Oxford: Clarendon Press (henceforth L&S), available online at: www.perseus.tufts.edu/hopper/resolveform?redirect=true&lang= Latin. We are deeply aware that some Latinized words that, on the basis of this methodological choice, turn out to be included in our category of ‘single-word switches’ are lemmatized in

296

Chiara Fedriani, Maria Napoli & Nadia Rosso

a less ‘standard’ spelling not registered in the same dictionary (cf. n. 4). In this chapter, all single-word switches taken from our corpus are also labelled Grecisms, intended as a broad category which, in our case, includes Greek material that is more or less well integrated into the Latin lexicon. In what follows, we will concentrate on two main aspects, focusing, respectively, on functions and forms of Grecisms. In the first part of the chapter we will offer a functional and semantic account of Greek words in Late Latin texts (Section 2). We first provide a quantitative survey of the occurrence of singleword switches in different texts and genres, also discussing what are the Greek lexical sources which Late Latin authors most frequently resorted to (Section 2.1). Drawing on these data, we turn to a qualitative analysis of the main functions performed by single-word switches in three distinct texts which are representative of different genres, namely Pelagonius’ Ars ueterinaria, Macrobius’ Saturnalia, and Servius’ In Uergilii Aeneidos Libros, also comparing different degrees of lexical variation regarding the employment of Grecisms in the various texts (Section 2.2). In Section 3, we focus on the specific morphological category of compounds, offering an overview of the types attested in the corpus along with an analysis of their distribution across texts and genres. Section 4 summarizes the results and reassesses the role of single-word switches within the broader phenomenon of language contact between Late Latin and Greek, also discussing its link to the related strategy of code-switching.

2 Single-word switches in Late Latin literary texts In Section 2, we aim to present the results of our research on single-word switches in Late Latin, which, for our purposes, should also provide evidence for the advantages of using a corpus-based approach for the investigation of textual bilingualism. As we will seek to show here, quantitative and qualitative analysis based on a large annotated corpus may shed light on the relationship between the use of a second language and the specific communicative aims of literary works, offering

more comprehensive tools like the Thesaurus Linguae Latinae. In our view and in consideration of the methodological choices one has to make when building an electronic resource, this is a compromise allowing the exclusion of fully acknowledged loanwords (e.g. Lat. epistula from Gr. ἐπιστολή) and, on the other hand, guaranteeing the inclusion of a considerable amount of words of Greek origin. Further analysis may help clarifying the status of specific Greek items in the Latin lexicon.

Forms and functions of Greek words in Late Latin literary texts

297

the possibility of studying textual bilingualism from a general perspective (looking at Late Latin literature as a whole, in terms of particular genres: see Section 2.1), or from a more restricted perspective (looking at individual authors and texts).

2.1 Frequency, distribution and semantic characterization of single-word switches In this section we offer a quantitative account of the distribution of single-word switches in the TBL corpus. Table 1 reports their absolute occurrences in the texts surveyed, categorized according to the textual genres they belong to. As is readily apparent, texts with a clear metalinguistic function such as commentaries, scholiastic texts, and rhetorical works contain the higher number of Greek words. This should come as no surprise, since these texts often contain etymological explanations based on Greek, bilingual glosses, erudite references to Greek sources, and a general attention to lexical precision: what Adams (2003: 337–338) calls seeking the mot juste.

Table 1: Frequency of single-word switches in different textual genres in the TBL corpus. Genre Commentary Scholiastic texts Rhetorical texts Encyclopaedic texts Medicine Veterinary texts Physiognomics Astronomy Mythography Grammatical texts Historiography Philosophy Theatre criticism Architecture Epistolography Biography Geography

Number of Greek words                 

298

Chiara Fedriani, Maria Napoli & Nadia Rosso

A note of caution must be inserted at this point, however; namely that some of these textual genres gather the highest number of total words. Put differently, it could also be the case that we have the highest number of single-word switches in these works precisely because they are the longest in the corpus, and therefore the probability of finding Greek increases accordingly. Therefore, given the inevitable discrepancy between the texts belonging to different genres in terms of number of words, in order to make the different occurrences of single-word switches comparable, all the frequencies were normalized for occurrences per 10,000 words. The results of this normalization are provided in Figure 1, which shows some interesting changes in the frequency distribution just presented. Firstly, the textual genres which seem to be most dependent on Greek are those related to the medical domain, including veterinary medicine (Pelagonius’ Ars ueterinaria), medicine proper (Caelius Aurelianus’ Celeres passiones), and the theories of temperament and character pursued in the realm of physiognomy (De physiognomonia liber). Next come rhetorical and grammatical works, followed by specialized manuals on architecture and geography, and by Nonius’ lexicographic work De compendiosa doctrina. These data show that the texts which contain the highest frequency of single-word switches in relation to the overall size of each work, that is to say, the texts which show the highest ‘density’ of Greek, are practical treatises which are closely linked to the Greek tradition and its prestigious technical lexicon.

M

Ph ys io

gn

om ic s e Rh et dici or ne Ve ical te te xt Gr rina s ry am t m ex at ts ic al te Ar ch xts it Le ect ur xi e co gr ap Ge h og y ra M p En y cy tho hy gr cl op ap ae hy d Sc ho ic te lia xt s st Co ic t ex m t m en s ta r ie As tro s no m Ph ilo y so Bi phy og Ep ra is ph to y lo gr Hi ap st or hy io gr ap hy

100 90 80 70 60 50 40 30 20 10 0

Figure 1: Normalized frequency of single-word switches in different genres.

The distribution of Greek words across different genres goes hand in hand with their classification in lexical fields: the larger semantic class is that of technical terms pertaining to rhetoric, grammar and textual criticism (606 tokens, i.e. 28% of the total); next we find medical terms (355, corresponding to 16%) and

Forms and functions of Greek words in Late Latin literary texts

299

geographical and astronomical terminology (233, i.e. 11%). The general picture emerging from this quantitative analysis is that highly specialized genres contain the most Greek and, accordingly, Greek is often selected to mention technical terms without adding the filter of translation. Proper names and epithets deserve a separate mention, also considering their particular nature. They constitute the semantic class which is next in line after the three just discussed, being instantiated by 189 tokens in the corpus (9%). These lexemes are found above all in etymological explanations: arguably, referring to the Greek lexical root is crucial when reconstructing (putative) relations with other words (example 3). Alternatively, proper names and epithets are used by Late Latin authors in calling forth fresh and erudite displays of their deep knowledge of the Greek language and of the Greek literary sources (example 4): (3) Ἡρακλῆς enim quid aliud est nisi Ἥρας (id est aëris) κλέος? (MACR. Sat. 1, 20, 10) (4) unde et Homerus Τροίην εὐρυαγυῖαν dixit, id est Troiam latas plateas habentem. (FVLG. myth. 1, 21) In conclusion, the examination of the Greek words from a quantitative perspective together with a semantic account of them – in terms of a distinction between lexical fields – allows us to ‘measure’ lexical bilingualism in Late Latin literature. However, in order to measure and evaluate these types of phenomena in individual texts, we need to inspect the function of such forms more closely. With this aim in mind, we will now turn to Section 2.2.

2.2 A functional overview of single-word switches in three Late Latin texts In this section we shall provide a functional account of single-word switches based on three texts which represent different genres: Pelagonius’ Ars ueterinaria (late fourth century CE), a technical and practical manual on veterinary medicine regarding mostly horses and relying on the Corpus hippiatricorum Graecorum; Macrobius’ Saturnalia, an encyclopaedic work containing historical, mythological and grammatical discussions, full of opinions and citations from earlier Latin and Greek writers (fifth century CE); and Servius’ In Vergilii Aeneidos Libros, a detailed commentary of Vergil’s epic poem (fourth to fifth centuries CE). These texts have been selected because they contain a good amount of Greek words (162 tokens in

300

Chiara Fedriani, Maria Napoli & Nadia Rosso

Pelagonius Ars ueterinaria, 114 in Macrobius’ Saturnalia, and 282 in Servius’ commentary), thus enabling a detailed analysis. Starting from Ars ueterinaria, a first, basic point emerged from our analysis, that there exists a clear difference between the use of Greek on the side of Pelagonius, on the one hand, and of Macrobius and Servius, on the other. The motivation behind the insertion of single-word switches, indeed, rests on very different factors. Pelagonius employs Greek technical terms exclusively: names of plants and other botanic labels, vegetable oils, resins and other substances, such as hammoniacum, a resinous gum originally distilled from a tree near the temple of Ἄμμων, an Egyptian and Libyan deity. Therefore, the kind of Greek words found in Pelagonius’ manual involves an ‘instrumental’ use of the language. Single-word switches employed by Pelagonius refer to entities firstly discovered or used by Greeks, and whose translation would have been difficult or not necessary. Greek plants and substances were probably known through their Greek labels and referred to as such. Further evidence for the fact that Greek is first of all a necessary source for filling lexical gaps comes also from an important observation, namely that five tokens alone cover the 65% of the total occurrences of Greek words in Pelagonius’ text. This points towards a very low degree of lexical variation. Let us briefly describe these five single-word switches, in order to get a better understanding of the functions of Greek in Pelagonius’ work. The most frequent Greek word, occurring 27 times, is tracanthum ‘tragacanth’, a dried exudation obtained from the branches of natural strains of the Astragalus tree (< Gr. τραγάκανθα < τράγος ‘male goat’ + ᾄκανθα ‘thorn’, then a ‘goat’s thorn’: cf. Genaust 2013: 367). Next come ysopus, a kind of oregano (< Gr. ὕσ(σ)ωπον, probably, in turn, of Semitic origin, cf. Hebrew ʼēzōb), occurring 23 times, and ypopanacis, always in the Genitive (< Gr. ὀποπάναξ), referred to the roots of a plant commonly known as ginseng (see the discussion in Niedermann 1944: 231), occurring 22 times. We then find terebentina, an oleoresin (17 times), and troglitis (murra), a kind of myrrh from the land of Troglodytes, on the west coast of the Arabic gulf (see Ortoleva 1999: 197), used 16 times.4 The type/token ratio resulting from the frequency of these five Greek words indicates a relatively

4 We thank an anonymous reviewer for pointing out that Fischer’s edition of Pelagonius, selected for the TBL corpus, relies on a very late redaction, which incorporates numerous vulgar and non-standard spellings. The Grecisms mentioned above are attested before Pelagonius with different spellings reproducing more closely the Greek source: e.g., tracanthum as tragacanthus/dragantum (L&S s.v.); ysopus as hysoppus (L&S s.v.); ypoponacis as opoponax (L&S s. v.); terebentina as terebinthinus (L&S s.v.); troglitis as troglodytis (L&S s.v.).

Forms and functions of Greek words in Late Latin literary texts

301

restricted vocabulary, since these five types occur 102 times, over a total of 162 tokens. According to Adams (1995: 336), Latin-speaking veterinarii could insert Greek words in their treatises for different reasons: (i) ‘servile’ transfers from Greek sources; (ii) the use of labels for diseases which were first identified and described by Greeks, who also established the relative terminology; (iii) filling lexical gaps; and (iv) providing the text with ‘scientific impressiveness’. Our analysis shows that in Pelagonius Greek words meet urgent lexical needs and can fulfil functions listed from (i) to (iii) above. Erudite displays seem to be less relevant in this text. Let us now compare this picture with the data emerging from Macrobius’ Saturnalia and from Servius’ commentary. In these texts, Greek serves two main functions, namely that of providing etymological explanations, often resulting in creative excursus, and metalinguistic comments. Let us first consider etymologies. Both authors often venture into reconstructions of the lexical origin of Greek words: in such cases, citing words in Greek is clearly functional to highlight derivational relationships among words. Examples are (5) to (8) below: (5) Haec (scil. Medea) apud Graecos θεὸς γυναικεία dicitur, quam Varro Fauni filiam tradit adeo pudicam ut extra γυναικωνῖτιν numquam sit egressa, nec nomen eius in publico fuerit auditum. (MACR. Sat. 1, 12, 27) (6) Ἀπόλλωνα Δέλφιον vocant, quod quae obscura sunt claritudine lucis ostendit ἐκ τοῦ δηλοῦν τὰ ἀφανῆ aut, ut Numenio placet, quasi unum et solum. Ait enim prisca Graecorum lingua δέλφον unum vocari. Vnde et frater, inquit, ἀδελφός dicitur quasi iam non unus. (MACR. Sat. 1, 17, 65) (7) Neque minus Romani, ut pleraque alia ex Graeco, ita et lucem videntur a λύκῃ figurasse. Annum quoque vetustissimi Graecorum λυκάβαντα appellabant τὸν ὑπὸ τοῦ λύκου id est sole βαινόμενον καὶ μετρούμενον. (MACR. Sat. 1, 17, 39) (8) Dicti autem sunt Myrmidones propter hanc causam: Aeacus cum in arbore fici formicas, id est μύρμηκας vidisset, optavit tot sibi socios evenire. (SERV. Aen. 2, 7) Interestingly, these excerpts allow us to gain a better understanding of the level of competence in Greek mastered by the authors considered. Firstly, the etymological explanations provided often imply the reference to rare (γυναικωνῖτις

302

Chiara Fedriani, Maria Napoli & Nadia Rosso

‘women’s quarter’, example 5) or archaic words (δέλφος, a word of prisca Graecorum lingua according to Macrobius: example 6), this pointing to a deep knowledge of the Greek lexicon. Secondly, Macrobius in particular shows a high level of morphological and syntactic awareness. At a morphological level, he skilfully creates bilingual prepositional phrases featuring a Latin preposition governing a Greek noun in the case required by the Latin head (extra plus the accusative γυναικωνῖτιν, example 5), and is also able to provide a morphological parsing of the word ἀδελφός, which he interprets as being formed by a privative ἀ- prefix and δέλφος ‘one’ (i.e., a brother is who is ‘not one’).5 Evidence for Macrobius’ competence in Greek syntax comes in turn from the passage reported under (7), where he uses the Greek article τόν with an anaphoric function linking back to the Latin noun annum; moreover, when switching to Greek in using two Greek participles, he conjoins them with καί, probably in a mechanic way (cf., however, also example (16) below). This ‘spontaneous’ use of the Greek conjunction may testify, once again, the author’s fluency in written Greek. Thirdly, change of code can constitute in itself a strategic metalinguistic device signalling the specific contextual function of Greek, namely, that of explaining the meaning of a given root, as in example (8) from Servius. Such a skilfull insertion of Greek is based on the phonetic resemblance between the two lexical roots involved in the discussion (Myrm- and μύρμη-) and witnesses a very good mastering of the language. A similar case is found in bilingual glosses provided by Servius to explain the meaning of words or phrases. To arrive at a better definition, the author often combines Latin and Greek in synonymic chains, where the use of Greek may seem groundless at first glance. However, Greek insertions are sometimes motivated in terms of morphological relationship between Greek and Latin roots, as in example (9), where the two adjectives share the same root (< i.e. *g̑enh1-). (9) DIA CAMILLA generosa, εὐγενής. (SERV. Aen. 11, 657) In other bilingual glosses, the reason for the addition of Greek equivalents is even more subtle. An example is (10), where Servius paraphrases the meaning of the imperative sine – as occurring in Virgil’s verse quoted here as (11) – with three distinct Latin synonyms, to which he adds a fourth Greek analogue:

5 This is, by the way, a wrong etymology. The origin of the word lies in the univerbation of the copulative prefix ἁ- + δελφύς ‘womb’. Brothers are those who come from one and the same womb (see, e.g., Beekes 2010: 20).

Forms and functions of Greek words in Late Latin literary texts

303

(10) SINE permitte, relinque, patere, συγχώρησον. (SERV. Aen. 10, 598) (11) Vir Troiane, sine hanc animam et miserere precantis. (VERG. Aen. 10, 598) In examining the gloss in (10), the question arises whether there was a real need for the additional Greek word, even implying a switch, and what was this need. In order to provide an answer, we have to look at the context in which sine was used by Virgil: in (11), crucially, the imperative governs a direct object in the accusative, thus realizing an argument structure which is not usually instantiated by this verb. As a matter of fact, sino is typically construed with an object clause, with the subjunctive (especially in the imperative form)6 or is characterized by absolute use, whereas it occurs only ‘rarely with an accusative’ (L&S s.v.). Given that, one could hypothesize that the Greek verb συγχωρέω, usually taking the accusative case when used under the meaning ‘concede, to give up’, is selected as an illustrative counterpart for this peculiar syntactic use of sino in Latin (for which, however, there are other examples in Vergil). Another important function of single-word switches in Servius is that of resorting to appropriate technical terms pertaining to the specific vocabulary of rhetorical and textual criticism (which, needless to say, is paramount in a literary commentary). Examples (12) to (14) report passages in which the author prefers to use Greek terminology to refer to technical notions: (12) PINEVS ARDOR ignis περίφρασις. (SERV. Aen. 11, 786) (13) ‘Messapus’ autem ut diceret, vitavit ὁμοιοτέλευτον. (SERV. Aen. 11, 464) (14) ET TAMEN ἀνακόλουθον est; non enim praemisit ‘quamquam vicina sit’. (SERV. Aen. 3, 478) We can close this section by commenting the passage given as example (15), which is interesting because it combines in a very short comment the two main functions performed by single-word switches discussed so far, namely a seek for terminological precision in the lexical field of rhetoric (κατὰ ἀντίφρασιν) and an etymological intention (ἀχαίρων as an explanation for Charon): (15) CHARON κατὰ ἀντίφρασιν quasi ἀχαίρων. (SERV. Aen. 6, 299)

6 Compare, for instance, TER. Andr. 900 sine me expurgem.

304

Chiara Fedriani, Maria Napoli & Nadia Rosso

All the examples mentioned in Section 2 show how a close inspection of Grecisms in different texts allow to assess different reasons for using the Greek language, probably reflecting also different degrees of bilingualism by Latin authors (cf. also Section 4).

3 Greek compounds in the TBL corpus In this section we will concentrate on the specific category of compounds7 created with Greek lexical material, which are particularly relevant to illustrate Late Latin authors’ competence in Greek, as emerges, in particular, from examples found in Fulgentius, Ausonius and Macrobius, to mention some of them. Firstly, a morphological classification will be provided, focusing on the types of lexical categories involved in compounding and on their different combinations; secondly, we shall offer a quantitative account of the distribution of Greek compounds across literary genres, looking briefly also at the their functions. In order to classify Greek compounds from a morphological perspective, in general two criteria have been applied in the relevant literature: (i) the distinction between Determinativkomposita and Kopulativkomposita (Schwyzer 1939); (ii) the presence or absence of the head of a compound and its position. However, these criteria cannot be easily used for assessing the structure of ‘bilingual’ compounds, especially when dealing with forms corresponding to hapax legomena, since the interpretation of their meaning and, then, of their structure, is often challenging. Moreover, there are many forms which do not obey the canonical rules of compounding at least in the text of the reference editions. This may be illustrated quoting the following example from Fulgentius,8 in which the Greek compound tarematadamonta ‘(lit.) ruling the words’ is created by combining a noun phrase which incorporates the definite article and a participle:

7 For the purposes of our analysis we have adopted a broad definition of compounds, to be intended not only as forms which “involve a FC [first constituent] that does not correspond to a full ‘word’ but to a stem, and a SC [second constituent] which may consist of either a stem or an independently attested word, but which usually displays special derivational suffixes when used in compounding” (Tribulato 2015: 18), as Greek compounds typically do, but also as forms resulting from the process of juxtaposition, where the first constituent may be inflected. 8 The philological problems emerging from Fulgentius’ texts and possible new interpretation will be not dealt here. The reader is referred to Battezzato and Rosso (forthcoming).

Forms and functions of Greek words in Late Latin literary texts

305

(16) Radamantum enim Grece quasi tarematadamonta, id est uerbum domantem. (FVLG. Virg. cont. p. 101) For these reasons, we have decided to adopt Grandi and Pompei’s (2010) morphological classification based on the identification of the lexical categories which are combined to form the compound.9 More precisely, Greek compounds occurring in the TBL corpus are distributed across two distinct macro-groups, namely (i) nominal compounds ([[X]X [Y]Y]N) and (ii) adjectival compounds ([[X]X [Y]Y]A). Before presenting our classification in more detail, a brief remark is in order here with regard to the number of compounds attested in the TBL corpus, that is 300: among them, 153 forms correspond to the lexical category of nominal compounds, whereas the other 147 belong to the lexical category of adjectival compounds. As mentioned before, some of these compounds have never been attested previously, but are created ex nouo,10 although, in some cases, analogy with existing forms may have played a role: an example is sciomantia ‘necromancy’ (SERV. Aen. 6, 149), made up of the nouns σκιά ‘shadow’ and μαντεία ‘divination’, presumably modelled on the Greek compound necromantia having the same meaning. Coming back to the category of nominal compounds, we may observe that they are subdivided into three subgroups: (i) [[X]X [Y]N]N compounds, (ii) [[X]X [Y]V]N compounds and (iii) [[X]N [Y]Adv]N compounds. The different types belonging to the three subgroups are illustrated in Table 2 along with their absolute occurrences. As shown in Table 2, in the first subgroup ([[X]X [Y]N]N), the second member of the compound is always a noun, whereas the first member is typically a noun (84 occurrences) or an adjective (27 occurrences), or, less frequently, a verb, an adverb, a prefix, a numeral or a pronoun. Interestingly enough, the [N N]N type embraces many nouns from the lexical field of animal (such as cynoperdices), including both living species and fantastic beasts: e.g., arcoleontes ‘bear-lions’ (HIST. AVG. Gord. 33, 1), elephantocamelos ‘elephant-camels’ (NON. p. 120, 19), camphippi ‘camel-horses’ (NON. p. 120, 19), and the hapax nyctalopecas ‘nocturnal foxes’ (IVL. VAL. 3, 21 l. 467).

9 See also Tribulato (2015). 10 In particular, Fulgentius and Ausonius seem to create compounds ex nouo as a form of literary lusus. However, we must take into account that we do not always have evidence for attributing authorship of compounds found in Late Latin authors, which means that we are not always sure that compounds corresponding to hapax legomena were not previously attested as part of the Greek literary tradition.

306

Chiara Fedriani, Maria Napoli & Nadia Rosso

Table 2: Nominal Greek compounds in the TBL corpus. Compound type

Example

[N N]N type [A N]N type [V N]N type [Adv N]N type [Pref N]N type [Num N]N type [Prn N]N type [N V]N type [A V]N type [N Adv]N type

cynoperdices ‘dog-partridges’ (IVL. VAL. ,  l. ) μονοναύτην ‘captain’ (COMMENT. Lucan. , ) lipothymiam ‘faint’ (CAEL. AVR. chron. , , ) παλιλογία ‘repetition’ (PS. IVL. RVF. schem. lex.  p. , ) ἡμιθέους ‘demigods’ (COMMENT. Lucan. , ) duodecatemorion ‘twelfth part’ (FIRM. math. , , ) ταυτολογία ‘tautology’ (AQVILA rhet.  p. ,  E.) buleforunta ‘advisor’ (FVLG. myth. , ) homotomiam ‘identical incision’ (CAEL. AVR. chron. , , ) pegaseon ‘eternal fountain’ (FVLG. myth. , )

Number of Greek compounds          

The second subgroup ([[X]X [Y]V]N) is well represented by the [N V]N type (23 occurrences), having a noun as its first member and a verb as its second member, whereas only in two cases we find a compound made up of an adjective and a verbal root. Both types are often encountered in medical texts: this is the case of homotomia, from the adjective ὁμός ‘similar’ and the root of τέμνω ‘cut’. Finally, only two occurrences are found within the third subgroup ([[X]N [Y]Adv]N), both attested in Fulgentius. We shall examine now adjectival compound types, which may be subdivided into four subgroups: (i) [[X]X [Y]N]A compounds; (ii) [[X]X [Y]A]A compounds; (iii) [[X]X [Y]V]A compounds; (iv) [[X]X [Y]Adv]A compounds. Table 3 displays the different types belonging to the four subgroups and their absolute occurrences. The first subgroup ([[X]X [Y]N]A) is the most productive: as is evident from Table 3, in the major part of cases, the first element is an adjective (44 times), but it may also correspond to a noun, a numeral, an adverb, a verb or a pronoun. Quite frequently, the second element is a nominal-root rather than a free element, according to Grandi and Pompei’s model of compounds with ‘hybrid’ structure, namely [X]X [ROOT]N -ος,-ον / -ης, -ες: for example, hecatonpylos ‘with hundred doors’ (AMM. 17, 4, 2). Moreover, as exemplified by feronedon, formed by the participle φέρων ‘bringing’ and the noun ἡδονή ‘pleasure’, the nominal root may appear without the expected morphological ending. In the second subgroup ([[X]X [Y]A]A), the adjective-adjective combination (15 occurrences) is quite frequent in forming compounds which seem not have been attested before, such as olonxenos, made up of an adjective inflected in the genitive (ὅλων ‘all’) and the adjective ξένος ‘stranger’. In a few cases, the first element of the compound may also be a prefix or a pronoun.

Forms and functions of Greek words in Late Latin literary texts

307

Table 3: Adjectival Greek compounds in the TBL corpus. Compound type

Example

[A N]A type

πολιοκροτάφους ‘with white-haired temple’ (MACR. Sat. , , ) draconipedes ‘with snake's feet’ (SCHOL. Stat. Theb. , ) ἑπτάπυλος ‘with seven doors’ (SCHOL. Stat. Theb. , ) opisthotonos ‘stretched backwards’ (PELAGON. ) feronedon ‘bringing pleasure’ (FVLG. myth. , ) αὐτόχθονες ‘autochthonous’ (SERV. Aen. , ) olonxenos ‘stranger to all’ (FVLG. myth. , ) hemitritaicis ‘demi-triangular’ (CAEL. AVR. chron. , , ) αὐτοκίνητον ‘moving by himself’ (MACR. somn. , , ) ἡλιοβλήτους ‘beaten by the sun’ (MACR. Sat. , , ) ὁμοιόπτωτον ‘with the same ending’ (MACR. Sat. , , ) διχότομος ‘bipartite’ (FAV. EVL. , ) medenidon ‘knowing nothing’ (FVLG. myth. , ) meidusam ‘sightless’ (FVLG. myth. , ) gliconmeden ‘not at all sweet’ (FVLG. myth. , )

[N N]A type [Num N]A type [Adv N]A type [V N]A type [Prn N]A type [A A]A type [Pref A]A type [Prn A]A type [N V]A type [A V]A type [Adv V]A type [Prn V]A type [Conj V]A type [A Adv]A type

Number of Greek compounds               

The third subgroup [[X]X [Y]V]A, having a verbal root as its second element, includes noun-verb combination, adjective-verb combination, adverb-verb combination, pronoun-verb combination and conjunction-verb combination. The first type, which is the most common (23 occurrences), may be illustrated with the following case: (17) Denique inustos morbo Ἀπολλωνοβλήτους καὶ ἡλιοβλήτους appellant, et quia similes sunt solis effectibus effectus lunae in iuvando nocendoque, ideo feminas certis adflictas morbis σεληνοβλήτους et Ἀρτεμιδοβλήτους vocant. (MACR. Sat. 1, 17, 11) Example (17) shows that Macrobius uses four compounds with the same structure (the second element of these compounds is the verbal root of βάλλω ‘to hit’) in the same sentence, i.e. Ἀπολλωνοβλήτους ‘beaten by Apollo’, ἡλιοβλήτους ‘beaten by the sun’, σεληνοβλήτους ‘beaten by the moon’ and Ἀρτεμιδοβλήτους ‘beaten by Artemis’. Interestingly enough, the first two compounds are conjoined by the Greek form καί, the other two by Latin et ‘and’, maybe suggesting Macrobius’ intention of playing with the two languages. The last subgroup ([[X]A [Y]Adv]A), where an adjective and an adverb are combined, consists of only one form (gliconmeden) attested in Fulgentius.

308

Chiara Fedriani, Maria Napoli & Nadia Rosso

We have not included four special cases in Tables 2 and 3, which may be classified as ‘multiple’ compound types (not further represented in the TBL corpus): (1) [Det N V]A type, corresponding to tarematadamonta (see ex. (16)); (2) [Prn Adv V]A type, attested as autenunoe ‘not knowing herself’ (FVLG. myth. 2, 12); (3) [A Prp A]A type, documented by ἀλλοπρόσαλλον ‘being volte-face’ (PHYSIOGN. 73), which contains a prepositional phrase formed by πρός ‘to, towards’ and ἄλλος ‘another’, with the repetition of the last element as first member of the compound; (4) [N N N V]N type, corresponding to σαλσοστιχονυγοποιητής ‘poetaster’ (AVSON. 27, 6 [401 S.], 17). Finally, it may be noted that, apart from the Greek compounds described above, six hybrid forms – that is, forms resulting from the combination of Greek and Latin lexical material (Magni 2016) – are also attested in the TBL corpus, namely centippi ‘with hundred horses’ (FVLG. myth. 2, 14) and the following forms drawn from Ausonius’ macaronic epistle 6: gelidoτρομεροι ‘trembling for cold’ (6); teneroπλοκάμων ‘with tender curls’ (7); πολυcantica (13) ‘with many songs’; pinnoστέφανοί ‘crowned with feathers’ (14); πολυrisae (15) ‘with many laughs’. As mentioned at the beginning of this section, we have also investigated the quantitative distribution of Greek compounds across literary genres. The results of this investigation are reported in Figure 2, where the frequencies were normalized for occurrences per 10,000 words:

25

20

15

10

5

Ve te rin ar yt M ex yt ts ho gr Ph ap ys hy io gn om ic s Bi og Rh ra ph et or y ic al te xt s M Gr e am di ci m ne at En ic cy al cl t ex op ts ae di ct Ep ex is ts to lo gr Hi ap st hy or io gr ap Co hy m m e nt Sc ar ho ie lia s st ic te xt Ph s ilo so ph Le xi y co gr ap hy As tro no m y

0

Figure 2: Normalized frequency of compounds in different genres.

Forms and functions of Greek words in Late Latin literary texts

309

As can be observed from Figure 2, compounds are most frequently attested in veterinary texts and, next, in mythography. However, it must be remarked that Fulgentius’ mythological treatises present the greatest number of compounds (49 tokens), almost all appearing as hapax legomena (see n. 10). Moreover, it is worth stressing that the quantitative data concerning medical and rhetorical texts are consistent with the frequencies emerging from the analysis of Greek words (see Section 2). In sum, on the basis of our corpus-data and in view of the functional analysis just provided, we may conclude that there are different motivations which lead Late Latin authors to employ compounds. Such motivations partly overlap with what observed on the function of Greek words in Section 2, such as the insertion of technical terms according to a specific lexicon, their use in bilingual glosses, especially in the translation of Latin compounds, or their use in the explanation of the etymological origin of Greek lexemes, and, more in general, in contexts in which an author shows off his erudition and linguistic creativity, sometimes playing a sophisticated literary lusus.

4 Conclusions The analysis carried out in this chapter has shown that, generally speaking, Greek is often employed in Late Latin literary texts belonging to different genres, being attested with the highest frequency in medical treaties (including veterinary practice and physiognomics), followed by rhetorical and grammatical works. Interestingly, the data emerging from the normalized frequencies of Greek words, on the one hand, and, more specifically, of Greek compounds, on the other, show a very similar distribution: the only difference between the quantitative import of Greek in the two types of words in the TBL corpus is constituted by Fulgentius’ works, which contain a very high number of compounds. This is basically due to the fact that Fulgentius creatively switches to Greek, introducing many compounds to provide new etymological interpretations, often with an allegorical nuance, in order to distance himself from the previous tradition and at the same time to surprise the reader with an unexpected explanation, precisely conveyed by an uncommon, but yet transparent and understandable, word. Compounds, as we have seen, nicely accommodate the need of linguistically expressing the ambivalent nature of invented and fantastic creatures, often made up of two distinct animals. Along with lexical inventiveness, another major function of compounds is that of expressing technical concepts such as diseases: also in this case, the juxtaposition of two

310

Chiara Fedriani, Maria Napoli & Nadia Rosso

lexical roots allows to provide a more detailed and transparent description. This is in accordance with one of the main reasons underlying the employment of Greek words in general, which are often used precisely with the aim of introducing technical terminology, especially if depending on the Greek cultural tradition. The ludic dimension evoked in the discussion of Greek compounds allows us to touch upon another interesting point emerged from our analysis, namely the fact that many Late Latin authors skilfully played with Greek at different linguistic levels: from the graphic and phonetic domain, with strategic associations of Greek and Latin words in view of their resemblance, to that of morphological parsing, to lexical creativity and (more or less reliable) etymological reconstructions. In our view, playing with a foreign language in this way requires an advanced competence and constitutes in its own an important piece of evidence to assess a high degree of metalinguistic awareness. In conclusion, the data commented in this chapter can help us gain further understanding of a proficient reading and knowledge of Greek as a written language mastered by Latin authors, thus highlighting aspects and mechanisms of the specific phenomenon of textual bilingualism. We can therefore conclude that the ability of morphologically parsing Greek words, the suggestion of derivational relationships between Greek and Latin lexical roots, the creation of etymological analyses, of new compounds, along with other literary plays, all constitute a piece of evidence to assess the specific characteristics of bilingualism in terms of a written, educated and literary phenomenon in an ancient scenario.

Bibliography L&S = Lewis, C. & Short, C. (1879). A Latin dictionary. Oxford: Clarendon Press. www.perseus. tufts.edu/hopper/resolveform?redirect=true&lang=Latin (last accessed: 22 February 2019). TBL = Textual Bilingualism in Latin: tbl.uniupo.it (last accessed: 22 February 2019). Adams, J. N. (1995). Pelagonius and Latin veterinary terminology in the Roman Empire. Leiden/ New York/Köln: Brill. Adams, J. N. (2003). Bilingualism and the Latin language. Cambridge: Cambridge University Press. Battezzato, L. & Rosso, N. (forthcoming). I composti greci in Fulgenzio: nuove proposte filologiche. Beekes, R. (2010). Etymological dictionary of Greek. Leiden/Boston: Brill. Fedriani, C. & Napoli, M. (2018). Which type of bilingualism? A corpus-based approach to the use of Greek in Late Latin. Lingue e Linguaggio, 17(1), 129–147.

Forms and functions of Greek words in Late Latin literary texts

311

Fedriani, C., Napoli, M. & Rosso, N. (forthcoming). Un nuovo corpus per lo studio del bilinguismo testuale in latino tardo. Aspetti metodologici e primi risultati. In G. Haverling (Ed.), Latin vulgaire latin tardif XII: Actes du XIIe Colloque international sur le latin vulgaire et tardif (Uppsala, August 2016). Genaust, H. (2013). Etymologisches Wörterbuch der botanischen Pflanzennamen. Basel/ Stuttgart: Springer. Grandi, N. & Pompei, A. (2010). Per una tipologia dei composti del greco. In I. Putzu, G. Paulis, G. F. Nieddu & P. Cuzzolin (Eds.), La morfologia del greco tra tipologia e diacronia (pp. 209–232). Milano: Franco Angeli. Magni, E. (2016). Notha uerba: l’interferenza e il mutamento attraverso le formazioni ibride latino-greco. Lingvarum Varietas, 5, 133–144. Niedermann, M. (1944). Vegetianus. Museum Helveticum, 1(4), 231–233. Ortoleva, V. (1999). Publii Vegeti Renati Digesta artis mulomedicinalis, liber primus, introduzione, testo critico e commentario. Catania: Dipartimento di Studi antichi e tardoantichi. Schwyzer, E. (1939). Griechische Grammatik. vol. I: Allgemeiner Teil, Lautlehre, Wortbildung, Flexion. Munich: Beck. Tribulato, O. (2015). Ancient Greek verb-initial compounds. Berlin/Boston: De Gruyter.

Robert Maltby

Greek in Donatus’ Terence commentaries Abstract: The chapter discusses the use of Greek in Donatus’ Terence commentaries. Particular emphasis is placed on Greek in the manuscript tradition, the use of Latin or Greek scripts, Greek glosses on Latin words, comments on Hellenisms and Greek morphology, quotation from Greek authors, authorial comments in Greek and Greek critical and rhetorical terms, including adverbs in -ῶς. A concluding section draws comparisons between the Greek in Donatus’ Terence commentary and that of Servius and Servius auctus (Danielis) in the Virgil commentaries. Keywords: Latin linguistics, Donatus, ancient commentators, Servius, Greek loanwords, rhetorical terminology

1 Introduction In an earlier study (Maltby 2013) I investigated the use of Greek in Servius’ Virgil commentaries, especially his use of Greek critical terminology. As pointed out by Adams (2003: 323–329) this kind of code switching in Latin has a long history in letters and rhetorical works reaching back to the second century BC. It has its roots in the convention that the rhetorical schools at Rome made extensive use of a Greek critical vocabulary. The present chapter will compare Donatus’ use of Greek with that of Servius and Servius Danielis not only in the use of Greek technical terms derived from the Hellenistic Greek scholia and grammatical tradition, but also in the curious practice of using Greek terms to gloss Latin words. It will also take into account the choice of Greek or Latin scripts, comments on the morphology of the Greek loanwords, the use of Greek adverbs in -ῶς, and will touch briefly on problems concerning the manuscript transmission of Greek, as discussed in more detail in Bureau and Nicolas (2013). A final concluding Section 10 sums up the findings of Sections 2–9 and compares Donatus’ use of Greek with that of Servius and Servius Danielis.

Robert Maltby, University of Leeds https://doi.org/10.1515/9783110647587-020

Greek in Donatus’ Terence commentaries

313

2 The transmission of Greek in the manuscripts of Donatus The problem of identifying Donatus’ Greek is a difficult one (see Wessner 1902: vol. I, xliv); especially so since most of our witnesses are datable to the 15th century, and since the authentic text of Donatus himself had been dismembered, scattered in the margins of Terence manuscripts and reconstructed in the middle ages (see Reeve 1983). Most of these manuscripts make no attempt to reproduce Greek, but simply leave a lacuna where they encountered Greek or an earlier lacuna in their original. Two older manuscripts, the 11th century A (Parisinus lat. 7920) and the 13th century B (Vaticanus Reginensis lat. 1559), make valiant attempts to reproduce the Greek of their originals, as do two later editions, Stephanus (Paris 1529) and Lindenbrog (Paris 1602), which drew on important manuscripts no longer available to us (Wessner 1902: vol. I, xv–xvi). But Stephanus, Robert Estienne, tells us on his title page that he restored the Greek Graecis etiam repositis; so that most of the Greek it contains could simply be the result of brilliant humanist conjecture (Wessner 1902: vol. I, xxxv). Lindenbrog is more accurate in indicating his sources, but still in many cases the grounds for his restoration of Greek remain uncertain (Wessner 1902: vol I, xxxv–xxxvi). Besides these, our best sources from the 15th century are Vaticanus Reginensis lat. 1496 (V), Vaticanus Reginensis lat. 1673 (G) and Caesenas Malatestianus S.XXII.5 (M) (in which a later hand, M4, supplied the Greek, again possibly based on his own conjecture). Wessner (1902: vol. I, xliv) gives his own rationale in restoring Greek as follows: qua in re eam potissimum secutus sum rationem, ut omnibus locis Graecam restituerem formam, ubi aut optimi libri eam exhibent aut in quibusdam lacuna exstat, cum alii Latinam formam praebeant. However, in a number of cases where only transliterations are found in the codices Wessner restores the Greek, as for example with δεικτικῶς (following Stephanus and followed by Cioffi) at DON. Ter. Andr. 30, 2 and 333, 2, as discussed below in Section 9. For the purposes of this chapter, I will take my examples mainly from the Andria commentary, for which we are now fortunate to have a new edition (Cioffi 2017). In some special cases, such as Greek glosses on Latin words (Section 3), comments in Greek by Donatus (Section 7) and adverbs in -ῶς (Section 9) the full set of commentaries on all five plays discussed by Donatus will be taken into account.

314

Robert Maltby

In order to illustrate the way in which Greek is used in the commentaries and presented in the manuscripts I set out below some of Donatus’ comments on TER. Andr. 301: SO. nempe ut curentur recte haec? SI. immo aliud. SO. quid est quod tibi mea ars efficere hoc possit amplius? (TER. Andr. 30–31)

(1) HAEC δεικτικῶς. (DON. Ter. Andr. 30, 2) δεικτικῶς Steph. dic(h)ticos codd., cf. DON. Ter. Andr. 333, 2

(2) IMMO ALIVD bene ἀντέθηκεν τῷ ‘nempe’ τὸ ‘immo’. (DON. Ter. Andr. 30, 4) ἀντέθηκεν τῷ ‘nempe’ τὸ Wessner: ante ΕΗRΕΝΤLΕΝΕ peto A: ante peto K sp. interp: om. Sp. rel. Θ

(3) ARS ars ἀπὸ τῆς ἀρετῆς dicta est per συγκοπήν. ἀρετή autem uirtus est. (DON. Ter. Andr. 30, 5) ἀπὸ τῆς ἀρετῆς G2 Q M4 ft z: ΑΠΟΥΤΗCΑLΕΤΗ A: ΑNΟΥΤΗC ΑRHΤΗ P: apotu arres quod est uirtus μ: om. sp. rel. K Σ cyΝΚΟΝΝ A: syncopem KP Σ aΡΕΤΗ AP: arete K: om. sp. rel. Θ: arretis μ: ar(e)s Λ

[For other etymologies from Greek in Andria, see DON. Ter. Andr. 85, 1 DIC SODES . . . ‘sodes’. est autem ‘si audes’ . . . nam delirat, qui σῶος ζῇς interpretatur ‘sodes’; DON. Ter. Andr. 855, 7 CATVS ardens παρὰ τὸ καίεσθαι.] Apart from ἀπὸ τῆς ἀρετῆς in (3), none of the Greek is transmitted correctly in the manuscripts. Where Greek stood in the original the normal pattern is for A to make an attempt at it in capital letters and for the majority of manuscripts like T and C (group Θ in Cioffi) to mark a lacuna. In the case of (1) δεικτικῶς, it could be that a transliterated form had already stood in their archetype. The Greek script has been restored by Stephanus where only transliterations are found in the codd., as also at DON. Ter. Andr. 333, 2. In Servius there are four examples of this adverb = ‘in a deictic manner’, three in Greek script and one a curious hybrid between Latin and Greek at SERV. auct. Aen. 1, 106 hi pro illi uel alii, et bene deicticῶs.2 Wessner’s text probably underestimates the number of these transliterated forms in Donatus and replaces the majority with pure Greek, even when there is no manuscript evidence for it.

1 The sigla used will be those of Cioffi (2017: xxix–xxxi), the Terence quotations will be from Kauer and Lindsay (1958). 2 Wessner restores a similar hybrid form at DON. Ter. Eun. 313, 1 ἐγχωριarum.

Greek in Donatus’ Terence commentaries

315

I list in Section 8 below those rhetorical terms which occur mainly in Greek script in the manuscripts, those which occur in some passages in Latin and in some passages in Greek and those which occur in Latin script only, and Section 9 does the same for adverbs in -ῶς. All these are printed in Greek script in Wessner and Cioffi. As far as the content of the Greek is concerned, technical adverbs of the δεικτικῶς type (1) are common; other common ones found in Donatus are εἰρωνικῶς ‘ironically’ (11 occurrences, as opposed to 1 in Servius), ἠθικῶς (13 occurrences) ‘in accordance with his character’ and ἐλλειπτικῶς (12 occurrences) ‘elliptically’. These adverbs in -ῶς are frequent in all commentators, were readily transferred into Latin in either script, and (according to Adams 2003: 327 and n. 58) were a characteristic feature of the rhetorical register. They may often have Latin equivalents, for example moraliter in Donatus is used in exactly the same sense as ἠθικῶς; the two seem to be used interchangeably with no discernible difference in context. Section 9 below discusses the full set of these adverbs in Donatus’ commentary on all the plays he discusses.3 (2) illustrates two characteristic features of Donatus’ use of Greek. The form ἀντέθηκεν ‘he contrasted’ shows his willingness to conjugate Greek verbs within the framework of a Latin sentence (more examples are given in Section 7 below). Such freedom with comments in Greek is not found in Servius. Donatus (or his sources) may have had fewer qualms about reproducing the phraseology of Greek scholia and did not restrict himself simply to individual lexical items. Similarly the forms τῷ and τό here show him declining Greek articles. The use of the Greek articles in this way, to refer to particular words in the text, is again a feature of Donatus’ commentary4 not found in Servius. These forms confused later scribes and were often interpreted as parts of Latin words, cf. Andr. 30, 4 τό: peto A; Andr. 404, 2 πρὸς τό: poeto A: praesta(t) Θ; Andr. 803, 2 πρὸς τό: profecto Θ, Andr. 855, 7 παρὰ τό: parato A K. In (3) the phrase per συγκοπήν again illustrates the declension of Greek nouns in a Latin sentence. In the commentaries of Porphyrio and Servius the term syncope is always transliterated, showing the word was naturalised into Latin by the third century AD. In Donatus it occurs once in Latin and once in Greek script (see Section 8).

3 There is no commentary on Heautontimoroumenos. 4 Either to introduce a Latin word, as at DON. Ter. Andr. 235. 242, 1. 380, 1. 384, 1. 404, 2. 667, 2. 668, 2. 803, 2, or to introduce a Greek neuter form, as at 447, 2 and 798, 2.

316

Robert Maltby

Finally in (3) the phrase ἀπὸ τῆς ἀρετῆς illustrates the common Greek technical formula denoting an etymology. This is particularly common in the Latin grammatical tradition, where, as here, a Latin word (here ars) is derived from Greek. In the bracket after example (3) I list the other etymologies from Greek found in the Andria. These are all based on the ancient tradition, which appears first of all in Cato’s Origines and in Varro’s lost De Origine Linguae Latinae, that the people of Romulus’ time had learnt Aeolic Greek from Evander and the Greek Arcadians who came with him to Italy.5 This explained why Latin words could be derived from Greek, but not vice-versa. Having used this collection of notes containing Greek from Donatus commentary on TER. Andr. 30 by way of an introduction, I will now discuss in turn the varying contexts in which Donatus makes use of Greek, again illustrated mainly from the Andria commentary.

3 Greek glosses on Latin words A curious feature of Donatus’ commentary is the use of Greek to gloss Latin words (Craig 1948: 105). This oddity is also found in Servius’ Virgil commentary, where it is much more frequent in the extended version of this commentary known as Servius auctus or Servius Danielis, which probably included material from Donatus’ lost Virgil commentary.6 No doubt the aristocratic Roman students for whom Donatus’ commentary was intended would have had a good grounding in Greek, but why Greek glosses were needed in a Latin commentary intended for native Latin speakers is something of a mystery. One possible explanation is that Donatus’ commentary would contain material from earlier Roman grammarians, such as Probus of Beirut, who were teaching in a Greek-speaking environment where such glosses would make more sense. That the Greek words used could simply be the Greek found in Terence’s Menandrian or Apollodoran original, seems unlikely, as exactly the same technique is used in the Servian Virgil commentaries where there was no such exact Greek original. In fact there is little evidence that Donatus had access to the full Greek originals of Terence’s plays, but where he does quote Menander he is careful to name him (see Section 5 below). Of course, this does not exclude the possibility that some Greek quotations where Menander is not named could

5 See Maltby (1993: 49–51 and n. 11). The attribution to Cato and Varro is found in LYD. mag. 1, 5, 3 (p. 11, 17–21 W). 6 See Maltby (2013: 441–443).

Greek in Donatus’ Terence commentaries

317

also have come from Menander (see, e.g., (18) below, Andr. 543, 1) especially in view of the dismembered and scattered nature of the comments as we now possess them (see Section 2). In many of the examples the Greek gloss seems to clarify which of a number of possible meanings a particular Latin word is being used in, so: (4) QVIN TV VNO VERBO DIC ‘uno uerbo’ uno ἀξιώματι, una sententia, nam ἀξίωμα sententia est . . . ἀξίωμα enim constat ex nomine et uerbo. (DON. Ter. Andr. 45, 1) Cf. QVOD VERBVM AVDIO ‘uerbum’ pro ἀξιώματι more suo posuit, id est ‘uxorem dare’ (238) quod ait. (DON. Ter. Andr. 240) VERBVM pro dicto, sed proprie ἀξίωμα, id est sententia uel enuntiatio, quae uno stringitur et ligatur uerbo. (DON. Ter. Eun. 175, 2) ‘uerbum’ dictum intelligas, quod uerbo complectitur completae sententiae pronuntiationem, quod ἀξίωμα nominabatur. (DON. Ter. Eun. 178, 1) (5) me infensus seruat nequam faciam in nuptiis fallaciam. (TER. Andr. 212) SERVAT φυλάττει. (DON. Ter. Andr. 212, 2) So in (4) the Greek ἀξίωμα is used to show that in the Latin phrase uno uerbo the term uerbum is not being used in its normal sense of ‘word’, but rather, like the Greek ἀξίωμα and the Latin sententia, it is used to mean ‘sentence’, as is made clear by the phrase ἀξίωμα enim constat ex nomine et uerbo. A similar explanation is needed again at DON. Ter. Andr. 240 (see (4)) where uerbum refers to the phrase uxorem ducere noun + verb, and the comment more suo shows that Donatus takes this particular usage to be characteristic of Terence, as the further examples at DON. Ter. Eun. 175, 2 and 178, 1 show, (see (4)). Similarly in (5) the single Greek gloss SERVAT φυλάττει serves to make clear which of the possible meanings of seruat ‘save’, ‘serve’ ‘guard’ is operative here, namely ‘he keeps watch on me’. In the vast majority of other examples the Greek seems to clarify the meaning of words in Terence that have become archaic and fallen out of use in Donatus’ time. In this sense they provide a useful guide to elements of Latin vocabulary that were no longer current in the mid fourth century AD. So: (6) CIRCVITIONE τῇ περιφράσει. (DON. Ter. Andr. 202, 3) (7) VXORE EXCIDIT quod Graeci dicunt ἀπέτυχεν. (DON. Ter. Andr. 423, 4) (8) SCITVS elegans, pulcher, quem Graeci κομψόν dicunt. (DON. Ter. Andr. 486, 5)

318

Robert Maltby

(9) ‘puerpera’ πρωτοτόκος. (DON. Ter. Andr. 490, 6)7 (10) MVTIRE . . . quod Graeci φθέγγεσθαι. (DON. Ter. Andr. 505, 3) (11) PRORSVS . . . quasi ‘porro uersus’, quod Graeci dicunt μακράν. (DON. Ter. Andr. 510, 1) (12) ILICO quod Graeci dicunt αὐτόθεν, αὐτῇ, αὐτοῦ; nam loci significatio est, etiam breuitatem temporis notans. (DON. Ter. Andr. 514) (13) ERILEM FILIVM τρόφιμον. (DON. Ter. Andr. 602, 3) It is uncertain whether (6), circuitione in the sense of ‘periphrasis’, should be included here or with (4) and (5), since the Greek περιφράσει is simply clarifying that circumitio is being used in its transferred grammatical sense of ‘periphrasis’, rather than its literal meaning of ‘going round’. Certainly scitus, mutire, prorsus, ilico and erilem in erilem filium would have been archaic by Donatus’ time. In the case of uxore excidit (7) the Greek simply clarifies a possibly obscure phrase, as also perhaps with the technical puerpura (9), though the word was still current in the mid fourth century and Donatus uses this word himself in his commentary on Ter. Andr. 483, 3. Perhaps πρωτοτόκος is intended to give a more precise explanation of the word here as ‘giving birth for the first time’ (see TLL s.v. puerpera vol. X, 2, p. 2528, 46–49). Why a Greek gloss is required on perii in the following example is uncertain: (14) PERII ἀπόλωλα. (DON. Ter. Andr. 346, 2) In fact, perii occurs often enough in Roman comedy and elsewhere for it to be familiar to Donatus’ audience. Examples of such Greek glosses from commentaries on the other plays include: DON. Ter. Ad. 69, 2 QVI SVVM OFFICIVM FACIT ὁ τὸ καθῆκον ποιῶν; DON. Ter. Ad. 249, 3 SALTEM τὸ ἔσχατον; DON. Ter. Ad. 466, 1 AEQVALEM aequaeuum, ὁμήλικα; DON. Ter. Hec. 210, 2 haec exaggeratio est, quae δεινότης appellatur.

7 But cf. DON. Ter. Andr. 483, 3 quidam ‘ista’ ipsam puerperam dicunt where Donatus uses the word without explanation in his own commentary.

Greek in Donatus’ Terence commentaries

319

4 Greek in discussions of Hellenisms in Terence Donatus often mentions Hellenisms (Graecum, figura Graeca, ἑλληνισμός) in Terence’s Latin.8 The following examples occur in the Andria commentary: (15) EFFERTVR ‘efferre’ Graecum est (cf. Gr. ἐκφέρω), quod uitans Vergilius (georg. 4, 255–256), ne diceret ‘efferent’, ‘exportant tectis’ inquit ‘corpora luce carentum’. (DON. Ter. Andr. 117, 1) (16) NAVIGAT AEQVOR figura Graeca est; nos enim dicimus per aequor nauigat. similiter etiam alio loco (Aen. 12, 197) ‘terram, mare, sidera iuro’, cum Latinitas exigat, ut addatur praepositio ‘per’. (SERV. Aen. 1, 67) (17) NIHIL ME FALLIS figura ἑλληνισμός· οὐδέν με λανθάνοις ἄν. (DON. Ter. Andr. 204, 5) Cf. NIHIL ME FALLIS . . . sic Menander (Andr. frg. 35 K.-A) νῦν δ'οὐ λέληθάς με. (DON. Ter. Andr. 204, 4) (18) AH NE ME OBSECRA τῷ ἑλληνισμῷ· μὴ λιτάνευε, μὴ μάχου (Menander?, Andr. frg. 41 K.-A.) . . . pro ‘ne obsecres’. (DON. Ter. Andr. 543, 1) As the comment on effertur (15) shows, these were features which, although acceptable in poetry, were in Donatus’ view best avoided. Donatus sees efferre as a calque on the Greek ἐκφέρειν ‘to carry out the dead’ which he says is avoided by Virgil, who in the Georgics passage quoted uses exportare instead. As Servius’ comment on Virgil’s omission of a preposition in nauigat aequor (16) shows, such Greek figures were felt by the commentators to be alien to correct Latinitas. A common concern of both Donatus and Servius is that their audience should not be led astray by the poetic works they are reading into using Greek based and other phraseology which in prose would fail the test of pure Latin.9 The double accusative in nihil me fallis (17) and the ne + imperative construction in ne me obsecra (18) are simply stated to be Hellenisms, without any value judgement. Interestingly in his note 204, 5 (17) Donatus does not quote Menander’s Greek to illustrate his point, but makes up his own translation, which note 204, 4 (17) shows is different from Menander’s Greek. Elsewhere in the commentaries the phrase τῷ ἀττικισμῷ may be used for the same purpose, e.g. DON. Ter. Ad. 272, 1 HOC MIHI DOLET τῷ ἀττικισμῷ ‘mihi dolet’ pro ‘doleo’; DON. Ter. Eun. 45 QVID SIBI EVNVCHVS VELLET τῷ ἀττικισμῷ ‘sibi’.

8 For lexical Hellenisms in Latin, see Biville (1989). 9 See Maltby (2013: 443).

320

Robert Maltby

5 Quotations from Greek authors I list below (19–28) quotations from Greek authors in the Andria commentary. These come mostly, by way of comparison with Terence, from Menander’s Greek original, so (19–26), but occasionally Homer is quoted to illustrate linguistic or literary points, as in (27–28): (19) NIHIL ME FALLIS . . . sic Menander (Andr. frg. 35 K.-A.) νῦν δ'οὐ λέληθάς με. (DON. Ter. Andr. 204, 4) (20) uenit meditatus alicunde ex solo loco. (TER. Andr. 406) Menander (Andr. frg. 37 K.-A.) εὑρετικὸν εἶναί φασι τὴν ἐρημίαν οἱ τὰς ὀφρῦς αἴροντες. (DON. Ter. Andr. 406, 1) (21) NVNC PRIMVM FAC ISTA VT LAVET imperitiae notantes Menandrum aut Terentium ipsi ultro imperiti inueniuntur, nam et ille (Andr. frg. 39 K.-A.) λούσατ' αὐτήν dicens a consuetudine non recessit, cum λοῦσαι σαυτόν [se lauisse] a toto partem significet, et Terentius propius ad significationem accessit ‘ista’ dicendo, ne pudenda nominaret. (2) . . . ‘ista’ quae ex puerperio sordebant. quidam ‘ista’ ipsam puerperam dicunt – sic enim et Menander (Andr. frg. 39 K.-A.) λούσατ' αὐτὴν αὐτίκα – sed imperitiae accusantur, quod non continuo solent post puerperium lauare, sed diebus omissis. (DON. Ter. Andr. 483, 1–2) (22) LAC MIHI NON AESTATE NOVVM, NON FRIGORE DEFIT: multo melius quam Theocritus, ille enim ait (idyll. 11, 36) τυρὸς δ'οὐ λείπει μ'οὔτ' ἐν θέρει οὔτ' ἐν ὀπώρῃ. (SERV. ecl. 2, 23) (23) QVIDNAM AVDIO legitur et ‘audiam’; Menander enim sic ait (Andr. frg. 42 K.-A.) τί ποτ' ἀκούσομαι; (DON. Ter. Andr. 592, 1) (24) POSTHAC INCOLVMEM SAT SCIO FORE ME . . . Menander sic (Andr. frg. 43 K.-A.) ἂν θεὸς θέλῃ, οὐκ ἂν ἀπολοίμην. (DON. Ter. Andr. 611, 3) (25) EX ARA SVME HINC VERBENAS TIBI ‘ex ara’ Apollinis scilicet, quem Λοξίαν Menander (Andr. frg. 44 K.-A.) uocat. (DON. Ter. Andr. 726, 1) (26) SIC CRITO EST HIC MITTE hic Chremes traducit illum ab iracundia dicendo sic eum esse. Menander (Andr. frg. 48 K.-A) οὕτως αὐτός ἐστιν. (4) et recte, quia naturae ignoscitur, uoluntati non. (DON. Ter. Andr. 919, 3–4) (27) PVERVM AVTEM NE RESCISCAT . . . ‘puerum’ pro quolibet sexu . . . (3) . . . ut Graeci [pueros] παῖδας? Homerus (Il. 1, 255) Πριάμοιό τε παῖδες, et

Greek in Donatus’ Terence commentaries

321

Horatius (carm. 1, 12, 25) ‘dicam et Alcidem puerosque Ledae’. (DON. Ter. Andr. 400, 2–3) (28) summum bonum esse erae putabam hunc Pamphilum, amicum, amatorem, uirum in quouis loco paratum. (TER. Andr. 717–719) IN QVOVIS LOCO PARATVM . . . et maxime ad Homericam sententiam respexit (Il. 6.429–430) Ἔκτορ, ἀταρ σύ μοί ἐσσι πατὴρ καὶ πότνια μήτηρ / ἠδὲ κασίγνητος, σὺ δέ μοι θαλερὸς παρακοίτης. (DON. Ter. Andr. 718, 3) Menander and Homer are the only two Greek authors quoted in the Andria commentary.10 As pointed out earlier, we may doubt whether Donatus had access to the full texts of Terence’s Greek originals, but at least he must have had access to quotations of the Greek in earlier Latin commentators. What use, then, does he make of these comparisons with the Greek original? In (19) the quotation from Menander illustrates the linguistic point on nihil me fallis discussed in connection with (17) in Section 4 above, namely that this construction was, in Donatus’ view, a Hellenism. In (20) Menander’s words ‘proud men say the desert is good for making discoveries’ are simply given, without further comment, as Menander’s original for Terence’s less colourful version uenit meditatus alicunde ex solo loco. The comment in (21) with its quotation from Menander is more interesting and conforms to a type also exemplified in Servius. The midwife Lesbia is giving instructions after the girl Glycerium has just given birth. Donatus’ comments turn on the meaning of ista in Terence, as compared with αὐτὴν in Menander. In Terence Lesbia says ‘make sure these things (ista) are washed’, while Menander has ‘make sure you wash her (αὐτὴν)’. Donatus says that those accusing either Menander or Terence of ignorance of midwifery are themselves ignorant. Both Terence’s ista ‘those things’ and Menander’s αὐτὴν (a synecdoche: using the whole for the part) refer not to the girl herself, who should not be washed until several days after the birth, but to the pudenda. Terence, however, is described as coming closer to the required meaning (propius ad significationem accessit) by using the pronoun ista to refer directly to the pudenda. Similarly (22) in Servius on ecl. 2, 23 the claim is that the Latin text, referring to supplies of fresh milk in both summer and winter, is superior to its original in Theocritus, referring to supplies of milk in summer and

10 In commentaries on the other plays Donatus quotes the following Greek authors: Menander at Eun. 46, 4. 689, 1. Ad. 43, 1. 43, 3. 199, 1; Homer at Eun. 236, 5. 1072, 2. Ad. 480. 766, 1. 790, 1. Hec. 361, 2. 380; Aristophanes at Eun. 689, 1; Theocritus at Ad. 537, 2; Euripides at Hec. 214, 3; Apollodorus at Hec. 214, 5. 286, 5. 380. Phorm. 87, 4. 668, 1; Demosthenes at Phorm. 68, 2; Isocrates at Phorm. 252, 3.

322

Robert Maltby

autumn. Passing on to (23), Menander’s original with its future ἀκούσομαι is quoted to explain an alternative reading for audio in the Latin, namely the future audiam. Next (24), like (20) above, simply juxtaposes Terence’s Latin with Menander’s Greek without further comment. Possibly again the Menandrian version is more colourful than its Terentian adaptation. In (25) reference to the Greek is made to show that the temple referred to in Terence simply as ara was the temple of Apollo, as is shown by Menander’s use of the term Λοξίαν, an epithet of Apollo referring to the ‘ambiguity’ of his oracles, or possibly Ἀγυιαῖον, an epithet of Apollo guardian of the roads.11 Both epithets are suggestions originally of Meineke for the mangled Greek or lacunae of the manuscripts. In (26) Terence’s phrase sic Crito est is shown to be the exact equivalent of Menander’s οὕτως αὐτός ἐστιν ‘that is the way he is’, the point made by both being that a man can be excused for a character defect, but not for an act of will, as is made clear in note 4 on the same line, et recte quia naturae ignoscitur, uoluntati non. We pass on now to quotations from Homer (27–28). The first (27) is used to illustrate the lexical point that puer like the Greek παῖς can be used to refer to either a boy or a girl as at Iliad 1, 255 Πριάμοιό τε παῖδες, and to this he adds a Latin example from Horace’s Odes: puerosque Ledae. Finally, in (28) the famous quotation from Iliad 6, 429–430 of Andromache’s words to Hector, that he is her father, mother, brother and husband, is adduced as a parallel for the maid Mysis’ description of how suitable a match she thought Pamphilus was for her mistress Glycerium. The Homeric examples, of course, may well have originated in Donatus’ Virgil commentary, as the frequency of Homeric quotations in Servius’ Aeneid commentary shows.12

6 The morphology of Greek loanwords Occasionally, a note in Donatus discusses the morphology of Greek loanwords, usually nouns, giving the correct Latin termination: (29) PLATEA Graeci πλατεῖαν dixerunt, quam nos plateam dicimus. secundum formam eiusmodi et Μήδειαν Medeam, σπονδεῖον spondeum scribimus. (DON. Ter. Andr. 796, 3)

11 At DON. Ter. Eun. 85, 1 (where there is no mention of Menander) Wessner prints the same adjective in Roman script Agyiei, although there is evidence that the form was in Greek script in the mss. and Sabbadini’s Ἀγυιαίου is probably nearer the truth. 12 See Maltby (2013: 445–446).

Greek in Donatus’ Terence commentaries

323

Similar examples to those given here by Donatus, where the Greek ei diphthong becomes a long e in Latin, are also discussed in Servius’ note on Aeneid 1, 257: (30) CYTHEREA omnia quae apud Graecos ει diphthongon habent apud Latinos in e productum conuertuntur, ut Κυθέρεια Cytherea, Αἰνείας Aeneas, Μήδεια Medea. (SERV. Aen. 1, 257)

7 Comments in Greek by Donatus I list here cases in the Andria commentary like that bene ἀντέθηκεν τῷ ‘nempe’ τὸ ‘immo’ (32), also discussed as (2) in Section 2 above, where Donatus inserts a comment in Greek or in a mixture of Greek and Latin. Generally they speak for themselves without further need for explanation. It is however worth pointing out that the Greek τῷ δευτέρῳ συλλογισμῷ and λήμματι are found later in their Latinised forms at DON. Ter. Andr. 569 (see (37)). Such phrases tend to occur more frequently in Donatus than in Servius. (31) . . . sine officio prologi uel θεῶν ἀπὸ μηχανῆς. (DON. Ter. Andr. 28, 2) (32) IMMO ALIVD bene ἀντέθηκεν τῷ ‘nempe’ τὸ ‘immo’. (DON. Ter. Andr. 30, 4) (33) AMOR a necessario, MISERICORDIA ἀπὸ τοῦ ὁσίου. (DON. Ter. Andr. 261, 2) (34) τὸ ‘ego’ ἔμφασιν habet. (DON. Ter. Andr. 384, 1) (35) EX EA RE QVID FIAT VIDE ἀπὸ τῆς ἐκβάσεως. (DON. Ter. Andr. 385) (36) INVENIET INOPEM ὑποφορὰ εὔστοχος. (DON. Ter. Andr. 396, 1) (37) SI IN REM EST VTRIQVE VT FIANT τῷ δευτέρῳ συλλογισμῷ praue proposuit . . . nam in secundo λήμματι (Stephanus: λήμματe Wessner) negat, quod prius dixit. (DON. Ter. Andr. 546, 1) (contrast Latin syllogismus and lemmata at DON. Ter. Andr. 569 alter est syllogismus per contraria lemmata). (38) SI EVENIAT QVOD DI PROHIBEANT παρένθεσις per εὐφημισμόν. (DON. Ter. Andr. 568)

324

Robert Maltby

(39) EGO ME AMARE HANC FATEOR ἀμφιβολία τῆς ἐρωμένης (DON. Ter. Andr. 896) In commentaries on the other plays the longest example of this type occurs at DON. Ter. Eun. 167, 6, where Donatus appears to repeat a complete Greek scholion on the etymology of the word eunuchus: εὐνοῦχος εἴρηται ὡς εὐνὴν ἔχων, τοῦτ’ ἔστιν φυλάττων, ὡς ἡνίοχος ῥαβδοῦχος σκηπτοῦχος· εὐνὴν οὖν γυναικὸς κἀνδρός. At DON. Ter. Hec. 606 the Greek technical phrase οὐκ ὀρθῶς is used to question the reading of the text, while at DON. Ter. Phorm. 22, 1 and 270, 3 mention is made of the Greek rhetorical principle of ἀπὸ τῆς ἐπιεικείας τοῦ λέγοντος. Other longer comments in Greek are found at DON. Ter. Eun. 274, 1. 607, 3. Ad. 42, 1. 72, 2. 275, 2. 301, 3. Phorm. 17, 2. 295, 2.

8 Single-word critical and rhetorical terms in Greek I list here single-word critical and rhetorical Greek terms in Donatus’ Andria commentary, with an indication in brackets after each of the Latin author in whom they are first attested outside Donatus.13 Wessner prints them all in Greek script, but I separate them into three categories as found in the manuscripts: (a) Greek script in the best manuscripts and lacunae or transliterations elsewhere, as defined by Wessner (1902: vol. I, xliv) quoted in Section 2 above, (b) some passages in Greek script and others in transliteration and (c) all passages in Latin script. Where the same word occurs in some manuscripts in Greek script and in others in transliteration in the same passage, as with ἔλλειψις at Andr. 285, 1, the category it is assigned to will depend on how it is written elsewhere in the commentary: in this case (b), since of 14 different passages 10 are of type (a) (good manuscript evidence for Greek script) and 4 are of type (b) (no manuscript evidence for Greek script). (a) Greek script: ἀμφίβολος (PORPH.), ἀνάμνησις, ἀνθυποφορά (SEN.), ἀπόδοσις (PORPH.), ἀποστροφή (QVINT.), ἀρχαϊσμός (SERV.), ἀρχαίως, ἀσύνδετον, γνωμικός, δεινότης, δύο δι’ ἑνὸς (ἓν διὰ δυοῖν SERV.), ἐλάττωσις, ἐλλειπτικῶς (2), ἐμφατικώτερον (Gk. comparative in CIC., GELL., SERV.; see Adams 2003: 328), ἐνθύμημα (TIRO),

13 For a similar list of such terms in Servius, see Maltby (2013: 449–456).

Greek in Donatus’ Terence commentaries

325

ἐπανόρθωσις, ἐρωτηματικῶς, ἐφεξήγησις (SERV.), εὕρημα, ἐν ἤθει, ἠθικῶς (2) (SEN.), ἰδιωτισμός (2) (SEN.), μεταλημπτικῶς (2), μίμησις (4) (PORPH.), παρέλκον (10) (SERV.), παρέλκεται (2), παρόμοιον (6), παροιμία (2) (VARRO), τὸ πιθανόν (IVST.), τὸ πρέπον (3) (CIC.), συντομία, σχῆμα (PLAVT.), ὑπερβολή (QVINT.), ὑπερβολικῶς (SERV.), ἐν ὑποκρίσει (5) (SERV.), ὑποφορά, ὑφέν (3) (PORPH.). (b) Latin and Greek script: anacoloutcommenhon [Lat. 3, Gr. 4] (SERV.), antitheton [Lat. 1, Gr. 1] (PORPH.), aposiopesis [Lat. 3, Gr. 7] (QVINT.), auxesis [Lat. 1, Gr. 2] (PORPH.), ironia [Lat. 3, Gr. 1] (CIC.), ellipsis [Lat. 4, Gr. 10] (QVINT.), euphemismos [Lat. 1, Gr. 3], zeugma [Lat. 1, Gr, 1] (PORPH.), metaphoricos [Lat. 1, Gr. 1] (FEST.), oeconomia [Lat. 3, Gr. 3] (QVINT.), paradoxos [Lat. 1, Gr. 2] (CIC.), protaticon prosopon [Lat. 1, Gr. 1], syncope [Lat. 1, Gr. 1] (PORPH.), sylle(m)psis [Lat. 2, Gr. 5] (PORPH.). (c) Latin script only: amphibolia (2) (CIC.), anastrophe (2) (PLIN.), antiphrasis (QVINT.), apodicticos (adv.), asyndetos (adv.) (SERV.), catastrophe (PETRON.), character (VARRO), dicticos (adv.) (2) (SERV.), dicticos (adj.) (2), dilogia, ironicos (CIC.), emphasis (3) (QVINT.), epitasis (GELL.), hypocorisma, metaphora (QVINT.), metalempsis (QVINT.), metaplasmus (QVINT.), parasceue, parenthesis (QVINT.), protasis (GELL.), synecdoche (QVINT.), tmesis (SERV.).

9 Greek adverbs in -ῶς On the importance of these adverbs in the rhetorical tradition see the discussion of δεικτικῶς in Section 2. The aim of the present section is to identify all such adverbs as they appear in Greek script in Wessner’s text of Donatus’ full set of Terence commentaries (which treat all the plays except Haut.) and to identify (a) which occur in Greek script in the mss.; (b) which are conjectural, based on a lacuna in the text marking a lost Greek word; (c) which occur in the mss. in Latin script only. (a) Greek script is attempted in at least one manuscript: I list below those adverbs which are transmitted in Greek script in at least one manuscript and where a lacuna in others may also point to the presence of Greek in the text. It is clear that these should be printed in Greek script, as Wessner has them.

326

Robert Maltby

The indication (b) and/or (c) in brackets after a word shows that it also occurs in categories (b) and/or (c) below: ἀμεταπείστως Ad. 662; ἀπροδοκήτως Eun. 779, 1; ἀρχαιῶς Andr. 106, 4. Hec. 185, 1; ἀσυνδετῶς (b, c) Eun. 969, 4. 1057, 5. Phorm. 36, 5. 946; αὐξητικῶς (b) Hec. 319, 2; δεικτικῶς (c) Eun. 595, 3; διαλεκτικῶς Hec. 311, 1; εἰρωνικῶς (c) Eun. 237, 3. 288, 1. 908, 2. Ad. 187, 3. Hec. 436; ἐλλειπτικῶς (b, c) Andr. 361, 2. 468. Eun. 44. Hec. 270, 4. 875, 1. Phorm. 194, 1. 964, 2; ἐμπαθῶς (b) Hec. 198, 2; ἐμφατικῶς (c) Eun. 626, 1. Phorm. 112, 3. 200, 3; ἐρωτηματικῶς Andr. 794, 3; ἠθικῶς (b, c) Andr. 560. 637, 2. Eun. 214, 1. 310, 4. Ad. 61, 1. Hec. 550, 1. Phorm. 134, 1. 137, 1; ἰδιωτικῶς (b) Eun. 1001, 4; ἱλαρῶς Eun. 310, 4; καταχρηστικῶς Eun. 816, 3; μεταλημπτικῶς (c) Andr. 533, 4. 674; μεταφορικῶς (c) Andr. 192, 2. Eun. 312, 5. 354, 3; ὀρθῶς Hec. 606, παραδοξῶς Andr. 674, 4, περιφραστικῶς Phorm. 17, 2; πιθανῶς Ad. 654, 1; συντομῶς Hec. 805, 1; ὑπαλλακτικῶς Hec. 307, 1; ὑπερβολικῶς Andr. 676, 3. Eun. 231, 4. 310, 2. Phorm. 43, 4; φιλοσοφικῶς Eun. 588, 1. Of these Servius has only deicticῶς (see Section 2), periphrasticos (2) and hyperbolicos (3) in Latin script and none in Greek script. (b) A lacuna replacing Greek is marked in at least one manuscript: When a lacuna marked in at least one manuscript indicates missing Greek, but no Greek words are transmitted in any manuscript (either in Greek or in Latin script), then the Greek adverb printed in Wessner is purely conjectural; the majority are supplied by Stephanus (St.), but others are supplied by Lindenbrog (L), both of whom may have had access to manuscripts no longer available to us, or by Wessner (W) himself: ἀσυνδετῶς (a, c) Ad. 470, 3(St.). 482, 2(St.); αὐξητικῶς (a) Ad. 307, 3(St.). 561(St.). 682(St.). 994, 1(St.); ἀφελῶς Ad. 729, 1(St.); γνωμικῶς Ad. 164, 1(W.: νομικῶς St.); διασυρτικῶς Ad. 427(St.); ἐλλειπτικῶς (a, c) Eun. 849, 3(St.); ἐμπαθῶς (a) Ad. 301, 3(L). Hec. 205(L); ἐσχηματισμενῶς Hec. 311, 3(St.); εὑρητικῶς Ad. 938, 1(L); εὐσχημονῶς Ad. 666, 3(St.); ἠθικῶς Hec. 223, 1 (St.); ἰδιωτικῶς (a) Ad. 68, 2(St.). 559, 3(St.). 713, 4(L: Stephanus has ἰδιωτισμός), 849, 1(St.). 899, 1(St.). 916, 1(St.). Hec. 236, 2(St.). A further three cases should be added to this group, where a conjectural adverb does not fill a marked lacuna but replaces a Latin phrase in the text: ἁπλῶς Phorm. 914, 1 (Schoell: qui hunc codd.); ἐλλειπτικῶς Eun. 279 (W.: εἰρωνικῶς St.: est amicos / est eunuchos codd.); ἠθικῶς Eun. 14, 2 (Schoell: uos / nos / post codd.). The fact that different adverbs are suggested by Stephanus and Wessner at Andr. 164, 1 and Eun. 279 stresses the conjectural nature of this group. This needs to be kept in mind in cases where a single adverb could substantially affect our interpretation of the text, as with Lindenbrog’s εὑρητικῶς at Ad.

Greek in Donatus’ Terence commentaries

327

938, 1 emphasising Terence’s independence of Menander. Perhaps editors should mark a lacuna in the text and mention these conjectures only in the apparatus to make this point clear. (c) The adverb occurs only in transliterated form in the manuscripts: Although the majority of these adverbs also occur elsewhere in the text in Greek script, the fact that only transliterated forms occur in the locations listed below suggest that a form in Latin script may have occurred here in the original. In all cases Wessner or Cioffi prints in Greek script: ἀποδεικτικῶς Andr. 794, 3 (Cioffi, for ἐπιδεικτικῶς Wessner), ἀσυνδετῶς (a, b) Andr. 938, 1. Ad. 97, 1; δεικτικῶς (a) Andr. 30, 2. 333, 2; δραματικῶς Hec. 349, 2; εἰρωνικῶς (a) Eun. 660, 1. 861, 1. 1032, 2. 1091, 1. 1093, 1; ἐμφατικῶς (a) Phorm. 120, 1; ἐπιδεικτικῶς Andr. 794, 3 (Wessner, for ἀποδεικτικῶς Cioffi); μεταλημπτικῶς (a) Eun. 809, 3; μεταφορικῶς (a) Andr. 480. Eun. 1023, 2. In the case of a further four adverbs Latin script in some manuscripts is found alongside a marked lacuna in manuscript C, Ox. Canon. Lat. 95, suggesting missing Greek, which leaves the question of Latin or Greek script open in these cases: εἰρωνικῶς Ad. 395, 2; ἐλλειπτικῶς Ad. 70, 3. 330, 3. 758; ἠθικῶς Ad. 304, 3. 407, 1. 413, 2; μεταφορικῶς Ad. 835, 1. Of this group only deicticῶs occurs in Servius; ethicos occurs three times in the Elder Seneca (Contr. 2, Suas.1) and Porphyrio has four examples of hironicos and five of metaphoricos.

10 Conclusions Donatus follows in general the Roman rhetorical tradition in his use of Greek. Apart from comments in Greek (Section 7), where he is much freer than Servius, and the use of the Greek definite article to refer to Latin words (Section 2), which is avoided by Servius, all the main categories discussed here are to be found again in Servius’ Virgil commentary. One odd feature, the glossing of Latin words by Greek (Section 3), is more common after Donatus in Servius auctus than in Servius himself and may have been taken by this Danielis version from the lost Virgil commentary of Donatus. No adverb in -ῶς is found in Greek letters in Servius or Servius auctus and these authors are much more sparing in their use of such adverbs than is Donatus. In general the single word rhetorical technical terms (Section 8) and adverbs in -ῶς (Section 9) are more likely to be written in Latin script in Donatus than his modern editors are willing to admit.

328

Robert Maltby

Bibliography Adams, J. N. (2003). Bilingualism and the Latin language. Cambridge/New York: Cambridge University Press. Biville, F. (1989). Grec et latin: contacts linguistiques et création lexicale. Pour une typologie des héllénismes lexicaux du Latin. In M. Lavency & D. Longrée (Eds.), Actes du Ve colloque de linguistique latine: Louvain-la-Neuve/Borzée, 31 mars–4 avril 1989 (pp. 29–40). Leuven: Peeters. Bureau, D. & Nicolas C. (2013). Le grec de Donat: de quelques problèmes ecdotiques dans le commentaire au Phormion. In A. Garcea, M.-K. Lhommé & D. Vallat (Eds.), pp. 335–350. Cioffi, C. (2017). Aeli Donati quod fertur commentum ad Andriam Terenti. Berlin/Boston: De Gruyter. Craig, J. D. (1948). Specilegium Donateum. In Mélanges de philologie, de littérature et d’histoire anciennes offerts à J. Marouzeau par ses collègues et élèves étrangers (pp. 103–112). Paris: Les Belles Lettres. Garcea, A., Lhommé, M.-K. & Vallat, D. (Eds.) (2013). Polyphonia Romana, hommages à Frédérique Biville. Hildesheim: Olms Verlag. Kauer, R. & Lindsay, W. M. (Eds.) (1958). P. Terenti Afri Commoediae (Oxford Classical Texts). Oxford: Clarendon. Maltby, R. (1993). Varro’s attitude to Latin derivations from Greek. Papers of the Liverpool Latin Seminar, 7, 47–60. Maltby, R. (2013). Greek in the Virgil commentaries of Servius. In A. Garcea, M.-K. Lhommé & D. Vallat (Eds.), Polyphonia Romana, hommages à Frédérique Biville (pp. 441–458). Hildesheim: Olms Verlag. Reeve, M. D. (1983). Aelius Donatus. In L. D. Reynolds (Ed.), Text and transmission: A survey of the Latin classics (pp. 153–156). Oxford: Oxford University Press. Wessner, P. (1902–1905). Aeli Donati commentum Terenti: Accedunt Eugraphi commentum et scholia Bembina (vols. I and II. Repr. 1962–1963). Stuttgart: Teubner.

Šime Demo & Nathalie Tassotti

Verb conjugation selection in macaronic Latin: a corpus-based analysis Abstract: We propose a comparative analysis of basic verb morphology in macaronic Latin, a peculiar family of artistic idioms that were used as a literary device in various parts of Europe starting in the Italian Renaissance and lasting until the present time. In macaronic Latin, a significant portion of the vocabulary is hybrid, generated by attaching Latin endings to lexical bases from several modern languages. In the present study we observe the frequencies of individual verb conjugations in a corpus of macaronic works and in a selection of ancient Latin texts. The results suggest a twofold conclusion. Firstly, the pattern of verb conjugation selection in hybrid verbs differs significantly from that in Latin verbs; secondly, among the hybrid verbs, the selection preferences are affected by the morphonological properties of the language involved in combining with Latin. Keywords: Latin linguistics, macaronic Latin, language mixing, Latin morphology, verb conjugation, correspondence analysis

1 Introduction 1.1 Conjugation classes in Latin The conjugation class of verbs is not a phenomenon that is normally researched in its own right. Unlike verb dimensions such as aktionsart, aspect, mood, number, person, tense, and voice, the conjugation class does not have direct syntactic,

Acknowledgement: This research was cofunded by the Austrian Academy of Sciences and the University of Zagreb. Šime Demo wants to thank Mislav Stjepan Žebec and the members of the Ludwig Boltzmann Institute for Neo-Latin Studies and of the Department of Languages and Literature (Classical Philology) at the University of Innsbruck for their warm hospitality and invaluable help during his research stay in Innsbruck in 2017/2018, which was crucial for the work on the present chapter. Šime Demo, University of Zagreb Nathalie Tassotti, University of Vienna https://doi.org/10.1515/9783110647587-021

330

Šime Demo & Nathalie Tassotti

semantic, or pragmatic implications.1 Consequently, it is a research topic of very limited relevance. Some of its features include the following: it is only sporadically and weakly connected with the verbʼs syntactic and semantic behaviour in a text, it is undeterminable or unclear for some verbs (e.g. sum and volo), there are overlaps between the classes (e.g. capio-type verbs), some verbs (such as tergeo vs. tergo) can oscillate between conjugation classes, and the division into conjugation classes fades when one goes beyond the present stem (as in cubui vs. praebui vs. procubui). The conjugation class is mainly viewed as representing a superficial phenomenon, particularly providing a didactic aid, and is often perceived as being a mere afterimage of an earlier linguistic state or of a semantic grouping. As a result, verb conjugation as a concept is frequently treated only marginally or indirectly; for example, in relation to general etymological questions,2 phenomena in archaic Latin,3 morphological errors resulting from insufficient linguistic competence,4 substandard Latin and its transformation into Romance languages,5 or morphologically and semantically grouped verb clusters.6 Apart from being able to serve as pointers to internal processes of language change (such as in explaining morphological analogy or the relationship between a daughter language and its parent language), conjugation classes can be involved in contact linguistic research. When a word enters a language, an inflectional class is assigned to it.7 The class choice depends on multiple factors, including the phonetic settings and quantitative prevalence of individual classes in the source and target languages. Therefore, it is in the area of lexical innovation that the division into conjugation classes can be fruitfully used for research.

1 We use the term to refer to the fact that in some languages verbs are divided into morphological types – for example, that Latin has four conjugations. 2 E.g. Meiser (1998: 179–202) on the PIE source of the conjugations. 3 E.g. De Melo (2007: 224–225, 298, 307, 315, etc.) on archaic Latin forms limited to certain conjugation classes. 4 E.g. Adams (2004: 741–750) on verbs in Claudius Terentianus and Vindolanda texts. 5 E.g. Grandgent (1907: 16–17) on verbal derivation, and (1907: 166–170) on verbal inflection in substandard Latin, Blaise (1955: 70) on the Latin of the Christians, Väänänen (1966: 87–88, 103–104) on the verbs in Pompeian inscriptions, Herman (2000: 70) on the verbs changing class in substandard Latin, Klausenburger (2000: 37–66) on the transition of verbs from Latin to Italian and French, Alkire and Rosen (2010: 127–144) on Latin and Romance verb morphology. 6 E.g. Haverling (2000) on verbs in -sco, Martzloff (2015) on -Cāre/-Cuī verbs, Viti (2015) on frequentative verbs. 7 E.g. a family of verbal loans from Greek in -izo (Burton 2000: 147), entering the Latin aconjugation.

Verb conjugation selection in macaronic Latin: a corpus-based analysis

331

1.2 Macaronic Latin One of the most peculiar contact linguistic phenomena involving Latin is represented by macaronic literature. It is written in macaronic Latin, which is an early modern parody of Classical Latin, consisting linguistically of regular Latin text and lexical hybrids composed of modern vernacular stems and Latin endings, as in the following example: (1)

Angla Flochosque canam, qui wachsunt pulvere schwarzo E Waßroque simul fließente et Schweißide warmo. (Anonymous, Floia [16th century] 1–2. In Schade 1855: 31)8 ʻI will sing of barbs and fleas, which rise from the black dust As well as from running water and warm sweat.ʼ9

All the function words are Latin, as are the inflectional endings and some of the bases, while the remaining bases in this example are German (typeset here in bold). The content is usually humorous and satirical.10 The macaronic tradition originated in Renaissance Italy and lasted from the late 15th century until at least the 1990s. Several hundred thousand lines of such literature have been discovered to date. No less than eleven European languages are paired with Latin to produce these texts.11 The vast majority of the linguistic analyses of macaronic Latin address the work of Teofilo Folengo (1491–1544), who was the main model for almost all the later authors.12 Some research has been done on non-Italian

8 Although the original version of the poem is written in Low German, we use the High German version from 1689, which was more widely known than the original version (Dahl 1962: 37). The title of the version quoted here is Flochia. 9 The translation is our own. 10 The term ʻmacaronicʼ is sometimes used in a wider sense, comprising various kinds of linguistic mixtures. In the present chapter we consider only macaronic works that originated in the Renaissance tradition and are based on lexical hybridisation. For a more detailed discussion on the definition of the macaronics, see Demo (2014). 11 In all cases of early modern macaronics the prestige language (Latin) is the frame language, providing functional morphemes (and most lexemes), while the other language (in most cases it is the L1) typically provides only lexemes. Other constellations of languages in linguistically mixed texts can, of course, be found, but not in the texts belonging to humanist macaronic tradition, characterised by high frequency of lexical hybrids and parodying Classical Latin poetry. 12 E.g. Bonora (1956), Paoli (1959), Chiesa (1972).

332

Šime Demo & Nathalie Tassotti

macaronics13; however, most of these studies investigate languages in isolation. In the present chapter, we perform a comparative analysis involving all macaronic linguistic combinations known to us.

1.3 Lexical productivity in the macaronics The period of the macaronics belongs to the last centuries of the development of Latin, in which lexical enrichment of the language has been limited, predominantly involving technical vocabulary required by various scientific domains.14 Macaronic Latin, on the other hand, consistently produced new lexical items without a practical need and purely to achieve literary goals. Such coinages were, to use Bauerʼs (2001: 36–41) terminology, existent, but not established (being item-familiar to the users) outside the limits of the genre. They fall under the heading of authorial creativity, but are not necessarily limited to one author.15 Some of them were adopted by later authors (but typically only within the macaronic genre).16 Readers of macaronic works did not possess item-familiarity with individual hybrid words, but they were aware of the pattern of hybridisation, which makes the process infinitely productive. All macaronic words are potential words within the macaronic system. Thus, every item constructed in accordance with the macaronic rules can be understood provided that the reader is familiar with both languages involved. Moreover, the routine creates humorous effect, particularly when additional strategies are applied (such as intralinguistic modifications, semantic novelty, or deliberately confused spelling). Without macaronic context, they remain nonce formations.

13 E.g. Giraud (1977) and Garavini (1978) on Provençal macaronics, Glidden (1982) on French tradition, Keipert (1988) and Ballester (1997) on Latin-Polish mixtures. 14 See e.g. Helander (2016) for a classification and many examples. 15 Bauer (2001: 57–58) distinguishes individual productivity from societal productivity, which is the only productivity that takes part in language development. 16 For example, macaronic word culamen ʻbuttʼ (from culus) is found in T. degli Odasiʼs Macaronea (632), in T. Folengoʼs Baldus (e.g. 1, 344; 1, 406; 1, 468), in R. Bellauʼs Dictamen metrificum de bello Huguenotico (18, 89, 126), in J.-B. Richardʼs Cagasanga Reistrosuyssolansqnettorum (97), in J. Kransfeltʼs reply to Richard Ad Caquasangam Joan. Baptistae Lichiardi Poetae Spaliporcini Reistrorum Macaronica defensio (19) (T. degli Odasi is from the late 15th century, while the others flourished in the 16th century).

Verb conjugation selection in macaronic Latin: a corpus-based analysis

333

2 Methodology 2.1 Research question Although the use of hybrid words that make up the core of macaronic strategy is strictly limited to the genre, they are formally no different from fully integrated loans such as phalanx or sabbata. In other words, they belong grammatically to Latin and behave like regular Latin words. However, whether and to what extent the strategy of mixing depends on the properties of the language that is coupled with Latin in the macaronics has remained unclear. In order to investigate this, we performed the first comprehensive and comparative analysis of macaronic Latin morphology. It has been our intention to test the hypothesis that the choice of the language that is intertwined with Latin influences the way the mixing is realised. The topic of the analysis is the selection of the verb conjugation class in hybrid verbs. We have wanted to see whether morphonological properties of involved languages affect the preferences in verb conjugation selection. In the present study, we analyse the distributional tendencies of four Latin verb conjugation classes across several subsets of data. Firstly, their frequencies are observed in a group of ancient Latin texts, as well as in macaronic literature. In the latter subcorpus, Latin and hybrid verbs are distinguished. We expect to find differences between hybrid verbs on the one hand and Latin verbs on the other. Secondly, we attempt to see whether and in what way the distribution of conjugation classes of hybrid verbs depends on the choice of language that is mixed with Latin. This is also done by observing the frequencies, but within the set of hybrid verbs. After analysing the correlation of the language involved and the conjugation class distribution, we offer our interpretation of the findings in the discussion section.

2.2 Corpus and data We have used two corpora for the present study. The corpus of macaronic poetry consists of samples from 60 macaronic works, representing all 11 language pairs, balanced according to the relative size of individual local traditions, and tagged according to the linguistic affiliation as Latin, hybrid, or embedded language words. The word count is 22,757 and the average sample length is

334

Šime Demo & Nathalie Tassotti

approximately 379 words. The main control corpus of ancient works comprises 11 samples from several ancient prose authors and poets, totalling 33,728 words and averaging around 3,066 words per sample.17 Since macaronic works contain both hybrid and Latin verbs, these groups have to be analysed separately. Latin verbs in macaronic texts serve as the additional control group, because their behaviour is not expected to diverge significantly from the behaviour of the verbs in regular Latin texts. The structure of the corpus is presented in Table 1. Table 1: Overview of the corpus used for the present study. Macaronic texts

Ancient Latin texts

 samples ( language pairs) Romance: French ( texts), Italian (), Portuguese (), Provençal (), Spanish (); Germanic: Dutch (), English (), German (); Slavonic: Croatian (), Czech (), Polish ()

 samples Caesar, Cicero, Jerome, Ovid, Petronius, Phaedrus, Propertius, Sallust, Suetonius, Tacitus, Virgil

Words tagged as Latin, hybrid, or embedded language

Words with full morphological annotation

, words

, words

, hybrid verbs

, Latin verbs

, (Latin) verbs

Total: , words, , verbs

We exclude from the analysis Latin verb forms that do not display conjugation class transparently, such as participles, gerunds, and gerundives, as well as verbs with peculiar or incomplete morphology that do not definitely belong to a major conjugation class (e.g. volo and its compounds, fio and its compounds, inquam, and aio). Likewise, all hybrid verb forms that do not unambiguously declare their conjugation class are excluded. A total of 7,632 verb items remain for the analysis.

17 The ancient texts are taken from Celano et al. (Eds.), The ancient Greek and Latin dependency treebank. As the verbs are lemmatised there, it was possible to extract conjugation classes automatically. 18 The inclusion of Jerome and Petronius made us use the wider term ʻancientʼ for the subcorpus, rather than ʻclassicalʼ.

Verb conjugation selection in macaronic Latin: a corpus-based analysis

335

3 Results 3.1 Comparison of the three verb groups The most basic way of examining the data is a direct observation of the item frequency. The raw counts of each verb class are presented in Tables 2, 3, and 4.19 Table 2: Frequency of verb conjugations in ancient works. Name

First

Second

Third

Fourth

Total

Caesar Cicero Jerome Ovid Petronius Phaedrus Propertius Sallust Suetonius Tacitus Virgil

          

          

          

          

          

Total











Table 3: Frequency of verb conjugations of Latin verbs in macaronic works. Language

First

Second

Third

Fourth

Total

Croatian Czech English Dutch French German Italian Polish Portuguese Provençal Spanish

          

          

          

          

          

Total











19 The frequencies in the macaronic texts are grouped together according to the language mixed with Latin.

336

Šime Demo & Nathalie Tassotti

Table 4: Frequency of verb conjugations of hybrid verbs in macaronic works. Language

First

Second

Third

Fourth

Total

Croatian Czech Dutch English French German Italian Polish Portuguese Provençal Spanish

          

          

          

          

          

Total











The relative distribution of the four classes within each sample is shown in Figure 1. One can observe that in the ancient corpus (left-hand panel) the third conjugation dominates, the fourth conjugation is the rarest one, and the remaining two are in the middle, with a moderate preference for the first conjugation. Looking at the Latin verbs in the macaronic corpus (middle panel), we encounter a similar picture. The ratios are sporadically somewhat warped (e.g. the fourth conjugation is absent from some texts, and there are two texts in which the first conjugation is extremely rare), but this can be accounted for by the smaller sample sizes and the presence of the hybrid verbs in the same passages. In general, the distribution does not seem to differ considerably from the one we observe in the ancient subcorpus. The hybrid verbs (right-hand panel), on the other hand, display a radically different pattern. The second conjugation is almost non-existent, while the fourth conjugation appears in less than one-third of the samples. Even more interesting are the two remaining classes. It seems that most texts strongly prefer the first conjugation, and others the third, sometimes to the point of the exclusion of the opposite class. The plots strongly suggest that hybrid verbs select conjugation classes differently from the way regular Latin verbs do. In order to see the differences and the similarities suggested by the bar plots more clearly, we use the correspondence analysis, which is a statistical technique that allows us to represent visually the relationships between several data series. This method is based on converting frequencies into distances, using the chisquared (χ2) distance. This makes it possible to represent the data in a coordinate

Proportion

0.00

0.25

0.50

Ancient Latin

Propertius

Phaedrus

Petronius

Ovidius

Hieronymus

Cicero

Caesar

Text

Vergilius

Tacitus

Suetonius

Sallustius

0.75

0.50

0.25

0.00

0.75

0.50

0.25

0.00

Text

1.00

Latin in macaronics 1.00

Figure 1: Bar plots of the relative proportions of the four verb conjugations in the three subcorpora.

First

Second

Third

Fourth

Conjugation

0.75

1.00

Text

Hybrid

Verb conjugation selection in macaronic Latin: a corpus-based analysis

337

338

Šime Demo & Nathalie Tassotti

system, called a biplot, in which the geometrical distance between the points representing observations corresponds to the differences between their respective counts. The biplots also display the distance of the observations from the categories to which they are related. The interpretation is straightforward – the closer the points on the plot are to each other, the more similar are the data they represent.20 If we apply correspondence analysis to the frequency of verb conjugation choice in our texts, plot the result onto the graph, and mark the points according to whether the verbs are from the ancient Latin texts, the Latin part of the macaronic texts, or the hybrid part of the macaronic texts, we obtain the following biplot (see Figure 2). Biplot of conjugation distribution: 3 subcorpora 0.8

fourth

Dim2 (20.2%)

0.4

second

0.0

first third

–0.4

–0.8 –0.5

0.0

0.5

1.0

1.5

Dim 1 (72.1%) Figure 2: Biplot of the correspondence analysis of verb conjugation distribution across the three subcorpora. Dark squares represent hybrid verbs in individual texts, medium grey dots refer to the ancient texts, and light grey triangles stand for Latin verbs in the macaronics.

20 About correspondence analysis see Greenacre (2007), Glynn (2014), and Beh and Lombardo (2014). For an application to Latin data see Passarotti et al. (2015).

Verb conjugation selection in macaronic Latin: a corpus-based analysis

339

The percentages on the axes show that 92.3% of variance is displayed on the map (of which 72.1% is on x-axis). This means that the reduction of the dimensions to a two-dimensional display resulted in loss of only 7.7% of the patterns on the plot (which means that the plot captures the patterns pretty accurately), and that 72.1% of the patterns is accounted for by the x-axes. Each point represents a text from our corpus. Hybrid macaronic verbs (which are represented by darker squares) are clearly separated from Latin verbs in the macaronics (light grey triangles) and from Latin verbs in ancient texts (medium grey dots, most of them close to the intersection of the axes). In addition, it seems that the centres of the distribution of Latin verbs in the macaronics and Latin verbs in the ancient texts are very close to each other, with Latin verbs in the macaronics being more dispersed. The density lines show a clear grouping in three areas, of which two belong to the hybrid verbs. Arrows lead to the points that represent conjugations. The biplot suggests, as do the bar plots discussed above, that most hybrid verbs in some texts strongly prefer the first conjugation, and that there is a group of texts that tends to choose the third conjugation. Correspondence analysis also reveals that there is a transitional strip between the two groups, consisting of half-adozen texts. The preferences of the texts represented in the middle of the biplot are not as clear. As hybrid verbs are prone to belong exclusively to either the first or the third conjugation in most texts, Latin verbs, which are not as extreme, are placed in the middle of the graph. However, their distances from the points representing conjugations are not an absolute indication of their preferences. The graph only shows that Latin verbs select more frequently second and fourth conjugations relative to the hybrid verbs. For example, the place of Jeromeʼs work (the dot next to the fourth conjugation arrowhead) only means that his sample contains a relatively higher incidence of fourth conjugation verbs, not that this is his preferred conjugation overall. In addition, as much of the variation is displayed on the horizontal axis, it has a greater weight in the interpretation. This means that the horizontal stretch of the data is more important (note that hybrid verbs are horizontally stretched), and that the vertical distances are larger on the plot than they are in reality (Jerome is, in fact, much closer to the first and third conjugations than he appears to be). We conclude that hybrid verbs tend to select conjugations differently from both Latin verbs in ancient texts and Latin verbs in macaronic texts, and that the latter two groups are relatively similar to each other. All this corroborates the suggestions of Figure 1. In order to observe the numerical relationships among the verb groups, we compare them with one another, using the chi-squared statistics. All occurrences

340

Šime Demo & Nathalie Tassotti

in a group have been combined. As the fourth and the second conjugations are extremely rare in hybrid verbs, we only use the first and the third conjugations in the tests. We get the following results (with Bonferroni correction)21: Ancient vs. Latin in macaronics: χ2 = 7.8893, df = 1, p-value = 0.015* Ancient vs. hybrid: χ2 = 278.35, df = 1, p-value < 0.001*** Latin in macaronics vs. hybrid: χ2 = 153.07, df = 1, p-value < 0.001*** While the difference between the Latin verbs in the two groups of texts is only moderately significant (at a common significance level of 0.05), the difference between the hybrids and either group of Latin verbs is highly significant (at the 0.001 significance level). This is an additional confirmation of the visual impression presented by the plots.

3.2 Conjugation selection in hybrid verbs Having observed some general trends in the distribution of macaronic hybrid verbs and the difference from the distribution of Latin verbs, regardless of their appearance in ancient or macaronic works, we take a closer look at the hybrid verbs to see whether there are any factors that could prove to be predictive of their distributional behaviour. As we saw, hybrid verbs display unusual behaviour and suggest a clear division into at least two groups. If we perform a correspondence analysis only on them and plot the results, we obtain the biplot presented in Figure 3. It is approximately a close-up version of the previous figure.22 The text points are shape-coded and colour-coded by linguistic groups: the dark squares represent texts consisting of a Romance language combined with Latin, the dots stand for Germanic texts, most of which cluster on the righthand side of the plot, and light grey triangles (mainly lined up on the left-hand side) indicate the Slavonic tradition.

21 The result is usually considered statistically significant if p-value is less than 0.05. In this case there is less than 5% chance that the differences within the pair of the group is due to chance. Higher levels of significance are frequently set to 0.01 and 0.001. The increasing levels of the significance are marked with one, two, or three asterisks, respectively. As more comparisons are performed, we use Bonferroni correction, multiplying the p-value by the number of the comparisons (here, three). 22 It turns out to be a mirror image because of the algorithm used by the computer system.

Verb conjugation selection in macaronic Latin: a corpus-based analysis

341

Biplot of conjugation distribution: Hybrid verbs (by language group) fourth

1.0 Dim2 (9.5%)

Second

0.5

0.0

third first

—0.5

0.5

0.0

1.0

Dim1 (87.3%) Figure 3: Biplot of the correspondence analysis of hybrid verb conjugation distribution. Dark squares belong to the Romance tradition, medium grey dots to Germanic texts, and light grey triangles to Slavonic languages.

The plot reveals that the texts preferring the first conjugation belong to the Romance and Slavonic traditions, while most Germanic texts have a strong inclination towards the third conjugation, with some Germanic texts (and one Slavonic text) lying halfway between the two dominant clusters. To observe the patterns at the level of individual languages and language groups, it is necessary to examine a finer-grained representation. Figure 4 displays six biplots representing selected texts from our corpus. In Panels a to c, we can see the distribution of hybrids belonging to texts from three Romance languages: Italian, French, and Provençal.23 Most of them are concentrated in the lower left quadrant, indicating a strong preference for first conjugation hybrids. Some apparently outlying points, leaning towards the fourth conjugation, are of very limited importance because as little as 9.5% of variance is represented by the y-axis, which means that the distribution in the left–right direction accounts for most of the data.

23 The remaining two – Spanish and Portuguese – behave very similarly to the rest.

Šime Demo & Nathalie Tassotti

342

b. Hybrid verbs: Provenal

a. Hybrid verbs: Italian

fourth

fourth 1.0

second

Dim2 (9.5%)

Dim2 (9.5%)

1.0

0.5 third

0.0

second 0.5

–0.5

first 0.0 0.5 Dim1 (87.3%)

–0.5

1.0

c. Hybrid verbs: French

Dim2 (9.5%)

Dim2 (9.5%)

1.0

0.5

third

0.0

third

0.0 first

0.0 0.5 Dim1 (87.3%)

–0.5

1.0

e. Hybrid verbs: German, Dutch

0.0 0.5 Dim1 (87.3%)

1.0

f. Hybrid verbs: English

fourth

fourth 1.0

1.0

Dim2 (9.5%)

Dim2 (9.5%)

second

0.5

first

second 0.5

second 0.5

third

0.0 0.0

1.0

fourth

second

–0.5

0.0 0.5 Dim1 (87.3%)

d. Hybrid verbs: Slavonic

fourth 1.0

third

0.0

first

third first –0.5

0.0 0.5 Dim1 (87.3%)

1.0

first –0.5

0.0 0.5 Dim1 (87.3%)

1.0

Figure 4: Correspondence analysis of hybrid verb conjugation selection for some groups of texts in the corpus.

Verb conjugation selection in macaronic Latin: a corpus-based analysis

343

In Panel d, all Slavonic texts are represented. They also tend to select the first conjugation. The two outliers are both Polish texts and represent only a small part of that group; thus, we do not notice any notable variation based on the language. An extremely particular spot is the point at the far bottom-left corner of most graphs. No less than 20 texts – 8 Polish, 5 Italian, 2 Provençal, 2 Spanish, 1 Czech, 1 English, and 1 French – which is one-third of the subcorpus – occur at this single point. It is the position of samples that contain only the first conjugation hybrid verbs. The texts with exclusive selection of the first conjugation hybrids are also clearly visible in the right-hand panel of Figure 1. A separate group is represented by German and Dutch texts – here, the third conjugation overwhelmingly dominates. On the other hand, the texts from the remaining Germanic language, English, present a scattered profile and act as a type of bridge between the Romance/Slavonic and the Germanic groups, without belonging to either of the two clusters. The statistical tests confirm the findings based on the plots. Again, we select only the first and third conjugation verbs for the analysis, and test those texts and groups of texts that satisfy an assumption of the chi-squared test requiring that all frequencies in the contingency tables be at least 5. Testing the homogeneity of the texts within Romance and Germanic linguistic groups results in highly significant variation only within the Germanic group, while the non-significance in the Romance texts suggests that they select conjugations very similarly to each other: Romance group: χ2 = 29.406, df = 32, p-value = 0.599 Germanic group: χ2 = 201, df = 14, p-value < 0.001*** The Slavonic texts do not have sufficient frequency of third conjugation verbs for the test to be viable, but the panel d of Figure 4 shows that they are even more homogenous than are the Romance texts. A further step is to find the source of the heterogeneity in the Germanic group. Figure 4 suggests that the English macaronic tradition is the most dispersed. As only one Dutch text was available, and Figure 4 shows that it behaves similarly to the German texts, we checked German and English language macaronics: German texts: χ2 = 9.7173, df = 6, p-value = 0.137 English texts: χ2 = 17.792, df = 6, p-value = 0.007**

344

Šime Demo & Nathalie Tassotti

According to the results, German texts do not differ from one another significantly, while English texts are highly significantly diversified and responsible for the heterogeneity of the Germanic group. This validates the findings in Figure 4.

4 Discussion The data clearly show that there is a marked difference in conjugation distribution between hybrid verbs on the one hand and Latin verbs on the other. Latin verbs in macaronic texts show a moderately significant difference from the verbs in the ancient corpus, and this might be the case not because there is an intrinsic difference between the two groups, but due to the fact that the presence of hybrid verbs has influenced the frequencies of Latin verbs in the macaronics. In most of the macaronic texts first conjugation hybrid verbs prevail, although there is a small group of texts that prefer the third conjugation. The fourth conjugation is rare, and the second is extremely rare. A more detailed analysis has shown that the choice of the language affects the selection preferences. In the following we attempt to provide an explanation of the findings. Romance and Slavonic traditions strongly prefer the first conjugation. This might have two causes. Firstly, the a-conjugation is the most frequent class in all these languages. The Latin first conjugation has a phonological similarity with these classes of verbs in the respective languages, and in the case of the Romance languages the connection is also etymological.24 Furthermore, the Latin first conjugation is the preferred class for creating loans, and is morphologically simpler than the third conjugation. In German and Dutch traditions, on the other hand, we observe an extremely high incidence of third conjugation hybrids. This seems to be accountable for on phonological grounds, too. German and Dutch do not have Romance-like division into four conjugations – in those two languages the ending -en dominates the verbal inflexion. Although it is qualitatively (and, even more, graphically) close to both second and third Latin conjugation infinitive ending -ere, its vowel is short and unstressed, just like the thematic vowel -e- of the Latin third conjugation. Moreover, the third conjugation is the only one in which the present infinitive (along with some present indicative forms) has the stress on the antepenultimate syllable, which seems to be the default stress

24 In French the dominant class ends in -er, which is nonetheless given rise to by first conjugation Latin verbs; here, -a- emerges in tenses other than present.

Verb conjugation selection in macaronic Latin: a corpus-based analysis

345

assignment position in German.25 This makes the Latin third conjugation the only one that does not compel the authors to depart much from what German and Dutch readership was accustomed to in their own languages (e.g. compare wáchsere ≈ wáchsen with wachsére ≈ *wachsén or wachsáre ≈ *wachsán). All authors – despite being spread across several centuries and a large territory – are remarkably consistent in this respect. This influence was apparently stronger than the centrality of the Latin first conjugation in producing loans. English is a special case, without a definite focus and sharing the properties of both previously analysed groups. This can be explained by the particular linguistic set-up of the English language. Genetically, it is a Germanic language, but lacks the characteristic verb endings and the division into conjugation classes. Moreover, it is relatively abundant in Romance lexical material, which makes it an atypical Germanic language. All these factors make it a good candidate for not fitting into either group. It is interesting that, once the preferred pattern is selected, texts tend to adhere to the same conjugation even in subjunctive forms, although characteristic vowels in the present subjunctive (-e- for the first conjugation and -a- for the third) do not match the respective thematic vowels (-a- for the first conjugation and -e- for the third).26 This means that the selection preference went beyond the mere phonological resemblance and entered the level of the grammatical system.

5 Conclusion We hope to have shown in the present study that the choice of the partner language in macaronic Latin has a significant influence on at least one morphological feature, namely the selection of conjugation. The choice can be directly related to the phonological properties of the embedded language, but its implications transcend the level of phonology. Although Italian macaronic literature was taken as a model for all other traditions, local vernaculars were still able to interfere with the morphological

25 See Domahs et al. (2014). The stress-based explanation for German and Dutch has been suggested to us by Martin Bauer and the anonymous reviewer for the present collection. The latter also drew our attention to the reference cited in this footnote. Here we wish to express our gratitude to both colleagues. 26 In our corpus there are 55 hybrid pres. and impf. subjunctive forms (22 in Romance, 30 in Germanic, and 3 in Slavonic texts). Only two of them (both in Portuguese texts) do not belong to the preferred conjugation.

346

Šime Demo & Nathalie Tassotti

layout of the macaronics. Verb conjugation selection seems to be one instance of this process. In more general terms, this fact presents an example of the influence linguistic structures can have on the process of cultural transmission.

Bibliography Adams, J. N. (2004). Bilingualism and the Latin language. Cambridge: Cambridge University Press. Alkire T. & Rosen, C. (2010). Romance languages: A historical introduction. Cambridge: Cambridge University Press. Ballester, X. (1997). A propósito de dos poemas macarrónicos de Ruiz de Moros. In J. M. Maestre Maestre, J. P. Barea & L. C. Brea (Eds.), Humanismo y pervivencia del mundo clásico (pp. 1107–1125). Cádiz: Excelentisimo ayuntamento de Alcañiz. Bauer, L. (2001). Morphological productivity. Cambridge: Cambridge University Press. Beh, E. J. & Lombardo, R. (2014). Correspondence analysis: Theory, practice and new strategies. Chichester: Wiley. Blaise, A. (1955). Manuel du latin chrètien. Strasbourg: Le Latin chrètien. Bonora, E. (1956). Le Maccheronee di Teofilo Folengo. Venezia: Neri Pozza. Burton, P. (2000). The old Latin gospels: A study of their texts and language. New York: Oxford University Press. Celano, G. G. A., Crane, G. & Almas, B. (Eds.). The ancient Greek and Latin dependency treebank. perseusdl.github.io/treebank_data (accessed: 21 October 2017). Chiesa, M. (1972). La tradizione linguistica e letteraria cristiano-medievale nelle ‘Macaronee’. Giornale storico della letteratura italiana, 149, 48–86. Dahl, J. (1962). Maccaronisches Poetikum, oder Nachtwächteri veniunt cum Spießibus atque Laternis. Mit Illustrationen aus alter und neuer Zeit. Ebenhausen bei München: Langewiesche-Brandt Verlag. Demo, Š. (2014). Towards a unified definition of macaronics. Humanistica Lovaniensia, 63, 83–106. Domahs, F., Grande, M., Huber, W. & Domahs, U. (2014). The direction of word stress processing in German: Evidence from a working memory paradigm. Frontiers in Psychology, 5, 574. Garavini, F. (1978). Gentigallantes sunt omnes instudiantes. In Testi e interpretazioni (pp. 357–407). Milano/Napoli: Ricciardo Ricciardi. Giraud, Y. (1977). La Babel facétieuse du macaronique: A. dʼArena décerveleur de langage. Discussion. Bulletin de lʼAssociation dʼétude sur lʼhumanisme, la réforme et la renaissance, 7, 60–65. Glidden, H. H. (1982). Latin, français, graphisme dans les jeux linguistiques de Tabourot des Accords. Bulletin de lʼAssociation dʼétude sur lʼhumanisme, la réforme et la renaissance, 15, 56–62. Glynn, D. (2014). Correspondence analysis: Exploring data and identifying patterns. In D. Glynn & J. A. Robinson (Eds.), Corpus methods for semantics: Quantitative studies in polysemy and synonymy (pp. 443–485). Amsterdam/Philadelphia: John Benjamins. Grandgent, C. (1907). An Introduction to Vulgar Latin. Boston: D.C. Heath & Co.

Verb conjugation selection in macaronic Latin: a corpus-based analysis

347

Greenacre, M. (2007). Correspondence analysis in practice. London: Chapman and Hall. Haverling, G. (2000). On Sco-verbs, Prefixes and Semantic Functions: A study in the development of prefixed and unprefixed verbs from Early to Late Latin (Studia Græca et Latina Gothoburgensia LXIV). Göteborg: Acta Universitatis Gothoburgensis. Haverling, G. V. M. (Ed.) (2015). Latin linguistics in the early 21st century: Acts of the 16th International Colloquium on Latin Linguistics, Uppsala, June 6th–11th, 2011. Uppsala: Uppsala Universitet. Helander, H. (2016). On neologisms in Neo-Latin. Renæssanceforum, 10, 9–34. Herman, J. (2000). Vulgar Latin. University Park, Pa.: Pennsylvania State University Press. Keipert, H. (1988). Sprachprobleme der makkaronischen Dichtung in Polen. Die Welt der Slaven, 33, 354–388. Klausenburger, J. (2000). Grammaticalization: Studies in Latin and Romance morphosyntax. Amsterdam/Philadelphia: John Benjamins. Martzloff, V. (2015). Domuī, secuī, cubuī. Sur les verbes latins en °Cāre formant un parfait en Cuī. In G. V. M. Haverling (Ed.), (pp. 59–72). Meiser, G. (1998). Historische Laut- und Formenlehre der lateinischen Sprache. Darmstadt: Wissenschaftliche Buchgesellschaft. Melo, W. D. C. de (2007). The early Latin verb system. New York: Oxford University Press. Paoli, U. E. (1959). Il latino maccheronico. Firenze: Le Monnier. Passarotti, M., McGillivray, B. & Bamman, D. (2015). A treebank-based study on Latin word order. In G. V. M. Haverling (Ed.), (pp. 338–350). Schade, O. (1955). Fercula macaronica, vol. I. Hannover: Carl Rümpler. Väänänen, V. (1966). Le latin vulgaire des inscriptions Pompéiennes. Berlin: Akademie Verlag. Viti, C. (2015). The use of frequentative verbs in Early Latin. In G. V. M. Haverling (Ed.), (pp. 170–182).

Words Ancient discourse on language

Cécile Conduché

Clausula : un terme technique de la métrique à la morphologie Abstract: In this chapter, we study the use of the term clausula in ancient Latin grammar, with particular focus on its morphological meaning. Firstly, we survey all the instances of clausula meaning “word-ending”, from Quintilian up to Eutyches (sixth century AD). Then, we set out the case for a concomitant development in the meanings of Latin clausula and Greek κατάληξις within the bilingual framework of ancient grammatical discourse. Building on that demonstration, we compare and contrast the construct of clausula in Eutyches with the modern notion of suffix, in order to value the accuracy of translating the former with the latter. Keywords: Latin linguistics, clausula, grammatical terminology, κατάληξις

1 Introduction À consulter la notice que Hey a consacrée au mot clausula dans le Thesaurus Linguae Latinae, deux aspects ressortent dans la répartition des emplois. En premier lieu, les contextes sont presque uniquement langagiers : la clausula est la conclusion d’un discours ou d’un écrit tandis que des sens comme « fin des temps » (n° II et III) sont tout à fait marginaux. En second lieu, le nom clausula a une affinité marquée avec la langue technique et, à côté du sens général de « conclusion », on le retrouve dans une série de domaines techniques (grammaire, métrique, rhétorique, dialectique, droit). C’est sur les emplois de clausula dans l’un de ces domaines, la grammaire, que nous souhaitons par cet article faire le point. En effet, à côté des emplois en métrique, fréquents et bien connus, on trouve un emploi en morphologie, au sens de « terminaison de mot ». Cet usage, quoique repéré par le Thesaurus (IB3), reste mal répertorié. L’article du Thesaurus lui-même ne fournit pas toutes les références, qui sont pourtant peu nombreuses. L’index de Lomanto et Marinone (1990) est complet pour son corpus, mais celui-ci est limité à la collection des Grammatici Latini de Keil. Surtout, le terme est absent du Lexicon of grammatical terminology de Schad (2007).

Cécile Conduché, Fondation Thiers https://doi.org/10.1515/9783110647587-022

352

Cécile Conduché

Or le terme revêt une certaine importance dans l’analyse morphologique de grammairiens parmi les plus tardifs de l’Antiquité latine et en particulier chez Eutychès, auteur dont nous préparons une édition. Deux questions nous ont donc guidée dans cette recherche, l’une pratique et l’autre plus théorique. La première consistait à peser la légitimité d’une traduction par « suffixe », soit un terme désignant spécifiquement l’élément intermédiaire entre la base de dérivation et la désinence portant la flexion. La seconde consistait à élucider, à travers cette étude de cas, la construction d’un emploi technique. Afin de répondre à ces deux préoccupations, nous examinerons successivement les emplois de clausula dans l’analyse morphologique des grammairiens latins, la part du contexte bilingue gréco-latin dans la construction de son acception morphologique et l’usage qu’Eutychès fait du terme.

2 Dossier latin 2.1 Analyse du terme La formation de clausula est claire et parfaitement analysable en synchronie.1 Il s’agit d’un substantif dérivé en -ula sur la base de supin claus- du verbe claudere. Le suffixe, forme de féminin de -ulus, a, um, est le représentant latin d’un ancien *-e-lo- bien attesté dans diverses langues indo-européennes.2 Sur le plan sémantique, les adjectifs et substantifs en -ulus, a, um sont traditionnellement considérés comme des diminutifs et, de fait, quoique tous ne soient pas à strictement parler des diminutifs, la palette de sens que présente la majorité d’entre eux est typologiquement associée aux formations de diminutifs, du moins dans les langues indo-européennes.3 Sur le plan morphologique, ces diminutifs ou termes assimilables sont majoritairement formés sur une base nominale. Clausula, nom d’instrument dérivé sur base verbale apparaît donc comme un type minoritaire. Il s’agit même d’une minorité dans la minorité car les noms d’agent en -ul- sont ordinairement dérivés directement sur la racine verbale, ou sur la base de l’infectum,

1 Le TLL s.v. cite le grammairien Rufinus citant Varron « clausulas quoque primum appellatas dicunt, quod clauderent sententiam ». 2 Sur l’étymologie du suffixe, on peut se reporter à Leumann, (19765 : 309–310). 3 Pour une présentation générale de ce suffixe associé à des bases nominales, voir Gaide (1992), repris avec des modifications dans Gaide (2002). La démonstration du lien entre les divers sens des dérivés latins en -ulus, a, um et les formations diminutives dans les langues indo-européennes a été faite par Zucchelli (1969 : 71–122).

Clausula : un terme technique de la métrique à la morphologie

353

comme radula « racloir » sur radere « raser ». Dans le cas de clausula, la base est celle de l’accompli, avec une diathèse active : la clausule est « ce qui conclut ». Si les parallèles stricts sont rares à l’intérieur du latin, en revanche, l’usage du suffixe pour former des noms d’agent ou d’instrument déverbaux est bien documenté tant en latin que dans la famille indo-européenne en général. A ce cadre général, nous ajouterons deux hypothèses personnelles. La première est que clausula peut fonctionner en synchronie comme le pendant intellectuel, ou du moins langagier (ce qui conclut un discours), de claustrum (ce qui verrouille un espace). La seconde hypothèse est celle d’une influence de la valeur diminutive généralement reconnue au suffixe -ulus. Comme on sait, clausula désigne une section de discours, précisément la section finale qui clôture l’ensemble. La valeur est très nette dans le sens technique de ‘clausule métrique’ comme dans celui d’‘article final d’une loi’. Le terme en est donc venu à désigner non plus tant l’instrument de la conclusion qu’un sousensemble final de ce qui est clôturé.

2.2 Emplois du substantif Les attestations de clausula au sens de ‘fin de mot’ ont une répartition chronologique irrégulière. On en trouve deux occurrences isolées dans l’Institution oratoire de Quintilien puis on le retrouve dans des traités grammaticaux tardoantiques dont les auteurs portent tous des noms grecs.

2.2.1 Quintilien Le sens particulier que prend clausula au livre 12 de l’Institution oratoire est repéré depuis longtemps déjà et cité dans l’article du TLL. Quid quod pleraque nos illa quasi mugiente littera cludimus, in quam nullum Graece uerbum cadit ? At illi ny iucundam et in fine praecipue quasi tinnientem illius loco ponunt, quae est apud nos rarissima in clausulis. (QVINT. inst. 12, 10, 31) ‘Bien plus : nous en terminons beaucoup par cette lettre qui meugle et sur laquelle, en grec, on n’achève aucun mot. Eux, à sa place, mettent le nu, gai carillon, surtout en finale, qui chez nous est des plus rares dans les terminaisons.’4

Un emploi comparable a été relevé plus récemment dans un passage du livre 9.

4 Sauf mention contraire explicite, les traductions sont de nous.

354

Cécile Conduché

Etiam monosyllaba, si plura sunt, male continuabuntur, quia necesse est compositio multis clausulis concisa subsultet. (QVINT. inst. 9, 4, 42) ‘De même plusieurs monosyllabes de suite ne feront pas bon effet, parce que l'ensemble, sans cesse interrompu par les terminaisons, aura l'air saccadé.’

Dans ces deux occurrences, clausula désigne la fin du mot de manière très neutre, sans impliquer d’analyse morphologique précise. En outre, dans les deux cas, la discussion porte sur l’euphonie. Il s’agit donc de la fin de mot considérée d’un point de vue phonétique, quoique ni prosodique ni métrique.

2.2.2 Diomède le grammairien Il est notoirement difficile d’établir une chronologie absolue des grammairiens tardo-antiques dont les ouvrages sont recueillis dans le corpus des Grammatici Latini et, parfois, même la chronologie relative fait défaut. Dans le cas de Diomède, les deux bornes sont constituées par Charisius (4e siècle) qu’il utilise et Priscien (premier tiers du 6e siècle) qui le cite.5 La grammaire de Diomède porte les marques d’une production en milieu hellénophone. À deux reprises, clausula, qui dans cet ouvrage est employé avant tout en son sens métrique, prend le sens de ‘fin de mot’. Le contexte est celui de la présentation des figures de style. Paragoge est cum ad ultimam simplicis dictionis clausulam aut littera adiungitur aut syllaba : littera, ut apud Plautum, « quo ted hoc tis » pro te ; syllaba, ut potestur pro potest. (DIOM. gramm. I 441, 17–19) ‘Il y a extension lorsqu'on ajoute au bout de la terminaison d'un mot simple une lettre ou une syllabe : une lettre comme chez Plaute « quo ted hoc tis » [‘où à cette heure tardive te . . .’, PLAVT. Curc. 1] au lieu de te ; une syllabe comme potestur au lieu de potest [on peut].’

Dans la conception de la paragoge ici exposée, ted n’est pas une forme archaïque d’ablatif ni potestur une forme de passif impersonnel également limitée à l’usage pré-classique. Il s’agit, dans l’analyse du grammairien latin, des formes classiques te et potest avec ajout respectivement d’une lettre et d’une syllabe de remplissage, dépourvues de sens et de fonction. Dans cette analyse, clausula désigne la véritable terminaison, celle des formes classiques te et potest, que l’on retrouve derrière l’extension.

5 Kaster (1988 : 270–272) et Holtz (1981 : 82–83) pour la synthèse des données de la question. Barwick (1922 : 8) pour la démonstration de l’utilisation de Charisius par Diomède.

Clausula : un terme technique de la métrique à la morphologie

355

homoeoteleuton est oratio similibus clausulis terminata, id est pari uerborum exitu finita, ut apud Ennium « eos reduci , deuehi quam deseri / malui » ; et Sallustius « in nuda in tecta corpora » ; Vergilius « bella horrida bella ». (DIOM. gramm. I 447, 5–10) ‘L'homéotéleute est un énoncé qui s'achève par plusieurs terminaisons semblables, c'est-àdire qui se finit par des mots de même extrémité, comme chez Ennius « eos reduci , deuehi quam deseri / malui » [‘j’ai préféré qu’ils soient ramené plutôt que délaissés, raccompagnés plutôt qu’abandonnés’, ENN. scaen. 390–391], Salluste « in nuda in tecta corpora » [‘contre des corps nus, vêtus ( ?)’, SALL. hist. frg. 4, 62], Virgile « bella horrida bella » [‘des guerres, d’affreuses guerres’, VERG. Aen. 6, 86].’

Dans la grammaire de Diomède, comme chez Quintilien, la terminaison est envisagée avec un souci de l’euphonie et de ses techniques. Le référent du terme clausula recouvre des marques flexionnelles, mais il s’agit vraisemblablement d’une coïncidence et ici, encore comme chez Quintilien, l’emploi de clausula au sens de ‘terminaison’ n’implique pas d’analyse fonctionnelle de la terminaison ainsi désignée.

2.2.3 Phocas L’œuvre du grammairien Phocas est à situer sensiblement dans la même fourchette chronologique que celle de Diomedes : Phocas connaissait l’œuvre de Donat (troisième quart du 4e siècle) et il est nommé par Priscien.6 Son œuvre grammaticale comporte une Vie de Virgile et un traité sur le nom et le verbe. Ce grammairien est le premier à faire un emploi régulier du terme clausula dans une acception morphologique. On en compte cinq occurrences dans le traité.7 La valeur de clausula y apparaît nettement fonctionnelle. Il s’agit de la lettre ou du groupe de lettres qui permet d’assigner un lexème à un modèle de flexion. Dans quatre des cinq occurrences, la discussion porte sur la déclinaison nominale. unum eadem clausula terminatum generis masculini, nomen serpentis apud Lucanum lectum est, « et natrix uiolator aquae » ; duo feminini, quae a masculinis non ueniunt, haec meretrix, haec cicatrix. (PHOC. gramm. V 421, 15–17 = de nom. 23, 3 Casaceli) ‘on lit un seul masculin achevé par la même terminaison, le nom d’un serpent chez Lucain [LVCAN. 9, 720], « et l’hydre (natrix) profanateur de l’eau », et deux féminins qui ne sont pas issus d’un masculin, la courtisane (meretrix), la cicatrice (cicatrix).’

6 Holtz (1981 : 231–232) ; Kaster (1988 : 339–341). L’article fondamental sur le grammairien Phocas est Mazzarino (1973–1974 : 505–527). 7 PHOC. gramm. V 412, 11 ; 417, 32 ; 418, 26 ; 421, 15 ; 433, 22 = De nom. 4, 1 ; 18, 4 ; 19, 4 ; 23, 3 ; De uerbo 47, 4 Casaceli.

356

Cécile Conduché

Quoique la clausula ici discutée coïncide avec un suffixe au sens moderne du terme, il s’agit bien d’une syllabe finale de nominatif singulier assez caractéristique pour fournir un critère de classement des noms. L’unique application du terme clausula à la flexion verbale ouvre la perspective d’un nouveau type d’analyse. pauca praeterea eandem clausulam habent nouissima littera praesentis temporis in i conuersa, uinco uici, rumpo rupi, soluo solui, uoluo uolui, cudo cudi. (PHOC. gramm. V 433, 22–4 = 47, 4 Casaceli) ‘en outre, quelques-uns ont la même terminaison, transformant la lettre finale de présent en -i, uinco uici [‘je vaincs je vainquis’], rumpo rupi [‘je rompts je rompis’], soluo solui [‘je délie je déliai’], uoluo uolui [‘je roule je roulai’], cudo cudi [‘je bats je battis’].’

Le grammairien a exposé les différents modes de formation du parfait, considérant qu’ils sont nécessairement dérivés du présent. La dernière catégorie est un reliquat qui présente, dit-il, ‘la même terminaison’. Il faut, semble-t-il, comprendre ‘la même que le présent’, étant entendu que la finale absolue propre à la première personne du parfait n’entre pas en ligne de compte. La clausula désignerait donc ici la terminaison du radical, voire sa consonne finale.

2.2.4 Eutychès En dépit de ces précédents, l’emploi de clausula au sens morphologique est surtout un trait de l’idiolecte technique du grammairien Eutychès. Ce dernier est mieux daté que les précédents du fait qu’il mentionne Priscien comme son maître et que Cassiodore utilise un de ses ouvrages dans son traité d’orthographe écrit vers 580. Eutychès nous est connu comme l’auteur de deux traités grammaticaux, l’un sur l’aspiration en latin, c’est-àdire en fait sur le graphique, dont une partie est conservée dans le De orthographia de Cassiodore, et l’autre sur la conjugaison, qui nous est parvenu par tradition directe. La problématique d’ensemble de ce traité De verbo porte sur l’assignation des verbes latins à un modèle de conjugaison qui permet de déduire l’ensemble des formes de la flexion verbale. Eutychès développe pour y répondre un exposé en deux livres. Le premier traite de la dérivation des noms sur base verbale et des verbes sur base nominale. Le second présente, par ordre alphabétique de la finale du thème d’infectum, un aide-mémoire des conjugaison accompagné d’observations sur les verbes irréguliers. C’est dans le premier livre, plus théorique et centré sur la

Clausula : un terme technique de la métrique à la morphologie

357

dérivation, que se lisent toutes les occurrences de clausula, au nombre de dix-sept.8 A tribus enim quasi fontibus uerbalia nomina deducuntur, id est aut ab indicatiuo praesenti aut a praeterito perfecto aut ab infinitiuo futuro siue participiis ; et singulis eorum certae sunt deriuationum clausulae tam generales quam speciales destinatae. (EVTYCH. gramm. V 452, 5–8) ‘En effet, les noms déverbaux sont tirés de trois sources : l'indicatif présent, le prétérit parfait, l'infinitif futur autrement dit les participes ; et des terminaisons de dérivations précises, génériques aussi bien que spécifiques, sont associées à chacune d'elles.’

Ici, de façon beaucoup plus explicite que chez Phocas, la clausula est un élément de l’analyse morphologique du nom. Du point de vue combinatoire, elle s’associe de façon plus ou moins productive à une base de dérivation, qu’Eutychès désigne ici par fons. Du point de vue catégoriel, elle peut transformer une base venue d’une partie du discours en un mot appartenant à une autre partie du discours. Dans l’extrait cité, il s’agit de transformer l’une des trois bases verbales (on reconnaît dans la liste d’Eutychès nos thèmes d’infectum, perfectum et supin) en nom. Du point de vue fonctionnel, elle organise le lexique en séries repérables et répondant à un modèle de flexion (les noms en -tor, les noms en -mentum, etc). La clausula, en revanche, n’est pas systématiquement associée à un sens.

2.3 Emplois des verbes À cette spécialisation de clausula correspondent des emplois de verbes de la famille lexicale au sens de ‘terminer un mot’. Deux verbes sont concernés : claudere que Quintilien associe en une sorte de figure étymologique à clausula dans l’extrait du livre 12 cité ci-dessus et, de façon très marginale, son composé concludere. Le verbe claudere signifie, comme on le sait bien, ‘terminer’, ‘finir’ au sens le plus général. C’est le contexte qui permet de définir l’emploi spécialisé, du terme, grammatical et en particulier morphologique.9 On le trouve surtout au passif car, en morphologie, le thème est en général le mot, qui se trouve sujet de la phrase exposant la façon dont il se termine. Le nom de la lettre ou de la syllabe se trouve alors à l’ablatif instrumental. Schad (2007) relève dans le

8 EVTYCH. gramm. V 452, 8 ; 453, 26 ; 454, 7 ; 454, 13 ; 455, 7 ; 455, 25 ; 456, 23 ; 459, 28 ; 460, 4 ; 460, 9 ; 460, 19 ; 460, 23 ; 461, 23 ; 462, 5 ; 463, 13 ; 463, 22 ; 464, 15. 9 Schad (2007) s.v. claudere. Ce qui suit s’appuie sur cette notice.

358

Cécile Conduché

corpus des grammairiens latins 13 occurrences d’une telle tournure réparties entre quatre auteurs : un pseudo-Probus, Charisius, Diomedes et Servius. Un seul traité grammatical, en revanche, fait usage de concludere en ce sens et dans ce type de contexte. Il s’agit des Instituta artium, dans lesquels on le lit à cinq reprises, toujours au passif.10 Le recours, occasionnel, à un verbe si exceptionnel en contexte morphologique peut s’expliquer par un souci de varier les formulations.

3 Dossier bilingue La grammaire est dans le monde latin antique une science d’origine grecque et qui se développe dans une confrontation constante au grec à la fois sur le plan linguistique, par la comparaison des deux langues, et sur le plan métalinguistique, par adoption et assimilation des théories grammaticales grecques. Ce second aspect de l’hellénisme des grammairiens latins englobe la configuration du vocabulaire technique. Il est donc légitime et pertinent de rechercher dans la terminologie grammaticale grecque un terme qui rassemble les traits que nous avons mis en évidence ci-dessus dans clausula : un dérivé sur une base verbale signifiant ‘achever’, une limitation aux contextes langagiers, un emploi technique majoritaire en métrique, et en particulier en métrique de la prose, et un emploi technique secondaire en morphologie. Le substantif qui nous a paru correspondre le plus étroitement à ces critères morphologiques et sémantiques est κατάληξις. Au niveau de la formation, κατάληξις et clausula diffèrent. Le verbe καταλήγειν, en effet, attesté depuis l’époque classique,11 signifie plutôt ‘s’achever’ par épuisement que ‘fermer’ et la dérivation est celle d’un nom de procès et non d’instrument. Cela dit, κατάληξις comme clausula apparaît limité au lexique savant, technique même, et en particulier grammatical. Sa première attestation remonte au corpus aristotélicien et la valeur de nom de procès portée par le suffixe y est pleinement réalisée.12 Par la suite,

10 PROB. inst. gramm. IV 77, 19 ; 79, 31 ; 80, 37 ; 85, 8 et 140, 9. 11 AESCH. Ag. 1479, avec pour sujet ἄχος, ‘la douleur’. 12 ARIST. hist. anim. 634b19–21 (livre X) Καὶ ἄρχονται μὲν ἐκ λευκῶν γαλακτοειδῶν, ἀνόσμων μενόντων· τὰ δὲ φοινικᾶ μέν, ἀπολήγοντα δὲ λευκότερα, ἐσχάτης καταλήξεως. ‘D’autre part, les règles commencent par des écoulements blancs qui ressemblent à du lait et restent inodores. Puis elles sont rouges, et à la fin blanchâtres, quand elles sont sur le point de cesser’, texte et traduction de Louis (1964–1969). La date de cette première attestation est difficile à fixer. Le livre X, en effet, est généralement considéré comme apocryphe. Louis (1964–1969 : III, 153), a proposé d’y voir l’ouvrage d’un médecin du début du 3e siècle av. J.-C.

Clausula : un terme technique de la métrique à la morphologie

359

les deux sens majoritaires de ce substantif dans la littérature conservée sont ‘clausule’ en contexte métrique et ‘terminaison de mot’.

3.1 Contexte morphologique La très grande majorité (124 occurrences sur 150) des emplois de κατάληξις dans le corpus du Thesaurus Linguae Graecae se trouve dans les œuvres d’Apollonios Dyscole et dans celles qu’on attribue à son fils Hérodien. Dans les traités d’Apollonios Dyscole, κατάληξις désigne la terminaison du mot, lettre, groupe de lettres ou syllabe finale. La terminaison ainsi désignée constitue une marque morphologique. τοῖς δὴ οὖν συγκριτικοῖς τῶν ἐπιρρημάτων παρεπόμενόν ἐστι τὸ ὑπερτίθεσθαι, ὡς πρόδηλον κἀκ τῆς καταλήξεως, καθὸ καὶ ἐπ’ ὀνοματικῆς συγκρίσεως ἡ αὐτὴ πάλιν κατάληξις ἐγγίνεται, γοργότερος γοργότατος, ταχύτερος ταχύτατος. καὶ πάλιν παρὰ τὸ ἀνωτέρω τὸ ἀνωτάτω. (Grammatici Graeci 2, 1/1, 168, 21–25) ‘Aux adverbes comparatifs échoit donc nécessairement le passage au superlatif, comme on le voit aussi d'après la terminaison : suivant ce qui se fait sur un comparatif nominal, on ajoute la même terminaison, gorgoteros gorgotatos [‘plus effrayant le plus effrayant’], takhuteros takhutatos [‘plus rapide le plus rapide’], et de même d'après anōterō [‘plus en haut’], anōtatō [‘le plus en haut’].’

Comme on le voit dans cet exemple, le terme κατάληξις renvoie à une terminaison régulière, soit qu’il s’agisse de flexion, soit de dérivation. L’usage qu’Hérodien a fait de ce terme technique, pour autant qu’on puisse en juger à travers les intermédiaires qui nous transmettent les fragments de son œuvre, semble avoir été très proche. Par exemple, dans le sommaire de la Prosodie générale (Grammatici Graeci 3, 1, 3–4), les livres sont divisés suivant la terminaison des noms dont ils traitent ; le terme employé pour ces terminaisons est κατάληξις. Le sens de κατάληξις et l’analyse interne du mot que ce terme engage paraissent donc très proches de ce que recouvre clausula chez Phocas et Eutychès. Or κατάληξις, comme clausula, est attesté auparavant et de façon contemporaine comme un terme de métrique.

3.2 Contexte prosodique et métrique La chronologie des sens de κατάληξις en contexte prosodique et métrique ne peut être qu’hypothétique, tant les traités techniques d’époque hellénistique nous manquent. La prudence commande de se limiter à deux constats. En

360

Cécile Conduché

premier lieu, le substantif est attesté avec son sens métrique plus tôt que dans l’analyse morphologique, dès le 1er siècle av. J.-C. au plus tard. En second lieu, l’une des premières attestations, peut-être la première, concerne l’analyse rythmique et métrique de la prose. Nous proposons l’exposé de l’histoire haute du terme suivant la chronologie généralement considérée comme la plus probable.

3.2.1 [Démétrios de Phalère] La question la plus épineuse est celle de la datation du traité Du style attribué à Démétrios de Phalère. Si l’on adopte l’hypothèse, moyenne sur le plan chronologique, de Chiron (1993), Démétrios aurait travaillé à Athènes dans les dernières décennies du 2e siècle av. J.-C.13 Nous aurions bien alors une occurrence hellénistique de κατάληξις pour désigner la clausule d’une période, avant toute attestation de clausula dans les textes latins.14 Il se trouve une seule fois, au paragraphe 19. τρία δὲ γένη περιόδων ἐστίν· ἱστορική, διαλογική, ῥητορική. ἱστορικὴ μὲν ἡ μήτε περιηγμένη, μήτ᾽ ἀνειμένη σφόδρα, ἀλλὰ μεταξὺ ἀμφοῖν, ὡς μήτε ῥητορικὴ δόξειε καὶ ἀπίθανος διὰ τὴν περιαγωγήν, τὸ σεμνόν τε ἔχουσα καὶ ἱστορικὸν ἐκ τῆς ἁπλότητος, οἷον ἡ τοιάδε· « Δαρείου καὶ Παρυσάτιδος γίγνονται » μέχρι τοῦ « νεώτερος δὲ Κῦρος »· ἑδραίᾳ γάρ τινι καὶ ἀσφαλεῖ καταλήξει ἔοικεν αὐτῆς ἡ ἀπόθεσις. (DEMETR. elocut. 19) ‘Il y a trois espèces de périodes : la période narrative, la période du dialogue et la période oratoire. La période narrative ne doit être ni trop arrondie ni trop lâche mais tenir un juste milieu ; elle évitera ainsi l'allure oratoire et la perte de force persuasive qui viennent de la rondeur. Quant à la noblesse et la vertu narrative, elles lui viendront de sa simplicité, comme dans le passage qui va de « Darius et Parysatis » à « le cadet Cyrus », où la terminaison fait l'effet d'une cadence stable et sûre.’ (Trad. de Chiron 1993)

L’éditeur et traducteur du texte, Chiron, suggère dans son commentaire que la dernière phrase est une observation d’ordre métrique. On peut avancer à l’appui l’adjectif dérivé καταληκτικός. Celui-ci est employé à deux reprises dans deux paragraphes successifs pour caractériser le péon quatrième, celui qui finit par une longue, et le différencier du péon premier, qui présente la longue à l’initiale.15

13 Chiron (1993 : xiii-xl et xxxvi-xxxvii). 14 Chez Aristote, qui a théorisé le phénomène, le terme désignant la clausule métrique de prose est τελευτή. C’est peut-être aussi bien τελευτή que Cicéron traduit par clausula comme le veut Mankin (2011 : 265) ; mais cela n’est plus forcément vrai de Quintilien. 15 DEMETR. elocut. 38–39.

Clausula : un terme technique de la métrique à la morphologie

361

3.2.2 Denys d’Halicarnasse C’est au 1er siècle a.C. que l’on peut lire la première attestation de κατάληξις au sens très précis de ‘syllabe finale hors décompte’, dans le traité sur la composition stylistique de Denys d’Halicarnasse.16 Denys donne des exemples de prose rythmique en l’espèce de périodes qu’il découpe en pieds métriques. La fin du découpage coïncide toujours avec l’avant-dernière syllabe, que suit « la catalexe », c’est-à-dire la finale, de quantité indifférente. Τοῦ δὲ τρίτου κώλου τοῦ· τοσαύτην ὑπάρξαι μοι παρ᾽ ὑμῶν εἰς τουτονὶ τὸν ἀγῶνα, ἄρχουσι μὲν ὑποβάκχειοι δύο, ἔπειτα δὲ κρητικός, ᾧ συνῆπται σπονδεῖος· εἶτ’ αὖθις βακχεῖος ἢ κρητικός, καὶ ὁ τελευταῖος πάλιν κρητικός, εἶτα κατάληξις. (DIONYS. compos. verb. 18) ‘Le troisième côlon, « qu'elle me vienne aussi abondamment de votre part en vue du procès actuel », commence par deux hypobacchées, puis un crétique, auquel est joint un spondée ; ensuite à nouveau un bacchée ou un crétique, et pour finir encore un crétique, puis la finale.’ (trad. de Aujac 2003)

Les trois autres occurrences du terme dans l’ouvrage ont le même sens et supposent également que ce sens technique est suffisemment établi et assez largement diffusé pour qu’on l’emploie dans un traité protreptique.

3.2.3 Héphestion Comme on sait, le plus ancien traitement systématique de la métrique grecque qui nous soit conservé en entier est le Manuel (Ἐγχειρίδιον) d’Héphestion, composé sous les Antonins.17 Le terme κατάληξις apparaît dans un dernier sens correspondant au latin clausula, celui de ‘vers conclusif’, clausule de strophe par un vers de même rythme que les précédents, mais plus bref.18

16 Leo (1889 : 287) relevait déjà cette acception particulière de κατάληξις chez Denys comme un trait qui l’éloigne des analyses de la métrique selon lui alexandrine représentée par Héphestion. 17 Pour une synthèse récente sur Héphestion et sa métrique, avec la bibliographie antérieure, on consultera la notice d’Ippolito (2006). 18 On voit que la spécialisation du substantif l’a éloigné subtilement de l’adjectif correspondant καταληκτικός, pourtant employé par Héphestion dans le même ouvrage pour qualifier un vers qui s’interrompt brutalement en cours de pied ou de mètre. La κατάληξις n’est pas l’action d’interrompre un rythme en cours mais la brève séquence finale conçue comme un ajout au rythme précédent.

362

Cécile Conduché

ἐκεῖ γὰρ μετὰ δέκα ἀναπαιστικά, λόγου χάριν, καὶ κατάληξιν ἐπάγουσιν εὐθὺς ὅμοια μὲν καὶ ἀναπαιστικά, οὐ μέντοι τῶν ἴσων συζυγιῶν. (HEPHAEST. poem. 6, 3) ‘car là, après dix vers anapestiques, par exemple, ils ajoutent encore comme clausule des anapestes, mais sans conserver le nombre des dipodies.’

Naturellement, étant donné le caractère parcellaire de la documentation, ce témoignage ne garantit qu’un terminus ante quem pour une telle acception. On constate cependant que tous les sens de κατάληξις en contexte métrique sont en place à l’époque où Apollonios Dyscole et Hérodien (sous le règne de Marc Aurèle) en développent l’emploi en contexte morphologique. Il est par conséquent vraisemblable que ce terme technique, appliqué d’abord pour désigner une clausule métrique, ait été transféré dans un second temps à la terminologie grammaticale morphologique, transfert facilité par l’identité des maîtres de l’une et l’autre discipline.

3.3 Histoire d’un concept technique Ce détour par le grec, en nous donnant la traduction de clausula, permet de compléter l’histoire du concept ainsi nommé et d’avancer, bien entendu à titre d’hypothèse, la reconstitution historique suivante. Le substantif κατάληξις s’est introduit au cours de l’époque hellénistique pour nommer la clausule métrique, probablement d’abord dans la prose d’art. Du côté latin, clausula s’impose comme désignation de la clausule métrique, et partant comme traduction de κατάληξις. À partir de ce moment, la terminologie demeure double, mais l’équivalence admise assure que les évolutions connues par le concept dans l’une des aires linguistiques se transmettent à l’autre aire linguistique, éventuellement avec retard mais sans distorsion majeure. Un peu avant l’ère chrétienne, toujours dans le contexte de l’analyse métrique, κατάληξις commence à désigner non plus seulement l’achèvement, voire la partie finale, d’une séquence rythmique, mais aussi bien un élément ajouté qui en marque la clôture. Telle est la notion qui apparaît très nettement chez Denys d’Halicarnasse et que l’on retrouve dans une moindre mesure chez Héphestion. La notion de κατάληξις/clausula recouvre donc à cette époque en métrique l’abstraction qu’est la terminaison, le segment final et l’élément qui clôt. Le passage au domaine morphologique se fait au plus tard au 1er siècle ap. J.-C. avec Quintilien. Le sens privilégié est alors celui de segment final. Cela dit, encore trois siècles plus tard, Diomède emploie clausula au sens d’‘ajout en fin de mot’, qui rejoint alors le sens du terme morphologique et stylistique de παραγωγή. C’est peut-être la notion d’ajout, l’idée que la clausule n’est pas simplement détachable par opération analytique mais bien une conclusion travaillée spécialement pour venir s’ajuster à la fin d’une séquence, qui a conduit

Clausula : un terme technique de la métrique à la morphologie

363

à associer κατάληξις au processus de dérivation, phénomène visible chez Apollonios Dyscole et Hérodien en grec et qu’on retrouve pour le terme latin dans les traités de Phocas et d’Eutychès. Nous aurions alors l’exemple de l’évolution sur la longue durée d’un concept technique, passant en particulier par le transfert d’un domaine, la métrique, à un domaine voisin, la morphologie, perceptible à travers l’histoire combinée des emplois de deux termes, l’un grec et l’autre latin. Un tel phénomène trahit l’unité de l’aire culturelle envisagée, du moins dans les domaines techniques, sous la dualité linguistique.

4 Pour la traduction d’Eutychès Une fois reconstituée, autant que possible, la profondeur historique de la notion exprimée dans la grammaire latine antique par le terme clausula, il nous est possible de revenir à un nouvel éclairage d’Eutychès. La question offre à présent deux aspects. Le premier est celui du sens précis de clausula dans la terminologie propre à Eutychès dans le traité sur le verbe. Le second est celui de l’usage de la notion et de ses conséquences sur la conception de la morphologie latine développée par Eutychès.

4.1 Spécialisation Dans le De verbo d’Eutychès, l’évolution de clausula vers la désignation d’un segment du mot soumis à l’analyse est complète et explicite. En témoignent les passages où le grammairien énumère les clausulae nominales ou adjectivales qui s’attachent aux différentes bases de dérivation issues du verbe. Ces segments sont tellement bien individualisés que leur manipulation s’accompagne d’une amorce de théorie de la voyelle de liaison. In terminationibus autem incipientibus a consonante consonans antecedens ipsam positionem uerbi inuenitur praecedens nominis clausulam, in prima quidem coniugatione a sequente, in reliquis autem i, in bulum uel in bilis, ut sto stas stabulum, uoco uocas uocabulum, pateo pates patibulum, infundo infundis infundibulum, item sto stas stabilis ; item in bundus, erro erras errabundus, ludo ludis ludibundus, morior moreris moribundus, furo furis furibundus. (EVTYCH. gramm. V 454, 5–12) ‘Quant aux terminaisons qui commencent par une consonne, la consonne qui précède la marque de forme de base du verbe s'y trouve devant le suffixe de nom, suivie de /a/ dans la première conjugaison, de /i/ dans les autres : en -bulum ou en -bilis comme sto stas stabulum [‘être debout étable’], uoco uocas uocabulum [‘nommer mot’], pateo pates patibulum [‘être évident fourche patibulaire’], infundo infundis infundibulum [‘verser

364

Cécile Conduché

entonnoir’] ; aussi sto stas stabilis [‘être debout stable’] ; aussi en -bundus, erro erras errabundus [‘errer errant’], ludo ludis ludibundus [‘jouer folâtre’], morior moreris moribundus [‘mourir mourant’], furo furis furibundus [‘délirer délirant’].’

Le terme clausula, avec ce référent concret, s’intègre pleinement à l’analyse morphologique de la dérivation. Il désigne l’un des éléments constitutif des nouvelles unités lexicales dont le nom d’ensemble est traductiones ou deriuationes. Et sic alias similiter traductiones uerborum antecedente consonante prima sibi tantum coniugatio uindicat, seu a uerbis seu a nominibus ueniant. Qualia sunt in co uel in to uel in lo uel in cinor uel in cito uel etiam in alias quasdam terminationes desinentia. In quibus etiam illud est obseruandum, quod per earum ipsarum clausularum incrementa deriuatio uerbi solet prouenire, ut uello uellis uellico . . . (EVTYCH. gramm. V 462, 1–6) ‘Et ainsi de même, les autres verbes dérivés à consonne pénultième, la première conjugaison seulement les revendique pour elle, qu'ils viennent de verbes ou de noms. Tels sont les verbes qui se terminent en -co, en -to, en -lo, en -cinor, en -cito, ou même en certaines autres terminaisons. Il faut également faire à leur propos cette observation que c’est par l'ajout de ces suffixes-là qu’on forme usuellement une dérivation sur verbe, comme uello uellis uellico [‘arracher déchirer’] . . .’ Adeoque uera uidetur esse obseruatio clausularum, quod ex una terminatione nominis duae uerbi factae deriuationes diuersam pro qualitate finalitatis coniugationem meruerunt : ut ab eo quod est meta metor quidem metaris primae est coniugationis, metior metiris quartae . . . (EVTYCH. gramm. V 463, 22–26) ‘Et la règle sur les suffixes est si vraie, visiblement, qu’à partir d’une seule terminaison de nom deux dérivations de verbes ont reçu des conjugaisons différentes suivant la nature de leur fin : ainsi, sur meta [‘borne’], metor metaris [‘arpenter’] est de la première conjugaison, metior metiris [‘mesurer’] de la quatrième . . .’

Comme le montre ce dernier passage, la spécialisation technique de la terminologie reste limitée. Ici, clausula apparaît encore comme interchangeable par simple souci stylistique de variété avec terminatio et finalitas. Il s’agit là d’une limite bien connue de la terminologie antique. Cependant, à condition de ne pas négliger cette limite, on peut identifier chez Eutychès un approfondissement de la notion de clausula qui conduit à une analyse originale des mots latins.

4.2 Amorce d’une analyse tripartite Si les grammairiens latins n’ignorent ni le phénomène de dérivation ni celui de la flexion, bien distingués à l’époque où écrit Eutychès, leurs analyses morphologiques sont en général étanches. D’un côté, on présente les dérivations classées sur des critères plutôt sémantiques. D’un autre côté, on expose, sous forme de règles de substitution et combinaison, les formations de déclinaisons

Clausula : un terme technique de la métrique à la morphologie

365

et conjugaisons. Dans les deux types d’exposés, les formes sont soumises à une analyse en deux parties, celle qui reste identique dans l’opération et celle qui varie. Eutychès paraît parfois amorcer une analyse en trois éléments (radical, suffixe de dérivation et désinence, pour employer les termes modernes). Dans cette analyse, la clausula et l’usage du terme clausula jouent un rôle pivot. Comme nous l’avons vu plus haut, la clausula se distingue de l’élément qui précède, la base de dérivation. De façon décisive, la clausula est également distinguée du modèle de flexion suivi par le verbe ou le nom qu’elle permet de dériver. Ici, la distinction n’est pas si nette qu’avec le radical : les éléments ne s’enchaînent pas en séquence stricte. Il n’en reste pas moins que clausula et flexion sont envisagées comme des réalités distinctes avant d’être associées. Nominum enim de uerbi positione traductorum absque paucis generalibus, quas superius ostendimus, reliquas speciales et tam clausulis quam declinationibus diuersas et in rarioribus exemplis inuentas terminationes, uerborum uero de nominibus factorum et generales finalitates et ad unam tantum pertinentes coniugationem et in exemplis innumerabilibus apparentes uidemus. (EVTYCH. gramm. V 460, 17–26) ‘Nous voyons en effet que les terminaisons des noms tirés de la forme de base d’un verbe, en dehors d'un petit nombre de génériques que nous avons indiqués plus haut, sont, pour le reste, spécifiques, différentes selon les suffixes et les déclinaisons et attestées dans des exemples assez clairsemés, tandis que les finales des verbes formés sur des noms sont génériques, relèvent d'une seule conjugaison et apparaissent dans d'innombrables exemples.’ Nec ad aliam potest huiusce modi uerborum deriuatio uenire coniugationem, nisi forte incidat in clausulam, iam ante cuidam coniugationi generaliter deputatam, ut in urio desinentia meditatiua esse quartae facit ipsa formae finalitas et inchoatiua tertiae. (EVTYCH. gramm. V 461, 22–25) ‘Et une dérivation des verbes de ce type ne peut aboutir à aucune autre conjugaison [scil. que la première], à moins qu'elle ne s'achève par un suffixe déjà attribué de façon générale à une conjugaison, comme la terminaison même de la forme fait que les désidératifs qui s’achèvent en -urio sont de la quatrième conjugaison et les inchoatifs de la troisième.’

Le terme de clausula est donc chez Eutychès strictement associé au phénomène de dérivation, aussi bien en vue de la formation des noms que de celle des verbes. Comme élément de dérivation, la clausula est évidemment opposée à la base qui se l’associe, base qui retient également l’attention du grammairien. Elle est également associée à un modèle de flexion sans que ces deux éléments se confondent. A cet égard, l’évolution est nette par rapport au sens imprécis de ‘terminaison’, voire ‘finale’ qui était celui de clausula dans ses premiers emplois métalinguistiques hors métrique chez Quintilien. On voit, à travers les emplois de clausula, se dessiner une conception du nom et du verbe en trois

366

Cécile Conduché

éléments, la base, l’élément de dérivation lexicale et la terminaison porteuse des marques de flexion.

5 Conclusion Dans de telles conditions, il nous semble légitime, dans la traduction d’Eutychès, d’employer le terme ‘suffixe’ pour rendre clausula. Le terme reste anachronique étant donné que le grammairien latin ne place pas son analyse dans le cadre d’une théorie des affixes, qui n’a émergé que de nombreux siècles plus tard. Audelà de la question ponctuelle de la traduction d’Eutychès, l’histoire du terme et de la notion de κατάληξις/clausula telle que nous avons proposé de la reconstituer indique une direction de recherche qui n’a pas, à notre avis, été assez explorée, celle de l’influence de la terminologie et des analyses métriques sur la grammaire gréco-latine dans son ensemble. On a en effet, depuis deux siècles, beaucoup souligné la dette de la grammaire envers la dialectique d’une part et la rhétorique d’autre part. En revanche, la métrique a généralement été considérée comme un appendice de la grammaire, au sens strict de l’étude des parties du discours, alors qu’elle pourrait bien, on le voit, en constituer l’une des sources.

Bibliographie Aujac, G. (2003). Denys d’Halicarnasse: Opuscule rhétoriques (Collection des Universités de France). Paris: Les Belles lettres. Barwick, K. (1922). Remmius Palaemon und die römische Ars grammatica. Leipzig: Dieterich. Casaceli, F. (1974). Phocas: De nomine et verbo. Napoli: Libreria scientifica. Chiron, P. (1993). Démétrios. Du style. Paris: Les Belles lettres. Gaide, F. (1992). Les substantifs « diminutifs » latins en . . . LVS, . . . LA ou . . . LVM. Revue de philologie, 66(1), 15–27. Gaide, F. (2002). Les substantifs « diminutifs » latins en . . . LVS, . . . LA ou . . . LVM. Dans C. Kircher-Durand (Ed.), Grammaire fondamentale du latin (vol. IX, pp. 111–123). Louvain/ Paris: Peeters. Holtz, L. (1981). Donat et la tradition de l’enseignement grammatical. Paris: Centre national de la recherche scientifique. Ippolito, A. (2006). Hephaestion. Dans Lexicon of Greek Grammarians of Antiquity, reference works.brillonline.com/browse/lexicon-of-greek-grammarians-of-antiquity (dernier accès: 8 mars 2019). Kaster, R. A. (1988). Guardians of language: The grammarian and society in Late Antiquity. Berkeley: University of California Press. Keil, H. (1857–1880). Grammatici Latini ex recensione Henrici Keilii (7 vols.). Leipzig: Teubner. Leo, F. (1889). Die Beiden Metrischen Systeme des Alterthums. Hermes, 24, 280–301.

Clausula : un terme technique de la métrique à la morphologie

367

Leumann, M. (19765). Lateinische Laut- und Formenlehre. Munich: Beck. Lomanto, V. & Marinone, N. (1990). Index grammaticus. Hildesheim/New York: Olm-Weidmann. Louis, P. (Ed.) (1964–1969). Aristote: Histoire des animaux. Paris: Les Belles lettres. Mankin, D. (2011). Cicero: De oratore, book III. Cambridge: Cambridge University Press. Mazzarino, A. (1973–1974). Appunti sul metodo: Intorno all’età e all’opera di Foca. Helikon, 13–14, 505–527. Schad, S. (2007). A lexicon of Latin grammatical terminology. Pisa/Rome: Fabrizio Serra. Thesaurus Linguae Graecae (2000–). stephanus.tlg.uci.edu (dernier accès: 15 March 2019). Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. Zucchelli, B. (1969). Studi sulle formazioni latine in -lo- non diminutive. Parma: Università degli studi di Parma.

Wolfgang David Cirilo de Melo & Panagiotis Filos

Linguistic terminology in Varro Abstract: Just as grammatical theory came from Greece, so did grammatical terminology, whether through direct loans or adaptations using Latin lexical material. Varroʼs terminology is strikingly different from what we find in later grammarians. He uses Greek words sparingly, not only for the names of the grammatical disciplines, but also for some highly technical elements. For most concepts between these extremes, he works with Latin lexical material rendering Greek terminology. But the overarching principles of Varronian linguistics are not expressed through calques or semantic shifts based on Greek. Here Varroʼs thought is as independent as his language. Keywords: Latin linguistics, grammatical terminology, borrowing, calque, Greek, loan shift

1 Introduction This chapter focuses on linguistic terminology in the first large-scale treatment of the Latin language, Varroʼs De lingua Latina. Originally, the work consisted of 25 books; an introductory volume was followed by 6 volumes on etymology, 6 on morphology, and 12 on syntax. The first two hexads on etymology and morphology were further divided into two triads each, one devoted to theory and one to practice. All that survives in direct transmission is books 5–10, that is, the text on practical etymologies and morphological theory. We shall restrict ourselves to these books transmitted directly and leave aside the fragments from other books of the De lingua Latina, or other grammatical works written by Varro, because these pose their own set of problems.1

1 To mention just one such problem, datiuus is almost non-existent in the directly transmitted portions, where (casus) dandi is preferred. The fragments contain datiuus regularly. Probably the relevant fragments are paraphrases rather than direct quotations, although the occasional datiuus in direct transmission makes us hesitate. Wolfgang David Cirilo de Melo, University of Oxford Panagiotis Filos, University of Ioannina https://doi.org/10.1515/9783110647587-023

Linguistic terminology in Varro

369

Given the influence of Greek grammarians on their Latin counterparts, and of Latin grammarians then on those working on modern languages, the most significant question is perhaps the following: to what extent are we dealing with loans, loan translations and calques, and independent innovations? The second question, which cannot really be separated from the first, is what should count as linguistic terminology. For later Latin, this may not be much of a problem, but for earlier treatises it is a serious issue, as the terminology has not yet become established and there are many usages that diverge from what we find later on.

2 Ways of adapting terminology The grammatical terminology we find in English and German, two principal languages of modern scholarship, but also two representative cognate Germanic languages influenced by Latin in their lexicon, is based on Late Latin terminology rather than on Varro; this later terminology is discussed extensively in Schad (2007) and Lomanto and Marinone (1990). English typically borrowed the later Latin terminology wholesale. Thus, from Latin casus, syllaba, and accusatiuus came about English case, syllable, and accusative; the corresponding German terms are Kasus, Silbe, and Akkusativ. Such terms have, for the most part, undergone only minor adaptations, but in syllable/Silbe we can see bigger morphological and phonological changes. Interestingly, German also has alternative loan translations for many grammatical terms. Fall ʻfallingʼ is an alternative for Kasus, but most speakers of German are not aware of the metaphor behind Fall, just as many Romans would not have been aware of the metaphor behind casus, which is a translation of Greek πτῶσις ʻfall, falling downʼ.2 For the Greeks, the nominative was the reference form and was treated as ʻuprightʼ (ὀρθή/rectus), like the stem of a tree, while the other cases were ʻbentʼ (πλαγία/obliquus) and thus falling (πτῶσις/casus). Even in Greek, the terminology became fixed and the metaphor was no longer perceived; hence terms like πτῶσις ὀρθή ʻupright caseʼ (lit. ʻfalling downʼ) for the nominative, which is a contradiction in terms. Latin adapts Greek in similar ways: sometimes we find direct loanwords, sometimes we find loan translations of various kinds. As we will see, for Varro the choice between these strategies is based on semantic fields.

2 For Greek terminology see especially Bécares Botas (1985).

370

Wolfgang David Cirilo de Melo & Panagiotis Filos

3 What is terminology and what is non-technical? Varroʼs grammatical terminology is quite different from what we find in later authors. There are two reasons for this. The first is that a great deal of Latin grammatical terminology consists of calques of Greek terms. The Romans developed an interest in grammar rather late, and mostly as a means of studying rhetoric. Ciceroʼs works on rhetoric provided a pleasant and elegant discussion of rhetoric in Latin, but for practical purposes they were not particularly useful. His contemporaries would instead read Greek treatises, or dry, technical introductions in Latin like the Rhetorica ad Herennium. This means that there was no fully developed and standardized Latin terminology in Varroʼs time. The second reason is that Varroʼs treatise is, to all intents and purposes, quite idiosyncratic. Latin grammar is dealt with as an end in itself, not as a means to prepare for rhetorical studies or to emend ancient authors. In his day, this was a novelty. And so he does things that other grammarians would not normally do. The entire discussion of analogy and anomaly has few parallels elsewhere. He is going his own way and is using his own terminology, which is not always consistent (contra Taylor 1978: 73–74). Quite a large part of Varroʼs grammatical expressions were not adopted by later authors. But this presents us with a major problem. What should count as grammatical terminology? When terms like casus or accusatiuus are used, it is easy to see that they are intended as technical. But what about exterere ʻto rub out, elideʼ? Is it intended as a technical term, or is Varro presenting technical facts in nontechnical language? There is no foolproof way of finding out. For this chapter, we have collected a complete list of grammatical terms, both real and potential ones. Anything that is clearly calqued on Greek and/or lives on in later Roman grammarians can be treated as a technical term, even if it is attested only once in Varro. Anything that is not calqued and does not live on later on is treated as a technical term if it is frequent enough to be statistically significant (that is, occurring at least five times), and attested with a consistent meaning. Finally, anything that is not calqued and does not live on, and is attested less than five times, receives special discussion. A few words should be said about our method of rendering Varroʼs terminology. In principle, one can distinguish between glosses, transpositions, and definitions. To give an example, appellandi pars could be glossed as ʻpart of addressingʼ, transposed as ʻnominal(s)ʼ, and defined as ʻa/any word with case, but without tenseʼ. For each term, we have tried to gloss, transpose, and define, but for practical reasons not always in this order; what stands in inverted commas can be a gloss, a transposition, a definition, or a combination of these elements. This procedure may not appear to be very

Linguistic terminology in Varro

371

systematic, but we felt that it would work better in practice than a more rigid method.

4 The actual terminology 4.1 Greek terms with no synonyms or synonyms that are restricted in use Somewhat surprisingly, Varro uses only 11 terms that are straightforwardly Greek, which marks them as technical. Just as what remains of Varroʼs work, these fall into two groups, those dealing with etymology and those dealing with morphology. We will ignore the spelling of these terms, as the scribe of the most significant manuscript, F, clearly struggled with the Greek script; not everything rendered in the Latin alphabet was necessarily written in this way by Varro. On the etymological side, we come across ἐτυμολογία and its synonym ἐτυμολογική ʻetymologyʼ. The former term is attested in 5, 2, where it is said that the Greeks use it. The latter term is found in 5, 1 and 7, 109 and is not treated as a purely Greek expression, which indicates that it had some currency in Latin; in a Greek context, one would sometimes find or have to understand τέχνη with it. In 5, 2, Varro states explicitly that the term refers to the formal side of things rather than to the meaning of words. What is remarkable is that on the first occurrence of these terms, Varro feels the need to explain them, even though they must have come up before in the preceding books on the theory of etymology. Connected with these terms is ἐτυμολόγος ʻetymologistʼ (5, 29 and 6, 39), and ἔτυμον ʻgenuine form, base form, elementʼ. The latter is attested seven times, and Varro may in fact provide us with the first attestation of this word as a technical term in both Greek and Latin, even though the Stoics are likely to have preceded him in that respect (cf. also his contemporary DIOD. SIC. 1, 11). Its meaning is consistent; in 7, 45, for instance, the ἔτυμον of Martialis ʻof Marsʼ is Mars. For Varro, such a base form is always a full word, not an endingless morpheme. Etymologia is occasionally translated into Latin, but not by Varro. CIC. top. 35 has notatio and ueriloquium, and QVINT. inst. 1, 6, 28 adds originatio as a third term. None of these words is used by Varro, at least not in the extant portions of our work, and although ueriloquium ʻtruth-speakingʼ is the closest translation of the Greek word, it is also the one that is used the least, probably because neither in Greek grammar nor in Latin is it the goal of etymology to

372

Wolfgang David Cirilo de Melo & Panagiotis Filos

find the ʻtrue meaningsʼ of words (Grebe 2001: 142); rather, etymology is meant to establish formal connections between words that sound similar. As already noted by Cavazza (1981: 214), Varro has much more as well as clearer and better terminology for morphology than he has for etymology. However, most of it is not Greek in an obvious way. On the Greek side, one comes across ἀναλογία ʻanalogy, regularityʼ recurrently. For Varro, analogy is the principle of regular inflection. Its opposite, ἀνωμαλία ʻanomaly, irregularityʼ, is almost as common as a term, and much of his work is devoted to the ultimately Hellenistic question of how regular language is. Connected with these concepts are λόγος ʻrelationshipʼ and ἀνὰ λόγον ʻin proportionʼ. The former can refer to the relationship between numbers; for instance, 5 is half of 10. Or it can refer to that between two case forms, like the nominative and genitive of a specific word. Ἀνὰ λόγον occurs only five times, and is normally rendered as pro portione, as Varro says explicitly in 10, 37. We shall now focus briefly on the Latin (near-)synonyms. Pro portione refers to what we today call ʻfour-part/proportional analogyʼ. As 1 is to 2, so 10 is to 20; or as equus ʻhorseʼ (nominative) is to equi ʻhorseʼ (genitive), so catulus ʻpuppyʼ (nominative) is to catuli ʻpuppyʼ (genitive). For Varro, the expression pro portione is still a prepositional phrase. It occurs 30 times in this form. Only once, as an accusative object, do we find proportionem (8, 57), but perhaps this passage is corrupt. Varro speaks of pro portione when he wants to stress the idea of a mathematical pattern underlying regular inflection, but he prefers analogia when he focuses on grammatical doctrine (Schironi 2007: 326). Similitudo ʻsimilarityʼ is used 21 times, always non-technically, in books 5–7, and 73 times, always technically, in books 8–10; because of the many nontechnical uses, Ramos Guerreira (1985: 263) expresses doubts about the technical value of the word, but the semantic distinction between the etymological books and the morphological ones is neat. Similarity is the precondition for analogy. Ratio ʻprinciple, method; proportionʼ is attested 73 times, often in nontechnical meanings. Thus, in 6, 63 the word means ʻaccountʼ or ʻbalanceʼ in a financial sense. In 6, 39, the use is only seemingly linguistic: we have basic forms, and of these no further ʻaccountʼ is to be given, which does not mean, though, that our word here is a technical term for ʻetymologyʼ; on the other hand, we could be dealing with a playful pun between the technical and the non-technical usage. However, while not all attestations of the word have a technical meaning, it is worth pointing out that 70 of the 73 attestations occur in the morphological books, where grammatical terminology is in general more frequent than in the etymological books. In 8, 6, ratio simply means ʻmethodʼ; in other words, if one learns to inflect one word through such a method, one can inflect many others. In 8, 57, ratio uerborum is technical; it is the principle of

Linguistic terminology in Varro

373

analogy, opposed to irregular usage. This opposition is quite common; ratio is the reasonable expectation that words inflect along the same lines. Elsewhere, as in 9, 79 or 10, 36, the word indicates extra-linguistic or linguistic ʻproportionʼ; the same ratio that we find in nominative equus ʻhorseʼ and genitive equi recurs in nominative erus ʻmasterʼ and genitive eri. This meaning ʻproportion, relationshipʼ is, as Varro points out in 10, 2 and 10, 37, his rendition of Greek λόγος. Aequabilitas ʻanalogyʼ is probably intended as a technical term, but is only used once, in 9, 1 in a translation of a book title, as a synonym for analogia. On the opposing side, one finds the word dissimilitudo ʻdissimilarityʼ. It is attested 17 times, and always in a technical meaning. Dissimilarity can exist in internal features like case, external ones like sound, or both. Inaequabilitas ʻanomalyʼ is only attested three times (8, 28; 8, 30; 9, 1), but probably intended as a technical term (in 9, 1, again as a rendition of a Greek book title). It is a synonym for anomalia, which is much more common. Only three other purely Greek terms are used, with fairly specialized meanings. Περ σημαινομένων ʻmeaning, semanticsʼ is only found at 5, 2, where it is ascribed to ʻthe Greeksʼ, which indicates that it had not become a standard term in Latin grammar. It refers to the meanings of words in the etymological process. Ὁμωνυμία ʻhomonymy, instance of homonymyʼ is found twice in 9, 89. The term is ascribed to others, but it is not specified whether these others are exclusively Greek or include Romans. Varroʼs usage of the term corresponds nicely to our modern usage. And finally, συνωνυμία ʻsynonymy, instance of synonymyʼ is only found in 9, 90 and is ascribed to others; we are not told whether these others include Roman grammarians, but at least it is clear that we are dealing with a technical term. Interestingly, Varro does not use the term for near-synonyms like buy and purchase, but for alternative inflections of the same personal name. What we can see quite clearly here is that Varro uses Greek only for general, overarching principles and for their opposite, namely highly specialized technical terms. For the latter, it makes life easier not to invent native terminology. For the former, Greek terms do have native synonyms, but the Greek terms are often preferred because the native ones have broader meanings and could be construed as non-technical.

4.2 Mixed terminology Terminology that is part Greek and part Latin is virtually absent. Syllaba ʻsyllableʼ, a direct loan from συλλαβή, is frequent, and forms various compounds: bisyllabum ʻdisyllableʼ (9, 91), trisyllabum ʻtrisyllableʼ (8, 77; 9, 91), and quadrisyllabum ʻquadrisyllableʼ (8, 77, conjectural).

374

Wolfgang David Cirilo de Melo & Panagiotis Filos

4.3 Terminology using the Latin lexicon 4.3.1 Words, word classes, roots Our modern-day ʻparts of speechʼ are the Greek μέρη λόγου. Varro has pars orationis, but this is not always technical. In 9, 45 we find in maiore parte orationis, and it is clear that the adjective modifies parte rather than parte orationis, so we have to translate as ʻin the majority of speechʼ. Elsewhere, we come close to the Greek ʻparts of speechʼ. Verbum is extremely common and often used non-technically as ʻwordʼ. Thus, in 7, 30 the preverb ambe- ʻaroundʼ is called a uerbum, and in the next chapter the noun adagio ʻproverbʼ is also called a uerbum. Not even in the morphological books is the word always used in the technical meaning ʻverbʼ; in 8, 1, homo ʻmanʼ is called a uerbum. The technical meaning ʻverbʼ is a loan translation of ῥῆμα ʻword, verbʼ. We can see this meaning clearly in 8, 11, where nouns and verbs (uerba) are opposed. It is unclear whether primigenium uerbum ʻbase wordʼ is technical; it is attested three times (and in 6, 36 we also find uerborum primigenia), and refers to words that have their own roots and are not derived from some other word. Praeuerbium, attested five times, corresponds to Greek ἐπίρρημα and to our modern ʻpreverbʼ. It is not obvious whether this technical term is restricted to prefixes used for verbs, or whether it can refer to any prefix. Radix ʻrootʼ is more complex. In 7, 28, we find the metaphor that Sabine extends its roots into the Oscan language. But the metaphor of roots is not restricted to genetic relationships between languages. In 5, 93 we learn that mederi ʻto healʼ is the root of medicus ʻmedical doctorʼ. This shows that Varroʼs concept of a root is quite different from our modern concept of a root, which is a form stripped of all affixes that may not occur on its own in a language – barring a few words that are nothing more than the root itself. Based on the presence or absence of case and tense, four parts of speech are distinguished. In 8, 44–45, these are referred to as appellandi (pars), ʻpart of addressing, nominalʼ, corresponding to ὄνομα ʻnounʼ; as dicendi (pars), ʻpart of saying, verbʼ, rendering ῥῆμα ʻthat which is said, verbʼ; as iungendi (pars), ʻpart of joining (case and tense), participleʼ, ascribed to others and translating μετοχή; and as adminiculandi (pars), ʻpart of supporting, uninflected part of speechʼ, referring to uninflected words. Further subcategorization is possible. Among words with case, the genuine nominals are called nominatus. These nominatus comprise nomina ʻpersonal namesʼ and uocabula ʻcommon nounsʼ, but these two designations can be used more loosely. Nomen seems to correspond to ὄνομα only at first sight. The Greek

Linguistic terminology in Varro

375

word also means ʻnameʼ in a non-technical sense, yet in grammatical treatises it indicates the noun more generally3 (but for the Stoics only the ‘proper noun’), so Varro is not simply taking over the general Greek usage here. In most instances, nomen as a technical term thus has the same meaning that it has in nontechnical passages, that of ʻpersonal nameʼ. However, sometimes Varro does not follow his own definitions strictly. In 7, 109, nomina ʻnamesʼ are given to things, and this looks like a non-technical usage repeated in 8, 5. But from there it is only a small step away from calling things nomina ʻnouns,ʼ and occasionally Varro does this, as in 8, 13, where nomen and uerbum are contrasted and can be translated as ʻnounʼ and ʻverbʼ, or as ʻsubjectʼ and ʻpredicateʼ. Similarly, the boundaries of uocabulum are sometimes blurred. For ʻpronounʼ, Varro has articulus. In Greek, ἄρθρον ʻjointʼ came to be used as the term for ʻarticleʼ, even though for the Stoics it also referred to pronouns. Articulus ʻjointʼ is used by later Roman grammarians as a translation of the technical meaning of ἄρθρον. For instance, Quintilian (inst. 1, 4, 19) points out that the Latin language does perfectly well without articles. However, Varro uses articulus 12 times (two instances are conjectural), for Latin words, without any reference to Greek articles. For Varro, articulus is simply the cover term for all pronouns, whether definite or indefinite. Pronomen stands for ʻdefinite pronounʼ. Varroʼs use is quite different from the use of pronomen in later times, where it means ʻpronoun in generalʼ; only Priscian follows Varro. Prouocabulum stands for ʻindefinite pronounʼ. This term is made up by Varro and occurs only in 8, 45. For ʻparticipleʼ, Varro has participium and participale, attested only three times in total. These terms are a better rendition of μετοχή ʻparticipleʼ, literally ʻsharing (properties of nouns and verbs)ʼ. Interestingly, in all three attestations of the term, Varro indicates that it is called like this by others. As Varro is mostly interested in inflection, there are not many terms for uninflected words. Copula ʻconnective, connectionʼ is among them. In modern terminology, a copula is a verb like to be, which links subject and predicate. Varro does not have this usage. For him, the copula is et ʻandʼ (8, 10). This usage does not survive in later grammarians, and it is unclear whether Varro intended it as technical. The word is attested five more times, four times in 9, 4 and once in 10, 33. Each time, a connection between two opposites is meant, for instance that between (regular) nature and (irregular) usage, or that between present stem and perfect stem. The word comes close to our expression ʻlinguistic oppositionʼ. Another type of uninflected word is called effutitum ʻinterjectionʼ; the word comes up only in 7, 93, but it is not certain whether

3 Also including adjectives; Varro does not distinguish between nouns and adjectives either.

376

Wolfgang David Cirilo de Melo & Panagiotis Filos

effutitum naturaliter est should be taken as a phrase with a periphrastic perfect, ʻ(it) has been poured out naturallyʼ, or as a phrase that contains a nominalized participle and is thus in the present tense, ʻis a natural interjectionʼ. Only in the latter case does it make sense to treat effutitum as a potential technical term. Varro does not yet have the term interiectio ʻthrowing in, interjectionʼ, a translation of παρεμβολή,4 but effutitum could be treated as a less literal loan translation, and this is why we are inclined to see it as a technical term. For word classes, then, the connection between Varronian terminology and Greek is relatively loose.

4.3.2 Case, number, gender We have already discussed casus ʻcaseʼ, a term so common that Varro also derives an adjective casualis from it (8, 52; 10, 18). Casus rectus, rectus, and rectum, all used interchangeably, are loan translations, as seen earlier. The same is true of obliquus and obliquum. The names of the cases Varro uses are also based on Greek. Nominandi ʻ(case) of namingʼ and nominatiuus are used for the nominative, as translations of Greek ὀνομαστική. The term nominatiuus, which lives on in later grammarians, is only found in 10, 23, while nominandi is found nine times. The latter does not catch on, probably because it is a genitive with ellipsis of the head noun, and the case of the head noun cannot be seen in this short expression, while nominatiuus, which is equally elliptical, works better, as it agrees in case with the head noun that is missing. Patricus ʻgenitiveʼ, or ʻfatherʼs (case)ʼ, is a mistranslation of Greek γενική ʻ(case of) originʼ. Patricus is attested five times and may well be a Varronian coinage,5 while later grammarians use genitiuus as a more precise rendering of the Greek term. Dandi ʻdative, (case) of givingʼ, for δοτική, is also found five times. Later on, only datiuus is used, a term which Varro may have used himself, given that he also has nominatiuus and accusatiuus. For the accusative, Varro has accusandi (8, 66) and accusatiuus (8, 67), both translations of αἰτιατική, and bad ones at that. The Greek term means and is intended as ʻ(case of the thing) causedʼ, although in theory it could also mean ʻ(case) of accusingʼ. This latter meaning was not intended by Greek grammarians,

4 CHAR. gramm. p. 315, 14–15 seems to quote rather than paraphrase Varro and has interiecta particula. 5 But note also DIONYS. THRAX p. 31, 7, who has πατρική.

Linguistic terminology in Varro

377

but eventually gave rise to the standard Latin terms. For the vocative, we find uocandi, a translation of κλητική ʻfor callingʼ, attested five times. In 9, 43, the full form casus uocandi is found, but the elliptical use of uocandi is the norm. As Varro has accusatiuus beside accusandi, it is likely that he could have used uocatiuus as well, but no such form is attested. For the ablative, Varro says sextus, ʻsixth caseʼ. It is unclear whether this is meant as a technical term. It is obviously not a translation from Greek, as Greek only has five cases and the sixth case is ʻpeculiar to Latinʼ, as Varro says in 9, 62, the only passage where the term occurs. Varro does not yet have a term ablatiuus. For number, Varro has the obvious numerus, a rendition of ἀριθμός. The word is attested very regularly in Varro, but there are only three certain instances where grammatical number is meant. In 9, 66 and 9, 67, nouns with numerus contrast with mass nouns (mensura, not necessarily a technical term). And in 10, 65, numerus is treated as a grammatical category like case. Varro already has singularis ʻsingularʼ. This is a common word and can be used as a noun or as an adjective. Unlike its Greek equivalent ἑνικός, which is almost exclusively a grammatical term,6 singularis can have the non-technical meaning ʻindividualʼ as well, but is common as a grammatical term. For plural, Varro has multitudinis. The Greek equivalent πληθυντικός is a purely technical term. Latin multitudo ʻgreat numberʼ is an everyday word, but Varro, and almost only Varro, uses this word as the technical term for ʻpluralʼ. However, this term is common enough in Varro to see that the technical usage differs substantially from the non-technical ones. Varro uses this noun mostly as an elliptical genitive (ʻ[words] of the pluralʼ). After prepositions construed with the ablative or accusative, it is again the fixed, elliptical genitive that is used, although Varro can make a concession and say in multitudine ʻin the pluralʼ as well as in multitudinis ʻin [words in] the plural.ʼ This restriction in case may be one reason why Varroʼs usage did not catch on. Later grammarians prefer pluralis. For ʻgenderʼ, Greek has γένος; Varro uses both genus and sexus. Genus is the looser term. The word is extremely common and mostly used in the nontechnical meaning ʻkind, typeʼ. Even in linguistic contexts, the word mostly has non-technical uses: thus, there are four genera of words, those with case, those with tense, those with both, and those with neither (6, 36); then there are two genera of words, those that inflect and those that do not (8, 9); or there are two genera of words, and these are noun/subject and verb/predicate (8, 12). The fact that genus can be used in all these contexts shows that it is not a technical term as such. However, just like later grammarians, Varro also uses genus in

6 Provided that we leave aside some Neoplatonic, non-grammatical usages.

378

Wolfgang David Cirilo de Melo & Panagiotis Filos

the meaning ʻgrammatical genderʼ. We find this use in 9, 37 and 9, 57, while in 9, 62 and 10, 8, Varro uses the more explicit genus nominum ʻgender of nounsʼ. For sexus, the meaning ʻbiological sexʼ is very common in Latin, but not found in Varroʼs work under examination; here the word is attested only in 8, 46 and means ʻgrammatical genderʼ. The names of the genders are uirile, muliebre, and neutrum, corresponding to ἀρσενικόν, θηλυκόν, and οὐδέτερον. Masculinum and femininum are not used by Varro in this work, but interestingly, in rust. 3, 5, 6 femininum is used for the grammatical gender of the merula ʻblackbirdʼ. Varro knows the distinction between biological sex and grammatical gender, as he shows in 9, 40–41, but he uses uirilis for ʻmaleʼ as well as for ʻmasculineʼ. All in all, as one may see, terminology for core nominal concepts is heavily dependent on Greek.

4.3.3 Other terminology for nominals For Varro, Latin nouns can be definite, indefinite, or something in between (8, 45). For ʻdefiniteʼ, we find finitus, for ʻindefiniteʼ, infinitus. These may be loosely based on σαφής and ἀσαφής, which mean ʻ(un-)clear, (in-)distinctʼ, so that the later definitus and indefinitus are better renditions than Varroʼs terms, which have more to do with limitations. It is not clear whether certus ʻdefiniteʼ, found as a synonym of finitus in 9, 85, is meant as technical term. For ʻdegree of comparisonʼ, the Greek term is ἀξίωμα, which in nontechnical works means ʻrank, distinctionʼ. Varroʼs contentio ʻcompetitionʼ is thus not a translation; and since it only occurs in 8, 75 and is not used by later grammarians, we do not know whether it was intended as a technical term. Collatio is only attested in 8, 78 in Varro, but lived on in later grammarians. Declinationis genus augendi ʻthe inflection type of increaseʼ is attested only in 8, 52. Diminutives are also discussed in 8, 52, where Varro speaks of the declinationis genus minuendi ʻinflection type of diminutive formationʼ. The expression does not recur elsewhere, and Varro does not yet have deminutiuus/diminutiuus. It is unclear whether the gradus ʻsteps, degrees of diminutionʼ are meant as technical. Finally, as far as compounding is concerned, Varro makes use of compositus and compositicius. Although these are attested only three times, i.e. as an adjective and a neuter noun, we are clearly dealing with a technical term here. It is similar to, but not directly translated from Greek σύμμιγμα ʻmixing together, compoundʼ. From the examples adduced, it is unclear whether Varro classifies as compounds only words that contain two lexical elements, or also

Linguistic terminology in Varro

379

those with one lexical element and a prefix like com- ʻtogetherʼ. Concludere means ʻto compoundʼ in 5, 171, but the single attestation may not be technical. For these other nominal concepts, Varroʼs dependence on Greek is less marked. 4.3.4 Person, voice, tense, aspect, mood For ʻpersonʼ, Varro uses persona, a translation of Greek πρόσωπον ʻface, mask, (linguistic) personʼ. In 8, 20, Varro makes the same distinction that we have today: the first person is the speaker, the second person is the addressee, and the third person is neither. Varro does not use a term for ʻvoiceʼ and avoids terminology for ʻactiveʼ and ʻpassiveʼ. In 10, 33, he speaks of the (species) faciendi and patiendi, and instead of speaking of deponents, he refers to verbs that do not have an opposite. The later terminology of actiuum and passiuum does not yet occur; it is a clearer rendition of ἐνεργετικόν and παθητικόν. For ʻtenseʼ one comes across tempus, which is so common that one may also find an adjective temporalis, mostly used in uerbum temporale ʻtime word, verbʼ. Like the Greek equivalent χρόνος, tempus indicates both ʻtimeʼ and ʻtenseʼ. The Latin term is technical and could be a translation of the Greek term, although it is hard to see in which other possible way a Roman could have talked about verbal time. The chief tenses are expressed through the endings of finite verbs. For Varro, these are praeteritum ʻpastʼ (παροιχόμενος χρόνος), praesens ʻpresentʼ (ἐνεστώς χρόνος), and futurum ʻfutureʼ (μέλλων χρόνος), the latter only attested three times. Unlike other grammarians, Varro recognizes that the three tenses exist for each verb stem, infectum and perfectum; these two expressions may be Varronian coinages without clear Greek equivalent, but in modified form the terminology lives on (QVINT. inst. 1, 6, 26 has praeteritum perfectum for the perfect tense). Varro talks very little about mood. In 9, 101 we find the (convincingly conjectured) form indicandi (diuisio) for ʻindicativeʼ, probably a coinage on ὁριστικὴ ἔγκλισις. Altogether, for central verbal terminology Varro relies quite heavily on Greek. 4.3.5 Basic principles of analogy Materia and res are both used for ʻgrammatical substance, internal featuresʼ like case or tense. Gender is another such internal feature; it is seen from agreement patterns, but not always from the outward shape of a word. Such internal

380

Wolfgang David Cirilo de Melo & Panagiotis Filos

features are ʻunderlyingly presentʼ, for which Varro says subesse.7 For this semantic side of things, Varro has the verb significare, the adjective significabilis (only in 6, 52), and the noun significatio (only in 9, 40). One could of course argue that Varro is rendering σημαίνειν here, but it is hard to see how else Varro could be talking about meaning. This cannot be said of the Varronian adsignificare ʻto mean, indicate in additionʼ, a new coinage barely attested outside Varro. It seems to render Aristotleʼs προσσημαίνειν ʻto indicate in additionʼ (Hartung 1973: 296). Five times the concept indicated in addition is tense; Varro is speaking about verbs, which have a lexical meaning, but also mark time. Internal features contrast with figura and uox (adjective uocalis used only once securely), both of which stand for the ʻspoken formʼ of a word. Greek σχῆμα could in non-technical usages be translated as figura, but Varroʼs technical usage differs considerably from σχῆμα as a technical term, which refers to the (accentual) distinctions between simple nouns and compounds (Lomanto 2001: 174). Loquela ʻspoken formʼ is a synonym of figura, but ascribed to others (6, 57). Perfect analogy exists when a transition of internal features like number is matched by a transition in sound. For these transitions, Varro uses transire and transitus/transitio as well as flectere ʻto inflectʼ (10, 29) and flexura ʻinflectional pathʼ (10, 28). Transitions in sound occur especially in the exitus ʻendingʼ (9, 62, twice in 10, 21), but not every word is commutabilis ʻinflectableʼ (9, 99, where we also find incommutabilis). The ending consists normally of one or two litterae ʻletters, soundsʼ; Varro makes the distinction between letter and sound, but does not yet have the terminology for it, unlike PRISC. gramm. II 6, 14, who has littera and elementum. In sum, here we can see a fairly independent Varro at work.

4.3.6 More on analogy When we name a thing or an action, this act is called impositio (verb imponere, adjective impositicius). Naming takes place in the base form, which is the nominative singular for nouns and the first person singular for verbs. For ʻbase formʼ, Varro has caput, but also principium, although the latter can also refer to linguistic principles more generally (8, 5 or 10, 11).

7 One may wonder whether subiungere ʻto subordinateʼ is meant as a technical causative to subesse in 10, 66–67, where plurals are subordinated to singulars.

Linguistic terminology in Varro

381

When it comes to the derivation or inflection of words, Varro speaks of inclinare or, more commonly, of declinare; the nouns are inclinatio, declinatus, and most commonly declinatio. All of these are based on κλίσις ʻbending, inflectionʼ. Varro can draw a distinction between derivation and inflection; the former is declinatio uoluntaria/declinatus uoluntarius, based on human will and thus somewhat irregular, the latter is declinatio/declinatus naturalis, based on nature and thus fairly regular. Among the different forms of analogy, there are two types. One is deiunctus ʻdisjoinedʼ (10, 45; 10, 47), having four elements, like our four-part analogy (sing: sang:: ring: rang). The other type is coniunctus ʻconjoinedʼ, having four elements, but one occurs twice (past: present:: present: future). As a side note, coniunctus is also used for a simple verb combined with a preverb or for pluralonly nouns having constituent parts (scalae ʻstairsʼ). The Greeks organized inflectional forms in a παράδειγμα, an ʻexhibition side by sideʼ, a good description of what a paradigm actually does. Varro has forma or formula, indicating the outward shape of a thing or a word (10, 27), an individual form within a paradigm (8, 9), and then a paradigm itself (9, 105). Pure analogy should in practice be moderated by consuetudo/usus ʻusageʼ. In this area, Varro can be seen to be quite independent of Greek terminology.

4.3.7 Other terminology There are three words for phonological processes. Cogere ʻto contractʼ (10, 81) could potentially be based on συναγωγή ʻcontractionʼ. Exterere, attested four times in the perfect participle, is used for the loss of a single sound or of a syllable; and excludere, twice in the perfect participle (5, 133 and 9, 57), is its synonym. Although QVINT. inst. 1, 4, 14 also uses excludere, it is unclear whether any of these terms is intended as technical. For metaphorical expressions, Varro has tralatus/translatus/tralaticius/ translaticius, based on μεταφέρω ʻto take take over, use metaphoricallyʼ. The opposite is proprius ʻliteral, non-metaphoricalʼ. In 8, 80, on the other hand, propriae res are ʻproper namesʼ as opposed to common nouns. Other words are ambiguous. Species ʻtype, speciesʼ is not always technical. The types referred to can be different persons or numbers (6, 36), or agent nouns (8, 57), or word classes (10, 18 [20]). In 10, 73, Varro distinguishes three species of usage: old, current, and poetic. And finally, in 10, 79 there are different species of analogy, to be followed by different groups of people. Diuisio ʻdivisionʼ is equally vague: it refers to divisions into various word classes, but there are also various divisions within verbs, for instance into present and perfect stems.

382

Wolfgang David Cirilo de Melo & Panagiotis Filos

Similarly, communis ʻcommonʼ can modify consuetudo or usus, but there is also the common case seen in ouis ʻsheepʼ (nominative or oblique). In 8, 80, res communis is used for ʻcommon nounsʼ as opposed to proper names. And in 9, 89, nomen commune ʻcommon, shared nameʼ is Varroʼs gloss for ὁμωνυμία ʻ[instance of] homonymyʼ. Duplex ʻdoubleʼ is used for analogy if it exists in both form and meaning, for analytic verb forms consisting of two words, and for count nouns that have both singular and plural. Simplex ʻsimpleʼ is used for synthetic verb forms as opposed to analytic ones; for verbs that are basic citation forms and do not contain prefixes; finally, for nouns that are not compounds containing two lexical elements. In 10, 24, the word is used for nouns that have singular and plural, while plural-only nouns are for objects with individual parts which are joined together and which are therefore not simple. Finally, in 10, 63 and 10, 68, analogy is considered simple if it is simple only in form or only in meaning, but not in both. Many of these terms may not be technical, especially those that have such variable meaning.

5 Conclusions Given the strong dependence of Latin grammarians on their Greek predecessors, it comes as a surprise that Varro uses relatively few direct loans; where such loans occur, they are used in principle for the names of the linguistic disciplines and for some highly technical expressions. For anything between these extremes, Varro employs Latin lexical material to render Greek terminology. However, the overarching principles of Varroʼs linguistic scholarship are not expressed through calques or loan shifts; the language is as independent as Varroʼs thought here.

Bibliography Bécares Botas, V. (1985). Diccionario de terminología gramatical griega. Salamanca: Ediciones Universidad de Salamanca. Calboli, G. (Ed.) (2001). Papers on grammar, VI. Bologna: CLUEB. Cavazza, F. (1981). Due note sulla terminologia tecnica etimologico-grammaticale di Varrone. Giornale Italiano di Filologia, 33, 209–219. Grebe, S. (2001). Views of correct speech in Varro and Quintilian. In G. Calboli (Ed.), pp. 135–164. Hartung, H.-J. (1973). Παρεπόμενα ῥήματος bei Varro? (De lingua Latina 10, 31–33). Glotta, 51, 293–311.

Linguistic terminology in Varro

383

Lomanto, V. (2001). Nomi a confronto. In G. Calboli (Ed.), pp. 165–190. Lomanto, V. and Marinone, N. (1990). Index grammaticus: An index to Latin grammar texts. Hildesheim: Olms-Weidmann. Ramos Guerreira, A. (1985). Similitudo en el metalenguaje del De lingua latina de Varrón. In J. L. Melena (Ed.), Symbolae Ludovico Mitxelena septuagenario oblatae (vol. 1, pp. 255–263). Veleia: Instituto de Ciencias de la Antigüedad, Universidad del Pais Vasco. Schad, S. (2007). A lexicon of Latin grammatical terminology. Pisa/Rome: Fabrizio Serra. Schironi, F. (2007). Ἀναλογία, analogia, proportio, ratio: loanwords, calques, and reinterpretations of a Greek technical word. In L. Basset, F. Biville, B. Colombat, P. Swiggers and A. Wouters (Eds.), Bilinguisme et terminologie grammaticale gréco-latine (pp. 321–338). Louvain/Paris/Dudley, MA: Peeters. Taylor, D. J. (1978). ʻOrdoʼ in book X of Varroʼs De lingua Latina. In J. Collart (Ed.), Varron, grammaire antique et stylistique latine (pp. 71–74). Paris: Les Belles lettres.

Anna Novokhatko

The typology of linguistic metaphor in first-century CE Roman thought Abstract: This chapter, belonging to the field of the history of Roman linguistics, builds on my previous analysis of the theoretical linguistic treatment of metaphor in Roman thought from first century BCE to first century CE. It argues that metaphor from the very beginnings has been analysed not only within rhetoric but also within both linguistic and philosophical domains. Aristotelian cognitive approaches are significant for understanding Quintilian’s treatment of metaphor, which Quintilian develops along semantic and pragmatic lines. Seneca’s approach to metaphor contributes towards an understanding of the intellectual context of Quintilian’s first-century CE Rome. The iconicity of metaphor and metaphorical mapping developed in Seneca’s semiotic use of ‘mental pictures’ in their interplay with metaphor constituted a particularly important contribution to theoretical debates on metaphor. Keywords: Latin linguistics, conceptual metaphor, semantic theory of metaphor, metaphor and iconicity

1 Introduction This chapter builds on my two previous publications on linguistic approaches and theoretical treatments of metaphor in the first-century BCE and firstcentury CE Rome.1 For my project the corpus of Latin2 authors has been analysed with the aim of revealing the uses of the term metaphora (or tra(ns)latio,

1 Novokhatko (2016) and Novokhatko (2017). 2 As the issue concerns intellectual debates in Rome, Greek texts written at the same time and in the same context are also relevant, thus when necessary contemporary or source Greek texts have been considered as well. But the focus here is how the debates have been translated into Latin, which Latin patterns have been chosen and what is new in Latin – constructions, content, quoted examples, etc. – in the sense that we do not have similar attestations in Greek texts. For obvious reasons – as many of texts have been lost – we can never be sure that the idea is indeed originally Latin and has not been borrowed from a Greek original. In some cases such imitation and borrowing is attested, in other cases the question should remain open. Anna Novokhatko, Albert-Ludwigs-Universität Freiburg https://doi.org/10.1515/9783110647587-024

The typology of linguistic metaphor in first-century CE Roman thought

385

tra(ns)lata verba or such like). The results of this investigation reveal that although the Roman treatment of metaphor had always been regarded as merely rhetorical and stylistic, and this fact had repeatedly been emphasized by scholars (as opposed to Aristotelian cognitive approaches to metaphor!), the Romans do not discuss metaphor simply as a decorative figure, but also and deliberately within the linguistic and philosophical domains. I will focus on the first century CE exclusively in order to consider the typology of linguistic metaphor in various texts of the second half of the first century CE, such as Seneca’s letters (written before 65 CE) and his De beneficiis (finished perhaps by 64 CE) which reflect certain contemporary theoretical views on metaphor and some passages, not sufficiently analysed before, from Quintilian’s Institutio oratoria (published around 95 CE). I will argue that both linguistic manifestations of conceptual metaphors (having the form A is B/A as B, the concept A being understood in terms of concept B) and semiotic concepts were significant in Roman theory of language. In modern theory conceptual metaphors can be classified according to their conventionality (conventionalized vs. unconventional or novel), cognitive function (structural, orientational, ontological), nature (propositional knowledge and images of various kinds) and level of generality (specific level vs. generic level).3 The typologies of conceptual and linguistic metaphor that stem from this treatment reveal a greater depth to Roman theory of language.

2 Conceptual/cognitive metaphor theory and the creation of new word meanings I would first like to focus on the cognitive aspect, and in particular how linguistic metaphors can be analysed to better understand cognitive metaphor theory. In her 2005 monograph Metaphor and corpus linguistics Alice Deignan emphasizes: “It is fairly rare for a completely new lexeme to appear in the language. Gaps in the lexicon that arise through changes and developments in society and technology are more often filled by adapting existing lexemes. This happens through a range of processes, the most important of which is probably metaphor.”4 This role of metaphor has long been recognized, metaphor being the

3 See in particular Kövecses (2010: 33–46), Prandi (2010), Prandi (2004: 388–404); cf. Deignan (2005: 34–39). 4 Deignan (2005: 25).

386

Anna Novokhatko

most important force in the creation of new meanings in language,5 and this aspect is emphasized by Quintilian in book 12, where he compares Greek and Latin and argues that Greek vocabulary is richer than Latin. itaque tanto est sermo Graecus Latino iucundior ut nostri poetae, quotiens dulce carmen esse voluerunt, illorum id nominibus exornent. his illa potentiora, quod res plurimae carent appellationibus, ut eas necesse sit transferre aut circumire: etiam in iis quae denominata sunt summa paupertas in eadem nos frequentissime revolvit. (QVINT. inst. 12, 10, 33–34) ‘Thus Greek is so much more agreeable than Latin that our poets whenever they have wanted to make a sweet song adorn it with the words of the Greeks. More importantly, many things have no designations, so that one is obliged to use either metaphor or periphrasis; even with those which have names, the great poverty drives us back to the same (names) again and again.’6

Thus one can either use a metaphor (transferre) or periphrasis (circumire) and describe the same object in other words. Though both substitute the proper name through a new one, only metaphor is an associated word, whilst periphrasis (literally ‘to go around’) is a descriptive construction. The necessity and thus centrality of metaphor for the development of language is important here: metaphors are not listed as is the case in Cicero alongside unusual words (inusitata verba) and neologisms (novata verba),7 but metaphor is described as motivating new word meanings altogether. As many objects have no designation (res plurimae carent appellationibus), the metaphor has to be used (transferre necesse est), says Quintilian, and thus the same name can be reused for various objects. A slight contradiction within Quintilian’s own terminology might be noted here. In book 8 chapter 6, the main treatment of metaphor in Quintilian’s work, he draws a distinction between catachresis and metaphor based on exactly this criterion. discernendumque est hoc totum tralationis istud genus, quod abusio est ubi nomen defuit, tralatio ubi aliud fuit. (QVINT. inst. 8, 6, 35) ‘This whole category of metaphor should be distinguished from that, because catachresis appears when no other designation previously existed, whilst metaphor where there was a different designation.’

Both metaphor and catachresis are used out of necessity, but catachresis is used where there was no other designation, whilst metaphor is used where there was

5 Ullmann (1962: 202). 6 All translations are mine own. 7 CIC. de orat. 3, 152. On Cicero’s analysis of metaphor, see Novokhatko (2016: 400–403).

The typology of linguistic metaphor in first-century CE Roman thought

387

previously another designation. Thus in the latter passage (book 8) the metaphor is said to be used substituting a previous designation whilst in the former passage (book 12) the same metaphor is said to be used for an ‘empty’ object without designation. It is noteworthy that a meaning-bearing function of metaphor remains important in both cases: metaphor provides a new meaning to an object, independent of its previous name or, indeed, whether there was a name. My solution to the contradiction in Quintilian is that although Quintilian insists on distinguishing between metaphor and catachresis in detail in book 8, he is aware of the linguistic and cognitive force of metaphor in creating new meanings. Whenever he discusses this phenomenon he speaks of metaphor in general, and catachresis should thus be categorized for him as a type of metaphor. Thus in the same book 8 he says: metaphora enim aut vacantem locum occupare debet aut, si in alienum venit, plus valere eo quod expellit. (QVINT. inst. 8, 6, 18) ‘Metaphor should either occupy an empty place/room, or, if it enters the others’ place/ room, it should have a more effective meaning than what it banishes.’

The conceptual consistent of metaphor is accentuated by the all-embracing noun locus (cf. Greek τόπος), which can mean sedes argumenti or ‘topic, subject, concept’.8 If locus is understood not literally as ‘room’ but as concept, then metaphor

8 On such use and definition of locus/loci, see CIC. de orat. 2, 291 cum et argumenta causae et eos locos, quibus animi iudicum conciliantur, et illos, quibus permoventur, vidi atque cognovi (‘when I saw and ascertained the arguments belonging to the case and also those topics which win the mood of the judges and those which arouse their emotions’); top. 7 cum pervestigare argumentum aliquod volumus, locos nosse debemus; sic enim appellatae ab Aristotele sunt eae quasi sedes, e quibus argumenta promuntur. Itaque licet definire locum esse argumenti sedem (‘if we want to track some argument, we have to know places/topics; for thus were called by Aristotle those kind of places from which arguments are drawn. Therefore we may define locus as the place of an argument’); QVINT. inst. 5, 10, 20 excutiamus nunc argumentorum locos, quamquam quibusdam hi quoque de quibus supra dixi uidentur. Locos appello non, ut uulgo nunc intelleguntur, in luxuriem et adulterium et similia, sed sedes argumentorum, in quibus latent, ex quibus sunt petenda (‘now let us analyse the places of arguments. Although some people think that they are the same with those which I have discussed above, I call loci not how they are now commonly understood, against luxury, and adultery and such like, but the places of arguments, in which they lie hidden and from where they are supposed to be drawn’); QVINT. inst. 12, 8, 13 multa etiam quae litigator nihil ad causam pertinere crediderit patronus eruet, modo per omnis quos tradimus argumentorum locos eat (‘there are also many points which a client would hold as irrelevant for his case, and the advocate will unearth, only should he go through all the places of arguments which I have described’).

388

Anna Novokhatko

either occupies an empty concept (vacantem locum occupare) or enters the concept of others (in alienum venit). Quintilian contrasts two conceptual domains of metaphor, in modern terminology ‘source’ and ‘target’. Whilst target domain provides more effective meaning (plus valere) than source domain, Quintilian uses the Greek term metaphora (not the much more common Latin translatio!), and thus it is metaphor that is used for the creation of new meanings. Further, when one should distinguish between two kinds of linguistic metaphor, the creation of new names for unnamed objects and the replacing of a previous name with a new one for the same object, Quintilian is ready to accept the terminology of catachresis for the first category (creating a new name) and metaphor purely for the second one (replacing a previous name). Metaphor thus has a crucial role for Quintilian’s understanding of language development.

3 Semantic metaphor theory and re-working word meanings Another very important aspect of Quintilian’s understanding of the deeply cognitive value of metaphor should also be emphasized. Discussing the issue of designation and meaning at the very beginning of his work, Quintilian considers proper words (propria verba) and metaphors (tralata) as an introductory constituent part of linguistic preliminaries. propria sunt verba cum id significant in quod primo denominata sunt, tralata cum alium natura intellectum, alium loco praebent. (QVINT. inst. 1, 5, 71) ‘Words are proper words when they signify that which they were first designed to name; metaphorical when they reveal one meaning as understood by nature and another at the place/context.’

Whilst a proper name shares the nature of the thing it originally designates, signifying the thing that was first designed to name (id significant in quod primo denominata sunt), a metaphor keeps both layers (alium – alium), the original ‘natural’ one (natura intellectum) as well as the transferred one (loco). In book 8 chapter 2 he comes closer to this crucial question of the twofold semantic layers of metaphor. tralatio quoque, in qua vel maximus est orationis ornatus, verba non suis rebus accommodat. quare proprietas non ad nomen sed ad vim significandi refertur, nec auditu sed intellectu perpendenda est. (QVINT. inst. 8, 2, 6)

The typology of linguistic metaphor in first-century CE Roman thought

389

‘Metaphor too, in which we find the greatest ornament of speech, does not fit words to their own things. Propriety therefore refers not to the word, but to its semantic value, and is to be judged not by hearing but by understanding.’

Metaphorically employed words do not fit the objects they belong to, they have one meaning understood by nature alium natura intellectum, the other meaning understood at the place or in context, alium loco. The criterion of propriety crucial for metaphor should be evaluated (perpendenda est) not by referring to the designation (ad nomen) itself, but to the value/meaning of the word (ad vim significandi).9 Only keeping in mind this twofold nature of metaphor can one perceive its new meaning. In hearing (auditu) one hears only one part, to understand the other part and the whole one needs comprehension/sense/intellect (intellectu). It is thus clear that the semantic and cognitive value of metaphor is pivotal to Quintilian’s analysis. He returns to this point not once or twice at random, but rather analyses the problem repeatedly, returning to this issue on a number of occasions in his work. That Quintilian is not inventing semantic theories himself, but rather reflects contemporary debates on the relationship of objects and their designations, debates that first developed in early Greek studies of language and proceeded through Plato, Aristotle, and the Stoics in Greek and Roman linguistic thought, can be proved through citation of an example from Seneca. In his Letter 9 Seneca quotes Epicurus and argues that in quoting, exact words were less important than meanings: vel si hoc modo tibi melius enuntiari videtur (id enim agendum est ut non verbis serviamus sed sensibus) . . . (SEN. epist. 9, 20) ‘Or if to put in in the following way seems better to you (for it should work that we use meanings and not words) . . .’

The Stoic opposition non verbis sed sensibus (σημαῖνον/expression for verbum versus σημαινόμενον/λεκτόν/meaning for sensus in Stoic terminology) recalls the classical sophistic statements from the period of the first linguistic exercises and experiments in analysing language, such as Socrates at the beginning of Plato’s Ion: ἅμα δὲ ἀναγκαῖον εἶναι ἔν τε ἄλλοις ποιηταῖς διατρίβειν πολλοῖς καὶ ἀγαθοῖς καὶ δὴ καὶ μάλιστα ἐν Ὁμήρῳ, τῷ ἀρίστῳ καὶ θειοτάτῳ τῶν ποιητῶν, καὶ τὴν τούτου διάνοιαν ἐκμανθάνειν, μὴ μόνον τὰ ἔπη. (PLATO Ion 530b)

9 On the spectrum of connotations for the noun vis as the translation of the Greek noun δύναμις, see Zanker (2016: 66–69).

390

Anna Novokhatko

‘and at the same time it is necessary for you to deal with many other good poets, and especially with Homer, the best and the most divine of the poets, and to examine his thought/meaning, not only his words.’

The opposition καὶ τὴν διάνοιαν, μὴ μόνον τὰ ἔπη which a rhapsode should examine in Homeric text, corresponds to Seneca’s non verbis serviamus sed sensibus and to Quintilian’s non ad nomen sed ad vim significandi, pointing to the crucial duality in language in general and in the theory of metaphor in particular, that of content and expression. Seneca considers this issue in other letters as well. Thus commenting on the Greek term dialecticē, he states that it deals with utterances and meanings. ῥητορικὴ verba curat et sensus et ordinem; διαλεκτικὴ in duas partes dividitur, in verba et significationes, id est in res quae dicuntur et vocabula quibus dicuntur. (SEN. epist. 89, 17) ‘Rhētoricē deals with words, meanings, and order; dialecticē is divided into two parts, utterances and meanings, that is into things which are said, and the words through/in/by which they are said.’

The terminology used here is significant as it reveals a self-confident semantic approach to language: Seneca distinguishes between verba, significationes, res and vocabula. This semantic approach is applied further down to metaphor specifically. cum adsuevit animus fastidire quae ex more sunt et illi pro sordidis solita sunt, etiam in oratione quod novum est quaerit et modo antiqua verba atque exoleta revocat ac profert, modo fingit †et ignota ac† deflectit, modo, id quod nuper increbruit, pro cultu habetur audax translatio ac frequens. (SEN. epist. 114, 10) ‘When the mind has got used to scorning the usual things, and regarding as vulgar what was customary, it searches for what is novel in speech; at one time it summons and displays ancient and old-fashioned words, at another it coins †unknown and† misshapes them, at another – something that has spread recently – a bold and frequently used metaphor is being employed.’

Within the framework of traditional treatments of metaphor, Seneca reflects on the nature of meaning and metaphor, and juxtaposes the creation of new meaning (fingit) to the re-use of known vocabulary (antiqua verba atque exoleta revocat). The characteristic of metaphor as audax (‘bold’) and frequens (‘frequently used’) underlines Seneca’s awareness of the traditional theoretical discourse on metaphor. Thus, according to Aristotle and his successor Theophrastus, metaphor should not be bold (θρασεῖα

The typology of linguistic metaphor in first-century CE Roman thought

391

and τὰ τολμηρά), but apologetic (αἰσχυνομένη).10 This characteristic of metaphor is reused in theoretical texts several times during the first century BCE, and becomes a kind of theoretical background when employed in Seneca’s letters.11 The same idea of metaphor, a sign of the συνήθειαconcept, actual language usage serving as a criterion in Hellenistic thought, is recalled by Seneca in a further passage. invenio tamen translationes verborum ut non temerarias ita quae periculum sui fecerint. (SEN. epist. 59, 6) ‘I discover metaphors, both those that are not over the top and those that have stood up to the test.’

The attribute non temeraria for translatio has a long tradition behind it.12 Thus the author of the Rhetorica ad Herennium (written around 84 BCE, the earliest Latin source for the theory of metaphor) used Hellenistic sources and argued that ‘metaphor is said to have to be apologetic in order to make the transit to a similar thing with good reason, and not to appear as though it had made a bound to the dissimilar in an unselective, over the top/rash, and impulsive manner’ (translationem pudentem dicunt esse oportere, ut cum ratione in consimilem rem transeat, ne sine dilectu temere et cupide videatur in dissimilem transcurrisse).13 This adverb temere (‘over the top, rash, recklessly’) here corresponds perfectly to the Senecan adjective temerarius from his translationes verborum non temerarias. The syntactic use of the word is strikingly similar as well: both authors employ the negative form of a precept explaining how metaphor should not be used: ne temere uideatur in the Rhetorica ad Herennium corresponds to non temerarias in Seneca. Whether the treatise Rhetorica ad Herennium was a direct source for Seneca or, alternatively, whether there were more texts with the same common source, all reusing the same terminology, remains unclear. Yet this may serve as a reminder that Seneca’s passage should be read in the broader diachronic context of Roman reflections on metaphor.

10 THEOPHR. frg. 689A–690 FHSG. Guidorizzi and Beta (2000: 70–73, 163–164), Novokhatko (2014: 416). 11 See Zanker (2016: 171). Cf. PHILODEM. Rh. 1, 174; DEMETR. eloc. 80; PS. LONG. 32, cf. ARIST. rhet. 1408b2 (not concerning metaphors) and in Latin pudens or verecunda (CIC. de orat. 3, 165; fam. 16, 17, 1 [Latin translation of Theophrastus’ term]); cf. RHET. Her. 4, 34, 45; QVINT. inst. 8, 3, 37; 8, 6, 11 et al. 12 Novokhatko (2016: 399–400, 404). 13 RHET. Her. 4, 34, 45.

392

Anna Novokhatko

4 The iconicity of metaphor, or image schemata and metaphorical mapping Seneca remains a significant source for another, this the third function of metaphor. Seneca’s approach to metaphors has been much commented upon on in the scholarship. In the last decades work has been completed on Seneca’s use of metaphor and on Seneca’s claims about his own use of metaphor.14 It has been argued that “Seneca elides almost entirely the aesthetic role of figural language to insist on its cognitive and communicative force, and indeed its ability to ‘bring home’ to us the lived meaning of an abstract idea”.15 However, though it is clear that for Seneca as philosopher and tragedian metaphor plays a central role, and Seneca himself was fully aware of this fact, all these studies focus on searching for metaphors in Seneca’s work, in other words, metaphor in Seneca. What I am interested in here, is Seneca on metaphor, in other words, Seneca’s views on metaphor in the context of Greek and Latin concepts on metaphor more generally. On a number of occasions he comes close to discussing the theory of metaphor, which may be an indication of the significance of this issue for contemporary theoreticians. In book 4 chapter 12 of his De Beneficiis Seneca discusses the meaning of the term beneficium: ‘dicitis’ inquit ‘beneficium creditum insolubile esse, creditum autem non est res per se expetenda.’ cum creditum dicimus, imagine et translatione utimur; sic enim et legem dicimus iusti iniustique regulam esse, et regula non est res per se expetenda. ad haec verba demonstrandae rei causa descendimus; cum dico creditum, intellegitur tamquam creditum. vis scire? adicio insolubile, cum creditum nullum non solvi aut possit aut debeat. (SEN. benef. 4, 12, 1) ‘“You say, he said, a favour is a loan that cannot be repaid, but a loan is not something desirable for its own sake.” When we say “loan” we are using a mental picture and a metaphor; for in the same way we say that a law is the rule of just and unjust, and a rule is not something desirable for its own sake. We resort to these words in order to make the subject clear; when I say “loan”, you should understand “like a loan”. Do you wish to know? I add “not repayable”, whereas there is no loan that either cannot or should not be repaid.’

Seneca here returns to the metaphor he dealt with in the same work (book 2 chapter 18 paragraph 5), using the concrete and structured vocabulary of finance (a source domain) whilst speaking about an abstract and unstructured

14 Armisen-Marchetti (1989: 23–26), Bartsch (2009: 188–194), Sjöblad (2015). 15 Bartsch (2009: 193).

The typology of linguistic metaphor in first-century CE Roman thought

393

subject of favour (a target domain). He considers the favour to be a loan, the simile comparing the favour with money. et quidem diligentius quaerendus beneficii quam pecuniae creditor . . . et plus solvendum est. (SEN. benef. 2, 18, 5) ‘Indeed, I must think more carefully when it comes to a favour, than when it comes to money . . . I have extra payments to make.’

beneficium creditum insolubile in SEN. benef. 4, 12, 1 is thus an excellent example of a conceptual mapping in terms of Lakoff and Johnson, one concept being understood in terms of another concept. What is important in 4, 12, 1, however, is that Seneca marks and comments upon this conceptual shift: cum creditum dicimus, imagine et translatione utimur (‘When we say “loan” we are using a mental picture and a metaphor’). The term imago, a multiple and ambiguous concept, should here mean something like ‘mental picture’.16 It is difficult to say exactly which Greek term is being translated here either by Seneca or by his Latin source. Various terms such as εἴδωλον, εἰκών, φάντασμα, φαντασία would work for ‘mental picture’ in this context.17 Metaphor and mental picture are juxtaposed and discussed together, here for the first time. If we recall iconicity as a concept for the comprehensive

16 Similarly, but without the close connection to metaphor, the word imago is attested in Quintilian, when he discusses varieties of enargeia. QVINT. inst. 8, 3, 63–64 est igitur unum genus, quo tota rerum imago quodam modo uerbis depingitur . . . plurimum in hoc genere sicut ceteris eminet Cicero: an quisquam tam procul a concipiendis imaginibus rerum abest ut non . . . non solum ipsos intueri uideatur et locum et habitum, sed quaedam etiam ex iis quae dicta non sunt sibi ipse adstruat? (‘There is one kind [of enargeia], in which the whole picture of things is somehow painted in words . . . The most outstanding in this kind, as in the others, is Cicero: could someone be so far from perceiving images/mental pictures of things as not to seem that he is seeing the people, the place and the dress and to add something which was not said for his own part?’). The analysis of the vocabulary of imagery of Seneca by Armisen-Marchetti (1989: 19–66) is comprehensive; Armisen-Marchetti admits that she fails, however, to explain the use of the term imago in the two passages discussed here (benef. 4, 12, 1 and epist. 59, 6), precisely because she tries to analyse them rhetorically (Armisen-Marchetti 1989: 27). In my view, only the cognitive approach to imago and its translation as ‘mental image/icon’ makes sense in the passages discussed. 17 The anonymous author of a treatise contemporary to Seneca On the sublime discusses the ‘production of images’ (εἰδωλοποιία) as a necessary part of text-creating (PS. LONG. de sublim. 15, 1). See Cassin (2014: 245–249, especially 248–249) on curiosity towards the notion of mental image in the early modern European thought.

394

Anna Novokhatko

theory of metaphor,18 the metaphor functions iconically in representing the similarity of two objects or situations. As Marcel Danesi put it out, “the particular content of a metaphor can be said to constitute an interpretation of reality in terms of mental icons that literally allows us to see what is being talked about”.19 The iconicity of metaphor is a creative process concerning the relationship of form and meaning in language. It is based on the facts of experience and implies similarities whose perception depends on cultural codes. The creditum insolubile is culturally determined, the Roman reader of Seneca’s work perceiving this ‘cognitive instrument’ perfectly.20 Seneca’s theoretical statement that whilst using a metaphor ‘loan’ we make use of a mental picture and a metaphor, is in fact very close to Lakoff’s ‘metaphorical language makes use of conceptual metaphors’.21 In all probability this is not a new or original thought of Seneca’s, rather Seneca functions here as a reflector of the intellectual debates of his time. The iconicity of metaphor (expressed in terms of semiotic theory) and metaphorical mappings (image-schema and frame structures, expressed in terms of cognitive theory) seem to have been discussed at a theoretical level in mid-first-century CE Rome. In his letter 59, discussed above, Seneca discusses this issue again, putting metaphors and mental pictures on one level and revealing his theoretical interest in metaphor. invenio tamen translationes verborum ut non temerarias ita quae periculum sui fecerint; invenio imagines, quibus si quis nos uti vetat et poetis illas solis iudicat esse concessas, neminem mihi videtur ex antiquis legisse, apud quos nondum captabatur plausibilis oratio: illi, qui simpliciter et demonstrandae rei causa eloquebantur, parabolis referti sunt, quas existimo necessarias, non ex eadem causa qua poetis, sed ut inbecillitatis nostrae adminicula sint, ut et dicentem et audientem in rem praesentem adducant. (SEN. epist. 59, 6) ‘I discover metaphors, both those that are not over the top and those that have stood up to the test. I discover mental pictures, – if anyone forbids us from employing them and believes that only poets are permitted them – then he seems to me unfamiliar with any classical authors whose language has not yet been governed by the need to please good opinion. They who used to speak simply and with the purpose of making the subject clear, are full of comparisons/parabolae, which I consider necessary, not for the same reason as with the poets, but in order that they may work as support given our weakness, in order to lead both the speaker and the hearer to the point at hand.’

18 On the iconicity of metaphor, see Nöth (1985); Nöth (1990: 131–133); Eco et al. (1994). On the iconic character of metaphor on the basis of Charles Sanders Peirce’s semiotics, see Henle (1958: 177–181) and Gumpel (1984). 19 Danesi (1995: 266). Italics of the author. 20 Nöth (1990: 133). 21 Lakoff (2008: 24).

The typology of linguistic metaphor in first-century CE Roman thought

395

The quas existimo necessarias at the end of the passage is significant as it recalls the embedment of metaphor into the language and, also a traditional significant function of metaphor, that of necessity. Simple language is full of necessary comparisons, says Seneca, conventional or ‘dead’ metaphor in contemporary conceptualist terminology, this being the result of a semantic shift in the development of a language.22 Seneca, in fact, explains and expands this function: comparisons are necessary because they support in weakness in order to lead both the speaker and the recipient in rem praesentem. The manifestation of metaphor, first emphasized by Aristotle as ‘setting the matter before one’s eyes’ (τὸ πρᾶγμα πρὸ ὀμμάτων ποιεῖν or its Latin variant rem ante oculos ponere), was central to both the Greek and Latin concept of metaphor, with its concern with the horizon of expectations of the audience developed recently further by Paul Ricœur.23 Whilst emphasizing that comparisons help clarify things to both the speaker and the hearer (et dicentem et audientem in rem praesentem adducant), Seneca reveals himself to be aware of the role of metaphor in the dynamic process of communication and perception and the theoretical context and vocabulary of the theory of metaphor.24 Further, as emphasized through the anaphoric invenio . . . invenio, both objects translationes and imagines provide a structural analogy, considered syntactically. In the same letter further down in paragraph 9, Seneca blurs two terms deliberately, self-reflexively using the word imago for metaphor: nos multa alligant, multa debilitant. diu in istis vitiis iacuimus, elui difficile est; non enim inquinati sumus sed infecti. ne ab alia imagine ad aliam transeamus, hoc quaeram quod saepe mecum dispicio, quid ita nos stultitia tam pertinaciter teneat? (SEN. epist. 59, 9) ‘Many things hinder us, many weaken us. Long have we laid inside these vices, it is difficult to be washed clean; for we are not stained, we are corrupted. In order not to flit from one mental picture to another, I shall ask a question which I often consider within myself, why does stupidity keep such a steady hold of us?’

The imagines Seneca has in mind are conceptual metaphors, and here Seneca is the first attested author who comes close to a ‘constitutive’ role for images and

22 See Traugott (1985) and Novokhatko (2017: 314–315) with further parallels and bibliography. 23 ARIST. rhet. 1405b12–1411b25; DEMETR. elocut. 81; RHET. Her. 4, 34, 45; CIC. de orat. 3, 160; QVINT. inst. 8, 6, 19. See Ricœur (1975: 49–50). 24 Cf. Seneca’s arguments on the power of visuality formulated in similiar terms with those on the power of metaphor: in rem praesentem venias oportet, primum quia homines amplius oculis quam auribus credunt, deinde quia longum iter est per praecepta, breve et efficax per exempla (SEN. epist. 6, 5).

396

Anna Novokhatko

imagination, something that Paul Ricœur insisted upon when he posed a question on the boundary between a semantic theory of metaphor and a psychological theory of imagination and feeling: “There is a structural analogy between the cognitive, the imaginative, and the emotional components of the complete metaphorical act and that the metaphorical process draws its concreteness and its completeness from this structural analogy and this complementary functioning”.25 Seneca seems to have been well aware of this precisely structural analogy between the cognitive and the imaginative, between translationes and imagines.

5 Conclusion The aim of this chapter was to consider the typology of linguistic metaphor in the first century CE. The most representative authors for an understanding of this typology are Seneca and Quintilian. Both linguistic manifestations of conceptual metaphors and cognitive and semiotic concepts are significant in Seneca and Quintilian. The passages which have recently been discussed elsewhere26 point to the following conclusions for views on metaphor up until the first century CE: metaphor is considered semiotically (QVINT. inst. 8, 6, 7 and 35); metaphor is a cognitive process (QVINT. inst. 8, 6, 18 and 19); metaphor is considered from a sociolinguistic perspective (QVINT. inst. 8, 6, 6); metaphor is context-oriented (QVINT. inst. 8, 3, 38 and 11, 2, 2). To those four a further two parameters can now be added. First, the already known cognitive perspective has been emphasized by Quintilian through the capacity of metaphor to create new meanings und thus to use existing knowledge for the generation of new knowledge (QVINT. inst. 12, 10, 33–34). Second, a totally new perspective is found in Seneca: juxtaposing metaphors and mental images, Seneca argues that metaphor, being a mapping from a source to a target, implies iconicity manifested as image-schemata (SEN. benef. 4, 12, 1 and epist. 59, 6, 9). The interaction of metaphor and iconicity is significant for Seneca in the characterization of language as body and as process. This typology drawn from Seneca and Quintilian should be viewed both as a significant contribution to and as a constituent part of the broader linguistic and philosophic debates of the time. Metaphor in the first century CE was concerned with questions of cognitive significance within a broader comprehensive model of human communication. The results raise further questions, in

25 Ricœur (1978: 141). 26 Novokhatko (2017).

The typology of linguistic metaphor in first-century CE Roman thought

397

particular with regard to metaphorical correspondence and the structure of linguistics mappings, as well as questions concerning broader Roman approaches to the creation of meaning. Such questions deserve, however, a separate paper.

Bibliography Armisen-Marchetti, M. (1989). Sapientiae Facies: Étude sue les images de Sénèque. Paris: Les Belles Letres. Bartsch, S. (2009). Senecan metaphor and Stoic self-instruction. In S. Bartsch & D. Wray (Eds.), Seneca and the Self (pp. 188–217). Cambridge: Cambridge University Press. Cassin, B. (Ed.) (2014). Dictionary of Untranslatables: A philosophical lexicon. Transl. by S. Rendall, C. Hubert, J. Mehlman, N. Stein & M. Syrotinski. Princeton: Princeton University Press. Danesi, M. (1995). The iconicity of metaphor. In: M. E. Landsberg (Ed.), Syntactic iconicity and linguistic freezes: The human dimension (pp. 265–283). Berlin/Boston: De Gruyter. Deignan, A. (2005). Metaphor and corpus linguistics. Amsterdam/Philadelphia: John Benjamins. Eco, U., Niklas, U. & Edeline, F. (1994). Metaphor. In T. A. Sebeok (Ed.), Encyclopedic Dictionary of Semiotics, vol. 1 (2nd edition, pp. 534–549). Berlin/New York: Mouton de Gruyter. Guidorizzi, G. & Beta, S. (2000). La metafora: Testi greci e latini, trad. and comm. Pisa: Edizioni ETS. Gumpel, L. (1984). Metaphor reexamined: A non-Aristotelian perspective. Bloomington: Indiana University Press. Henle, P. (1958). Metaphor. In P. Henle (Ed.), Language, thought, and culture (pp. 173–195). Ann Arbor: University of Michigan Press. Jürgensen, H. (1968). Der antike Metaphernbegriff (Dissertation). Kiel. Kövecses, Z. (2010). Metaphor: A practical introduction (2nd edition). Oxford: Oxford University Press. Lakoff, G. (2008). The neural theory of metaphor. In W. Gibbs (Ed.), The Cambridge handbook of metaphor and thought (pp. 17–38). Cambridge: Cambridge University Press. Nöth, W. (1985). Semiotic aspects of metaphor. In W. Paprotté & R. Dirven (Eds.), (pp. 1–16). Nöth, W. (1990). Handbook of Semiotics. Bloomington: Indiana University Press. Novokhatko, A. (2014). Ancient theories of metaphor. In G. K. Giannakis (Ed.), Encyclopedia of ancient Greek language and linguistics, vol. 2 (pp. 414–418). Leiden/Boston: Brill. Novokhatko, A. (2016). The use of the term ‘metaphor’ in Latin linguistic discourse before Quintilian. In P. Poccetti (Ed.), Latinitatis rationes. Descriptive and historical accounts for the Latin language (pp. 395–409). Berlin/Boston: De Gruyter. Novokhatko, A. (2017). The linguistic treatment of metaphor in Quintilian, Pallas, 103, 311–318. Paprotté, W. & Dirven, R. (Eds.) (1985). The ubiquity of metaphor: Metaphor in language and thought. Amsterdam/Philadelphia: John Benjamins. Prandi, M. (2004). The building blocks of meaning: Ideas for a philosophical grammar. Amsterdam/Philadelphia: John Benjamins.

398

Anna Novokhatko

Prandi, M. (2010). Typology of metaphors: Implications for translation, Mutatis Mutandis 3(2), 304–332. Ricœur, P. (1975). La métaphore vive, Paris: Seuil. Ricœur, P. (1978). The metaphorical process as cognition, imagination, and feeling. In S. Sacks (Ed.), On metaphor (pp. 141–157). Chicago/London: University of Chicago Press Journals. Sjöblad, A. (2015). Metaphorical coherence: Studies in Seneca’s Epistulae Morales. Lund: Lund University. Traugott, E. C. (1985). ‘Conventional’ and ‘dead’ metaphors revisited. In W. Paprotté & R. Dirven (Eds.), (pp. 17–56). Ullmann, S. (1962). Semantics: An introduction to the science of meaning. Oxford: Blackwell. Zanker, A. (2016). Greek and Latin expressions of meaning: The classical origins of a modern metaphor. Munich: Beck.

Sounds

Francesca Cotugno

The use of initial h- in the writing-tablets from Roman Britain Abstract: This chapter is focused on the treatment of h in initial position in non-literary texts written on tablet from Roman Britain. The analysis highlights the variation concerning the treatment of h-. We consider the cases of h- insertion in initial position in the Vindolanda corpus, which targets specific areas of the lexicon: everyday language (Tab.Vindol. 622, hostrea) and, more importantly, personal names (Tab.Vindol. 184, Huettius). In contrast, the other nonliterary corpora of Londinium-Bloomberg, Carlisle and curse tablets show a different outcome, as there are only cases of h- deletion in initial position, which follows a more widely attested non-standard Latin development, which is eventually seen in the formation of the Romance languages (Allen 1965: 53). Keywords: Latin linguistics, Roman Britain, historical sociolinguistics, writingtablets

1 Introduction and description of the corpora Roman Britain represents an interesting area of study due to its geographical, socio-historical and linguistic perspectives. The Roman presence in this area coincides with the first historical period recorded for Britain and creates a mixture of different cultures which interacted over a time span of about four centuries. For this analysis, the following non-literary sources are considered:

Acknowledgement: I would like to thank Alex Mullen for useful advice. Any mistakes are my own. This chapter was presented at the 19th International Colloquium on Latin Linguistics (ICLL) held in Munich on 24–27 April 2017 and is also part of the LatinNow project (latinnow. eu) whose principal investigator is Alex Mullen. LatinNow aims to situate the phenomena of Latinization, literacy, bi- and multilingualism within broader social developments. The project is hosted by the University of Nottingham and based at the Centre for the Study of Ancient Documents (CSAD), University of Oxford. LatinNow receives funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 715626). Francesca Cotugno, The University of Nottingham https://doi.org/10.1515/9783110647587-025

402

– – – –

Francesca Cotugno

Londinium-Bloomberg stylus tablets (50–80 CE) Carlisle ink-written wooden writing-tablets (79–105 CE) Vindolanda ink-written wooden writing-tablets (85–205 CE) Curse tablets incised on metal from diverse locations (175–400 CE)

The Londinium-Bloomberg tablets comprise the oldest corpus available and represent important evidence in reconstructing the first decades of the Roman conquest (Tomlin 2016). All the documents are written in the so-called ‘Old Roman Cursive’ script which was widely in use from the first century BCE to the third century CE. The bulk of these documents was written from 50 to 80 CE and they present a picture of the more formal, official, legal and economic aspects of life in Roman London, immediately before and after its destruction by Boudicca. However, not all of the 409 documents collected carry legible texts. Only 185 of them can currently be analysed. The authors of these documents are varied: according to the analysis led by Tomlin (2016: 51), the authorship – when traceable and only in a few cases – belongs to Vangiones, Nervi, Lingones and people coming from Noricum. Not only the provenance of the writers varies, but also their social status. There are merchants, brewers and slaves together with prefects and soldiers. The only edition available for this corpus is the one recently published by Tomlin in 2016. It should be noted that the Vindolanda corpus remains unparalleled in its complexity of text types, and number of tablets (772 writing-tablets against the 405 tablets from Londinium-Bloomberg). The Carlisle corpus, despite its meagre size (77 ink-written tablets), is a good source of comparison to the Vindolanda corpus as both these corpora are composed of documents written in auxiliary forts alongside the Stanegate road. The Carlisle corpus is a mixture of accounts and letters which are unfortunately preserved in a highly fragmentary state. At the moment, there is only one edition available, published in the journal Britannia (Tomlin 1998). Most of the tablets can be dated to between the late first and mid-second centuries CE. Even if the number of the items collected at Carlisle is relatively low, this corpus shows some unique evidence (especially Tab.Luguv. 1 and 16), focusing on which it is possible to investigate the nonliterary Latin used by the men of the ala Gallorum Sebosiana, garrisoned at Carlisle in the second century CE. Conversely, the Vindolanda auxiliary fort has a long and rich history whose legacy derives both from its writing-tablets and from extensive archaeological investigations that underscore its historical and linguistic importance. The Vindolanda writing-tablets offer an unparalleled source of evidence of garrison life at the northernmost border of the Empire and of linguistic variation among its writers. This military fort has a long archaeological history starting from the mid-

The use of initial h- in the writing-tablets from Roman Britain

403

70s CE, but the Vindolanda corpus is dated at a precise period because all the documents so far published are ascribable to the Batavian and Tungrian cohorts – together with their civilian associates – garrisoned at Vindolanda between the end of the first and third centuries CE. Specifically, the bulk of the Vindolanda corpus belongs to the period when the fort was occupied by the 9th cohors Batavorum (roughly from 95 to 103 CE). In this way, together with the Carlisle and, perhaps to an extent, the Londinium-Bloomberg corpora, it is a source of non-literary documents written on tablet from individuals from the provinces of Gallia Belgica and the Rhine frontier and does not (perhaps with one or two possible exceptions) represent the Latin spoken by the native population of Britain. The text types of the Vindolanda corpus are the most multifaceted among the corpora considered; as a matter of fact, this is the only corpus in which documents written by both men and women are available, and the topics and types of texts range so broadly that it is easy to reconstruct aspects of the daily life of the garrison. It is possible to recognize a few major categories: personal correspondence, official correspondence, writing exercises, miscellaneous and descripta. The category of miscellaneous includes the documents that cannot be subsumed under the other labels since it encompasses those documents whose real text type cannot be identified. In addition, the categories of personal and official correspondence can be subdivided in order to obtain finer-grained distinctions (Table 4). The Vindolanda writing-tablets have added a lot of information, not only according to the linguistic perspective but also regarding onomastics and the lifestyle of the Roman auxiliary. Indeed, many of the letters contain touching moments, such as greeting friends and ‘messmates’ (Tab.Vindol. 310, 346), New Year wishes (Tab.Vindol. 261), invitations to birthday parties (Tab.Vindol. 291), and regular requests to send more beer (Tab.Vindol. 628), socks and underwear because of the horrible weather (Tab.Vindol. 234, 346). This corpus has an editio princeps (Bowman and Thomas 1983), a re-edition and three further updates (Bowman and Thomas 1994, 2003; Bowman, Thomas and Tomlin 2010, 2011). Finally, we have to mention the 300 curse tablets found all over Britain, among which 200 are published legible documents. These finds are spread over 39 different sites and grouped into 10 different UK counties with a particularly high number from Uley and Bath (Aquae Sulis), bearing witness to differential Latinization in this Province. The chronological range in which we can place the curse tablets is between 175 and 400 CE. They provide an invaluable source of non-literary Latin, written by inhabitants of Roman Britain from various social environments, including the local population, the auxiliaries and the legionaries garrisoned in the forts, and people from other provinces (Adams 1992: 24).

404

Francesca Cotugno

Compared to the other corpora analysed, these curse tablets represent a completely different type of non-literary document, first because they belong to a later period, but above all because they seem to the product of the local population which merged its indigenous features with those of the Latin culture. These tablets provide an insight into the religion and the writing habits of the Romano-Britons. The language used is influenced by local cultural features, structured around an apparatus of magical terms and formulae originating in the Latin tradition, and echoing other linguistic domains such as legal, liturgical and commercial language. The curse tablets from Roman Britain are available in several publications (Tomlin 1988; Kropp 2008; and the annual publication of the journal Britannia).1

2 H- in Roman Britain This chapter will consider the use of initial h- as marker of two distinct phenomena which are not directly correlated but are both ascribable to the use of the aspirate in initial position. On the one hand, 26 cases of initial h- deletion have been collected among all the corpora considered. This phenomenon consists of dropping the aspirate sound and not replacing it in the written text. In general, there was already an incorrect use of the aspirate sound, since the aspirate started to be lost in sub-elite varieties at an early date, i.e. the third century BCE (Allen 1965: 53–54), giving rise to a period of transition in which hypercorrect forms were abundant, both in speech and writing, to compensate for this lack of pronunciation (Sturtevant 1947: 56; Leumann 1977: 144). Only 2 cases of h- deletion are attested at Londinium-Bloomberg (Tab.Lond. 37; 55), 16 cases at Carlisle, in just one document which is a nonformal account concerning the administration of the garrison (Tab.Luguv. 1), 6 cases at Vindolanda (Tab.Vindol. 649; 691) and 2 cases in the curse tablets (Ratcliffe-on-Soar and Hamble Estuary). Another use of the aspirate, the insertion of h- in unexpected positions from a Classical Latin point of view, was also attested. Depending on the word targeted by this phenomenon, this can be interpreted as a hypercorrect form, based on the phenomenon of h- deletion, as noted above, or it can highlight another possible use of the aspirate, i.e. in personal names which appear to be of non-Latin origin. The examples of this initial h- number only 10, but,

1 Roger Tomlin is currently working on a new publication of Uley’s curse tablets.

405

The use of initial h- in the writing-tablets from Roman Britain

interestingly, they appear in a localized area (Vindolanda) and target specific lexical items.2

2.1 H- deletion We first draw attention to the phenomenon of h- deletion (Table 1). In this table, the occurrences with initial h- in which the words were correctly spelled can be compared with the cases of h- deletion. [-H] indicates the phenomenon of h- deletion whereas [H] indicates the correctly spelled words. Table 1: Comparison between correctly spelled h- and h- deletion. Londinium-Bloomberg

Carlisle

Vindolanda

Curse tablets









[-H]

[H]

[-H]

[H]

[-H]

[H]

[H]

[H]

















%

%

%

%

%

%

%

%

For each corpus considered, there is not one single writer, since many different hands are recognizable (Cotugno 2015: 194–195). Moreover, in a group of texts such as the curse tablets, the documents are written by a large number of different authors, and the variety of hands and misspellings in a busy sanctuary such as Bath is a clue as to the potentially relatively limited use of the scribes (though it is possible that some texts may have been copied from templates: Mullen 2007: 42). For this reason, it is very helpful to consider the different cases one by one, as they can be signifiers of language variation which slip through the orthographic awareness of these writers (Adams 1995: 92). In the Londinium-Bloomberg corpus, there are only 2 cases of h- deletion. Tab.Lond. 37 is a personal letter whose topic is the visit of a certain Atigniomarus on 25 December (61–65 CE), whereas Tab.Lond. 55 is a fragmentary loan-note

2 Only the initial position has been considered because in the corpora analysed is not common in internal position. In fact, considering the Vindolanda writing-tablets, -h- occurs only in the word mihi where it is deleted only 14 times against 49 correctly written cases. It is also available in the forms chortis (Tab.Vindol. 127), in the abbreviation chor. (Tab.Vindol. 396) and in cohors (26 occurrences, often abbreviated coh., e.g. Tab.Vindol. 210, 245, 248). Besides these, there is also the form exibe in which it has been deleted (Tab.Vindol. 282).

406

Francesca Cotugno

addressed to Narcissus, the slave of Rogatus the Lingonian (60–75 CE). They were written approximately in the same period and in both documents the spelling that diverges from the classical norm is abere. In this corpus is correctly placed in initial position in only two other occurrences: hanc (Tab. Lond. 55) and hominem (Tab.Lond. 41). Moving to the northern frontier, the fort of Luguvalium offers 16 cases of initial h- deletion that can be compared to 5 cases in which the initial h- is correctly placed. However, in the specific case of the Carlisle corpus, all the hdeletion belong to one single tablet which is also an account written by the curator of the Ala Gallorum garrison and involves one single word, i.e. hordei. The document is Tab.Luguv. 1 and, since it is an account, it is a relatively nonformal document concerning the administration of the fort itself. Concerning h- deletion, in the Vindolanda writing-tablets there are 6 cases scattered among personal correspondence and miscellaneous texts (Table 2). Table 2: h- deletion in the Vindolanda corpus. N.

Tab. No.

Text type

Period

Word





Correspondence (Male)

– CE

abet





Correspondence (Male)

– CE

abent





Correspondence (Male)

– CE

abet





Miscellaneous

– CE

abebis





Miscellaneous

– CE

aud





Correspondence (Male)

– CE

abeas

It is noticeable that also in this corpus, 5 of the 6 cases involve the verb habeo. This was already noticed in the Londinium-Bloomberg documents. Taking into account the different text types in which h- deletion occur, it can be observed that this phenomenon, even if scarce, may also depend on the text type as 4 of them are in personal correspondence, whereas 2 occur in documents labelled as ‘miscellaneous’. The topic dealt with in Tab.Vindol. 707 suggests that it is probably a document pertaining to personal correspondence as well, but there is no certainty in this, because the text and writing support are too fragmentary. All the letters considered seem to belong to the non-formal end of the stylistic distribution of the documents: Tab.Vindol. 649 is probably a draft and also contains non-classical features such as a few cases of vowel syncope (e.g. singla, singlos). The other documents are unfortunately too fragmentary to offer more linguistic information.

The use of initial h- in the writing-tablets from Roman Britain

407

In the curse tablets, it was only possible to find two cases of h- deletion. To be specific, these cases are found in one tablet from the Hampshire foreshore of the Hamble Estuary (Hassall and Tomlin 1997: 455–456), and one from Red Hill, Ratcliffe-on-Soar, in Nottinghamshire (Mullen 2013: 266). Both tablets – like the vast majority of the Romano-British tablets – are related to a request for justice after a theft, but this is the only feature that these two tablets share. The words involved are ominem (Hamble) and ospitio (Ratcliffe-on-Soar). Both words are from everyday language, but h-deletion does not seem especially common in the curse tablets. Moreover, the word hominem is correctly spelled in Uley 78 (hominis) whereas hospitio is correctly spelled with the initial h- three times: hospitio (Tab.Sulis 99), hospitiolo (Uley 3 and 72). In the other non-literary corpora, these words are never the target of h-deletion: at Vindolanda, hospitium is correctly spelled three times (Tab. Vindol. 157; 633; 880), whereas homo is correct nine times in seven tablets (homo, Tab.Vindol. 155; 157; 311; 344; 876; homines, Tab.Vindol. 155; 157). Londinium-Bloomberg features only hominem, correctly spelled in Tab.Lond. 41. In conclusion, in the curse tablets the evidence is meagre but it seems that the targeted words are not related to the verb habeo – as in the other corpora – and the everyday words involved usually appear with correct spellings in the other corpora. To summarize, at Londinium-Bloomberg 2 of the 4 cases in which initial h- is supposed to occur show h- deletion. The same phenomenon is apparently more frequent at Carlisle with 16 cases out of a total of 21, corresponding to 68%, but this is entirely thanks to the evidence collected from Tab. Luguv. 1. At Vindolanda 6 (i.e. 8%) of the 76 words starting with h- were spelled without initial h-. The curse tablets show a low number of cases: only 2 (i.e. 5%) of the words in which initial h- is supposed to occur are actually written without initial h-.

2.2 H- insertion As mentioned above, only in the Vindolanda writing-tablets is it possible to notice the phenomenon of h- insertion, 10 cases in total, which can be explained in different ways, as we shall see (see Table 3). The data collected in Table 3 suggest that this phenomenon is not text-typedependent as it does not occur in a specific text type but has a scattered distribution. As a matter of fact, the phenomenon of h- insertion is prevalent in the ‘account’ documents, which are notes and lists written without heightened

408

Francesca Cotugno

Table 3: h- insertion in the Vindolanda corpus. N.

Tab. no.

Text type

Period

Word





Account

– CE

Huetti[us]





Account

– CE

Huete[ris]





Account

– CE

halicae





Correspondence (female)

– CE

haue





Account

– CE

halicae





Account

– CE

hálicae





Account

– CE

hálicae





Correspondence (male)

– CE

hostrea





Correspondence (male)

– CE

Hario





Military report

– CE

Huennius

concern for linguistic accuracy. The three items of personal correspondence featuring this phenomenon are Tab.Vindol. 292, 622, 670.4 This phenomenon of h-insertion might depend on multiple factors: on the one hand, some of these tokens can be interpreted as hypercorrect forms, which may possibly represent the pronunciation of the authors, or mistakes of scribes, and h-insertion is a well-documented response to the non-standard loss of initial h-. On the other hand, 4 of these occurrences are personal names whose origin is not necessarily Latin. We may not therefore be dealing with h- insertion as these names may originally have possessed an initial aspirate. Table 4 lists the occurrences of words with initial h- under examination. They have been arranged according to their type: words pertaining to everyday language, and personal names. In the following analysis, each type is

3 The reading Huetti[us] is available in the appendix of Bowman and Thomas (2003: 156). The original edition (Bowman and Thomas 1994) had a different reading. 4 Tab.Vindol. 670 is from a later period and was written in a variant form of Old Roman Cursive. It may also be interpreted as a transitional form leading to New Roman Cursive. It is the only Vindolanda document from this period (165–205 CE) written in such a style. The other two personal letters are both related to Flavius Cerialis of the 9th cohort of the Batavians: Tab. Vindol. 292 is the birthday invitation to Cerialis’ wife, Sulpicia Lepidina, from her friend Claudia Severa, whereas the other letter (Tab.Vindol. 670) is from Brocchus, Claudia’s husband and Flavius Cerialis’ friend and fellow prefect.

The use of initial h- in the writing-tablets from Roman Britain

409

Table 4: Occurrences of initial h- under examination. Everyday language

Non-classical form

Classical form

Total

halica



%



%



have



%



%



hostrea



%



%





Personal name



Non-classical form



Classical form

Total

Hario



%



%



Huete[ris]



%



%



Huennius



%



%



Huetti[us]



%



%









compared with the cases in which the same word was correctly written in the Vindolanda corpus. As seen in Table 4, most of the occurrences occur only once in the whole Vindolanda corpus. Moreover, the only token that also occurs in other corpora and in its alleged classical form is the name Huennius: it appears in Tab.Lond. 61 as the patronymic Vannii. Six of the 10 cases involving initial h- are everyday words, and the remaining 4 are personal names. In Tables 4 and 5, it is noticeable how the word alica ‘gruel’ occurs in 2 tablets – a total of 4 tokens – with h- insertion, whereas there is a third tablet in which the word seems correctly spelled. Tab.Vindol. 233 is a letter draft from Cerialis to Brocchus, prefect of Briga. Unfortunately, the reading alica is only conjectural as the text is very faded (Bowman and Thomas 1983). If we accept the reading alica, it can be surmised that at least one writer from the household of Cerialis correctly wrote alica. Conversely, there are at least two writers who wrote the same word with h- (Tab.Vindol. 193 and 586). Both of them are accounts: the first is a note from the household administration, the second is a summary account that first records dispensations – either sales or loans – to various named individuals, while in the second part it records the supplies received. The incorrect use of h- may be interpreted as a case of hypercorrection adopted by the writers of Tab.Vindol. 193 and 586. The cause of this phenomenon cannot be

410

Francesca Cotugno

Table 5: Distribution of the occurrences of the classical and non-classical words with h- insertion. Token

Translation

Occurrence

alica

‘gruel’

Tab.Vindol. : halicae Tab.Vindol. : halicae, hálicae, hálicae Tab.Vindol. : alicas

ave

‘hail’

Tab.Vindol. : have

ostreum

‘oysters’

Tab.Vindol. : hos[t]rea Tab.Vindol. : ostria

Arius

Personal name

Tab.Vindol. : Hario

Vannius

Personal name

Tab.Vindol. : Huennius

Veteris

Personal name

Tab.Vindol. : Huete[ris]

Vettius

Personal name

Tab.Vindol. : Huetti[us] Tab.Vindol. : Vettius

explained straightforwardly, but one interpretation is that they added the aspirate as it mirrored the word as it was heard/spoken by the writer.5 Similarly, Tab.Vindol. 291 is the only document in which there is the occurrence of the word have in the whole Vindolanda corpus. Specifically, this letter was written by Claudia Severa, wife of Brocchus, to invite Sulpicia Lepidina to her birthday party. Personal correspondence was often copied down by professional scribes, and the closing formulae were usually added by the sender (Halla-Aho 2009: 61). The word have belongs to the part added by Claudia Severa herself and may be interpreted as a hypercorrection.6 hos[t]rea (Tab.Vindol. 622) occurs in a letter written by Brocchus to Cerialis. In Tab.Vindol. 299, however, there is the variant form ostria, written without h-. All these occurrences may be interpreted as possible cases of hypercorrection. The problem is understanding which was the standard form used as a referent, since for these cases it is possible to refer to examples from literary texts (e.g. SEN. epist. 122, 16; SCRIB. LARG. 104 for halica; CAEL. Cic. fam. 8, 16, 4; MART. 5,

5 An example in this sense is in Tab.Vindol 225. Here there is an et hiem erased and replaced by etiam. This could be a dictation error but also it possibly shows that scribes were conditioned to look out of loss of h- and to reinsert it, sometimes getting it wrong. 6 Even if forms like have are attested in classical texts, they will be considered because they can give us information about the possible collocation of the writers across the social spectrum (Adams 2013: 126).

The use of initial h- in the writing-tablets from Roman Britain

411

51, 7 for have, also hostrea was written with initial h- in a later period, see TLL vol. IX, 2, p. 1159, 2–1160, 47). We now turn to a consideration of the personal names. Here there is no absolute certainty that the evidence collected are names of Latin origin. Onomastics can be tricky in relation to identifying the writers’ origins and for this reason it must be supported with historical-archaeological and other linguistic data. Considering how the names are rendered in written form can be a clue to the writer’s background, because there is not necessarily a classical norm governing them. The names in the Vindolanda corpus generally follow Latin phonology and morphology. Nonetheless, non-Latin features sometimes intrude. This is perhaps due to a gap in the classical guidelines, or appreciation of them, that means writers are not properly guided towards Latin renditions. In relation to personal names, an overview of the attested forms is therefore offered in the following paragraphs. – Hario (Tab.Vindol. 670) was interpreted by Bowman and Thomas as the dative form of Arius (Bowman and Thomas 2003). If we accept this, Hario can be interpreted as the hypercorrect form of Ar(r)ius.7 Unfortunately, the text is fragmentary before and after this name, and it is not possible to state whether the name is the dative of Harius or the nominative of Ario, a different personal name. The name Ario can be recognized as both Celtic and Germanic in origin. On one hand, the form Ario has been reconstructed as a Celtic form by David Stifter, director of the Nichtmediterrane Namen in Noricum project (NNN). On their website,8 the proposed etymology for Ario is *ario- ‘noble/lord, freeman’. On the other hand, the argument for its Germanic origin, with possible initial h-, is based on Walde and Pokorny’s (1927–1932: I 353, 462) and Lehman’s (1962: 178) reconstruction: PIE *kor-yo-s, Got. harjis, ON herr, OE here ‘army’, OHG as. heri ‘army, multitude’. Whatever its origin, whether Germanic or Celtic, this name was widely used without initial h-. However, there are no attestations of this name in the other British corpora, whether Harius or Ario. We are also aware that the Latin sources are insufficient to give a satisfactory explanation of this name. – Vettius/Huettius is found twice only in the Vindolanda corpus. The first form, Vettius (Tab.Vindol. 214), refers to the aquilifer of the Legio II Augusta. This tablet is a letter sent to the fort of Vindolanda and its writer is not from the same community as the 9th Batavian cohort. Conversely,

7 The name Harius is not attested with h- in the Thesaurus Linguae Latinae as the Onomasticon of the TLL only goes as far as the letter D. However, it is possible to make reference to the gentilicium Arrius commonly written as Arius (TLL vol. II, p. 643, 76–644, 31). 8 www.univie.ac.at/austria-celtica/personalnames/details.php?id=100.

412

Francesca Cotugno

the form Huettius from Tab.Vindol. 184 has a different tone as it is from an account recording the expenses of the different centuriae garrisoned at Vindolanda. It may be that these two forms represent two different levels of Latinization of a same name: Vettius, which adheres entirely to Latin standards, and Huettius which shows a non-Latin aspiration before [w].9 – Veteris/Hueteris is a linguistic conundrum. On the one hand, it can be interpreted as a Latin name, whereas, on the other, it can be considered a Latinized form of a Germanic form related to Old Norse hvítr ‘white’.10 The instantiations in Old Norse, Old English and Old High German are based on a full grade form *hwīta, but those in Old Frisian and Dutch (wit) – connected with the Batavians and the Tungrians – have a short vowel as they are based on the zero-grade variant *hwitta (Kroonen 2003: 267). The fact that [i] was ĭ may explain the change towards [e] in forms like Veteris recorded by writers from Gallia Belgica, homeland of a large number of the Vindolanda writers and eventually also of the Frisians and Dutch. In this sense, Hueteris can be interpreted as a Germanic form with an initial aspirated element reflecting Germanic pronunciation. Forms with initial h- have also been found in areas near Vindolanda, at the minor forts of Brocolitia (Carrawburgh, RIB 1549, Huiteribus) and Vercovicium (Housesteads, RIB 1602 Hueteri). At both forts, the presence of Germanic auxiliaries has been recorded (RIB 882, 1576, 1580, 1586, 1594, 1618, 1619). – Huennius (Tab.Vindol. 862), is found in an account with reference to a faber. It has been linked to the name Vannius which is considered to be of Germanic origin (Reichert 1987: 766), and attested in the genitive form in a loan-note from the Londinium writing-tablets where it appears as a patronymic. The form Vannius is probably a Latinized form of a Germanic name. Tacitus passes on details about the king of the Germanic tribe Quadi, Vannius, who came to power following the defeat of the Marcomannic king Catualda (ann. 2, 63, 6). Due to the possible interpretations of this name,11 it is difficult to pinpoint the meaning for Vannius, and therefore its possible 9 There is a cognomen of Latin origin Vettius (CIL VI 3353, AE 1978, 564). In OPEL IV: 164 it is attested as a gentilicium elsewhere in the Roman Empire, especially in Italy. However, also in the Vindolanda writing-tablets Vettius from Tab.Vindol. 214 is a gentilicium. 10 From the PIE *hwuīta - *hwuitta (adj.) ‘white’ and therefore ON hvìtr, OE hwīt, OHG wīz, hwīz, but OFri. wit Du. wit. 11 Ernst Förstemann (1856: 1257) lists the name Vannius under the name element ‘Vand’ that does not have a clear meaning but is believed to be etymologically related to Slavic and Germanic tribes, the Wends and Vandals. As already supposed by Förstemann himself, the meaning of that element is not wholly certain. He thinks that it is etymologically related to the names of two tribes, namely the Wends (Slavic) and the Vandals (Germanic). Nonetheless, it is

The use of initial h- in the writing-tablets from Roman Britain

413

Germanic correspondent. However, the attestation of the personal name Huennius, in the light of the Germanic umlaut of palatal vowels before geminates and nasal,12 can be seen as an alternative form of the same name: on the one hand, Vannius, attested in classical text, such as the Annales of Tacitus, is the Latinized and literary form, on the other hand, Huennius may be a more speech-related form. The use of h- in personal names throws some light on its use in words which were not clearly regulated by classical norms. It seems that the writer used initial h- in order to maintain a feature of the original spoken form that, in the specific cases considered here, seems to be Germanic.

3 H- outside Britain The texts from the garrisons of Vindolanda and Carlisle and those from London represent different realizations of sub-varieties whose roots are linked to Continental varieties of Latin. It is not possible to identify a clear-cut variety of British Latin, as it is composed of many different sub-varieties from the Continental provinces. For this reason, a comparison with the continental nonliterary documents is fundamental. If the h- deletion seems to be a very common feature in the Roman Empire at large, e.g. in words like habeo, this is not the same for the phenomenon of hinsertion. According to Clackson and Horrocks, the dropping of h- was also frequent in Pompeian graffiti for the word habere but h-insertion was rare (Clackson and Horrocks 2007: 241). Moreover, taking into consideration other non-literary documents written on tablet, the focus on the verb habeo as a cross-province feature is also confirmed by the letters of Claudius Terentianus: h- deletion here occurs four times (P. Mich. 8 468 abes, 8 470 abet, 8 471 abuit, abere). Nonetheless, there are no cases of h-insertion comparable to those found at Vindolanda. For Vindolanda, the areas of reference are the Celto-Germanic provinces of Germania Inferior and Gallia Belgica, whereas for Londinium-Bloomberg it is possible to add, at least for a few of the tablets the Noricum and Gallic provinces, and the same goes for the writers at Carlisle. For this reason, the analysis can be extended to the alleged homelands of these writers, analysing the inscriptions

possible to relate this element to the name ‘Van’, also related to the Gothic vêns ‘resources’, and the Old High German wân ‘hope’ as well as wan ‘deficient, imperfect’ and wâni ‘need’. 12 See Saibene and Buzzoni (2006: 127–130).

414

Francesca Cotugno

from CIL XIII. The names of Germanic origins show the same use of h- in names for which a corresponding standard does not exist in Classical Latin or has been reworked so extensively that it is no longer recognizable in the writers’ orthographic consciousness. In a certain sense the situation from Vindolanda mirrors that of these provinces. For the sake of clarity, a few examples will be quoted: in CIL XIII 7994, 8611, 8661, 8830, the name of the Germanic goddess Hludana and the name Hrannon are attested (Battaglia 2007: 193; AE 1988 n. 896). These, together with the alternation between Arii and Charinii in Tacitus and Pliny, indicate that strong aspiration occurred at the beginning of these Germanic names. For the writers from Vindolanda, where language contact was common, the Old Roman Cursive script was a unifying medium for written communication and reciprocal understanding. Nevertheless, these people might have tried variable spelling strategies in order to represent different phonetic habits that were linked to their identities (Cotugno 2015). Since local scribes adopted the Latin orthographic system in Roman cursive and capital letters, the difficulty of representing German sounds could easily cause spelling inconsistencies (König and Van der Auwera 1994: 90). With regard to h- in the Celto-Germanic linguistic environment, as highlighted by Battaglia (2007: 203), it is highly possible that such inconsistencies may be a precocious attempt to adapt the graphemic inventory to the phonological inventory of the writer, flavoured by a non-Latin language. Concerning onomastics, the attestations found in Roman Britain and the Continental provinces show that there was not a clear norm to guide the writers towards the creation of well-formed words from a Latin perspective and because of this, there can be interference of non-Latin features.

4 Conclusions In this chapter, it was possible to provide an analysis of the two different types of use of initial h- in non-literary documents written on tablet from Roman Britain. The overall frequency of these occurrences in respect of the whole number of words for each corpus is not quantitatively relevant, nonetheless, each corpus is able to provide relevant information about the linguistic situation from this province. Concerning h- deletion, available in all the corpora considered, there are 26 cases in total (see Section 2.1). For Londinium-Bloomberg, this phenomenon targets the verb habeo, similarly to what happens in Vindolanda and other Roman provinces. It can be said that this habeo h- deletion is a pan-Roman feature, as it occurs not only in Gallia Belgica and Germania Inferior, but also

The use of initial h- in the writing-tablets from Roman Britain

415

in the Latin letters of Claudius Terentianus and in Pompeian graffiti (Adams 1978; Adams 2013: 125). In contrast, at Carlisle, h- deletion seems an idiolectal phenomenon as it is related to only one tablet and one writer (Tab.Luguv. 1). Nonetheless, the overall number of occurrences of words starting with h- is very low and this does not allow a definitive diagnosis. It should be noted that even if hordei occurs in non-standard forms 15 times, the form habeo (differently conjugated) is correctly spelled 3 times. It has been seen that the verb habeo is the most common word targeted by h- deletion in the corpora under consideration. However, this did not happen in Tab.Luguv. 13, 33, 63. This may be a further confirmation that the h- deletion is a feature characteristic of the curator author of Tab. Luguv. 1. Concerning the curse tablets, the cases with initial h- deleted do not seem to be related to any of the cases found in the other corpora. This may be related to the different provenance of the writers, but the evidence is too meagre. Conversely, the collected data from Vindolanda are in the same vein as the results from Londinium-Bloomberg and CIL XIII, as the large majority of the cases involve the verb habeo. The initial aspirate is a sociolinguistically marked feature. This is particularly evident in the different form of the verb habeo written without h-. As already highlighted by Allen (1965: 53–54) and more recently by Clackson and Horrocks (2007: 240–241) the aspiration was a sociolinguistic marker. On the one hand, those with a higher level of education maintained the aspirate in the correct contexts as witnessed by St. Augustine (AVG. conf. 1, 18, 29); on the other, the presence of the aspirate was perceived as a feature of prestige, and thus was subject to hypercorrection by those less well educated (CATVLL. 84). This hypercorrection may explain the form of several of the lexemes discussed above. However, some of the words with initial h- may not be a result of the same kind of process of hypercorrection (i.e. adding an aspirate where it does not etymologically exist). Rather, the initial h- in personal names may reflect a second language – namely the writer’s mother tongue – which is influencing his spoken and written Latin, as the /h/ may have an etymologically sound phonetic value in the language of origin. Vindolanda is the only corpus from Roman Britain showing all three phenomena, h-deletion, hypercorrect h- insertion and contactlanguage initial h-. Maybe this could be linked to the presence of Batavian and Tungrian writers, who are not present in the other corpora. Moreover, the variety of text types of this corpus showed that the occurrence of these phenomena is not dependent on the text type as it occurs in personal correspondence, accounts and military reports. Nonetheless, a shared feature is the non-formal style, which can be linked to a lower concern with accuracy. The expedient of initial h- for representing an initial aspirated sound which was not typical of Latin is not

416

Francesca Cotugno

uncommon and shows alternate forms, like Arii and Charinii and occurs elsewhere in a Celto-Germanic environment (see the forms Hludana and Hrannon found CIL XIII, see Section 3). In this normative void, the spoken language presumably makes an appearance through the use of the aspiration, which was no longer commonly used in Latin, according to the contemporary testimony. On the one hand there is the feature of h- insertion which may represent a case of hypercorrection in Latin words which never had it and is not attested only in Roman Britain but in Latin language at large (Adams 2013). On the other hand, there is another phenomenon at Vindolanda, initial h- which involves mainly Latinized personal names, which is not directly linked to h- deletion since it may pertain to the Celtic or Germanic pronunciation of these names. In conclusion, the corpora considered show different treatment of h- according to the alleged provenance of the writers themselves. In this way, it is possible to confirm that, in a certain sense, the corpora are ascribable to people from Gallia Belgica and Germania Inferior as they mirror the epigraphic behaviour of their homeland. Only the curse tablets have a different outcome as hospitio(lo) and homo occur only once in the non-classical cases of h- deletion. If the overuse of h- in initial position was intended as a sociolinguistic marker in Catullus’ epigram 84, in which the rusticus Arius uses hinsidias for insidias, it also possible to emphasize that this feature does not pertain only to the diastratic variation, but it may possibly mark the difference provenance of the writers. Even if in Latin the use of the aspirate was not maintained, in Celtic and Germanic languages it still had a place in the phonological inventory. If this were not the case, the writers from Vindolanda, Germania Inferior and Gallia Belgica would not have adopted h- to write Latinized forms of their personal names. For this reason, it is appropriate to see the phenomenon of initial h- in personal names (and perhaps also to an extent the hypercorrect forms, though this is much less certain) as an effect of the transfer from the non-Latin languages of the writers. Moreover, it is important to stress that the collected data regarding this specific phenomenon cannot be seen as a casual coincidence and must be interpreted, even though the data set is quite limited, especially when all the cases occur in a single corpus.

Abbreviations Du. Got. OE OFri.

Dutch Gothic Old English Old Frisian

The use of initial h- in the writing-tablets from Roman Britain

OHG ON PIE Tab.Lond. Tab.Luguv. Tab.Sul. Tab.Vindol.

417

Old High German Old Norse Proto-Indo-European Tabula Londinensis (see Section 1) Tabula Luguvaliensis (see Section 1) Tabula Sulis (see Section 1) Tabula Vindolandensis (see Section 1)

Bibliography AE = Année épigraphique. CIL = Corpus Inscriptionum Latinarum (1862–). Berlin/Boston: De Gruyter. NNN = Non-Mediterranean Names in Noricum, www.univie.ac.at/austria-celtica /personalnames. OPEL = Lőrincz B. & Redő, F. (1994). Onomasticon Provinciarum Europae Latinarum (vols. I–IV). RIB = Collingwood, R.G. & Wright, R. P. (1965), The Roman inscriptions of Britain. TLL = Thesaurus Linguae Latinae (1900–). Leipzig: Teubner; Berlin/Boston: De Gruyter. Adams, J. N. (1978). Two unexplained misspellings in Claudius Terentianus: Greek interference in Egyptian Latin? Zeitschrift für Papyrologie und Epigraphik, 31, 135–137. Adams, J. N. (1992). British Latin: the text, interpretation and language of the Bath curse tablets. Britannia, 23, 1–26. Adams, J. N. (1995). The language of the Vindolanda writing-tablets: an interim report. Journal of Roman Studies, 85, 86–134. Adams, J. N. (2013). Social variation in the Latin language. Cambridge: Cambridge University Press. Allen, W. S. (1965). Vox Latina: A guide to the pronunciation of Classical Latin. Cambridge: Cambridge University Press. Battaglia, M. (2007). Diis deabusque Germanorum. In E. Fazzini & E. Cianci (Eds.), I Germani e la scrittura: atti del XXXIII Convegno dell'Associazione italiana di filologia germanica, Pescara 7–9 giugno 2006 (pp. 187–208). Alessandria: Edizioni dell’Orso. Bowman, A. K. & Thomas, J. D. (1983). Vindolanda: The Latin writing-tablets (Britannia Monographs, 4). London: London Society for the Promotion of Roman Studies. Bowman, A. K. & Thomas, J. D. (1994). The Vindolanda writing-tablets (Tabulae Vindolandenses II). London: British Museum Press. Bowman, A. K. & Thomas, J. D. (2003). The Vindolanda writing-tablets (Tabulae Vindolandenses III). London: British Museum Press. Bowman, A. K., Thomas, J. D. & Tomlin, R. S. O. (2010). The Vindolanda writing-tablets (Tabulae Vindolandenses IV part 1). Britannia, 41, 187–224. Bowman, A. K., Thomas, J. D. & Tomlin, R. S. O. (2011). The Vindolanda writing-tablets (Tabulae Vindolandenses IV part 2). Britannia, 42, 113–144. Clackson, J. & Horrocks,G. (2007). The Blackwell history of the Latin language. Malden: WileyBlackwell. Cotugno, F. (2015). I longa in iato nel Corpus Vindolandense. Studi e Saggi Linguistici, 53(2), 189–206.

418

Francesca Cotugno

Förstemann, E. (1856). Altdeutsches Namenbuch. Personnamen. München: W. Fink. Halla-Aho, H. (2009), The non-literary Latin letters. A study of their syntax and pragmatics. Helsinki: Societas Scientiarum Fennica. Hassall, M. W. C. & Tomlin, R.S.O. (1997). Roman Britain in 1996. Britannia, 28, 455–457. König, E. & Van der Auwera, J. (1994). The Germanic Languages. London: Routledge. Kropp, A. (2008). Defixiones: Ein aktuelles Corpus lateinischer Fluchtafeln. Speyer: Kartoffeldruck-Verlag. Kroonen, G. (2003). Etymological Dictionary of Proto-Germanic. Leiden: Brill. Lehman, W. (1962). Historical Linguistics: An Introduction. London: Routledge. Leumann, M. (1977). Lateinische Laut- und Formenlehre. München: Beck. Mullen, A. (2007). Evidence for written Celtic from Roman Britain: a linguistic analysis of Tabellae Sulis 14 & 18. Studia Celtica, 41, 29–43. Mullen, A. (2013). New thoughts on British Latin. A curse tablet from Red Hill, Ratcliffe on Soar. Zeitschrift für Papyrologie und Epigraphik, 187, 266–272. Reichert, H. (1987). Thesaurus Palaeogermanicus. Lexikon der altgermanischen Namen. Teil 1. Texte. Wien: Akademie der Wissenschaften. Saibene, M. G. & Buzzoni, M. (2006). Manuale di linguistica germanica. Milano: Cisalpino. Sturtevant, E. (1947). An introduction to Linguistic Science. New Haven: Yale University Press. Tomlin, R. S. O. (1988). Tabellae Sulis: Roman inscribed tablets of tin & lead from the sacred spring of Bath. In B. Cunliffe (Ed.), The temple of Sulis Minerva at Bath, vol. 2, The finds from the Sacred Spring (pp. 59–269). Oxford: Oxford Univ. Comm. Archaeol. Monogr. 16. Tomlin, R. S. O. (1998). Roman manuscripts from Carlisle: the ink-written tablets. Britannia, 29, 34–84. Tomlin, R. S. O. (2016). Roman London’s first voices. Writing tablets from the Bloomberg excavation, 2010–2014. London: MOLA Museum. Walde, A. & Pokorny, J. (1927–1932). Vergleichendes Wörterbuch der indogermanischen Sprachen. Berlin/Boston: De Gruyter.

Éloïse Lemay

The merger of /ō/ and /ŭ/: a computational take on the inscriptions of Gaul (330–730 CE) Abstract: This chapter uses statistical and philological methods to trace the history of the merger of /ō/ and /ŭ/, as evidenced by spelling confusions between and in the inscriptions of late antique and early medieval Gaul. I map out decade by decade the onset and development and confusions, tying their distribution pattern to the S-Curve model of linguistic feature diffusion. I then contrast this distribution pattern to another S-Curve distribution, that of the and confusions that stem from the much earlier merger between /ĭ/ and /ē/. This second distribution showcases the flattening out that accompanies mature features that are reaching maximal diffusion. The two S-Curves converge upon the same prevalence, indicating that both features have been equally adopted by the language. I then present and analyze the forms that feature and , in order to isolate finer-grained distribution, proposing original conditioning environments and subsequent generalization. A possibly related process, that of /ŏ/ written as , which produces a handful of and confusions, is accounted for either as purely graphic misspellings arising secondarily to the merger, or as a distinct linguistic process. Keywords: Latin linguistics, Latin phonology, linguistic feature diffusion, late antique and early medieval epigraphy, computational linguistics

1 Introduction One of the major Late Latin perturbations to the Classical Latin vowel system is the merger of Classical Latin /ō/ and /ŭ/ to /o/ and of /ĭ/ and /ē/ to /e/, brought Acknowledgement: I would like to thank Brent Vine, Craig Melchert, Stephanie Jamison, Nigel Holmes, and the reviewers for their expertise and meticulousness. This work was conducted as part of my dissertation, prior to joining Amazon as a computational linguist. It has been since reviewed for publication. Éloïse Lemay, Amazon Alexa https://doi.org/10.1515/9783110647587-026

420

Éloïse Lemay

about by the Late Latin loss of contrastive vowel length. The general development is illustrated in Figure 1.1 CL

ī

LL

ḭ

ĭ

ē ȩ[e]

ĕ

ā

ȩ[ε]

ă a

ŏ ϙ[ͻ]

ō

ŭ

ū

ọ[ᴏ]

u

Figure 1: Late Latin vowel mergers.

It has been claimed that the /ō/ and /ŭ/ merger occurred significantly later than the merger of /ĭ/ and /ē/ (Väänänen 1981: 30, n. 42 and n. 43). However, the evidence presented so far for its onset and early development has proven to be too scant and ambiguous to be convincing, as the very early evidence affects forms that would be especially liable to etymological archaizing spellings or that may be accounted for as Oscan influence (Powell 2011: 115; Adams 1977: 9–11, Adams 1995: 91–99 and Adams 2007: 669–670). The one attempt (Herman 1971) to examine the merger’s onset and early history has been rebuffed (Adams 2007: 669–670), on account of the paucity of the evidence and of the methodological choice of the author to conflate four centuries of linguistic development. Prinz’s dissertation (1932) provides a general survey of the epigraphic evidence available at his time, but since then, much as been discovered, and, due to the general nature of his studies, it is not clear what precise epigraphic sources Prinz analyzed. My contribution is to trace the early history of the /ō/ and /ŭ/ merger through a statistical, comparative and philological study of 600 late antique and early medieval inscriptions from three regions of Gaul: (1) Trier, (2) Vienne, and (3) Clermont-Ferrand and a few smaller city centers of the Roman province of Aquitania Prima.2 The Trier epigraphic evidence covers the onset of the merger as well as its early development, all the way to maturity. It also provides numerous attestations

1 This figure has been adapted from Väänänen (1981: 30 n. 42). 2 I am basing myself upon the text of the editions of the Recueil des Inscriptions Chrétiennes de la Gaule (RICG); the relevant volumes are Gauthier (1975) for Trier, Descombes (1985) for Vienne, and Prévot (1997) for Aquitania Prima. Trier and Vienne are major city centers that yield a considerable number of inscriptions, to the order of two to three hundred for each city. However, the smaller Clermont-Ferrand yields far fewer inscriptions, leading the RIGC editor to join to it the few inscriptions of neighboring cities Bourges, Limoges, Cahors and Le Puy, as these were all part of the same administrative Roman province of Aquitania Prima.

The merger of /ō/ and /ŭ/

421

of the merger, most of which are unambiguous. The other regions do not offer as wide a chronological spread; for this reason, they are used as supplementary evidence. I tracked the confusions between and as a measurable, quantifiable manifestation of the merger. The first sections of this chapter deal with the statistical distribution of these spelling confusions, detailing the evidence region by region. The second half of this chapter deals with the particular forms that have been observed, although concentrating less on the minutiae of ambiguous forms, which are detailed in Lemay (2017), and more on motivating and detailing the merger.

2 Chronological distribution of o-u confusions The evidence for the merger between /ŭ/ and /ō/ consists of confusions in the spelling of Classical Latin and . This results in spellings such as tumolo or tomolo for Classical Latin tumulum, tolit for tulit, iouenim for iuuenem, annus for annos, mondo for mundo, and on the other hand matrune for matrone.

2.1 Trier: chronological distribution of o-u confusions Trier offers excellent chronological distribution: it covers 400 years uninterruptedly with a rather even distribution. In particular, it offers numerous data points for late antiquity, which is a crucial period in the history of the merger. For this reason, Trier is my starting point. I retrieved all the forms featuring spelling confusions between Classical Latin and and established their chronological distribution,3 shown in Figure 2. Figure 2 shows which of the Trier inscriptions contain spelling

3 The Trier inscriptions do not bear dates; instead, they have received dating approximations by their editors. Thus, each inscription on Figure 2 receives a horizontal line that represents its date range. Gauthier (1975: 95–104) provides a systematic chronology of the Trier inscriptions, which she crosschecked with the dissertation work of Krämer (1974). She uses an assortment of grounds to establish an inscription’s date range: epigraphic, archaeological, diplomatic, stylistic, linguistic and art historical. For the purposes of establishing the chronological distribution of the o-u confusions, I have excluded any inscriptions for which dating rests upon linguistic grounds or remains uncertain. However, I am considering these inscriptions’ forms in the second portion of this chapter. Thus, out of the 237 Trier inscriptions, I consider 125 in my chronology of the o-u confusions, but all of the forms are treated in my inventory.

422

Éloïse Lemay

120

100

Chronological Sequencing

80

60

40

20

0 300

400

500

600

700

800

Years

Figure 2: Trier: spelling confusions between and .

The merger of /ō/ and /ŭ/

423

confusions between and , marked in the lighter grey color. The inscriptions that include unaffected and are left in black. These forms are inventoried below under Section 3. The number to the right of each inscription’s date rage is that of the Gauthier edition. As can be seen at a glance: (1) the earliest inscriptions never feature o-u confusions, (2) the earliest affected inscriptions are from the turn of the fifth century (all seven bear the date range 390–440 CE). There are relatively few inscriptions that are affected until the sixth century or so, after which there is a very marked, sharp increase. Most later (sixth and seventh centuries) inscriptions are affected, displaying o-u confusions. This development can be quantified further.

2.1.1 Trier: statistical analysis of distribution of o-u confusions I computed the frequency of occurrence of and confusions decade by decade, in order to isolate the onset of the merger in the Trier inscriptions, and to track its progression. First, I determined which inscriptions have a date range that covers a particular decade. I then computed the frequency of occurrence of o-u confusions among these inscriptions. In other words, I determined for each decade the ratio of “affected inscriptions” (inscriptions displaying one or more instances of the merger) to the “total of inscriptions”, allowing me to trace the evolution of the merger decade by decade. I take into consideration how broad each date range is, as an inscription with a narrow date range is a much more accurate predictor of what happens in a certain decade, than one with a very broad date range. This is what I mean by “weighted frequency of occurrence”.4 In Figure 3, each grey diamond indicates the ratio by decade of

4 I computed the weighted frequency of occurrence using the inverse of the range of possible dates for a given inscription as its weight. In this way, the date range of each data point is factored in. The rationale for the use of weights is this: an inscription with a narrow date range is a much more reliable indicator of what happens at any point encompassed by its date range, than an inscription with a broad date range. Without the use of weights, the inscriptions with the broadest date ranges would get disproportionately factored in, as they would not only be considered for every year of their range (while inscriptions with narrower date ranges would be considered only for a much shorter time span), but also as the broader inscriptions would be held to be equally good predictors as the narrower inscriptions. Thanks to the use of weights, broader inscriptions are held to be less reliable predictors, but over a broader set of years, which narrower inscriptions are held to be more reliable predictors, but over a narrower set of years.

424

Éloïse Lemay

Weighted frequency of occurrence of confusions between u and o, by decade Affected/total no of inscriptions

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 300

350

400

450

500

550

600

650

700

Years Figure 3: Weighted frequency of o-u confusions by decade.

how many inscriptions feature the o-u confusions out of the total number of inscriptions. The curve is a polynomial regression that indicates the trend. As can be observed from Figure 3: (1) The earliest instances are dated 390–440 and these are followed by a steady steep increase in frequency. (2) In Trier, the onset of the attestations might have happened in the second half of the third century CE or around 350 CE, depending on the statistical model used to extrapolate from the data trend; this is discussed just below. (3) There is no watershed moment at the fall of Rome or at any other point; there is no sudden sharp increase (or “spike”). (4) This new feature was integrated over some 300 years. This is shown by the trend slowing down and stabilizing in the seventh century; the curve flattens out. Here, we observe something striking. The number of attestations of o-u confusions increases rapidly over time, rising from the extremely rare around 390 CE, to the ubiquitous around 600 CE. It then affects one inscription out of two. We are observing the effects from a new development in the Latin language. The curve presented in Figure 3 is in fact an example of the S-shaped curve associated with the spread of linguistic innovation.5 According to the S-curve model, linguistic change starts slowly, accelerates rapidly, and ends slowly, forming the shape of the letter S. Effectively, this means that the development of a given

5 This is a common model used to account for the diffusion of linguistic features. It is presented and detailed in Blythe and Croft (2012), Denison (2003), Labov (1994), Kroch (1989) and Ghanbarnejad et al. (2014). Recently, Nevalainens (2015) noted its limitations, and illustrated the need for it to be applied with flexibility.

The merger of /ō/ and /ŭ/

425

linguistic is akin to a logistic function. In the case of Figure 3, the S-curve is flatter than typical. However, as recently pointed out in the scholarship, the adoption curves of different features may in fact display more or less steep S-curves.6 The earliest portion of the S-curve, which precedes the acceleration, does not find attestations in my corpus; this can be expected as the frequency of occurrence at such an early period is extremely low, and as the datasets from Gaul are relatively small compared to the Early Modern and Modern English datasets usually used in order to illustrate the effects of the S-Curve. We can use the S-Curve model to extrapolate an onset for o-u confusions in Trier. Examining only the steepness of the curve of Figure 3, one may be tempted to extrapolate an onset of 350 CE for the earliest attestations of o-u confusions in Trier; this would presuppose a constant rate. However, according to the S-Curve model, the speed at which a linguistic feature diffuses varies, and its early development is drawn out, just like its tail end. Consequently, it is preferable to extrapolate a much earlier onset, perhaps in the second half of the third century.

2.1.2 Comparanda: statistical analysis of i-e confusions in Trier It is important to distinguish the S-curve presented in Figure 3 as showing the early stage of an innovation, versus the almost flat trajectory that is consistent with a feature’s maturity. One example of a linguistic innovation that has reached its maturity is the case of the /ĭ/ and /ē/ merger, evidenced by spelling confusions between and . The /ĭ/ and /ē/ merger predates the Trier data, as it is already well established in the Latin of the letters of Claudius Terentianus (Adams 1977: 7–8), which date from the early second century CE. Thus, we should not expect to see the onset of merger in the Trier data. Instead, Figure 4 indicates that this merger is commonly encountered even in the earliest inscriptions of Trier, such that we can say that it is a well-implanted feature of the language even in the fourth and fifth centuries. The development is a slow, progressive increase, similar to the flattened out trend line of the o-u merger in the seventh century, indicating that it is a mature change. Figure 4 can be read thus: each grey diamond indicates for each decade the ratio of the inscriptions that feature i-e confusions to the total number of inscriptions. The black curve is a polynomial regression that shows the trend.

6 Ghanbarnejad (2014) and Nevalainens (2015).

426

Éloïse Lemay

Weighted frequency of occurrence of confusions between i and e, by decade Affected/total no of inscriptions

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 300

350

400

450

500 Years

550

600

650

700

Figure 4: Weighted frequency of i-e confusions by decade.

Comparing Figures 3 and 4, we can observe that by 600, frequency of occurrence for both i-e and o-u confusions is the same. It is about 0.47, meaning that the inscriptions feature each of the confusions half of the time. This means that by then we are equally likely to encounter i-e confusions as o-u confusions in an inscription. We can conclude that by 600, the /ŭ/ and /ō/ merger is as wellintegrated a feature of at least Gaulish Latin as the /ĭ/ and /ē/ merger.

2.2 Aquitania Prima: chronological distribution This is a less promising dataset, as there are very few Aquitania Prima inscriptions that are earlier than 500 and thus could inform us about the merger’s onset and early history. Indeed, the earliest instances are seven cases whose dates range from 502 to 550 CE. As for its more mature development: Aquitania Prima yields a steady number of o-u confusions throughout the sixth century. This fits with the Trier findings but does not tell us anything new.

2.3 Vienne: chronological distribution This is a much more promising dataset as it covers extensively the period that corresponds to the merger’s onset. To investigate o-u confusions here I looked at two words, annus and tumulus, which provide the majority of instances.

The merger of /ō/ and /ŭ/

427

I established the chronological distribution of the forms of annus for annos (acc. pl.) and the forms in tom- and -mol- for tumulo (abl. sg.); there are 80 such forms all in all. This examination yields: (1) three securely dated (consular dates) fifth century inscriptions (153 dated 495; 259 dated 488; 287 dated 486), (2) three estimated “late fifth/early sixth century” forms (122, 182, 196), (3) eighteen instances all in all from the first half of the sixth century. Thus, the Vienne corpus offers a few late fifth century instances and numerous sixth century instances. That there are no later cases is not especially surprising, as the Vienne inscriptions date for the most part from the fifth and sixth centuries; there are few in fact that are later. The earliest Vienne cases (~485-500) postdate those of Trier (~390-440) by some decades, possibly as much as a century. This is surprising, as from such a large corpus as that of Vienne, I would have expected early fifth or even perhaps late fourth century cases. It could be that looking up all of the o-u confusions would yield the missing very early cases. However, this is somewhat unlikely, as annus and tumulus yield by far the majority of forms. It could be that we are dealing with a dating bias from the Trier editor. The Trier inscriptions receive only relative dates, approximated by the editor, while many of the Vienne inscriptions bear precise and accurate consular dates. It may be that the Trier dating estimates skew a little early, by a few decades, and that the inscriptions dated from the early fifth century should really be pushed to the later fifth century. However, since there is no other evidence that the Trier dating skews early, this explanation is not very satisfying. It appears then that for now the data must be taken at face value: the Trier cases are simply earlier than the Vienne cases by a few decades. This may be due to regional variation in the language, in education, or in writing practices.

2.4 Observations about the chronological distribution of o-u confusions The earliest securely dated instances of o-u confusions consist in three instances found in Vienne in the second half of the fifth century. The earliest approximately dated instances consist in seven instances found in Trier and dated from the tail end of the fourth century or first half of the fifth century, and three additional instances found in Vienne dated from the second half of the fifth century.

428

Éloïse Lemay

According to the S-Curve model of language diffusion coupled with the trend of the frequency of occurrence of o-u confusions, it is possible to predict an onset for these confusions. I proposed the second half of the third century for the merger’s onset.

3 Inventory of forms Here are the forms that feature o-u confusion in Trier, Aquitania Prima and Vienne. The editors of the inscriptions do provide lists of forms, but I found in them several mistakes and omissions, detailed in my dissertation (Lemay 2017: 41–44, 157–162, and 198–211). The older inventory by Prinz (1932) was based on the 19th-century editions; a lot of data was not yet available. Indeed, for Trier alone the RICG edition feature twice as many inscriptions.7 It was thus necessary for me to compile an updated and corrected list based on the most recent editions by the RICG. The o-u confusion affects forms that contained /ō/ or /ŭ/ in Classical Latin, as one would expect, but there are also a few cases affecting forms with /ŏ/ by extension, as hypercorrection or as symptoms of general spelling confusion, or as part of a separate linguistic process; see Section 3.3. For some of these forms, there are alternative explanations to the merger, such as etymologizing or archaizing spelling (e.g. adoliscens for adulescens, suos and tuos for suus and tuus), assimilation effects,8 or case usage confusion (e.g. ad domino, for ad dominum). However, the fact that we encounter such a large number of straightforward cases indicates, in my view, that the merger was the leading motivator behind o-u confusions, and as such, even when an alternative explanation is possible, the form should still be considered as a valid attestation to the merger, at least in part. For this reason, I consider these ambiguous forms in my inventory. A granular philological discussion of ambiguous forms is presented my dissertation.

7 Gauthier notes in her introduction that her edition of the Trier editions contains twice as many inscriptions as the older edition by Le Blant; for discussion of sources, see Lemay (2017: 11). Nonetheless, Prinz’ observation that -ol- instead of -ul- had become rare by the imperial period (Prinz 1932: 23) fits with my results; still, for some of the later cases, alternative explanations to general o/u confusion can be advanced (such as secondary spread of stem vocalism, vowel harmony, case leveling); see Lemay (2017: 41–42) for discussion. 8 Accounting for sequences such as o . . . u > o . . . o (e.g. monomento for monumento, uolontas for uoluntas), o . . . u > u . . . u (e.g. dulure for dolure), u . . . o > o . . . o (e.g. mondo for mundo).

The merger of /ō/ and /ŭ/

429

3.1 /ō/ written as The single most common case is annus for annos (10 instances in Trier, 17 in Aquitania Prima, 44 in Vienne). Trier: annus (25, 29A, 33, 62, 107, 119, 134, 135, 153, 180) for annos, matrune (138) for matrone, numine (147) for nomine, ustiarius (165) for ostiarius, amure (147) for amore. Aquitania Prima: annus (9, 16, 23, 24, 25, 27, 30, 32, 37, 38, 42, 45, 48, 52, 53, 54, 55, 57) for annos, annu (24) for anno (abl.), denus (23) for denos, octubres (50) for octobres. Vienne: amure (265, 269) for amore, anenus (264) for animos, annus (44 cases; 17, 19, 21, 33, 41, 42, 46, 53, 55, 62, 68, 90, 91, 93, 94, 98A, 100, 103, 104, 115, 153, 156, 157, 159, 162, 171, 173, 176, 182, 189, 196, 220, 223, 245, 246, 249, 254, 259, 263, 270, 284, 285, 287, 292) for annos, clarissimu (258) for clarissimo, dulure (265, 269) for dolure, duus (103, 268) for duos, emeritus (147) for emeritos, eurum (269) for eorum, ferus (264) for feros, huc (283) for hoc, ignuscendum (162) for ignoscendum, ignuscendo (106) for ignoscendo, indicciune (21) for indictione, liuertus (258) for libertos, mure (264, 265) for more, nuuelis (264) for nobilis, nuuilior (264) for nobelior, octubres (121) for octobres, patruciniis (98B) for patrocinium, prumtus (162) for promptus, subrius (229) for sobrius, sulum (112) for solum, uutis (98B) for uotis. Along with this, there are a handful of personal names: Flurinus (68) for Florinus, Petrunia (160) for Petronia, Populunia (237) for Populonia, Scupilio (258) for Scopilio.

3.2 /ŭ/ written as Most examples affect tumulus and titulus, there are also a few examples affecting the perfect of fero, famulus and famula, and adulescens. Trier: adoliscens (147) for adulescens, coniox (160, 184) for coniux, ad domino for ad dominum (134, 193), Fedola for Fedula (a woman’s name; 21), Francola for Francula (a woman’s name; 54), titolo (7), titolum (127, 160) or tetolum (25, 45, 50, 51, 69, 72, 76, 77, 84A, 86, 107, 135) for titulum, nomero (135) for numero, sous for suus (145), tomolo (191) for tumulo, Ursolus for Ursulus (a man’s name, 72). Aquitania Prima: tumolo (41, 51), [.]olo (44), tomolo (4, 7, 16, 23, 29, 36, 37, 38, 39, 46, 52, 53, 57), thomolo (48) and tomulo (55) for tumulo. There is also famola (9) for famula, to[lit] (16) and abstolit (23) for tulit and abstulit, iouenim (23) for iuuenem, and mondo (25) for mundo.

430

Éloïse Lemay

Vienne: adoliscens (25, 158, 278) for adulescens, console (261) for consule, consolato (91) for consulatum, famolus (5, 173, 239) for famulus, famola (45, 69, 237, 284) for famula, infantola (103) for infantula, iogalem (218) for iugalem, monomento (176) for monumento, occopat (264) for occupat, omneuos (267, 269) for omnibus, oxsur (119) for uxor, resorge (48) for resurget, secolo (17) or [. . .]col [. . .] (158, 220, 220) for saeculum, suos (289) for suus, tetolo (119) or [. . .]tol[. . .] (98B) for titulo, tuos (264) for tuus, uolontas (69, 269) for uoluntas. Numerous forms of tumulus are affected, featuring tom- (10, 12, 21, 66, 67, 68, 69, 100, 103, 130, 135, 158, 171, 208, 217, 218, 219, 235, 237, 245, 246, 249, 270, 285, 286), -mol(17, 19, 21, 30, 45, 66, 67, 68, 69, 73, 91, 96, 100, 122, 136, 153, 158, 160, 171, 173, 207, 217, 219, 229, 237, 239, 246, 249, 256, 285, 290), -lom (219). In addition, there are a few personal names: Iouenales (182) for Iuuenalis, Lopa (48) for Lupa, Santolus (33) for Sanctulus.

3.3 /ŏ/ written as There are only a handful of cases, all of which are found in inscriptions from the middle of the fifth century or later. Trier: pupulo for populo (135), pusuerunt (18) for posuerunt. Aquitania Prima: doctur (25, 34) for doctor, pastur (25) for pastor, amatur (34) for amator, dolur (35) for dolor, senatur (61) for senator. Vienne: dulure (265, 269) for dolore, curpure (269) for corpore, oxsur (119) for uxor, pecture (269, 274) for pectore.

3.4 /ū/ written as In the three corpora, there is only one case of /ū/ written as , dated from the sixth century Vienne: notrit (92) for nutrit.

4 Phonological observations 4.1 Distribution of forms In the Trier and in the Aquitania Prima inscriptions, I found 34 instances of /ō/ written as spread over 6 lemmata, and about 48 instances of /ŭ/ written as

The merger of /ō/ and /ŭ/

431

spread over 15 lemmata. In Vienne, I found 73 instances of /ō/ written as spread over 25 lemmata, and 90 instances of /ŭ/ written as spread over 21 lemmata. /ŭ/ written as yields more forms than /ō/ written as . This may be explained, partially at least, by lemma frequency. Two words containing /ŭ/, tumulus and titulus, are extremely common in funerary epigraphy, unsurprisingly enough.

4.2 Dark /l/ interference? The affected /ŭ/ are in accent-bearing syllables as well as in accentless ones, but the merger targets forms with liquids and nasals disproportionately. Many examples of /ŭ/ written as affect l-forms; potentially, we may be dealing with the conditioning effects of dark /l/ (Weiss 2009: 117; Sen 2015: 15–31). It may be that a dark /l/ had a lowering effect. This would account for forms such as titolum, tumolo/tomolo, famola, tolit, Fedola, Ursolus, Francola. Looking outside of Gaul, there may be some earlier evidence of /ŭ/ written as in l-forms and nasal-forms; the CIL yields various Republican forms of Herculus written as Hercolus (CIL I2 1427, 1428, 1579, 1697, 1698, 2645), Fuluius written as Foluius (CIL I2 635, 643, 644), and one instance each of consultus written as consoltus (CIL I2 634), Pulc(er) written as Polc(er) (CIL I2 640), and Numuleius written as Numoleius (CIL I2 678).9 These forms range from the second century BCE to the second century CE. The antique forms from the CIL and the late antique and medieval forms of Vienne, Trier and Aquitania Prima would then go directly against the second century BCE wave of vowel raising said to have been brought about by dark /l/, which had also affected /o/ in open and closed syllables (Sen 2012: 472–473). If we accept the conditioning influence of dark /l/, then the forms that do not contain this environment are left unaccounted for. It may be that the merger was originally conditioned by dark /l/, but later became generalized, perhaps affecting nasal-forms first.

4.3 /ŏ/ written as , and the case of /ū/ Cases of /ŏ/ written as are rare, even at a late period, and the few cases skew late. Two alternative accounts are possible in my view.

9 These forms are supplied by Tamponi (2016).

432

Éloïse Lemay

The first is that the few cases of /ŏ/ written as are misspellings that are not indicative of any underlying sound changes, as the Romance languages do not support a view that /ŏ/ was ever merged with /u/. These misspellings may be the result of a generalized confusion between and brought about by the merger between /ō/ and /ŭ/. That is to say, after /ō/ and /ŭ/ merged, the inherited spellings of /ō/ and /ŭ/ eventually faded, resulting in spelling fluctuations between and . This fluctuation came to affect words that had originally had a /ŏ/, despite the fact these words did not play a part in the merger (and find different reflexes in the daughter languages). However, the fact that /ū/ remains almost always unaffected throws a wrench into this account. We would have expected its spelling to fluctuate as much as that of /ŏ/ at the later periods, if there was indeed a generalized spelling confusion between and , regardless of the inherited vowel length. That it does not requires an alternative narrative that may account for /ŏ/ forms. Nearly all of the forms of /ŏ/ written as presented above present unaccented vowels in final position before /r/. All of the nouns are third declensions in /-or/. This may be a continuation of a regular process in Latin, that of the raising of /o/ to /u/ in unaccented closed syllables, as part of vowel weakening; in a word-final position, VC# behaves like a closed syllable. Such a process would have occurred by the second century BCE, although it was not conditioned by _r/.# (Meiser 1998: 70, Sen 2012: 474). We would be dealing here with a much later wave, this time conditioned by /r/. The rule would take the form of: /o/ ˃ /u/ / _r/.#.

5 Sociolectal distribution About 10% of the Trier inscriptions inform us about the elite status of the dedicator or of the recipient; these inscriptions can be attached to the clergy, to the nobility, to veterans, and to city magistrates. I found that the Trier inscriptions that are attached to elite members of the community are better able to avoid certain types of vulgarisms, such as syncope, prothetic vowels, palatalization, and the various types of consonant cluster simplifications (Lemay 2017: 84, 118–120). This indicates that certain Vulgar Latin features were stigmatized; they were certainly part of the language, but they weren’t perceived as desirable in the written language of educated, high-status people. In addition, some sound changes are exclusively found in personal names, such as prothesis, and at all but the latest periods, palatalization (Lemay 2017:

The merger of /ō/ and /ŭ/

433

49–52). These may well have been attempts by the stonecutters to reproduce the way a person’s name was pronounced. By contrast, that these vulgarisms are not found anywhere else than in personal names is an indicator that they were so stigmatized as to be avoided systematically, even in the most modest of inscriptions. However, both vocalic mergers (/ō/ and /ŭ/ and /ĭ/ and /ē/) remain a constant feature of the language throughout the corpus. Once they start appearing in number, they are not avoided even in high status inscriptions. They are seen in the inscriptions of high-ranked clergy members, of nobles, of the military and of magistrates. This informs us that the mergers are an integral part of the written language, as commonly written and as written by the elite; they are perfectly accepted as part of “normative Merovingian Latin”. There is only the handful of inscriptions attached to the highest ranked clergy, bishops and the monastic orders, that are systematically able to avoid all of the Vulgar Latin features, including mergers. These inscriptions are the only ones that follow the Classical Latin norm. This is unsurprising, as bishops and monks would have been the most educated members of the society. Their epitaphs reflect the quality of their education through extremely conservative, archaizing Latin. But these inscriptions are outliers.

6 Conclusion I used the epigraphic evidence of Gaul and the S-Curve statistical model to determine (1) the onset of the o-u merger, (2) its development, and (3) its integration to the Latin language. In order to do so, I determined the development of two mergers, between /ō/ and /ŭ/, and between /ĭ/ and /ē/. In both cases, the data supports the S-Curve pattern of linguistic diffusion, as they indeed feature significant portions of the expected sigmoid distribution. However, the distribution patterns of the o-u and i-e confusions indicate that the two mergers are in different phases of their history. The diffusion of the /ō/ and /ŭ/ merger can be witnessed nearly from start to finish, while the /ĭ/ and /ē/ merger has already taken hold and spread. Both distributions converge upon the same frequency of occurrence, indicating that they are equally integrated beyond that point. I looked at the sociolinguistic distribution of both types of spelling confusions, examining their frequency of occurrence in the inscriptions that bear no status indicators and in the inscriptions that can be tied to elite members of the community. I conclude that, unlike other linguistic features that remained heavily marked, spelling confusions arising from both mergers had become an

434

Éloïse Lemay

integral part of the written Latin language, even in the formal, educated language of the elite, and that only a very small portion of the population had the classical education required to avoid it. It is in my view not conclusive that confusions affecting Classical Latin /ō/ are somewhat less common than confusions affecting /ŭ/. Likewise, that there are fewer affected lemmata is not especially noteworthy. These traits can be accounted for by the peculiarities of epigraphic language, which features a relatively poor lexicon, with some very high-frequency words. Accent does not appear to bear a role in the distribution of the confusions. However, confusions affecting /ŭ/ are disproportionately in forms featuring liquids and nasals, and are especially common in l-forms. This leads me to posit that dark /l/ may have been a conditioning factor at a very early phase in the development of the merger, before it eventually spread to forms with nasals, and then to other environments. It is unclear at which stage /ō/ would have been affected. The small number of late cases of /ŏ/ written as may be secondarilyarising misspellings due to the merger. However, it is possible that they may be the result of a late wave of /o/ to /u/ raising in unaccented closed syllables, conditioned by /r/.

Bibliography CIL = Corpus Inscriptionum Latinarum (1862–). Berlin/Boston: De Gruyter. RCIG = Recueil des Inscriptions Chrétiennes de la Gaule. Adams, J. N. (1977). The vulgar Latin of the letters of Claudius Terentianus (P. Mich.VIII, 467–72) (Publications of the Faculty of Arts of the University of Manchester 23). Manchester: Manchester University Press. Adams, J. N. (1995). The language of the Vindolanda writing tablets: An interim report. Journal of Roman studies, 85, 86–134. Adams, J. N. (2007). The regional diversification of Latin, 200 BC–AD 600. Cambridge/ New York: Cambridge University Press. Blythe, R. A. & Croft W. (2012). S-curves and the mechanisms of propagation in language change. Language, 88(2), 269–304. Denison, D. (2003). Log(ist)ic and simplistic S-curves. Motives for Language Change, 54, 70. Descombes, F. (1985). Recueil des inscriptions chrétiennes de la Gaule antérieures à la Renaissance carolingienne. vol. 15: Viennoise du Nord. Paris: Éditions du Centre National de la Recherche Scientifique. Gauthier, N. (1975). Recueil des inscriptions chrétiennes de la Gaule antérieures à la Renaissance carolingienne. Première Belgique. vol. 1. Paris: Centre national de la recherche scientifique.

The merger of /ō/ and /ŭ/

435

Ghanbarnejad, F. et al. (2014). Extracting information from S-curves of language change. Journal of The Royal Society Interface, 11(101), https://doi.org/10.1098/rsif.2014.1044. Herman, J. (1971). Essai sur la latinité du littoral adriatique à l’époque de l’Empire. In E. Coseriu & W.-D. Stempel (Eds.), Sprache und Geschichte. Festschrift für Harri Meier zum 65. Geburtstag (pp. 199–226). Munich: Fink. Kroch, A. S. (1989). Reflexes of grammar in patterns of language change. Language Variation and Change, 1(3), 199–244. Krämer, K. H. (1974). Die frühchristlichen Grabinschriften Triers: Untersuchungen zu Formular, Chronologie, Paläographie und Fundort. Trier: Rheinisches Landesmuseum Trier. Labov, W. (1994). Principles of linguistic change. vol. 1: Internal Features. Oxford: Blackwell. Lemay, É. (2017). Studies in Merovingian Latin epigraphy and documents (PhD thesis). Los Angeles: University of California. Meiser, G. (1998). Historische Laut- und Formenlehre der lateinischen Sprache. Darmstadt: Wissenschaftliche Buchgesellschaft. Nevalainen, T. (2015). Descriptive adequacy of the S-curve model in diachronic studies of language change. In C. Sanchez-Stockhammer (Ed.), Can we predict linguistic change? vol. 16. Studies in variation, contacts and change in English. Helsinki: Varieng. Powell, J. G. F. (2011). The appendix Probi as linguistic evidence: A reassessment. In R. Ferri (Ed.), The Latin of Roman lexicography (Ricerche sulle lingue di frammentaria attestazione 7, pp. 75–119). Pisa/Rome: Fabrizio Serra. Prévot, F. (1997). Recueil des inscriptions chrétiennes de la Gaule, antérieures à la Renaissance carolingienne: Aquitaine première (vol. 8). Paris: Centre national de la recherche scientifique. Prinz, O. (1932), De o et u vocalibus inter se permutatis in lingua latina quaestiones epigraphicae (PhD thesis). Halle. Sen R. (2012). Reconstructing phonological change: Duration and syllable structure in Latin vowel reduction. Phonology, 29(3), 465–504. Sen, R. (2015). Syllable and segment in Latin (vol. 16). Oxford: Oxford University Press. Tamponi, L. (2016). Allografi nell’epigrafia latina: Indici di variazione sociolinguistica (PhD thesis). Università di Pisa. Väänänen, V. (1981). Introduction au latin vulgaire. Troisième édition revue et augmentée. Paris: Klincksieck. Weiss, M. (2009). Outline of the historical and comparative grammar of Latin. Ann Arbor/ New York: Beech Stave Press.

Giovanna Marotta & Irene De Felice

Patterns of prosodic distribution of Latin long vowels Abstract: In Classical Latin, vowel quantity (VQ) plays a fundamental role in stress assignment, since it contributes to the assignment of lexical stress in polysyllabic words. However, no Romance languages inherited the original VQ contrast. In this contribution, we aim to find quantitative evidence for the tendency of Latin long vowels to occur in stressed syllables more frequently than in unstressed syllables, a tendency that possibly correlates with the progressive loss of the length contrast in the evolution of the language. Our research is based on the Latin lexicon annotated with a number of prosodic features, related to vowel length, syllable structure, syllable weight, and lexical stress. The analysis of our data clearly reveals a strong preference for long vowels to occur in stressed syllables, and in particular in the open ones, i.e. in the same context in which many Romance languages show long vowels. Quantitative evidence therefore indicates that in Classical Latin, the contrast of VQ, although still prevalent, was already prefiguring the prosodic system of the Romance languages, in which the inherited vowel quantity is no longer phonologically relevant and vowel length is largely determined by stress. Keywords: Latin linguistics, Latin lexicon, vowel length, syllable structure, stress, corpus annotation

1 The contrast of vowel length in Latin Even if the standard Latin orthography does not systematically distinguish between short and long vowels (Leumann 1977: 3–15; Weiss 2009: 29; Loporcaro 2015: 3–5), there is evidence supporting the relevance of vowel length and

Note: This paper is the result of the close collaboration of both authors. For the specific concerns of the Italian Academy, Giovanna Marotta is responsible for Sections 1, 2 and 6, and Irene De Felice is responsible for Sections 3, 4 and 5. Giovanna Marotta, Irene De Felice, University of Pisa https://doi.org/10.1515/9783110647587-027

Patterns of prosodic distribution of Latin long vowels

437

phonological contrast between short and long vowels in the classical era (at least at some sociolinguistic level, cf. infra). That Romans actually pronounced long and short vowels in different ways is reflected in the various strategies with which they sporadically notated long vowels in epigraphic texts, such as geminatio vocalium (e.g. CIL I2 1299, Rome, 130–100 BC: VAARUS), I longa (e.g. CIL I2 721, Rome, 82–79 BC: VICUS), or apex (e.g. Gordon 1, 61, Veii, 25 AD: UT AUGUSTÁLIUM NUMERÓ HABEÁTUR).1 The relevance of vowel quantity also emerges from metrics, since Latin poetry, as is well known, is based on patterns of syllable quantity. In this regard, we know that syllable length was not only relevant to the written poetry but also to its oral performance: Roman writers tell us that syllable length was in fact conveyed by actors, and clearly perceived by the audience. Therefore, even common people, who probably lacked any conscious knowledge of metrics or rhythm, were able to distinguish segmental length during theatrical performances. Some testimonies by Cicero are particularly interesting in this respect.2 Vowel quantity contrast in Latin may be indirectly reconstructed through interlinguistic comparison (e.g. PIE *máh2ter > Gr. μήτηρ, Lat. māter ‘mother’ vs. PIE *h2áǵ-e/o- > Gr. ἄγω, Lat. ago ‘to drive’), or through considering the borrowings into or from other languages (e.g. Lat. cattus > Old High German kazza ‘cat’ vs. Lat. strāta > strāz(z)a ‘street’; Sihler 1995: 73; Loporcaro 2015: 4–5). Moreover, contrasts at the level of vowel length distinguish different lexical entries, as many minimal pairs demonstrate, such as lēvĭs ‘smooth’ ~ lĕvĭs ‘light’, ōs ‘mouth’ ~ ŏs ‘bone’, pŏpŭlŭs ‘people’ ~ pōpŭlŭs ‘poplar tree’. Vowel quantity also creates morphological contrasts. This regularly happens in the nominal system, as for the -ă/-ā endings of the I decl. (pŏētă ‘poet’, nom./voc. sing. ~ pŏētā, abl. sing.), or for the -ŭs/-ūs endings of the IV decl. (lăcŭs ‘lake’, nom./voc. sing. ~ lăcūs gen. sing./nom./acc./voc. pl.). Morphophonological contrasts in terms of vowel quantity exist in the verbal system too, since the present and perfect stems are sometimes distinguished only by quantitative vowel gradation (ablaut), as in the case of the verbs vĕnit ‘he/she comes’ ~ vēnit ‘he/she came’, ĕmit ‘he/she buys’ ~ ēmit ‘he/she bought’, lĕgit ‘he/she reads’ ~ lēgit ‘he/she read’. Interestingly, Roman grammarians were aware of the fact that the difference in vowel length created a phonological contrast in Latin, as testified by the comments they made on some minimal pairs. For instance: Eadem littera alium atque alium intellectum, prout correpta vel producta est, facit: ut ‘malus’ arborem significet an hominem non bonum apice distinguitur, ‘palus’ aliud priore syllaba

1 Lazzeroni (1956); Allen (1978: 64); Weiss (2009: 29); Loporcaro (2015: 3–4). 2 See CIC. orat. 173; parad. 3, 26 and the recent comments by Marotta (2018).

438

Giovanna Marotta & Irene De Felice

longa, aliud sequenti significat, et cum eadem littera nominativo casu brevis, ablativo longa est, utrum sequamur plerumque hac nota monendi sumus. (QVINT. inst. 1, 7, 2–3) ‘The same letter produces different senses if it is long and if it is short. Thus, in malus, an apex indicates that it means ‘apple tree’ and not ‘bad man’; palus also means one thing if the first syllable is long and another if the second is long; and when the same letter is found as short in the nominative and as long in the ablative, we commonly need to be reminded which interpretation to choose.’3

Vowel length, however, is not only relevant to Latin phonology (for its contrastive value) but also to prosody, since it affects syllable weight (Mester 1994; Marotta 1999, and Marotta 2000), and consequently plays a central role in assigning word stress in polysyllabic words. Latin is a quantity-sensitive language: according to its stress algorithm, in a word of three or more syllables, the accent falls on the penultimate syllable if this is heavy; otherwise it falls on the antepenultimate (e.g. prṓvĕnĭt ~ prōvḗnĭt). As is well known, syllable weight is entirely determined by the structure of its rhyme:4 a syllable is heavy if it has a heavy nucleus (i.e. a long vowel or a diphthong), or a coda in the rhyme; syllable onsets (i.e. consonants that precede the nucleus) have no effect on syllable weight.

2 Vowel length in Romance languages While in the phonological system of Classical Latin vowel length had great importance for the reasons discussed in Section 1, no Romance languages inherited the original contrast between short and long vowels. The vowel length contrast occurring in some varieties (e.g. Friulan) does not descend from the Latin contrast, but developed more recently for different reasons (Loporcaro 2015). In fact, many Romance languages have long and short vowels, but this difference does not have a phonological status. Notably, in those varieties that present long vowels, it is lexical stress that has a phonological value, i.e. a contrastive function, whereas vowel length is generally a consequence of stress (i.e. stressed vowels lengthen, under some conditions; cf. Weiss 2009: 508; Loporcaro 2011a, Loporcaro 2011b, and Loporcaro 2015; Marotta 2016). For instance, in Standard

3 Trans. by Russell (2002: 185–187). 4 A syllable rhyme is composed of the phonemes from the vowel to the end of the syllable, that is the nucleus (i.e. a vowel or a diphthong) and the following consonants belonging to the same syllable, which constitute the so-called coda; see Marotta (1999) and the theoretical references there quoted.

Patterns of prosodic distribution of Latin long vowels

439

Italian, vowels in stressed open syllables are longer than unstressed vowels or vowels in closed syllables (especially in the penultimate position: see Marotta 1985), regardless of the original quantity of the vowel (e.g. Lat. pătrem > It. [ˈpa: dre] ‘father’; Lat. rŏtam > It. [ˈrwɔ:ta] ‘wheel’). Allophonic vowel lengthening depending on stress and syllable structure is also found in other Romance languages and varieties, such as Portuguese, Sardinian, Catalan, central and southern Italian dialects (Marotta 2016). Therefore, at some point in the history of Latin, short vowels must have begun to lengthen in stressed open syllables, and long vowels to shorten in unstressed ones; the phenomenon is known as Open Syllable Lengthening (OSL, in Loporcaro 2011a, Loporcaro 2011b, and Loporcaro 2015). Some passages from Pompeius, Consentius, and Augustine5 have been interpreted as indicating African Latin as the variety from which the loss of length contrast first spread (cf. Herman 1982), but the phenomenon cannot be understood as confined to Africa (as discussed in Adams 2007: 260–270 and Adams 2013: 43–51; Loporcaro 2015: 20–25). When vowel length ceased to be contrastive, the distribution of long vowels, which could occupy any position within a word, began to be determined by stress and by syllable structure; the original contrast of quantity was progressively lost, whereas differences in the degree of openness in mid vowels became more relevant. The fact that there was a quality distinction parallel to the quantity distinctions of short and long vowels (except /a/) in Latin is testified by grammarians, for instance by Servius and Victorinus, who explicitly write that and sound differently when long or short.6 Also, inscriptions show

5 POMP. gramm. V 285, 5–7 est alter [barbarismus], qui fit in pronuntiatu. plerumque male pronuntiamus et facimus vitium, ut brevis syllaba longo tractu sonet aut iterum longa breviore sono ‘there is another (barbarism), which is committed in pronunciation. Often we utter a bad pronunciation and commit the fault of sounding a short syllable long or, again, a long syllable short’; CONSENT. gramm. V 392, 3 ut quidam dicunt piper producta priore syllaba, cum sit brevis, quod vitium Afrorum familiare est ‘as some people say piper with a long first syllable, when it is short, a vice which is characteristic of Africans’; AVG. doctr. christ. 4, 10, 24 cur pietatis doctorem pigeat imperitis loquentem ossum potius quam os dicere, ne ista syllaba non ab eo quod sunt ossa, sed ab eo quod sunt ora, intellegatur, ubi afrae aures de correptione vocalium vel productione non iudicant? ‘Why should a teacher of piety when speaking to the uneducated have regrets about saying ossum (‘bone’) rather than os in order to prevent that monosyllable [i.e. ŏs ‘bone’] from being interpreted as the word whose plural is ora [i.e. ōs ‘mouth’] rather than the word whose plural is ossa [i.e. ŏs], given that African ears show no judgment in the matter of the shortening of vowels or their lengthening?’ (trans. by Adams 2007: 261–264). 6 SERV. gramm. IV 421, 16 Vocales sunt quinque, a e i o u. Ex his duae, e et o, aliter sonant productae, aliter correptae ‘There are five vowels, a e i o u. Two of these, e and o, sound in one way when long, in another when short’ (trans. by Sturtevant 1940: 111); PS. MAR. VICTORIN.

440

Giovanna Marotta & Irene De Felice

a growing confusion between the long mid vowels and the short high ones (cf. Marotta 2015), starting from the third century BC but spreading from the third century AD (e.g. MENOS for MINUS, CIL VIII 9984; VERTUTE for VIRTUTE, CIL V 6244; 2 ERODITA for ERUDITA, CIL I 1214/2; COM for CUM, CIL IV 39359). There are different hypotheses indicating when (and how) contrastive vowel length ceased to exist in spoken Latin, and when a quality distinction acquired a phonological status in large parts of Romània (for a general overview, cf. Loporcaro 2015: 18–19). We believe that this change did not affect all the levels of the Latin diasystem at the same time. Following the opinion of other scholars (e.g. Pulgram 1975; Vineis 1984; Giannini and Marotta 1989), we think that vowel length could have already lost its phonological status, and have been already substituted (or at least complemented) by a tenseness contrast since the Republican age (probably from the third century BC) in the basilectal (i.e. lower) varieties of Latin (Benedetti and Marotta 2014; Marotta 2015 and Marotta 2017). In this regard, it is worth noting that the tendency to reduce the functional load of the vowel length contrast can be traced back to changes that took place in the (pre)history of Latin, such as the shortening processes that variously affected unstressed (especially final) syllables (e.g. correptio iambica, shortening of final vowels followed by a consonant unlike -s; Loporcaro 2015: 9–12; Marotta 2017). In this contribution, we argue that the seeds of the changes that led to the definitive collapse of the vowel length contrast were already taking root in the lexicon of Classical Latin: although long vowels could occur in any position of the word, vowel length was largely conditioned by lexical stress and syllable structure. Therefore, in this study we aim at finding quantitative evidence for a tendency for long vowels to occur in stressed syllables more frequently than in unstressed syllables, and thus for a general drift towards the progressive loss of the length contrast in Latin.

3 Previous quantitative studies Among the previous studies that address the relation between vowel length and stress in Latin with a quantitative approach (e.g. Gaeng 1968; Omeltchenko 1977; Herman 1982 and Herman 1985), the analysis proposed by Herman (1968) is the closest one to our research, in terms of method and purposes. The author

gramm. VI 33, 3 O, ut e, geminum vocis sonum pro condicione temporis promit ‘O, like e, produces two vowel sounds according to the quantity’ (trans. by Sturtevant 1940: 119).

441

Patterns of prosodic distribution of Latin long vowels

examined the distribution of ca. 11,000 vowels and diphthongs, extracted from the Epistolae Selectae of Cicero (ed. by Dietsch 1911 [1854]), with respect to lexical stress (stressed vs. unstressed syllable) and the position within a word (initial, internal, or final position). In light of the purpose of our study, the most important result of Herman’s analysis is that even if short vowels turn out to be more numerous than the long ones (70.5% out of the total number of vowels and diphthongs extracted), the proportion of long vowels is significantly higher in stressed syllables than in the unstressed ones. Table 1 shows that long vowels or diphthongs occur in 41.8% of stressed syllables, but in only 22.5% of unstressed syllables.

Table 1: The distribution of vowels and diphthongs in stressed and unstressed syllables, expressed in percentage values (adapted from Herman 1968: 244).

Stressed syllable Unstressed syllable

ă

ā

ĕ

ē

ĭ

ī

ŏ

ō

ŭ

ū

ae

au

.

.

.

.

.

.



.

.

.

.

.



.

.

.

.

.

.

.

.

.

.

.

These data suggest that the contrast of vowel quantity in Classical Latin was already strongly limited to stressed syllables. Moreover, Herman (1968) analyses the distribution of Latin vowels considering not only their length but also their quality. As previously indicated in Table 1, there are relevant differences between the phonemes. For instance, ī is more frequent in unstressed syllables as opposed to the other long vowels. However, we shall not comment further on this point, since in this chapter we are only focusing on the distribution of long vowels and diphthongs without considering their quality, leaving this more detailed analysis for future research. Although Herman’s study still represents a model of quantitative analysis on a corpus of data relative to an ancient language, he did not consider syllable structure. In our opinion, this is a relevant factor that cannot be neglected within the general framework of Latin prosody. In other words, we believe that the interplay of syllable structure and lexical stress could have significant effects on the distribution of long and short vocalic segments in Latin.

442

Giovanna Marotta & Irene De Felice

4 Corpus and method of analysis Following Herman’s legacy, our goal is the research of distributional patterns of vowel segments in the Latin lexicon. Our quantitative analysis is based on a preexisting lexicographical resource, the Pocket Oxford Latin Dictionary (Oxford University Press, 2012). This dictionary covers more than 11,000 entries that represent the essential vocabulary of the Latin language.7 Compared to larger dictionaries, such as the Oxford Latin Dictionary, it has the advantage of excluding many rare forms, most words of foreign origin, and technical terms; moreover, it provides information about vowel length, marking long vowels with a macron.8 First of all, from the total number of dictionary entries (tot. 11,052), we deleted the forms that did not constitute a lemma (e.g. the entry adl-, which simply points to all-). Moreover, in order to avoid redundant information in the corpus, we reduced the pairs of homographs, when words were also identical in terms of vowel quantity (e.g. truncus_1: noun, ‘trunk, bust’; truncus_2: adjective, ‘cut, imperfect’). The total number of forms that constitute the corpus thus obtained is 10,874. Secondly, all lemmas were automatically divided into syllables using a specific set of rules9 (e.g. lin.gua, auc.si.li.um, pa.tri.a, cas.trum, pos.tu.lo; tot. syllables: 34,905). In the case of compounds or derived forms, hyphenation complies with morphological boundaries (e.g. ab.a.vus, and not a.ba.vus; ad.i.ci.ō, and not a.di.ci.ō). All lemmas were then automatically classified according to their number of syllables (cf. Figure 1). Finally, for each syllable of the corpus, we annotated the following phonological and prosodic features: – number of syllables constituting each lemma; – position of the stress in the lemma;10

7 The dictionary mainly presents the lexicon of the Classical period, with only some additions from the works of Plautus and Terence, and from the period of Silver Latin (but not from Medieval Latin). The online version of the dictionary can be accessed at www.oxfordreference.com/view/10. 1093/acref/9780191739583.001.0001/acref-9780191739583 (accessed: 30 July 2018). 8 However, this dictionary does not explicitly distinguish short vowels from those whose length cannot be determined for certain (cf. infra, Section 5.3). For the purpose of our study, and since we focus on the distribution of long vowels, we decided to consider as short all vowels that are not definitely marked as long. 9 This operation was conducted in collaboration with the Department of Humanities of Ca’ Foscari University of Venice (with Paolo Mastandrea, Luigi Tessarolo, and Silvia Arrigoni), and following the general framework of the Latin syllable as described in Marotta (1999). 10 Since the position of Latin stress is governed by the so-called penultimate rule, we were able to individuate for each lemma the exact position of stress, by knowing the word length and the weight of syllables. For the few lemmas that do not comply with this rule (cf. Allen

Number of syllables

Patterns of prosodic distribution of Latin long vowels

1 2 3 4 5 6 7

443

199 2313 4585 2745 859 161 12 Number of lemmas

Figure 1: Classification of the lemmas with regard to the number of syllables.

– prosodic status of each syllable in the lemma (stressed, post-tonic, pretonic); – syllable structure, i.e. open vs. closed syllable; – syllable structure in relation to the nucleus, i.e. short vowel vs. long vowel or diphthong.11 A preliminary caveat should be pointed out: since the current analysis is based on the paradigmatic entries of the dictionary, morphological alternation is not taken into account. We are perfectly aware that this constraint represents a limit, since Latin is a language with a very rich and pervading inflection. However, we do believe that the data collected, despite this limitation, could supply relevant cues regarding the prosodic drift taking place in the history of the language, because they refer to the prototypical forms present in the mental lexicon of Latin speakers. At the same time, we plan to couple these data concerning lexical entries with those occurring in written texts (cf. Section 6).

5 Results and discussion A first discrimination concerns monosyllabic and polysyllabic words. All of the 199 monosyllables of our corpus (e.g. vir, nunc, dē) consist of a heavy or superheavy rhyme, in accordance with the bimoraicity constraint on the minimal word in Latin (Mester 1994; Marotta 1999 and Marotta 2000). If the syllable is open, the nucleus always has a long vowel or a diphthong.

1978: 87; Weiss 2009: 109–111), such as oxytones resulting from apocope (e.g. illīc < illi + -ce), the position of stress was manually corrected. 11 We considered ae, au, oe, ei, eu, ou as Latin diphthongs (Sturtevant 1940: 123–138; Allen 1978: 60–63).

444

Giovanna Marotta & Irene De Felice

Since our goal is to find a possible correlation between lexical stress and the occurrence of heavy nuclei, we will describe the distributional patterns of long vowels in polysyllables only, i.e. in 10,675 lemmas. We will initially focus on the distribution of long vowels with respect to lexical stress (Section 5.1) and then on the distribution of long vowels with respect to the syllable structure (Section 5.2). In the last paragraph (Section 5.3), we will draw some general conclusions about the interaction between vowel length, stress and syllable structure.

5.1 Long vowels and stress Table 2 reports the distribution of short and long vowels and diphthongs in stressed and unstressed syllables, divided into pre-tonic and post-tonic ones. The data are presented in absolute as well as in percentage values. Table 2: Vowel length in stressed and unstressed (pre-tonic/post-tonic) syllables (tot. 10,675 lemmas, 34,706 syllables).

Stressed syllable

Pre-tonic syllable

Post-tonic syllable

Short V

Long V

Diphthong

Total

,

,



,

.%

.%

.%

%

,

,



,

.%

%

.%

%

,

,



,

.%

.%

.%

%

The data confirm that in Classical Latin both long and short vowels could occur in any position within a word, although short nuclei appear to be more frequent than long vowels, generally speaking. However, the different percentages of vowel occurrence show that the distribution is not entirely independent of lexical stress. In particular, the proportion of short nuclei is very high in the post-tonic position (72.1%), whereas long vowels are more frequent in the stressed syllables than in the unstressed ones. As Table 2 shows, long vowels occur in 41.1% of the stressed syllables (e.g. tāc.tus), but only in 23% of the pre-tonic (e.g. ō.rā.tor) and in 27.5% of the post-tonic ones (e.g. e.dō).

Patterns of prosodic distribution of Latin long vowels

445

Diphthongs are quite rare. Their percentages range from 3.1% in stressed syllables (e.g. poe.na) to 3.5% in the pre-tonic ones (e.g. lae.ti.ti.a), but are less than 1% in the post-tonic position, due to their sporadic occurrence in the morphological endings of the Latin lemmas (e.g. nū.gae).12 These results are largely consistent with those presented by Herman (1968), even if that study is based on a very different corpus (cf. infra), i.e. a prose text, whereas the present study is based on the lexicon. In particular, in Herman’s study, 38.2% of the stressed syllables contain a long vowel, and 3.6% contain a diphthong. The results emerged from our analysis are therefore very similar: 41.1% of stressed syllables present a long vowel, 3.1% present a diphthong. If we also include monosyllables in our analysis (as Herman did, in considering monosyllabic words as stressed syllables; cf. Herman 1968: 244), the percentages remain very similar (41.2% and 3.2%). The agreement between Herman’s data and ours confirms the legitimacy of a quantitative analysis based on the entries of the Latin dictionary for studying the distribution of prosodic features.

5.2 Long vowels and syllable structure We can now turn to consider the distribution of short and long vowels and diphthongs in open and closed syllables (Table 3). Table 3: Vowel length in open and closed syllables (tot. 10,675 lemmas, 34,706 syllables). Short V Open syllable

Closed syllable

Long V

Diphthong

Total

,

,



,

.%

.%

.%

%

,

,



,

.%

.%

.%

%

12 A chi-square test confirms that the distribution of long and short vowels and diphthongs in stressed and unstressed (pre-tonic/post-tonic) syllables is not due to chance (χ2 (4, 34706) = 1273.69; p < 0.001). The chi-square statistical test is used to compare the observed data (reported in Table 2) with the expected ones (which we would expect under the null hypothesis, i.e. that there is no significant difference between stressed, pre-tonic and post-tonic syllables with respect to the presence of a long or short vowel or a diphthong), and to determine whether the difference between the expected and the observed frequencies is significant. A p value < 0.001 means that the probability that the distribution observed is due to chance is less than one in a thousand.

446

Giovanna Marotta & Irene De Felice

Both in the stressed and in the unstressed position (cf. infra, Table 4), long vowels are much more frequent in open syllables (8,995 occurrences) than in the closed ones (1,626 occurrences), with a ratio of almost 6:1. In contrast, the distribution of short vowels is concentrated in closed syllables: out of 13,660 closed syllables, 11,959 of them, i.e. 87.5%, present a short vowel.13 The distribution of diphthongs is very similar to that of long vowels, since they most frequently occur in open syllables (614 vs. 75 occurrences), although the occurrence percentage of diphthongs remains quite low. In conclusion, the distribution of long and short vowels and diphthongs in a Latin word is sensitive to the syllable structure,14 in that long vowels (or diphthongs) mostly occur in open syllables.

5.3 General results: stress, syllables, and syllable weight In this section, we summarise the relevant results of our lexical and prosodic analysis on the Latin lexicon. In Table 4, we draw a general picture of our data showing the interaction between vowel quantity, lexical stress, and syllable structure over a total of 10,675 lemmas and 34,706 syllables. From the data reported in Table 4 we can observe that the great majority of stressed syllables (77.1%) present a heavy or super-heavy rhyme, which can be due to a long vowel (or diphthong) and/or a consonant in coda position. Long vowels are more frequent in the open syllables than in the closed ones (with a ratio of almost 6:1). In open stressed syllables only they are even more numerous than short vowels (35.1% vs. 22.9%). As for unstressed syllables, they generally (in more than 70% of cases) present short vowels. Despite the high presence of short vowels, a high percentage of unstressed syllables (64.5% of the pre-tonic and 61.5% of the post-tonic ones) present a heavy rhyme: however, in most cases, this is not due to the presence of a long vowel (as is the case in most of the stressed syllables), but rather to the presence of a short vowel followed by a consonant in coda. If we now consider the internal structure of the open stressed syllables, we see that the distribution of long and short nuclei is different, since the first ones are much more frequent than the second ones (cf. Table 5). The picture is specular in closed stressed syllables, where the number of long vowels remarkably 13 This fact may be partly due to the problem of the ‘hidden’ quantity, for which cf. Section 5.3. 14 As confirmed by a chi-square test conducted on the data reported in Table 3: χ2 (2, 34706) = 4162.69, p < 0.001.

Post-tonic syllable

Pre-tonic syllable

Stressed syllable

Long V

,

.%

,

.%

,

.%

Short V

,

.%

,

.%

,

.%

.%



.%



.%



Diphthong

Open syllable

.%

.%

.%

Long V + Diphthong

.%

,

%

,

.%

,

Short V

.%



.%



%



Long V

%



. %



. %



Diphthong

Closed syllable

Table 4: The interaction between vowel length, stress, and syllable structure (total 10,675 lemmas, 34,706 syllables).

. %

. %

.%

Long V + Diphthong

%

,

%

,

%

,

Total

Patterns of prosodic distribution of Latin long vowels

447

448

Giovanna Marotta & Irene De Felice

Table 5: Vowel distribution according to stress and syllable structure.

Open syllable Stressed Closed syllable

Open syllable Unstressed

Short V

Long V

Diphthong

,

,



.%

.%

.% , .% , .% ,

Closed syllable

.%

Long V + Diphthong

,

.% 



.%

% 

.%

.% 

.%

.%

% ,

.% 

% ,

.% ,

Total

% ,

.%

%

decreases. On the other hand, in the unstressed position, the percentage of short vowels increases, especially in the context of a closed syllable. It is worth commenting briefly on super-heavy syllables, which constitute only 5% of all syllables on average. In this position, the length of a vowel cannot be directly observed (i.e. it is ‘hidden’), because closed syllables are already prosodically heavy due to the consonant being in coda position. The occurrence of a long vowel in a closed syllable can mostly be reconstructed in an indirect way, i.e. with comparative analysis and etymology (even if some direct evidence can be found in some notations used in inscriptions and papyri, such as apex or I longa; cf. Section 1). As is well-known, metrics does not treat heavy and super-heavy rhymes differently, and in this position vowel quantity is irrelevant for stress assignment. In other words, in closed syllables, vowel length is a redundant feature. This makes it reasonable to speculate that the contrast of vowel length was first lost just in this position. To conclude the analysis of our results, we can state that stressed and unstressed syllables show a different distribution of light and (super-)heavy rhymes (Table 6; cf. also De Felice 2018):

Patterns of prosodic distribution of Latin long vowels

449

Table 6: Light and (super-)heavy rhymes in stressed and unstressed syllables. Light rhyme

(Super-)heavy rhyme

Total

Stressed syllable

,

,

,

%

.%

.%

%

Unstressed syllable

,

,

,

%

.%

.%

%

The difference between stressed and unstressed syllables becomes evident. The distribution of light and (super-)heavy rhymes in stressed and unstressed syllables, with a higher proportion of (super-)heavy rhymes in the stressed position, is also highly significant from a statistical point of view.15

6 Conclusion and future research The empirical evidence presented in this chapter clearly indicates that the distribution of Latin long vowels was not entirely independent from syllable structure (long vowels are mainly found in open syllables rather than in closed ones), nor from stress (long vowels are more frequent in stressed than in unstressed syllables). As stated in Section 1, vowel length still had a contrastive value in the age of Classical Latin (at least, in the higher varieties of the language), and hence in one sense should be independent of stress and syllable structure. However, our analysis clearly reveals a strong preference for long vowels to occur in stressed syllables, especially in the open ones, i.e. in the same context in which many Romance languages (like Italian) present long vowels. The lexicon of Classical Latin, therefore, already contained the seeds of the long drift towards the loss of vowel quantity along the typologically unmarked path already highlighted in various natural languages. The quantitative analysis presented here is based on a Latin lexicon (i.e. the Pocket Oxford Latin Dictionary) which has the advantage of providing evidence from a large number of lexical stems, although it does not account for the actual frequency of the different prosodic patterns in the real spoken and written language, where the morphologically inflected forms occurred too, with

15 χ2 (1, 34706) = 704.77; p < 0.001.

450

Giovanna Marotta & Irene De Felice

the consequence of increasing the number of words of three or more syllables. For instance, in the current research, only nominative forms like caput, ordo, or rex are included, whereas allomorphs like capitis, ordine or regem are excluded; the same for verbal forms like amo, lego (included) vs. amaverunt, legimus (excluded). For this reason, we plan to extend our research and to apply the same methodology to a large corpus of Latin literary texts, in order to verify whether the distributional patterns observed in the lexicon could be confirmed when dealing with different material. However, since the results obtained so far with regard to lexical data are largely consistent with those presented by Herman (1968), who based his quantitative study on Cicero’s letters, we do not expect that the outcome of our future analysis based on literary texts could differ significantly from what we have presented here.

Abbreviations Gr. Lat. PIE

Greek Latin Proto-Indo-European

Bibliography CIL = Corpus Inscriptionum Latinarum (1862–). Berlin/Boston: De Gruyter. Pocket Oxford Latin Dictionary: www.oxfordreference.com/view/10.1093/acref/ 9780191739583.001.0001/acref-9780191739583 (accessed: 30 July 2018) Adams, J. N. (2007). The regional diversification of Latin, 200 BC–AD 600. Cambridge: Cambridge University Press. Adams, J. N. (2013). Social variation and the Latin language. Cambridge: Cambridge University Press. Allen, W. S. (1978 [1965]). Vox Latina: The pronunciation of Classical Latin. Cambridge: Cambridge University Press. Benedetti, M. & Marotta, G. (2014). Monottongazione e geminazione in latino: Nuovi elementi a favore dell’isocronismo sillabico. In P. Molinelli, P. Cuzzolin, & C. Fedriani (Eds.), Latin vulgaire-latin tardif X: Actes du Xe Colloque international sur le latin vulgaire et tardif (vol. I, pp. 25–43). Bergamo: Sestante. De Felice, I. (2018). Strutture sillabiche nel lessico latino. In L. Costamagna, E. Di Domenico, A. Marcaccio, S. Scaglione & B. Turchetta (Eds.), Mutamento linguistico e biodiversità. Atti del XLI Convegno della Società Italiana di Glottologia (pp. 235–241). Roma: Il Calamo. Dietsch, R. & Klotz, R. (1911 [1854]). M. Tulli Ciceronis Epistolae selectae. Leipzig: Teubner. Gaeng, P. A. (1968). An inquiry into local variations in Vulgar Latin, as reflected in the vocalism of christian inscriptions. Chapel Hill: University of North Carolina Press.

Patterns of prosodic distribution of Latin long vowels

451

Giannini, S. & Marotta, G. (1989). Fra grammatica e pragmatica: La geminazione consonantica in latino. Pisa: Giardini. Herman, J. (1968). Statistique et diachronie: Essai sur l’évolution du vocalisme dans la latinité tardive. Word, 24, 242–251. Herman, J. (1982). Un vieux dossier réouvert: Les transformations du système latin des quantités vocaliques. Bulletin de la Société de Linguistique de Paris, 77, 285–302. Herman, J. (1985). Témoignage des inscriptions latines et préhistoire des langues romanes: Le cas de la Sardaigne. In M. Deanovíc (Ed.), Mélanges de linguistique dédiés à la mémoire de Petar Skok (pp. 207–216). Zagreb: Jugosiavenska Akademija Znanosti i Umjetnosti. Lazzeroni, R. (1956). La ‘geminatio vocalium’ nelle iscrizioni latine. Annali della Scuola Normale di Pisa, 25, 124–135. Leumann, M. (1977). Lateinische Laut- und Formenlehre. München: Beck. Loporcaro, M. (2011a). Syllable, segment and prosody. In M. Maiden, J. Smith & A. Ledgeway (Eds.), (pp. 50–108). Loporcaro, M. (2011b). Phonological processes. In M. Maiden, J. Smith & A. Ledgeway (Eds.), (pp. 109–154). Loporcaro, M. (2015). Vowel length from Latin to Romance. Oxford: Oxford University Press. Maiden, M., Smith, J. & Ledgeway, A. (Eds.). The Cambridge history of the Romance languages. Cambridge: Cambridge University Press. Marotta, G. (1985). Modelli e misure ritmiche: La durata vocalica in italiano. Bologna: Zanichelli. Marotta, G. (1999). The Latin syllable. In H. van der Hulst & N. A. Ritter (Eds.), The syllable: Views and facts (pp. 285–310). Berlin/New York: Mouton de Gruyter. Marotta, G. (2000). Sulla massimalità dei piedi trocaici: Il caso del latino. Lingua e Stile, 35(3), 387–416. Marotta, G. (2015). Talking stones: Phonology in Latin inscriptions? Studi e Saggi Linguistici, 58(2), 39–63. Marotta, G. (2016). Prosodic structure. In A. Ledgeway & M. Maiden (Eds.), The Oxford guide to the Romance Languages (pp. 484–494). Oxford: Oxford University Press. Marotta, G. (2017). Tra fonologia e sociofonetica: Il tratto di lunghezza in latino. In G. Marotta & F. Strik Lievers (Eds.), Strutture linguistiche e dati empirici in diacronia e sincronia (pp. 57–81). Pisa: Pisa University Press. Marotta, G. (2018). On Cicero’s fine-grained perception of prosodic features in Latin. Journal of Latin Linguistics, 17(2), 259–278. Mester, R. A. (1994). The quantitative trochee in Latin. Natural Language and Linguistic Theory, 12, 1–61. Omeltchenko, W. (1977). A quantitative and comparative study of the vocalism of the Latin inscriptions of North Africa, Britain, Dalmatia, and the Balkans. Chapel Hill: University of North Carolina Press. Pulgram, E. (1975). Latin-Romance phonology: Prosodics and metrics. Munich: Fink. Russell, D. A. (2002). Quintilian: The orator’s education, vol. I, books 1–2 (Loeb Classical Library 124). Cambridge, MA: Harvard University Press. Sturtevant, E. H. (1940 [1920]). The pronunciation of Greek and Latin. Philadelphia/Baltimore: Linguistic Society of America. Sihler, A. L. (1995). A new comparative grammar of Greek and Latin. New York/Oxford: Oxford University Press.

452

Giovanna Marotta & Irene De Felice

Vineis, E. (1984). Problemi di ricostruzione della fonologia del latino volgare. In E. Vineis (Ed.), Latino volgare, latino medioevale, lingue romanze: Atti del convegno della Società Italiana di Glottologia (pp. 45–62). Pisa: Giardini. Weiss, M. (2009). Outline of the historical and comparative grammar of Latin. Ann Arbor/ New York: Beech Stave Press.

Index abeo 114–115 ablative see ancient grammatical terminology abstract nouns. see nouns accurro 121 accusative 116–117, 118–121, 125–127, see also ancient grammatical terminology adeo 118–120 adjectives – deverbal 249, 253 – verbal 42, 135, 184 aer 290–291 African Latin see Latin analogy see ancient grammatical terminology ancient commentators 106, 312–328, see also Donatus, Servius ancient grammatical terminology 351–367, 369–372 – ablative 377 – accusative 376–377 – analogy 370, 372–373, 379–381 – compounds 378–379 – etymology 371–372 – gender 377–378 – genitive 376 – nominative 369, 376 – number 377 – prefix 374 – tense 379 argument 110–132 argument structure 38, 41, 110, 111, 115, 130, 166 aspect 169–170 – ingressive 174 – progressive 174 – resultative 170, 173–174 auctor (sum) 135–148 augeo 135–148 bilingualism 295–297, 299, 304 bracchium 288–289 Caland root 258, 262n8, 267n15, 270, 272 calque 208, 210, 220, 221, 283, 319, 369, 370 https://doi.org/10.1515/9783110647587-028

cannabis 281–282 castanea 283 chorda 286–287 clausula 351–367 coma 290 complement of place 110, 119 complementary distribution 37 compounds 228, 232, 234, 296, 304–310, see also ancient grammatical terminology concrete nouns see nouns concurro 129 consīdero 161–166 constituo 5–6 contero 153 convenio 129, 130 correptio iambica 440 Correspondence Analysis 336 cresco 150–152 curse tablets 30, 401–416 dative 43 decurro 116 deadjectival see verbs deerro 117 denominal see verbs derivational morphology 225, 228, 233, 241 desīdero 161–177 deverbal see adjectives, nouns, verbs diphthong 323, 441, 443–448 discerno 152 distributional semantics 71 Donatus 107n12, 312–328 dubenus, duenos, duhenus 178–190 emic approach 22 epigraphic texts 96, 419–434, 437, see also curse tablets, writing-tablets etymology 135–147, 149–158, 161–176, 178–189, 192–204, 206–221, 271, 289, 301, 316, see also ancient grammatical terminology, folk etymology exeo 114–115 extensité 52

454

Index

folk etymology 162, 286 fors 192–204 fortuna 192–204 genitive 43, see also ancient grammatical terminology glossaries and glosses 105–108, 179–180 grammatical texts 297–298, 354–357, 363–366, 368–382, see also ancient grammatical terminology Greek and Greek influence 6, 277–291, 294–310, 312–327, 351–366, 369–381 hypercorrection 404, 408–411, 415–416 iconicity 392–395 imago 393, 395 inchoative verbs see verbs incurro 121–122 Indo-European 144, 149–158, 162, 178–189, 195–197, 214 ineo 118–120 ingressive see aspect initial h- 401–416 – deletion 405–407, 413, 415 – insertion 407–413, 415 language contact 277–291, 401–416 Latin – African 7, 13, 18, 439 – agricultural 88 – Late 9–17, 294–310, 419–434 – legal 6–9, 140–143 – macaronic 329–346 – Vulgar 8, 176, 401–416, 432 Latinization 403, 412, 416 legal Latin see Latin letters of Claudius Terentianus 413, 415, 425 lexical roots 299, 302, 310 lexical semantics 192–204 lexicography and lexicology 36–48, 86–99, 101–108, 161–176 Liber glossarum 105–107 loan shift 368–382 locative 39

magia, magicus, magus 21–33 marmor 285–286 mass nouns see nouns Menander 316, 319–321 metaphor 87, 384–397 – conceptual 77–78 metaphora 384, 387, 388 metonymy 175 motion verbs see verbs mundus 206–221 nominative see ancient grammatical terminology nostalgia 161–176 nouns – abstract 60 – concrete 56 – deverbal 48 – mass 56 – noms de patrimoine 60 obambulo 123–124 obeo 121 occurro 121–122 orthography 401–416, 419–434, 436–450 percurro 116, 125 pervenio 125 phonology 401–416, 419–434, 436 Pliny the Elder 102–104 polysemy 86–99, 166 praeses 36–48 praesidens 36–48 praesideo 37–42 praesum 37–42 praetereo 127 prefix 110–131, 228–229, 234, see also ancient grammatical terminology – ab- 113 – ad- 113, 118 – com- 128 – de- 113, 164–165, 169 – ex- 113 – in- 113, 118 – ob- 113, 118 – per- 113, 124

Index

– prae- 37 – praeter- 113, 124, 127 – trans- 113, 124, 126 progressive see aspect prosody 436–450 quam multi 50–67 quantification 51–53 quantifier 50–67 quantus 50–67 quatio 156–158 Quintilian 353–354, 385–389, 393n16 quot 50–67 ratio 372 requiesco 153–154 resultative see aspect rhetorical terminology 312, 315, 324, 325 rhetorical texts 297–298, 309, 351–366, 384–397 Roman Britain 401–416 Romance languages 176, 277–291, 329–346, 438–440 saccus 284 S-Curve 424–425, 428, 433 secondary verbal derivation 149–158 semantic field 87, 165, 173, 176 Seneca the Younger 11, 385, 389–395 Servius 106–107, 301–303, 312, 314–316, 319, 321–323, 326–327 shortening of final vowels 440 single-word switches 295–304 sociolinguistics 401–416 spatha 287–288 stative verbs see verbs structural semantics 87 stuppa 282–283 suffix 140, 184, 228, 234, 244, 246, 250 – -bil- 232, 249 – -ēno- 181 – -ero- 165, 172, 174 – -sco 240–256, 258–273 – *-ti- 192–204 – -tor 225, 230 – *-tu- 192–204 – -ulus 352

455

syllable structure 436–450 syncope 150–152, 406 tantus 50–67 technical term 298, 303, 368–382 tempus 69–84, 379 Terence 312–327 textual genres 297–299 thematic vowel 259, 269, 344 Thesaurus Linguae Latinae 5n9, 23, 101–108, 162 tot 50–67 transeo 126–127 transfero, translatio, translata verba 381, 386, 388, 391, 393–395 transitivization 165 Varro 316, 368–382 Vedic periphrastic future 139 verbal adjective see adjectives verbal periphrasis 137, 139, 147 verbs – deadjectival 243, 247, 249 – denominal 159, 243, 247 – deverbal 240, 243, 247, 249 – hybrid 330–346 – inchoative 240, 258–273 – motion 110–131 – primary 149–158 – stative 243, 253, 260, 263–268, 270, 272 viesco 156 viridis 86–99 vocative 377 vowel quantity 436–450 Vulgar Latin see Latin word formation 166, 192–204, 224–237, 240–256 Word Formation Latin lexicon 224–237, 240–256 word stress 436–450 writing-tablets 401–416 – Carlisle 402, 404, 406, 407, 415 – Londinium-Bloomberg 402, 404, 405–407 – Vindolanda 402–403, 404, 406–413, 415–416