Latin Linguistics and Linguistic Theory (Studies in Language Companion Series, 12)

LATIN LINGUISTICS AND LINGUISTIC THEORY

STUDIES IN LANGUAGE COMPANION SERIES (SLCS) The SLCS series has been established as a companion series to STUDIES IN LANGUAGE, International Journal, sponsored by the Foundation "Foundations of Language". Series Editors:

John W. M. Verhaar Gonzaga University Spokane, WA

&

Werner Abraham University of Groningen The Netherlands

Volume 12

Harm Pinkster (ed.) Latin Linguistics and Linguistic Theory

LATIN LINGUISTICS AND LINGUISTIC THEORY Proceedings of the 1st International Colloquium on Latin Linguistics Amsterdam, April 1981

edited by HARM PINKSTER University of Amsterdam

JOHN BENJAMINS PUBLISHING COMPANY Amsterdam/Philadelphia 1983

© Copyright 1983 - All rights reserved ISSN 0165 7763 / ISBN 90 272 3011 0 No part of this book may be reproduced in any form, by print, photoprint, microfilm or any other means, without written permission from the publisher.

PREFACE The papers in this volume were presented at the First International Col­ loquium on Latin Linguistics, which was organized by the Department of Latin of the University of Amsterdam in April 1981. The Colloquium was de­ voted to theoretical and Latin linguistics in their mutual relationship. This vol­ ume also contains some of the prepared comments delivered at the Collo­ quium. Two of the papers that were read have been published elsewhere, viz. A.M. Bolkestein's 'Nominativus cum Infinitivo' (in: Bolkestein, A.M. e.a. 1981 Predication and Expression in Functional Grammar, London, Academic Press, p.90-106) and K. Strunk's 'Phänomene syn- und diasystematischer 'Selektion' im Latein' (in: Festschrift H. Stimm, Tübingen, Narr, 1982). Publication of this volume was facilitated by a grant from the "Faculteit der Letteren" of the University of Amsterdam. Martsen Hepkema did most of the work of preparing the papers for publication. I thank Philip Baldi, as well as the editors of the series and the publishers, for their kind cooperation. H. Pinkster University of Amsterdam Klassiek Seminarium Oude Turfmarkt 129 1012 GC Amsterdam

January 1982

T A B L E O F CONTENTS

PREFACE

v

INTRODUCTION by H. Pinkster

ix

I. METHOD R. Pfister. Kompetenz in der lateinischen Syntax

3

II. SENTENCE STRUCTURE Ph. Baldi: Speech perception and grammatical rules in Latin 11 A. Bertocchi/C. Casadio: Anaphoric Relations, Pronouns and Latin Complementation 27 G. Calboli: The development of Latin (Cases and Infinitive) 41 P. de Carvalho: Le système des cas latins: aperçu d'une théorie morpho59 sémantique R.G.G. Coleman: The structure of Latin complex sentences 73 H. Fugier. Organisation syntaxique des compléments de lieu et temps en Latin archaïque 95 (discussion: Ch. Guiraud) 115 Ch. Guiraud: L'importance du verbe dans la phrase latine 117 F. Hoff: Interrogation, interrogation rhétorique et exclamation en Latin 123 J.R. de Jong: Word Order within Latin Noun Phrases 131 (discussion: D. Panhuis) 142 Chr. Lehmann: Latin preverbs and cases 145 (discussion: P. de Carvalho) 161 M. Maraldi: New Approaches to Accusative Subjects: Case Theory vs. Raising 167 J.-P. Maurei: Les relatives en Latin: "Raising" ou "Matching"? . . . 177 (discussion: Chr. Lehmann) 196 H. Pinkster: Praedicativum (quantifying adjectives and adjectives denoting physical or mental state) 199

viii

TABLE OF CONTENTS

III. TEXT STRUCTURE M. E. Hoffmann'. Conversation openings in the comedies of Plautus . 217 IV. SEMANTICS A. Orlandini'. Une analyse sémantique et pragmatique des pronoms indéfinis en Latin 229 V. MORPHOLOGY/PHONOLOGY O.S. Pillinger: Latin Degemination: an Autosegmental Approach . . 243 (discussion: E. Vester) 260 Chr. Touratier: Analyse d'un système verbal (les morphèmes gramma­ ticaux du verbe latin) 261 (discussion: R. Coleman) 282 VI. LATIN LINGUISTICS AND LANGUAGE LEARNING E. Heilig: Zur Verwendung von dependenzgrammatischen Elementen im Lateinunterricht des Gymnasiums 287 G.M. Seligson: Latin at Michigan 1951-1981 295 INDEXES Author Index Subject Index

301 305

INTRODUCTION

H. PINKSTER University of Amsterdam

0. The articles in this volume are representative of the kind of research that is currently being carried out in the field of Latin linguistics. Most deal with syn­ tax (or rather: sentence structure), though they vary with respect to their em­ phasis on theory or description. They also vary with respect to the grammati­ cal framework within which they are formulated (with some preponderance of transformational generative approaches) and their degree of theoretical coherence. The papers are well-informed about the major developments in comtemporary linguistics and make intensive use of recent methods and types of argumentation. The reader may quickly ascertain this by taking a glance at the Author and the Subject Indexes. In this introduction I will not attempt to summarize the contents of the in­ dividual papers. For this the reader may consult the summaries that the au­ thors have attached to their papers. Instead, since this volume is published in a series of a general character, I will briefly review the present state of Latin lin­ guistics as compared with developments in contemporary linguistic theory and research. I will start with a section on the question whether it is possible to conduct up-to-date linguistic research for Latin at all. Next, I will briefly sketch the impact of recent linguistic theories on Latin linguistics in general. In the third section I will outline the possible interest the papers of this volume may have for linguists working on languages other than Latin. 1. Under the influence of generative grammar, in which much importance is attached to the intuitions of native speakers, syntacticians and semanticists of Latin and other extinct languages have endeavoured to prove that their re­ search can be done in a way that is not necessarily inferior to research on lan­ guages for which native speakers are available. Thus, R. Lakoff (1968: 3-8) states that all human languages by and large have the same basic semantic structure and possess comparable devices for expressing particular semantic

X

H. PINKSTER

relations, which explains why we can understand Latin sentences at all. This assumption is as appealing as it is difficult to prove. With regard to the subject of her book (on Latin Complementation), for example, she observes — with­ out much argument — that all languages have subordinating devices (: 6). Now, much diachronic work on Latin syntax is based on the assumption that Latin must have developed from a stage of language in which subordination did not exist, which implies that Indo-European was exceptional from a typo­ logical point of view and may constitute a glaring exception to her claim (see also Dressier (1969)). Some attention has been given to the matter in recent typological research. Thus, Givón (1978: 298) draws attention to 'some lan­ guages extant to this day ... where one could demonstrate that subordination does not really exist'. However, his entire argument has a strong phylogenetic bias, and he refers only to evidence from the Chuave language (see MallinsonBlake's critical comment (1981: 431-33). The illustration may show that in many instances our knowledge about what occurs in human language is hardly sufficient.1 But, even if our knowledge of what occurs in language would be more precise, the question remains which conclusions might be drawn for specific languages and specific historical processes. See Comrie (1981, chs. 1 and 10) and, from a different angle, Hawkins (1982) on this problem. Other linguists have argued that the lack of a native speaker's compe­ tence can be compensated for, to a large extent, by the use of a corpus of adequate size and variation (Ehlich 1981; compare also Pinkster (1972: ch. 1)). In fact, the Latin corpus is relatively accessible and has been investigated for linguistic purposes longer and perhaps more intensively than most modern languages. Outside Latin linguistics, the use of corpora is spreading rapidly.2 Moreover, the very concept of 'native speaker competence' has received con­ siderable criticism (see several of the articles in Coulmas, ed. (1981)). The conclusion might be, then, that the question of whether up-to-date linguistic research is possible for Latin can be answered in the same way as for any other language: Yes, it can, to the extent that the linguist can operationalize his questions in the light of the corpus available. Pfister's article (p. 3ff.) goes even further. He argues that the Latin syntactician-semanticist may with some confidence rely on his or her own competence, which, in­ deed, many Latinists have been able to do in the past. Most of the tests that are applied in modern language study are useful for a linguistic analysis of Latin as well. I mention the deletion and addition tests (which may be formulated in a way that makes them applicable to a corpus; compare Happ (1976: 347-428)), co-occurrence restrictions, coordination and

INTRODUCTION

Xl

juxtaposition, the use of question words, etc. I refer to the contributions in this volume by Fugier and Orlandini. Fugier advances empirical arguments for postulating three different classes of temporal adverbs based on a detailed analysis of distributional properties. Orlandini bases her analysis of, for in­ stance , quispiam ('someone') on the observation that this pronoun is absent in declarative sentences with a perfect tense. As far as I know the question of which linguistic model is most profitable for the description and analysis of this type of language has not received sys­ tematic attention. Baldi (p. 12) expresses some doubt about the usefulness of transformational grammar for syntactic research in Latin, in view of its re­ latively free word order. Boas (1975) goes in the same direction. From the point of view of continuity with earlier syntactic work, those linguistic models might be the easiest to apply which recognize syntactic, semantic and pragmatic functions. In this volume the relevance of pragmatic notions is demonstrated by De Jong. Compare also Heilig on the importance of pragmatics for the comprehension of Latin sentences (p. 290), as well as Bolkestein (1981). 2. As far as I know the influence of American Structuralism on Latin linguis­ tics has been very limited.3 European Structuralism, on the other hand, with its various branches such as the Prague, Copenhagen and Geneva schools, has been influential, and still is. I refer to several studies on the case, tense and mood systems in the well-known Jakobsonian tradition. A survey of studies on case can be found in Calboli (1972) and Serbat (1981). Calboli (1966) has a survey of studies on mood. Especially in Spain, this form of structuralism has had a long tradition (e.g. Rubio (1966) and Mariner Bigorra (1957; 1965)). Of the more recent approaches, Tagmemic Grammar seems to be negligible. In France, Guillaume's theory of 'psychomécanisme' has influenced several linguists. Here I mention only Vairel (1978) on the Latin perfect and imper­ fect . Another French theoretical linguist, Tesnière, has been rediscovered for Latin linguistics, it appears, by Dressier (1970). Both directly and indirectly, via the German Valenzgrammatik, Tesnière's influence is considerable, espe­ cially in Germany (Happ 1976). In France, on the other hand, there is some resistance to some of the Tesnièrian tenets, for example to the pivotal position attributed by him to the predicate (compare, for example, Guiraud in this vol­ ume) . Case grammar as such has been of little influence, although a number of studies have been published within this framework, especially in the U.S., e.g. the unpublished dissertation by Binkert (1970) (summarized in Calboli (1972: 268-95). See also Baldi (1976) and the Italian dissertation by Murru (1977). Elements of both Valency Grammar and Case grammar, together

xii

H. PINKSTER

with pragmatic elements, are, of course, present in the various forms for func­ tional grammar — e.g. Dik's Functional Grammar (1978) and the Prague School. Latin studies within these frameworks are Bolkestein (1981) and Panhuis (1982) on word order. The influence of Montague Grammar and re­ lated approaches seems to be very small. The typological approach, referred to already in section 1, seems to be rapidly growing in popularity. I refer to Adams (1976) on word order. As a corollary there is also increasing interest in diachronic syntax, especially the relationship between Indo-European and Latin and between Romance languages and Latin. An illustration of the latter is Plank's admirable study (1979) on the development of the Latin case sys­ tem. Transformational Generative Grammar, in its development over the years, has been the most influential of all in Latin linguistics (which, however, is not to say that it has also contributed most to new insights into Latin) both in the field of syntax and in the field of morphophonology. Examples in these fields are Lakoff (1968), Pepicello (1977), Calboli, ed. (1980), Zirin (1970) and Nyman (1977). This influence (both direct and indirect) is, of course, strong in the USA (there are many unpublished dissertations), but it is also considerable in Europe, especially in Italy and France. Apart from the above-mentioned studies, which are closely connected with specific linguistic models, there are, of course, several studies which be­ nefit from the findings in one or more of these models, but develop ideas of their own. Lehmann's Latein mit abstrakten Strukturen (1973) is a good exam­ ple as well as several Swedish studies (e.g. Bergh (1975), Helander (1977)). The same holds for publications by Latin linguists at the University of Amster­ dam, e.g. Bolkestein (1980) and Pinkster (1980).4 Several of the theories mentioned above have been used for the construc­ tion of educational material for secondary schools or universities. An Ameri­ can structural approach is Sweet (1957). A Tagmemic approach can be found in O'Brien (1965). Both works are intended for university students. The Cam­ bridge Latin Course in Britain (intended for secondary schools), which has also been translated into Dutch, is based on insights from an earlier form of Transformational Grammar, though not very consistently so. Happ and his colleagues have made some effort to improve teaching materials in Germany from a Dependency point of view (for instance Dönnges-Happ (1977)). Con­ tinuing Sweet and Happ's approach are Seligson's and Heilig's contributions to this volume (p.287ff. ; p.295ff.). A common problem for efforts such as these is that they are often introduced because teachers hope to use them as a defence

INTRODUCTION

xiii

against the erosion of the position of classical culture, applying the method without a thorough schooling in the pertinent linguistic model, and soon ar­ rive at the conclusion that these new methods are useless. This conclusion often parallels the view held by traditional philologists and by grammarians not keeping abreast of new developments in theoretical linguistics. Their in­ terest is mainly in new facts and not in new ways of describing well-known facts. However, Latin linguistics in this respect probably does not differ essen­ tially from modern language studies. 3. The relevance of this volume for contemporary theoretical linguistics is more obvious in the case of some contributions than for others. Some of the contributions are in fact formulated in terms of a particular theoretical framework or are critical with respect to (elements of) a particular theoretical framework. Other contributions deal with topics that have attracted attention within one or more linguistic models. Let me start with those studies that con­ centrate on sentence structure. First, there are a number of articles in a transformational framework, viz. Bertocchi-Casadio (p. 27ff.), Calboli (p. 41ff.), Maraldi (p. 167ff.) and Maurei (p. 177ff.). The first three of these are inspired by Chomsky (1981) and other recent generative work. Bertocchi-Casadio deal with anaphora, which has been a central topic in the current version of TG lately (e.g. Chomsky 1981: 183ff; 289ff). A textbook treatment can be found in Rad­ ford (1981: 364-95)). Maraldi explores how the Accusative and Infinitive con­ struction can be accounted for in Chomsky's latest model (1981:188). Calboli also devotes part of his article to this problem. Both approaches differ radically from an earlier transformational analysis of this construction in terms of rais­ ing (Pepicello 1977), which has been criticized by Bolkestein (1979) and Pil­ finger (1980) (see also Comrie (1981) and Wales (1982)). Maraldi and Calboli's proposals may be compared with Baldi's contribution in this volume (p. 11ff.). He argues that the accusative case form, being typically the case form for objects, cannot be explained as a trace of the objecthood of the constituent somewhere in deep structure (the constituent according to him is in fact a sub­ ject); instead, the accusative may be explained as due to the occurrence of the constituent in a linear position for which normally the accusative is used; that is, surface structure predominates over deep structure. It is well known that relative clauses have attracted the interest of lin­ guists of various persuasions. Maurei, for example, is interested in how rela­ tive clauses can be explained within the framework of transformational gram­ mar as developed, for example, in Milner (1978). He compares two different

xiv

H. PINKSTER

types of analyses, viz. the raising and matching analysis. He argues that the Latin facts cannot be accounted for by adopting the raising analysis. Calboli's approach to relative clauses is rather in a diachronic and typological vein. I refer to Mallinson-Blake (1981: ch. 5) as well as to Comrie (1979: ch. 7). Relative clauses are often explained diachronically by assuming that they have originated from former sentence types with indefinite and/or interroga­ tive pronouns (compare also Lyons 1977: 78-79). Hoff's article (p. 123ff.) bears on this in that he focuses on the distributional difference of wh-words in interrogative and exclamatory sentences. As for the other contributions that represent, or critically discuss, a par­ ticular school of contempory linguistics, I mention De Carvalho's contribu­ tion on the Latin case system in a Guillaumean framework; Guiraud's article on the role of the verb in the sentence, which, in his view, is overestimated in Valency grammar. He also gives a brief survey of the history of the concept of 'verb'/'predicate'. De Jong's article on Word Order in the Latin Noun Phrase is written in a functional framework (Dik 1978) and demonstrates the need to explain the relative order of attributes with respect to their head noun by using concepts such as 'focus' and 'topic'. Of the other syntactically oriented articles, Baldi's clearly has theoretical implications (p. llff.). He advances a number of observations in favour of assuming a more prominent role for surface structure. The use of the accusa­ tive for the subject of the Accusative and Infinitive-construction is just one example. Coleman (p. 73ff.) has developed a new method of describing Latin sentences in terms of complexity. Fugier's analysis of local and temporal adverbs is eclectic, using for the most part familiar ways of syntactic analysis such as distributional restrictions and possibilities of such words. Her treat­ ment of frequency adverbs as some sort of quantifiers of action deserves at­ tention (p. 95ff.). Lehmann's article on preverbs is, as far as I know, the first systematic treatment of the relationship between prefixing and case assign­ ment (p. 145ff.) It may also have relevance for typological studies on ergativity. My own article (p. 199ff.) may be interesting for linguists working on pre­ dication and quantifier floating. Latin seems to have had more freedom in this respect than, for example, English. Of the non-syntactic contributions Hoffmann's is inspired by recent dis­ course studies (p. 217ff.). Orlandini, in a logico-semantic approach, em­ phasizes the role of presupposition in describing the Latin system of indefinite pronouns. Pillinger (p. 243ff.) discusses the obscure phenomenon of degemination of double consonants in Latin, which is difficult to explain in a purely

INTRODUCTION

XV

segmental approach such as the one prevalent in generative grammar. In­ stead, Pillinger advocates handling problems like these in a suprasegmental (or rather: 'autosegmental') framework. Touratier (p. 261ff.) once again takes up the challenge of analysing the Latin verb system in such a way as to yield a one-to-one correspondence between formal units and meaningful ele­ ments, following Zellig S. Harris for the formal analysis and mainly recent French studies for the semantic elements (compare Matthews 1972). 4. The above paragraphs are not intended to suggest that this volume has little interest for the Latin philologists and does not contain hitherto unknown obser­ vations of a factual nature for the non-specialist. The volume as a whole also contributes to a refinement of our knowledge of the structure of the Latin lan­ guage, independently of theoretical preoccupations but, I should like to stress, thanks to the application of a variety of new methods.

NOTES

1) For the use of typological considerations for the reconstruction of extinct languages cf. also Dressier (1971) and Strunk (1977). 2) I refer to the corpus based work of Quirk e.a. (1972); Huddleston's study on The Sentence in Written English' ... 'based on an Analysis of Scientific Texts' (1971); Schiffrin's article on the his­ torical present (1981) and to such corpora as the English Brown Corpus and the Dutch 'Eindhovense Corpus'. 3) However, see Seligsor's article on Sweet (1975). Articles by R.A. Hall (1946) and F.W. Householder (1947) are exceptional. 4) The reader is also referred to Strunk's Introduction to a collection of papers on Latin linguis­ tics that are in one way or another illustrative of the evolution (1973). The best bibliography for Latin linguistics can be found in the journal Die Sprache.

BIBLIOGRAPHY Adams, J.N. (1976), 'A typological Approach to Latin Word Order', IF 81, 70-99 Baldi, Ph. (1976), 'Remarks on the Latin R-form Verbs', ZVS 90, 222-57 Bergh, B. (1975), On Passive Imperatives in Latin, Uppsala, Almquist & Wiksell Binkert, P.J. (1970), Case and Prepositional Constructions in a Transforma­ tional Grammar of Classical Latin, Ph.D. Michigan

xvi

H. PINKSTER

Boas, H.U. (1975), Syntactic Generalizations and Linear Order in Transfor­ mational Generative Grammar, Tübingen, Narr Bolkestein, A.M. (1979), 'Subject-to-Object Raising in Latin', Lingua 49, 1534 . (1980), Modal Verbs, Assen, van Gorcum . (1981), 'Embedded Predications, Displacement and Pseudoargument Formation in Latin', in: Bolkestein, A.M. e.a. (1981) Predication and Ex­ pression in Functional Grammar, London, Academic Press, p. 63-112 Calboli, G. (1966), Tmodi nel verbo greco e latino', Lustrum 11, 173-349; 13, 405-511 . (1972), La Linguistica Moderna e il Latino, I Casi, Bologna, Pàtron . (ed.) (1981), Papers on Grammar I, Bologna, CLUEB Chomsky, N. (1981), Lectures on Government and Binding, Dordrecht, Foris Comrie, B. (1979), Language Universals and Linguistic Typology, Oxford, Blackwells . (1981), T h e theoretical significance of the Latin accusative and infini­ tive: a reply to Pillinger', J . L . 17, 345-49 Coulmas, F. (ed.) (1981), A Festschrift for Native Speaker, The Hague, Mouton Dik, S.C. (1978), Functional Grammar, Amsterdam, North-Holland (repr. Dordrecht, Foris (1981)) Dönnges, U. & H. Happ (1977), Dependenzgrammatik und Lateinunterricht, Göttingen, Vandenhoeck & Ruprecht Dressier, W. (1969), 'Eine Textsyntaktische Regel der indogermanischen Wortstellung', ZVS 83, 1-25 . (1970), 'Comment décrire la syntaxe des cas en Latin', R.Ph. 44. 25-36 . (1971), 'Ueber die Rekonstruktion der indogermanischen Syntax', ZVS 85,5-22 Ehlich, K. (1981), 'Native Speaker's Heritage. On Philology of 'Dead' Lan­ guages', in: Coulmas, ed. (1981) 153-65 Givón, T. (1979), On Understanding Grammar, New York, Academic Press Hall, R.A. (1946), 'Classical Latin Noun Inflection', Cl.Ph. 41, 84-90 Happ, H. (1976), Grundfragen einer Dependenzgrammatik des Lateinischen, Göttingen, Vandenhoeck & Ruprecht Hawkins, J.A. (1982),'Cross-category harmony, X-bar and the predictions of 'markedness', J.L. 18, 1-35 Heiander, H. (1977), On the Function of Abstract Nouns in Latin, Uppsala, Almquist & Wiksell

INTRODUCTION

xvii

Householder, F.W. jr. (1947), 'A descriptive analysis of Latin declension' Word 3, 48-58 Huddleston, R. (1971), The Sentence in Written English, Cambridge, U.P. Lakoff, R.T. (1968), Abstract Syntax and Latin Complementation, Cam­ bridge Mass, M.I.T. Lehmann, Chr. (1973), Latein mit abstrakten Strukturen, München, Fink Lyons, J. (1977), Semantics, Cambridge, U.P. (2 vls.) Mallinson, G. &B.J. Blake (1981), Language Typology, Amsterdam, NorthHolland Mariner Bigorra, S. (1957), 'Estructura de la categoria verbal "modo" en latin clasico", Emerita 25, 449-86 . (1965), 'Notion básica de los modos en il estilo indirecto latino', Emerita 33,47-59 Matthews, P.H. (1972), Inflectional Morphology, Cambrigde, U.P. Milner, J.C. (1978), De la syntaxe à I'interpretation, Paris, Seriel Murru, F. (1977), Studio di un modello semantico-generativ o per l'insegnamento del latino. Il problemo dei Casi. Diss. Torino Nyman, M. (1977), 'Where does Latin sum come from', Lg. 53, 39-60 O'Brien, R.J. (1965), A Descriptive Grammar of Ecclesiastical Latin Based on Modern Structural Analysis, Chicago, Loyola U.P. Panhuis, D.G. (1982), The Communicative Perspective in Latin Word Order, Amsterdam, Benjamins Pepicello, W. (1977), 'Raising in Latin', Lingua 42, 209-18 Pillinger, O.S. (1980), 'The Accusative and Infinitive in Latin: a refractory complement clause', JL 16, 55-83 Pinkster, H. (1972), On Latin Adverbs, Amsterdam, North-Holland .(1980), 'Naamvallen in een valentiegrammatika', Lampas 13, 111-29 Plank, F. (1979), 'The functional basis of case and declension classes: from Latin to Old-French', Linguistics 17, 611-40 Quirk, R. e.a. (1972), A Grammar of Contemporary English, London, Longmans Radford, A. (1981), Transformational Syntax, Cambridge, U.P. Rubio, L. (1966), Introducción a la sintaxis estructural del latin. I Casos y preposiciones, Barcelona, Ariel Schriffrin, D. (1981), 'Tense variation in narrative', Lg. 57, 45-62 Serbat, G. (1981), Cas et fonctions, Paris, PUF Strunk, K. (ed.) (1973), Probleme der lateinischen Grammatik, Darmstadt, W.B.G.

xviii

H. PINKSTER

, (1977), 'Heterogene Entsprechungen zwischen indogermanischen Sprachen', ZVS 91, 11-36 Sweet, W.E. (1957), Latin: A Structural Approach, Michigan, Ann Arbor, U.P. Vairel, H. (1978), 'La valeur de l'opposition infectum/perfectum en Latin' REL 56, 380-412 Wales, M.L. (1982), 'Another look at the Latin accusative and infinitive', Lingua 56, 127-52 Zirin, R.A. (1970), The Phonological Basis of Latin Prosody, The Hague, Mouton

I. METHOD

KOMPETENZ IN D E R LATEINISCHEN SYNTAX

R A I M U N D PFISTER Universität München, Institut für Klassische Philologie ZUSAMMENFASSUNG Das Fehlen des native speakers macht die Anwendung von Methoden der modernen Linguistik, die die Kompetenz voraussetzen, problematisch, aber nicht unmöglich. Kompetenz ist die Voraussetzung auch für das heute noch praktizierte Lateinsprechen und Lateinschreiben, für didaktisch motivierte Übersetzungsübungen ins Lateinische und für die Textkritik. Einerseits begünstigen die Besonderheiten des lateinischen Corpus den Erwerb einer Kompetenz, anderseits schwinden durch die Entwicklung des Bildungswesens die Voraussetzungen für ihren Erwerb. Deshalb solite auch der Forscher kritisch gegenüber der eigenen Kompetenz und ihrer Anwendbarkeit sein und sich nach Möglichkeit auf das Corpus stützen. Die folgenden Überlegungen sind nicht auf abstrakte Theorie, sondern auf konkrete Praxis gerichtet. Ausgangspunkt ist die Frage, ob und wieweit Methoden der generativen Grammatik auf das Latein angewendet werden können, für das der native speaker fehlt. R. Lakoff meinte (1968: 75), daß auch der non-native speaker die Grammatikalität eines Satzes beurteilen könne, wenn er die Sprache gut kennt. J. Untermann meinte dagegen (1971: 180), daB bei toten Sprachen die Verfahren der generativen Grammatik komplizierten und letzlich vielleicht ganz unrationalen Zusatzüberlegungen unterworfen werden müßten. Doch die Forscher, die in der Folgezeit die gene­ rative Grammatik auf das Lateinische angewandt haben, vor allem unter der Ägide von Herrn Pinkster in Amsterdam und Herrn Calboli in Bologna, ha­ ben im Fehlen des native speakers kein unüberwindliches Hindernis gesehen. Es ware unfruchtbar, an die Spitze eine präzise Definition der Kompe­ tenz zu stellen. Die Kompetenz des idealen native speakers im Sinne Choms­ kys, der alle in einer Sprache möglichen Sätze generieren und über deren

4

RAIMUND PFISTER

Grammatikalitát entscheiden kann, ist ein ideales Konstrukt, das in keinem native speaker verwirklicht ist. Kompetenz wird in der angewandten Sprachwissenschaft und in der Didaktik für die jeweils verschiedene Sprachfähigkeit des einzelnen Individuums verwendet, und mit einer solchen Kompetenz haben wir es im folgenden zu tun. Die beschränkte Zeit erfordert eine Konzentrierung auf die Kompetenz des Syntaktikers. Anderes kann nur gestreift werden. So die Sprecherkompetenz. In der Festschrift for Native Speaker (F. Coulmas, ed.) (die für unser Problem kaum etwas ergibt) befindet sich für das Latein die Bemerkung (1981: 58): "No present day speakers exist." Dabei wird ausdrücklich nicht nur ein native speaker, sondern überhaupt ein speaker geleugnet. Es gibt aber Lateinsprecher in größerer Zahl. Auch auf die Lateinschreiber will ich nicht eingehen. Übersetzungen in das Lateinische werden in der Bundesrepublik Deutschland heute noch von den Studenten gefordert; auch an den Gymnasien wird teilweise noch in das Lateinische übersetzt. Hier muß von eigenen Erfahrungen berichtet werden. Denn wenn nach der Kompetenz in den ein­ zelnen Köpfen gefragt wird, dann kann man über seinen eigenen Kopf noch am besten Auskunft geben. Ich hatte bei meinen Übungen in den letzten 20 Jahren etwa 8000 Übersetzungen von Studierenden auf Grammatikalität und Akzeptabilitât zu prüfen. Die deutschen Textvorlagen sind meist manipulierte Kunstprodukte, die auch dem durchschnittlichen Studenten eine akzeptable lateinische Wiedergabe ermöglichen sollen. Diese Lehrtätigkeit wird gewöhnlich den unteren Rängen der akademischen Hierarchie überlassen, und ich muBte mir auch sagen lassen: Was du da machst, ist nicht wissenschaftlich; denn wir können die Akzeptabilität der von den Studenten produzierten Sätze nicht beurteilen. Von diesem Verdikt waren aber auch alle lateinischen Syntaktiker betroffen, die nicht rein strukturalistisch nur das Corpus analysieren, sondern generativ-transformationell arbeiten. Nun zu meinem Erfahrungen: In weiten Bereichen, vor allem in der Or­ thographie, in der Formenlehre und bei elementaren syntaktischen Erscheinungen wie der Kongruenz kann der Korrektor auf Grund seiner Kompetenz ohne weiteres den Rotstift ansetzen. Aber in vielen Fallen, etwa in der Rektionslehre oder bei der Wortwahl, muß der Korrektor oft mühsam die Handund Hilfsbücher oder die Texte selbst heranziehen, um über die Akzeptabili­ tät entscheiden zu können. SchlieBlich bleibt ein Rest, wo auch der Rückgriff auf das Corpus keine Entscheidung ermöglicht und der Korrektor nur Empfehlungen geben, aber die vorgelegte Formulierung trotz seiner Bedenken

KOMPETENZ

5

nicht mit durchschlagenden Gründen als inakzeptabel erweisen kann. Deshalb wird die Verfahrensregel eingeführt: Der Student hat eine Formulierung zu suchen, deren Akzeptabilităt er mit dem Corpus begründen kann, und er darf nicht dem Korrektor die Beweislast für die Inakzeptabilität zuschieben. Die Kompetenz des Korrektors nimmt zwar im Laufe der Jahre zu, aber sie hat ihre Grenzen. Bei der Annahme einer Inakzeptabilität wird man immer vorsichtiger. Ăhnliche Probleme hat man als Forscher gegenüber unbelegten Beispielsätzen bei modernen Syntaktikern. Bei manchen angeblich akzeptablen Sätzen kann man Zweifel bekommen, ohne diese stringent begründen zu können. Aber anders als gegenüber Studenten kann man Kollegen schwer schlicht und einfach sagen, daß man ihre Kompetenz anzweifelt. Bei als inak­ zeptabel bezeichneten Sătzen kann man mühsam nach einem Gegenbeispiel suchen; doch ist mit zufälligen Lücken im Corpus immer zu rechnen. Aber es gibt hier wie bei den Übersetzungsübungen einen groBen Bereich hinreichend evidenter Falle, und wenn die nötige Vorsicht gegeben ist, kann man dem Arbeiten mit unbelegten Beispielen ebensowenig die wissenschafliche Qualifikation absprechen wie den Übersetzungsübungen. In der Klassischen Philologie haben wir die Textkritik. Sie benützt die im Corpus gegebenen Argumentationsmittel, aber die letzte Entscheidung trifft doch die Kompetenz. Vom Kritiker werden gefordert Entscheidungen über die grammatische oder semantische Akzeptabilităt bei der Feststellung von Korrupturen und die Produktion von akzeptablen Sätzen bei der Emenda­ tion. Die an den Kritiker von den Lehrbüchern vor allem des 19. Jahrhunderts gesteilten Anforderungen erinnern an die für einen hochrangigen native speaker. Es wird gesprochen von Intuition, feinem Gefühl, einem künstlerischen Trieb, der ohne Reflexion sicher das Richtige trifft. Die Konjektural­ kritik hat oft der eigenen Kompetenz zu viel zugetraut und neben Glänzendem auch Wertloses hervorgebracht. Heute ist man eher ins Gegenteil verfal­ len und sieht Überliefertes auch dann für akzeptabel an, wenn es das schwerlich ist. Jedenfalls ist das Exercitium der Textkritik auch für den Syntaktiker eine gute Schulung, z.B. schon deshalb, weil bei den Konjekturen unterschieden wird zwischen evident, wahrscheinlich, annehmbar und möglich, was davor warnt, zu selbstsicher nur zwischen akzeptablen und inakzeptablen Sät­ zen zu unterscheiden. Den Syntaktiker interessiert besonders die Frage, wieweit die beschränkte Kompetenz des Latinisten ausreicht zur Anwendung bestimmter heuristischer Operationen (auch Tests genannt). Sie sind nicht speziell an die

6

RAIMUND PFISTER

generative Transformationsgrammatik gebunden. In geeigneten Fallen wurde im Sprachunterricht schon seit langem generiert und transformiert. Hans Glinz hat solche Operationen schon seit 1952 zunächst ohne Kenntnis der amerikanischen Linguistik auf das Deutsche angewandt. Der grundsätzliche Nutzen für die Wissenschaft wie für den Sprachunterricht ist kaum zu bestreiten; problematisch kann die Anwendung im Einzelfall sein. Wesentlich ist, daß der Forscher (oder Lehrer) mit eigener Kompetenz Sätze des Corpus ädert oder eigene Sätze produziert und dabei Entscheidungen trifft. Diese können unter anderem betreffen 1) die Akzeptabilität. Teilweise wird schlechthin von acceptable (oder wellformed) gesprochen. Mir scheint es — trotz der engen Beziehungen zwischen Syntax und Semantik — aus methodischen Gründen besser, mit Heinz Happ zu scheiden zwischen 'grammatisch' (d.h. syntaktisch annehmbar), 'seman­ tisch richtig' und 'logisch richtig'. Ich selbst unterscheide in meinen Vorlesungen zwischen 'grammatisch', 'logisch richtig' ("Dreiecke sind kongruent" ist ohne die nötigen Restriktionen logisch nicht richtig) und 'dem Kommunikationszweck adaquat' (ein Telegramm "Ich komme" ist grammatisch und lo­ gisch annehmbar, kann aber seinen Kommunikationszweck nur erfüllen, wenn auf beiden Seiten die nötigen Präsuppositionen gegeben sind). 2. die Ambiguität. Sie spielt im Unterricht seit je eine Rolle und wird gern mit zweideutigen Orakelsprüchen exemplifiziert. Der orthodoxe Strukturalist kann mit ihnen wenig anfangen. 3. die Identität in einer Tiefenstruktur bzw. das Vorliegen einer Paraphra­ se. Beides unterscheidet sich danach, ob man auf eine gemeinsame Tiefen­ struktur zurückgreift oder zwei Strukturen auf der Oberfläche vergleicht, etwa bei der Vergleichung eines Satzes im Aktiv mit dem entsprechenden im Passiv oder einer Partizipialkonstruktion mit einem entsprechenden Nebensatz. Hier kann die notwendigerweise gröbere Betrachtungsweise des Gram­ matikers und des Lehrers den Widerspruch des feinsinnigeren Philologen provozieren, dessen Blick sich mehr auf die Verschiedenheit der Struktur als auf eine Indentitat richten wird. Insbesondere müssen wir fragen, wieweit die beschränkte Kompetenz ausreicht zur Anwendung der an modemen Sprachen entwickelten Testverfahren. Hierher gehören die Deletion und die ihr entgegengesetzten Inser­ tion, die vor aliem für die Dependenzgrammatik und die Scheidung von Nu­ cleus und Peripherie bedeutsam sind, die Substitution, die zur Feststellung von Satzgliedkategorien dient, und die Permutation, die im Lateinischen kaum anwendbar ist. Fragetests werden im Lateinunterricht seit Jahrhunder-

KOMPETENZ

7

ten angewandt. Sie sind ein Spezialfall der Pronominalisierung, die in ihren anderen Formen mehr als üblich eingesetzt werden könnte. Möglichkeiten und Grenzen der Tests sind von H. Happ (1976) so eingehend untersucht, daß ich hier nicht weiter darauf einzugehen brauche. Hinzuweisen ist jedoch auf die grundsätzlichen Feststellungen Happs (1976: 349): die Tests sind nicht eine vom Sprachgefühl unabhängige Instanz. Für den sinnvollen Gebrauch ei­ nes Tests und für die Beurteilung seines jeweiligen Resultats ist die Kompetenz unentbehrlich. Hinzugefügt sei noch: Die Vorstellung, die Ergebnisse der sogenannten traditionellen Grammatik könnten jetzt mit einem untrüglichen Instrumentarium objektiv verifiziert oder falsifiziert werden, würde der sehr komplexen Sprachwirklichkeit nicht gerecht. Für einen seiner Natur nach unexakten Untersuchungsgegenstand, wie es die Sprache ist, ware eine Untersuchungsmethode, die völlige Exaktheit anstrebt, nicht adäquat. Der fehlende native speaker ist für den Latinisten ein Handikap. Doch wird dies durch besonders günstige Umstande gemildert. Das Corpus der Kunstprosa Ciceros und Căsars bietet einen synchronen Querschnitt, der stark normalisiert und ohne geographische und soziale Differenzierung ist. Die Inhalte sind zwar beschränkt, aber doch über Politik, Historie und Phi­ losophie breit gestreut. Bleibt man innerhalb dieses Rahmens, der, wie das gesamte Corpus des Lateinischen, durch eine Überfülle von Einzeluntersuchungen und Handbücher erschlossen ist, so ist, was auch das neulateinische Schrifttum zeigt, auch dem non-native speaker der Erwerb einer ausreichenden Kompetenz durchaus möglich, vielleicht sogar leichter als bei einer mo­ dernen Fremdsprache mit einer nicht mehr beherrschbaren Überfülle des Corpus. Die Erfahrung zeigt, daB der Lehrer beim Korrigieren von Übersetzungen in eine Fremdsprache im Lateinischen auf einer festeren Grundlage steht als etwa im Englischen, wo die Frage, was man gelten lassen soll und was nicht, oft kaum zu entscheiden ist. Dazu noch eine Bemerkung: Natürlich ist die Kunstsprache Ciceros und Căsars nicht das Latein schlechthin. Aber wenn in syntaktischen Untersuchungen Beispiele von Plautus bis zu Tacitus gemischt werden, so ist das dann problematisch, wenn man sich der Methoden bedient, die auf einer synchro­ nen Sprachbetrachtung begründet sind. Denn schon für Ciceros Briefstil und für livianisches Latein ist manches akzeptabel, was für die im engeren Sinn klassische Kunstprosa nicht akzeptabel ist. Anderseits ergibt sich in Hinblick auf die Kompetenz im Lateinischen bei der Betrachtung der Bildunsgeschichte eine Paradoxie. Die Berufung auf die Kompetenz als ein legitimes wissenschaftliches Verfahren auch für die Latinistik tritt in einem Zeitpunkt auf, in dem die Bedingungen für den Erwerb die-

8

RAIMUND PFISTER

ser Kompetenz sehr ungünstig sind. Ohne auf die Verschiedenheiten in den einzelnen Landern einzugehen, kann man wohl folgendes sagen: Der langwierige Erwerb einer lateinischen Sprecherkompetenz, der am besten schon vor der Pubertät einsetzen müBte, wurde schon im friihen 19. Jahrhundert weitgehend aufgegeben, der Erwerb einer Schreiberkompetenz gegen Ende dieses Jahrhunderts. Die Übersetzung in das Lateinische ist schon seit längerer Zeit stark umstritten und geht allgemein an den Schulen stark zurück. Die Ausbildung der Latinisten an den Universitäten folgt mit einer gewissen Verzögerung dem allgemeinen Trend. Der Referent mußte 1933 als Student in München für die Aufnahme in den Oberkurs des Seminars eine lateinische Abhandlung schreiben, und in den Sitzungen wurde lateinisch gesprochen. Das ist jetzt wohl überall vorbei. Die Frage muB erlaubt sein, wie der heutige Forscher zu seiner Kompetenz kommt. Beim Erwerb der Kompetenz in einer zweiten Sprache ist entsprechend den Untersuchungen Ervins und Osgoods zu fragen, wieweit durch den Unterricht ein unabhängiges, von Interferenz der Muttersprache freies coordi­ nate language system entstehen kann oder wieweit es bei einem compound language system bleibt, bei dem Kodierung und Dekodierung über das muttersprachliche Zentrum laufen. Die Gefahr ist groB, daB intuitive Entscheidungen über die Akzeptabilität durch die muttersprachliche Kompetenz beeinfluBt werden. AbschlieBend sei gesagt: Der Latinist braucht auf operationale Metho­ den, die eine Kompetenz zur Voraussetzung haben, nicht grundsätzlich zu verzichten. Aber der Grad der wissenschaftlichen Sicherheit, der mit solchen Methoden erreicht werden kann, darf nicht überschatzt werden. Der For­ scher solite deshalb gegenüber der eigenen Kompetenz skeptisch sein und sei­ ne Darlegungen soweit wie möglich auf das Corpus stützen.

BIBLIOGRAPHY Coulmas, F. (ed.). (1981), A Festschrift for Native Speaker. The Hague: Mou­ ton. Happ, H. (1976), Grundfragen einer Dependenz-Grammatik des Lateini­ schen. Göttingen: Vandenhoeck & Ruprecht. Lakoff, R.T. (1968), 'Abstract Syntax and Latin Complementation.' Research Monographs Nr. 49. Cambridge (Mass.) and London: M.I.T. Press. Untermann, J. (1971), 'Zu Raimund Pfisters 'Thesen zu Linguistik und Sprachunterricht". Gymnasium 78, 177-182. Heidelberg: Winter.

II. SENTENCE STRUCTURE

SPEECH PERCEPTION A N D GRAMMATICAL RULES IN LATIN

PHILIP BALDI Linguistics Program and Department of Classics The Pennsylvania State University

0.

Introduction

In the past decade there has been a virtual proliferation of linguistic the­ ories and models of grammar. Many of the new approaches are developments of earlier work in generative grammar; some are philosophically derivative of the generative school, but are quite different in organization and detail; still others are emphatically non-generative in both method and philosophy. An informal accounting of the approaches being utilized in current research yields the following list: Montague Grammar, Relational Grammar, Tagmemics, Corepresentational Grammar, Daughter Dependency Grammar, Stratificational Grammar, Equational Grammar, at least two kinds of Func­ tional Grammar (represented by Dik and Kuno), Role and Reference Gram­ mar, Universal Grammar and the methodologically allied Typological Syn­ tax, Valence-Oriented Dependency Grammar, and finally Generative Gram­ mar, currently in the form of the Revised Extended Standard Theory (REST). To this survey we might also add Case Grammar, which many linguists concerned with case-marking languages like Latin and Japanese adopted in the past decade, but which turned out as a theory to have insuperable inadeq­ uacies. We should also mention Generative Semantics, which had largely the negative value of highlighting the deficiencies in the Standard Theory of Chomsky 1965, but proved to be unworkable itself, especially in matters of syntax. And of course we must finally mention Traditional Grammar, a model of limited theoretical insight, but one to which we frequently turn when the theoretical needs turn to practical issues. And we must not fail to mention Pragmatics, Speech Acts, and Text Linguistics, even though they are of less concern to the classical linguist than to those whose concerns focus on spoken language situations. (See Moravcsik and Wirth 1980 for a recent survey of cur-

PHILIP BALDI

12

rent approaches to syntax in the U.S.). With such an array of approaches available, what is the ordinary working grammarian to do, especially the grammarian who sets for himself the less lofty but nonetheless noble task of putting the theories of others to work on the language or languages he knows best? How do we choose a model of gram­ mar? The answer, of course, is that we operate in the model which yields for us the most consistently positive results; this may, unfortunately, reduce to mat­ ters of taste rather than scientific judgment. But it seems clear that linguists working on classical Indo-European languages with complex inflectional sys­ tems will require an approach which is quite different from that developed for a morphologically simple language like English. Frameworks like generativegrammar (in all the stages of its development right up throught X Syntax, see Jackendoff 1977) and Relational Grammar (see Perlmutter 1980) define grammatical functions in terms of underlying phrase-structure configura­ tions. But languages like Latin have variable word-orders and complex sur­ face morphological agreement relations which cannot be captured in phrase structure without resorting to such devices as "scrambling rules." We can find a measure of stability in the midst of all this theoretical uncer­ tainty if we are willing to be flexible in our choice of models, and grant our­ selves the luxury of working in different frameworks to explain different kinds of grammatical phenomena. Generative-theory, based on underlying gram­ matical relations, provides a workable approach to grammatical constructions like relative clause formation, deletion, and other phenomena around which the model has been developed. But if our concern is the interrelationship of tense and aspect in the verb, agreement rules, or the description of condition­ al sentences or sequence of tense rules in Latin, we will have to turn elsewhere for a workable model. 0.1

Recent Developments in Generative Theory

In the past fifteen years there has been a steady evolution of generative theory, and the model has been reworked and reorganized along several dif­ ferent lines. The changes have been brought about by many factors; among them are the following: a. The recognition that transformations often do not preserve meaning, as Katz and Postal hypothesized in 1964. The Katz-Postal hypothesis had a profound effect on early generative theory, and its eventual disconfirmation brought about a weakening of the entire notion of the transformation. Of course doing away with transformations only placed the burden of syntactic

SPEECH PERCEPTION AND LATIN SYNTAX

13

(and semantic) proof somewhere else, so in later models of grammar the Phrase Structure Rules (which Chomsky and others, e.g., Postal 1964, had worked so hard to discredit) had to be enriched with various devices. b. As the notion of the transformation was weakened, the level of sur­ face structure achieved greater importance. Semantic interpretation inched progressively closer to the surface with each revision of the theory, so that now surface structure is held to be the only level that undergoes semantic in­ terpretation (see Fiengo 1980:11-19 for a formal review of the differences be­ tween the Standard Theory, Extended Standard Theory, and Revised Ex­ tended Standard Theory). Surface syntactic filters (see Chomsky and Lasnik 1977) assign * to structures overgenerated by syntactic rules. c. As the generative model has been reorganized and surface structure has figured more prominently in syntactic and semantic argumentation, there has been a concomitant turning away from formalistic theories towards communicately oriented, functionally based models. Such a development must be viewed as a positive step in the right direction. Linguists have begun to look seriously at the issue of why syntactic rules exist in a language as we find them. As Kuno has pointed out (1980:117): "Since nothing can be gained in analyz­ ing syntactic structures by ignoring, or being ignorant of, their communicative function, every good syntactician should also be a good functional syntactician." His claim is that every theory of grammar must have a place or places where various functional constraints on well-formedness can be stated. 0.2

Perceptual Processes in a Functional Perspective

I believe that Kuno is absolutely correct in his assessment of the situation in syntax. And I believe further that many of the functional requirements of a model of grammar will be stated at the level of surface structure, where the communicative needs of both speakers and hearers intersect. One of the ways in which communicative needs are met is by adjusting surface structures so as not to create perceptual asymmetries between speakers and hearers of an ut­ terance (see Janson 1979). Following Bever 1970 and Bever and Langendoen 1971, we may call these readjustments PERCEPTUAL CONSTRAINTS. Perceptual constraints assure analyzable, and hence comprehensible surface structures. The entire notion of perceptual constraints is that the FORM of a linguistic structure is not simply a function of mechanically applied underlying syntax. Sentences and other linguistic units must be structured in such a way that they are analyzable and decodable by hearers. These perceptual constraints are sensitive to the communicative needs of speakers and hearers. Consider the following sentence:

14

PHILIP B ALDI

(1) ?*The horse raced past the barn fell. Bever claimed (1970) that this sentence and others like it were judged to be unacceptable to speakers because of rules of performance (= surface struc­ ture) , not rules of competence (= deep structure). From the syntactic point of view, (1) is no different from (2): (2) The man thrown through the window died. Both (1) and (2) can be analyzed with underlying relative clauses which have been reduced by wh + be deletion, and their overall structures are identi­ cal. But example (1) contains a surface verb form (raced) which is ambiguous as to its preterite or participial interpretation. The unmarked reading of (1) assigns a verbal interpretation to the sentence, and confusion results when the real verb, fell, is encountered at the end of the sentence. No such confusion arises with (2), since thrown can have only the participial reading. Because of the opacity of sentences like (1 ), the syntactic rules are constrained by the sur­ face form of the sentence. In this way the rules of syntax are sensitive to the physical surface form of the utterance. Such perceptually motivated constraints are not confined to matters of syntax. And before moving to a consideration of the Latin data, I should like to provide one particularly clear example of a perceptually motivated expla­ nation which is superior to a traditional formal one. Our example deals with some well known data from English morphophonology in affixed forms with automatic alternations. In English possessive, plural, and third person verbal constructions we find three automatic alternants of {S}, viz. : (3) love-/z/, cat-/s/, church-/әz/, hit-/s/, judge-/әz/ A similar pattern is found in the past-tense form {D} : (4) love-/d/, walk-/t/, create-/әd/ Along with the participial and gerundival form /-in/, with its phonostylistic variant /-in/, this covers the list of productive inflectional suffixes in English. These suffixal forms have traditionally been advanced as solid examples of progressive (perseverative) assimilation: (5) The final phoneme of the root determines the shape of the suffix, left to right. Another familiar case of assimilation comes from English prefixation. English prefixes reflect regressive (anticipatory) assimilation, working right

SPEECH PERCEPTION AND LATIN SYNTAX

15

to left. In constructions of the type found in (6): (6)impossible (← in + possible); intolerable (←in + tolerable); incon­ sistent ~ inconsistent (←in + consistent) we find that the assimilation alters the form of the prefix as follows: (7) In prefixed forms, the form of the prefix is altered in anticipation of the first phoneme of the root, working right to left. Yet, despite their frequency in the handbooks, it appears that an EXPLA­ NATION of these processes has never been offered. That is, WHY does English have two different assimilatory types, one forward and one backward, even in the same word, as in (8): (8) imbibes, commits Why instead don't we have the pattern reflected in (9): (9)/in + possible/→ *innossible /con + fer/→ *Conner /dog + s/→ *doks /church + s/→ *churchs, *churss The answer to all this is simple: English generally obeys the isomorphic principle for root forms (one form, one meaning). Assimilations are produc­ tion constraints based on ease of articulation. But they are governed by per­ ceptual constraints on morphological relatability, and they work to maintain the phonological integrity of the root in its various forms. In other words, some minor allomorphy is permitted in the few suffixes and the easily recog­ nizable prefixes, but root allomorphy, which decreases analyzability and con­ tributes to suppletion, is generally not tolerated in English. Such notions as perceptual analyzability and morphological relatability are of great importance in the analysis of inflected languages like Latin. The reason for this is that the interaction of syntactic rules and surface inflectional morphology is very complex in languages which carry out much of their syntax in morphological terms. Both parts of the grammar must have a way of inter­ acting with each other. In what follows, I will present data from four Latin constructions which I believe will demonstrate the functional interrelatedness of these two aspects of Latin grammar. In different ways, each of these con­ structions is marked by a surface structure and a predicted underlying struc­ ture which do not match up. We will see that morphology will adjust the struc­ ture of the sentence in order to make it more orderly and analyzable.

16

PHILIP BALDI

1.0

The Partitive Genitive

The Partitive Genitive (PG) is used to designate that WHOLE of which only a PART is mentioned (cf. Kühner-Stegmann 1962:423-35; Bennett 1914:1337). Some typical examples are: (10) Plaut., Asin. 516: Ecqua pars ORATIONIS dabitur mihi? 'Will some share OF THE SPEECHMAKING be given to me?' (11) Plaut., Trin. 35: Pars HOMINUM. Tart OF THE MEN.' The PG is used with neuter pronouns such as quid, aliquid together with nouns, pronouns, and adjectives used substantively: (12) Plaut., Merc. 135: Quid est NEGOTI? (NOUN) 'What (OF AN) AFFAIR is there?' (13) Plaut., Aul. 802: Ibo intro, ud quid HUIUS verum sit sciam. (PRO­ NOUN) T will go inside, so that I might know what OF THIS is true.' (14) Plaut., Merc. 471: Quid mist in vita BONI? (ADJECTIVE) 'What (OF) GOOD is there in my life?' (15) Plaut., Aul. 735: Quid demerui MALI? (ADJ.) 'What (OF) EVIL have I deserved?' (16) Plaut., Rud. 647: Quid isINIQUIIfit?(ADJ.) 'What kind OFINJUSTICE IS being done to them?' (17) Plaut., Aul. 671: Ut aliquid BONI dicam. (ADJ.) 'SO that I might say something (OF) GOOD.'

By way of comparison with examples (14)-(17), note the following: (18) Paul. Nol., Carm. 10, 288: BREVE quidquid homo est. (not *BREVIS quidquid) 'A man is something SHORT-LIVED.' (19) Varro, L. L. 9, 60: Ut aliquid SINGULARE haberent. (not * aliquid SIN­ GULARIS) 'That they might have something INDIVIDUAL.' (20) Val. Max. 7, 10: . . . ne quid HOSTILE subito molirentur. (not *quid HOSTILIS) '[afraid] that they would attempt something HOSTILE.' (21) C i c , Inv. 1, 25: Aliquid TRISTE. (not *aliquid TRISTIS) 'Something APPALLING.'

What is of course peculiar and significant in examples (18)-(21) is that the rule of PG is not in effect. Instead of the neuter genitive of the adjective, which the rule predicts and which we find in examples (14)-(17), we find in these examples the neuter nominative of the adjective in agreement with its NP head, quid, aliquid, etc. Thus, instead of:

SPEECH PERCEPTION AND LATIN SYNTAX

17

(22a) *quid DIFFICILIS we find: (22b) quid

DIFFICILE

This pattern is in contrast with pairs like: (23a) quid

NOVI

(23b) *quid NOVUM

What these examples reveal is a syntactic asymmetry: a rule of syntax applies to one group, but not to the other. As we seek an explanation for this phenomenon, several facts need to be noted. The first concerns the rule of PG itself. PG is such a powerful rule of Latin grammar that it occurs even where its structural description is not met, as in the following adverbial expressions: (24) Plaut., Pseud. 402: Quaerit quod nusquam GENTIUMST. 'He seeks that which is nowhere ON EARTH.' (25) Plaut., Amph. 336: Ubi TERRARUM sim nescio. 'I do not know where IN THE WORLD I a m . '

A second point to be made concerns the nature of case-selection in Latin generally. As is well-known, Latin is not a language with liberal case-distribu­ tion, as is Sanskrit. There is typically only one case in which a given construc­ tion is realized, e.g., dative of indirect object, accusative of direct object, ab­ lative of means, and so on. There is no rule of PARTITIVE NOMINATIVE, as exam­ ples (18)-(21) suggest. The nominative and genitive cases simply cannot be freely exchanged in this manner elsewhere in the language. Our task, then, is to explain why this interchange between the two cases and the resulting asym­ metry exist in this construction. 1.1

The Genitive of Characteristic

A similar problem of a syntactic asymmetry between morphological classes is found in the so-called GENITIVE OF CHARACTERISTIC (GC). Since I be­ lieve that the solution to the problems posed by this construction and those found in PG is along similar lines, I will provide an explanation of both below. GC is used with the copula to express property, duty, or characteristic (cf. Kühner-Stegmann 1962:452-54). A few examples are: (26) Plaut., Capt. 583: Est MISERORUMUTmalevolentes sint atque invideant bonis. 'It is THE LOT OF THE WRETCHED that they be malevolent and that they envy the good fortunes of others.'

18

PHILIP B ALDI

(27) C i c , de Or. 2, 117: TARDI INGENII est rivulos consectari, fontes re­ rum non videre. 'It is CHARACTERISTIC OF A SLOW WIT to follow up the tiny rills, but fail to discern the source of things.' (28) Cic., Fam. 4,9,2: Tempori cedere semper SAPIENTIS est habitum. 'It has always been considered THE MARK OF A WISE MAN to yield to cir­ cumstances.' GC also reveals a syntactic asymmetry similar in type to that found in the PG rule. GC can be optionally converted to a construction in which the neuter nominative of the adjective corresponding to the genitive is used predicatively, provided that the adjective is of the first and second declension. Thus we find such pairs as the following: (29a)

STULTI est hoc facere. (GC)

Tt is CHARACTERISTIC OF A FOOL to do

this.' (29b) STULTUM est hoc facere. (neut. nom.) Tt is FOOLISH to do this.' (30a) BONI est patriam amare. (GC) Tt is the MARK OF A GOOD MAN to love his country.' (30b) BONUM est patriam amare. (neut. nom.) Tt is GOOD to love one's country.' However, the alternative illustrated by the previous two pairs is not avail­ able when the adjective is of the third declension and one-termination, e.g., sapiens or prudens: (31a) SAPIENTIS est hoc facere. (GC) Tt is THE MARK OF A WISE MAN to do this.' (31b) *SAPIENS est hoc facere. (neut. nom.) Tt is WISE to do this.' (32a) PRUDENTIS est earn amare. (GC) Tt is CHARACTERISTIC OF A PRU­ DENT MAN to love her.' (32b) *PRUDENS est earn amare. (neut. nom.) Tt is PRUDENTTOlove her.' Different types of third declension adjectives, those of two-terminations (e.g., hostilis, hostile 'hostile'; regalis, regale 'kingly') have exactly the oppo­ site distribution of the sapiens, prudens type. Note the following examples: (33a) *REGALIS est pro patria mori. (GC) 'It is THE SIGN OF A KING to die for one's country.' (33b) REGALE est pro patria mori. (neut. nom.) Tt is KINGLY to die for one's country.' (34a) *HOSTILIS est noctu oppugnare. (GC) Tt is CHARACTERISTIC OF AN ENEMY to attack at night.'

SPEECH PERCEPTION AND LATIN SYNTAX

19

(34b) HOSTILE est noctu oppugnare. (neut. nom.) 'It is ENEMY-LIKE to at­ tack at night.' In examples (33)-(34) we have a pattern where the GC cannot apply, and the neuter nominative used predicatively must be employed. Let us sum up the basic facts we have uncovered concerning both PG and GC: A. PG does not apply when the adjective is of the third declension: *aliquid DIFFICILIS (cf. aliquid NOvi) aliquid DIFFICILE (cf. * aliquid NOVUM) B. GC cannot be replaced by the neuter predicative if the adjective is third declension of one termination: SAPIENTIS e s t . . . (cf. STULTI e s t . . .) *SAPIENS e s t . . . (cf. STULTUM e s t . . .)

C. GC must be replaced by the neuter nominative if the adjective is third declension of two terminations:

*REGALIS est...(cf.{ STULTI est...}) STULTUM e s t . . . REGALE e s t . . . (Cf. { * S A P I E N S e s t ... })

1.2

An Analysis of PG and GC

We can resolve this apparent chaos if we look at these constructions in terms of their surface configurations and perceptual transparency and not in terms of rule violations and lexical or morphological constraints. Linguistic structures must be perceptible and analyzable to both hearers and speakers. Adjustments which are made in surface morphological structure in order to render sentences transparent and fully intelligible must in their turn become grammaticalized; i.e., they must become part of the production system. I be­ lieve that this is exactly what happened in the PG and GC rules outlined above: A. Latin sentences like * aliquid DIFFICILIS est are ungrammatical because of the formal identity of the genitive singular of all genders and the masculine (and feminine) of the nominative singular. What we see with * aliquid difficilis is an APPARENT violation of a gender agreement rule in Latin which specifies that adjectives must agree with their head NP in gender, number, and case. Such would especially be the case af­ ter esse, where the nominative masculine interpretation would be the unmarked reading of difficilis. So rather than present a structure of opaque surface form, Latin instead prefers to violate the rule of syn-

20

PHILIP B ALDI

tax, viz. quid FACILEest?Of course, in constructions such as quid NOVI or aliquid BONI, such a confusion is not possible, so the neuter genitive form remains. Two distributional facts further support this analysis. First is the fact that Latin speakers did not avoid such constructions as quid eius 'what of it,' where no gender confusion is likely. Second is the fact that plurals of third declension adjectives are routinely found in accord with the expected partitive rule: (35) Plaut., Pers. 55: quisquam MAIORUM MEORUM (not *quisquam maioris or *quiquam maiores) 'A certain one OF MY ANCESTORS.' (36) Ter., And. 963: Nemost, quern malim, OMNIUM. (not *nemo omnis or *nulli omnes) 'There is no one OF ALL OF THEM whom I would pref­ er.' These last two examples indicate quite strongly that lexical or formal indexing will not provide an adequate or efficient account of the dis­ tribution of these adjectives. B. Latin constructions like*SAPIENSest hoc facere (cf. STULTUMEST. . . ) are not available as alternatives to GC constructions like sAPIENTIS est hoc facere (cf. STULTI est. . .) because of the formal identity of the mascu­ line and neuter nominative forms in the sapiens, prudens group. STULTUM can only be read as a neuter; sapiens, on the other hand, has the unmarked reading of a nominative masculine. In surface morphologi­ cal form, this gives it the appearance of being structurally equivalent to a construction like (37): (37) *STULTUS est hoc facere. 'He is a fool to do this.' Sentences like (33) are ungrammatical in Latin, since subject-raising cannot take place in such a construction (see Bolkestein 1979:30 for some discussion). The problem of reading sapiens as a neuter when its unmarked reading would be as a masculine, equivalent to stultus, is clearly the basis of this restriction. Thus the APPARENT violation of the rule of subject-raising results in a misperception of the SAPIENS est. . . construction as an ungrammatical structure, and it is blocked out from the grammatical system. C. Similar to (B), the analysis of such sentences as *REGALIS est pro patria mori is once again resolved in terms of surface morphology. Regalis is of course identical in the masculine (and feminine) nominative and all genders of the genitive singular. Thus a sentence like *REGALIS est is

SPEECH PERCEPTION AND LATIN SYNTAX

21

read like *STULTUS est..., which is in violation of the raising rule. Thus the conversion to the unambiguous, perceptually transparent neuter predicative REGALE est... is obligatory. 2.0

The Gerundive with Dative of Agent

The gerundive is often used in Latin to express some future necessity; in such constructions, agents are in the dative case, as in the following examples: (38) Cic., Font. 36: Si Galii bellum facere conabuntur, excitandus NOBIS erit ab inferis C. Marius. 'If the Gauls try to make war, WE will have to rouse C. Marius from the dead.' (39) Plaut., Bacch. 625: Consulandus hic MIHIST. 'He must be consoled BY ME.'

(40) Sen., Ep. 36, 4: IUVENI parandum, SENI utendum est. A YOUNG MAN

must design things, AN OLD MAN must put them to use.' The dative-of-agent rule (DA) is of course a familiar one, and needs no further explanation. What is interesting about it, however, is that the rule is frequently violated when the sentence for which it is specified is unacceptably ambiguous or difficult to parse. Note the following examples: (41) C i c , Manil. 6: Aguntur bona multorum civium QUIBUS est A VOBIS consulendum. Tt involves the property of many citizens WHOSE IN­ TERESTS must be consulted BY YOU.' (not QUIBUS ... VOBIS) (42) C i c , Plan. 78: EI ego A ME referendam gratiam non putem? 'Am I not to think that the favor ought to be returned TO HIM BY ME?' (not EI . . . MIHl)

Such instances of the ablative in place of the expected dative seem to be motivated by one fact alone: the use of the double dative would result in an ambiguous structure, so the rules gives way. Putting a restriction on double datives seems to solve the problem, but indeed it is not the fact of two datives that is the basis of this restriction, as the following examples clearly show: (43) C i c , Verr. 3, 103: Moderandum

MIHI

est

ORATIONI MEAE.

T must

Check MY SPEECHMAKING.'

(44) C i c , Cluent. 158: Cuius VOLUNTATI MIHI obtemperandum est. 'To whose WISHES I am bound to defer.' One final example of this construction which is particularly interesting is the following:

22

PHILIP B ALDI

(45) C i c , Fam. 3, 11, 3: De mercenariis testibus A suis CIVITATIBUS notandis. 'That the hired witnesses must have been noticed BY THEIR COUNTRYMEN.'

If the more familiar dative of agent were used here, the resulting surface struc­ ture would be a perceptual muddle, viz. : (46) *De mercenariis testibus suis civitatibus notandis. However we choose to theorize our syntactic framework, we must have some functional mechanism for blocking such sentences as (46) from occur­ ring in Latin. Since it is the surface confusion of cases and the parsing and dec­ oding difficulties brought about by this confusion, we should place the burden on the surface morphological form of the sentence. 3.0

Gerund with Nominal Object and Gerundive Modifying Nominal

The gerund is a verb form which can take a nominal object, as in the fol­ lowing: (47) Plaut., Trin. 1048: Male FIDEM SERVANDO abrogant fidem, 'BY KEEP­ ING FAITH badly they destroy confidence. ' (48) Livy, 21, 54, 1: EQUITES TEGENDO satis latebrosum locum. 'A suffi­ ciently shady spot FOR CONCEALING HORSEMEN. ' As is well-known, constructions of this type could be replaced by the pas­ sive gerundive in agreement with the head noun in gender, number, and case. Indeed, by the time of Plautus the gerundive construction was the customary one: (49) Fide servanda (for fidem servando) (50) Equitibus tegendis (for equites tegendo) It is a legitimate question to ask why this shift took place in Latin, and there are certainly many factors which contributed to this development. I would suggest, however, that a strong motivating force in this exchange was the for­ mal identity of the gerund and gerundive in many inflectional forms (the mas­ culine and neuter oblique singular cases of the gerundive are identical to those of the gerund). Because of this overall similarity, a construction like (51) would be easily misread as containing a gender violation: (51) ad fidem servandum This would be the case even though, syntactically speaking, servandum is a ge­ rund governing fidem, not an adjective modifying it. The same would be true of the following hypothetical examples:

SPEECH PERCEPTION AND LATIN SYNTAX

23

(52) ad Gallos insequendum (ad Gallos insequendos) (53) epistulam scribendi studiosus (epistulae scribendae studiosus) (54) Utile bellum gerendo (utile bello gerendo) All of the above gerund-plus-nominal-object constructions seem to con­ tain violations of rules of concord; thus there is a marked tendency in the later language to avoid this problem and to conform to the rules of concord by using the modifying gerundival adjective. But even at that, if there is a potential for gender confusion, the gerund and not the preferred gerundive will be used: (55) Cic, de Or. 2,289: SUBABSURDA DICENDOETSTULTAREPREHENDENDO risus moventur (instead of SUBABSURDIS DICENDIS et STULTIS REPREHENDENDIS, where STULTIS REPREHENDENDIS is ambiguous as to gender). 'Laughs are raised BY SAYING INCONGRUOUS THINGS and BY CASTIGATING FOOLISH THINGS.'

(56) Livy, 1, 46, 7: Initium TURBANDI OMNIA a femina ortum est (instead of TURBANDORUM OMNIUM, which is opaque as to gender). 'The be­ ginning of THROWING ALL THINGS INTO CONFUSION came from a wom­ an.' It seems clear that the availability of the gerundive option is motivated at least in part by the opacity and ambiguity found in some of the gerund con­ structions. The very existence of options such as these, which we have come to call transformations, may be due in part to the need to preserve structural har­ mony in surface structure. 4.0

The Accusative with Infinitive

The final construction I will discuss is the Latin indirect statement or Ac­ cusative-Infinitive-Construction (AIC). This construction has been the source of constant debate since the publication of Lakoff 1968, and it has been dealt with by others in far more detail than I am prepared to go into at this time (e.g., Miller 1974, Pepicello 1977, 1980, Bolkestein 1976a, 1976b, 1979). AIC is a theoretically interesting and descriptively challenging construction. The relevant data involve such sentences as: (57) Dicunt Marcum venire. 'They say that Marcus is coming.' (58) Admonent Marcum venire. 'They warn Marcus to come.' (59) Marcum venire evenit. 'It happens that Marcus is coming.' Pepicello (1980) suggests a three-source analysis for the AIC construction; his argument hinges largely on the syntactic rule of RAISING; the subject of the

24

PHILIP BALDI

"lower" clause is "raised" into the position of object of the "higher" clause, much as in English such sentences as / consider him to be a friend contain "raised," derived objects. Pepicello's arguments have been strongly attacked by Bolkestein, and I do not wish to enter the argument. But it seems to me that both parties have missed one rather obvious point in the issue of whether or not Marcum is an object of dicunt. Pepicello claims that Marcum cannot be a deep-object of dicunt because (60) is ungrammatical. (60) *Dicunt Marcum. But what can this mean? In what sense is *dico Marcum ungrammatical? It is an intermediate structure only (if such exist), and there is no condition in the theory in which Pepicello is working (TG) which states that all stages in a deri­ vation must be potentially grammatical surface structures. A similar line of reasoning would force us to conclude that (61): (61) *I told to him. is an impossible source for (62): (62) I told him. because it does not occur on the surface. In her review of the raising issue Bolkestein (1979) argues against the raising analysis in a cogent and persuasive way. I should add to her arguments that such notions as "raising" and its partner "lowering" are rather unfortu­ nate metaphors based on the tree representation used in generative-trans­ formational grammar. Bolkestein also offers strong argumentation against the accusative of AIC as a grammatical object. She says that the accusative case is simply the case assigned to the subject of a non-finite verb form. But this begs the question. Why the accusative? While I agree that the accusative is not a grammatical object, the accusative case-form suggests strongly that the linear arrangement of word places that subject of the dependent clause in a position perceptually akin to the object function. In other words, while it may not be an object grammatically, it BEHAVES like one because it is in the frame commonly associated with object of a transitive verb. Compare the following English sentences: (63) (64) (65) (66)

Max gave a book to Mary (→ Passive) → A book was given to Mary by Max. Max gave Mary a book (→ Passive) → Mary was given a book by Max.

SPEECH PERCEPTION AND LATIN SYNTAX

25

It is clear that (66) is derived from (65) in the same way that (64) is derived from (63). (Example (65) is derived from (63) by rightward NP movement; see Smaby and Baldi 1981.) The passive rule picks up the first NP to the right of the verb, regardless of whether it is an object or not. What matters is that, due to the operation of the other rule, there is an NP next to the verb, even though it is an indirect object syntactically. The Latin situation is entirely parallel. The subject of the dependent clause occupies a perceptual slot associated with objecthood; thus, the accus­ ative case form is chosen. It is not strictly a grammatical choice. 5.0

Conclusion

I hope to have demonstrated in this paper that the interaction of syntax and morphology is a process frequently mediated by strategies of speech per­ ception. Needless to say there are many constructions in which an appeal to perceptual factors is unnecessary; but I believe that future research along these lines, perhaps involving other languages with complex syntax-morphol­ ogy relations, will reveal further instances of this interesting and pervasive lin­ guistic phenomenon.

NOTE Exceptions to the patterns outlined here are plentiful, e.g. Cic., Leg. 1, 22, 59: Aliquid se habere sentiet divinum; Livy 5, 3, 9: Quicquam non dico civilis, sed humani. I believe that most of these counterexamples have a rational explanation, and I deal with them in Baldi 1977: 102-63.

BIBLIOGRAPHY Baldi, Philip. (1977), 'Morphosyntax and the latin genitive.' Folia Linguistica XI, 1/2: 93-108. Bennett, Charles. (1914), The syntax of early latin. Boston: Allyn andBacon. Bever, Thomas. (1970), T h e influence of speech perception on linguistic structure.' In Advances in psycholinguistics, ed. by G.B. Flores d'Arcais and J.M. Levelt. Amsterdam: North-Holland. Pp. 21-50. , and D.T. Langendoen. (1971), 'A dynamic model for the evolution of human language.' Linguistic Inquiry 2.443-63. Bolkestein, A.M. (1976a.), 'A. c. I. and ut-clauses with verba dicendi in latin.' Giotta 545.263-91.

26

PHILIP B ALDI

. (1976b.), 'The relation between form and meaning of latin subordinate clauses governed by verba dicendi.' Mnemosyne 29.155-175; 268-300. . (1979), 'Subject-to-object raising in latin?' Lingua 49.15-34. Chomsky, Noam. (1965), Aspects of the theory of syntax. Cambridge: MIT Press. , and Howard Lasnik. (1977), 'Filters and control.' Linguistic Inquiry 8.425-504. Fiengo, Robert. (1980), Surface structure. Cambridge: Harvard University Press. Jackendoff, Ray. (1977), 'X syntax: a study of phrase structure.' Linguistic Inquiry Monograph Two. Cambridge: MIT Press. Janson, Tore. (1979), Mechanisms of language change in latin. Stockholm: Almqvist and Wiksell. Katz, J. and P. Postal. (1964), An integrated theory of linguistic descriptions. Cambridge: MIT Press. Kühner, Raphael and Carl Stegmann. (1962), Ausführliche Grammatik der lateinischen Sprache (rev. by A. Thierfelder). Vol. 1. Munich: Hueber. Kuno, Susumu. (1980), 'Functional syntax.' In Syntax and Semantics 13: cur­ rent approaches to syntax, ed. by E. Moravcsik and J. Wirth. New York: Academic Press. Pp. 117-35. Lakoff, Robin. (1968), Abstract syntax and latin complementation. Cam­ bridge: MIT Press. Miller, D. Gary. (1974), 'On the history of infinitive complementation in lat­ in and greek.' Journal of Indo-European Studies 2.223-46. Moravcsik, Edith, and Jessica Wirth. (1980), Syntax and Semantics 13: cur­ rent approaches to syntax. New York: Academic Press. Pepicello, William. (1977), 'Raising in latin.' Lingua 42.209-18. . (1980), 'An integrated approach to diachronic syntax: a case study from latin.' General Linguistics 20.71-94. Perlmutter, David. (1980), Relational grammar. In Syntax and Semantics 13: current approaches to syntax. New York: Academic Press. Pp. 195-229. Postal, Paul. (1964), Constituent structure. Bloomington: Indiana University. Smaby, Richard, and Philip Baldi. (1981), 'Particle and dative movement re­ considered.' Linguistic Analysis 7.171-86.

ANAPHORIC RELATIONS, PRONOUNS A N D LATIN COMPLEMENTATION

A. B E R T O C C H I University of Bologna Istituto di Filologia Latina e Medioevale

C. C A S A D I O University of Bologna Istituto di Filosofia

SUMMARY This paper is principally devoted to an inquiry into the status of Latin in­ direct reflexive se in the light of two recent theories of grammar. In conformity with current usage, we shall refer to them as OB and GB respectively. The former is the theory presented in Chomsky's On Binding, the latter is the al­ ternative proposal known as Government Binding Theory. In the following, we shall see how this kind of analysis leads to a new interpretation of indirect reflexive se reconciling its pronominal nature with its behaving as a bound anaphora. 0.

Preliminary remarks

The principles involved in the analysis that will be developed imply the model of Core Grammar assumed in Extended Standard Theory of transfor­ mational grammar according to which a grammar is made of a Base generating D-structures (=DS) which are mapped onto 'pre-surface' S-structures (=SS) by the transformational rule "Move a". 1S-structures, in turn, are mapped on­ to Phonetic Representation (=PR) and Logical Form (=LF): DS

ss PR

LF fig. 1

28

A. BERTOCCHI & C. CASADIO

Both DS and SS are conceived at a higher level of abstraction than earlier 'deep structure' and 'surface structure'. Phrase structure analysis provided by DS, under the restrictions of the set of principles known as X-bar Theory (cf. Chomsky 1970 and Jackendoff 1977), gives the structural information relevant to LF. On the other hand, SS includes abstract elements such as tra­ ces of movement rules which do not have phonetic realization and do not ap­ pear on the surface. Among a variety of interesting properties, this model offers the base for a full characterization both of referring expressions and of various kinds of anaphors by means of the explicit relation it states between such abstract S t r u c ­ tures and semantic representations in LF. 1.

The OB framework

Inside the OB theory, full noun phrases and pronominals are assigned their referential status by an indexing procedure applying in the mapping from SS to LF. The same indexing accounts for their uniqueness and distinctive­ ness, i.e. their being disjoint in reference from any other referring expression occurring in the same sentence. This is obtained by means of a referential in­ dex plus an anaphoric index in brackets: (1)

Camillus 2 mihi3{2} scripsit te4{2,

3}

locutum cum eo5{2,

3, 4 } .

2

Starting, for conventional reasons , with index 2, the procedure assigns to each NP an increasing referential index and a set of anaphoric indices marking its referential disjunction from every c-commanding NP, where: (2)

C-command NP i = a c-commands NPj= |3, if the first branching node dominating a also dominates β.

On the other hand, anaphors are related to their antecedents by a coin­ dexing procedure applying through interpretive rules such as Reciprocal, Bound Anaphora, etc., under the general condition that an anaphor is bound to an antecedent c-commanding it. The following are examples of coindexing under the rule of Bound Anaphora= Reflexive: (3)

a. Militesi sei defendunt. b. Alexander i vix a sei manus abstinuit. c. Pausaniasi consilia cum patriae turn sibii inimica capiebat.

Literally speaking, the coindexing procedure states that the anaphor takes the same referential index as its antecedent. Since not every NP position in SS is

ANAPHORIC RELATIONS

29

an appropriate position for anaphors, the OB theory constrains the domain of anaphoric relationships by means of two conditions referred to as binding con­ ditions: (4)

Opacity condition An anaphor cannot be free in the domain of the subject of NP or S3. Nominative Island Condition (NIC) A nominative anaphor cannot be free in S.

It is immediately clear that Binding conditions interact with another relevant part of this model: the Case theory. This theory defines three major cases: no­ minative, objective and oblique. As far as we are concerned, nominative case is assigned only to subjects of tensed sentences. So nominative anaphors are avoided by NIC since they cannot be bound to an antecedent in S (cf. (5a)) and non-nominative anaphors in subject position are admitted (e.g. accusative an­ aphors as Acl subjects; cf. (5b)). On the other hand, Opacity condition for­ bids the occurrence of a subject between an anaphor and its antecedent, i.e. it avoids the occurrence of an anaphor in a clause in which its antecedent does not occur (cf. (5c)): (5)

a. * John thinks himself is a fool. b. John believes himself to be a fool. c. * John hopes that Mary loves himself.

Before turning to a consideration of how these conditions work in Latin grammar, we have to recall their effects on another kind of anaphoric rela­ tion, the coreference relation between a pronoun and a noun phrase. Binding conditions remove from the anaphoric index of a pronoun all theindices of the NPs occurring outside the domain stated by each condition. So a pronoun oc­ curring in the subject position of a tensed clause and a pronoun occurring in the domain of the subject of a clause are free in reference, i.e. can be coreferential, to the NPs occurring outside that clause, via NIC and Opacity respec­ tively. For illustration's sake, let us look at some English examples: (6)

a. John 2 wanted [Mary 3 { 2 } to wash him4{3}] b. John 2 admits [that he3 is a fool] c. John 2 believes [him3{2} to be a fool]

While in (6a) coreference to the NP John is allowed via Opacity, in (6b) it is permitted by NIC: in both cases the index of the pronoun blocks coreference only to NPs occurring in the same clause, thus being free(i), where i is a varia­ ble ranging over referential indices, for the NPs occurring elsewhere (in the

30

A. BERTOCCHI & C. CASADIO

same sentence or in the context). Incidentally, it should be noted that the pro­ noun occupying the subject position in (6c) cannot be coreferential to the ma­ trix subject John, as stated by its anaphoric index; the reason is its being nonnominative since the clause is infinitive. In this position, coreference to the matrix subject should be marked by an anaphor (cf.(5b) above). So, that anaphors and pronouns have different referential functions fol­ lows directly from their syntactic properties through Binding conditions: ana­ phors must be bound to an antecedent inside the same clause (with the excep­ tion of subj ect positions of infinitival clauses) ; pronouns cannot be coreferen­ tial to an NP occurring in the same clause, behaving in this respect as fully re­ ferential NPs. See the difference between (7a) and (7b): (7)

a. Mary hopes that John shaves him. b. Mary hopes that John shaves himself.

Moreover, while anaphors cannot be bound to an antecedent occurring out­ side their clause, pronouns may be and usually are coreferential to NPs filling these positions (see the difference between (5c) and (6a)). So far the OB the­ ory seems to offer an adequate account of the patterns exhibited by pronouns and anaphors, at least for a consistent set of modern languages. Applying Binding conditions to Latin, we single out a range of examples exhibiting the expected patterns. Those are the cases in which the domain of the Binding conditions is limited to a single clause or to an AcI construction: (8)

a. Militesi sei defendunt. b. Italici Adherbali suadent [uti (ille)i sei Iugurthae tradat] c. AriovistuSi respondit [non sesei Gallis bellum intulisse]

In (8a) and (8b), the reflexive se is bound to an antecedent occurring in the same clause — where in (8b) this antecendent is not lexically realized — as predicted by Opacity condition; in (8c) the reflexive sese can be bound to the higher NP Ariovistus since it fills the subject position of an AcI construction, as predicted by NIC. Moreover, if the same positions were filled by the pro­ noun is, e.g.: (9) Militesi eosj defendunt. the pronoun would be disjoint in reference from the NPs occurring in the same clause, again as predicted by Binding conditions. We then might infer from their behaviour under Binding conditions that se is an anaphor, while is is a pronominal. This assumption seems to be con­ firmed by the patterns where is can be interpreted as coreferential to an NP oc-

ANAPHORIC RELATIONS

31

curring outside its clause, as in (10): (10) a. (Antonius) non commisisset ut ei senatus denuntiaret. b. (Ille) petiit ab oppidanis ut ei liceret ad Caesarem proficisci. But we find in Latin another well-known set of cases which contradict our assumption: (11) a. Ariovistusi conclamavit [quid ad sei venirent] b. Caesari mandavit legatis suis [ut quae diceret Ariovistus ad sei referrent] Here se occurs within an embedded clause, bound to the subject of the main clause. If we choose to account for se as an anaphor, we are facing an overt vio­ lation of the Opacity condition: the anaphor does not occur in the same clause as its antecedent, thus being free in the domain of the subject of the clause containing it. The term indirect reflexive is employed by traditional grammars to define such occurrences of se. An indirect reflexive, it is held, occurs within an em­ bedded clause making reference to the subject of the main sentence, when such an embedded clause expresses the 'thought' of this main subject. We pro­ pose to define these subjects as 'logical subjects' even if the term is not quite appropriate: we will accept it provisionally, until we reach the statement of a more adequate definition. Stating his Opacity condition, Chomsky holds that subjects are 'operators' that make certain domains opaque in analogy with modal operators or verbs of propositional attitude so that "a variable within such a domain cannot be bound outside of it" (1980:10). Now, the evidence provided by the indirect re­ flexive se suggests that Latin 'logical subjects' analogously determine an opaque domain including the main sentence along with its system of em­ bedded clauses. This domain of the logical subject annuls the 'opacizing' func­ tion of the single embedded subjects, allowing reflexive anaphora to occur freely within an embedded clause. As we stated more extensively in an earlier work (cf. Bertocchi-Casadio 1980), indirect reflexive gives rise to a conflicting situation between the condi­ tions defined by the OB theory for each local subject and its opaque domain, and the conditions holding for the opaque domain determined by a logical subject. To this extent, the OB theory does not overcome the contradiction thus taking place, under the assumption that the reflexive se is an anaphor. In other words, we cannot maintain that se is an anaphor if we accept the Opacity

32

A. BERTOCCHI & C. CASADIO

condition, and we have therefore to choose between two options: (i) to try to give a new definition of the Opacity condition, in some way including the no­ tion of logical subject, or (ii) to explore more carefully the status of the indi­ rect reflexive se. The remaining part of this paper is concerned with this se­ cond alternative. 2.

The new perspective in the GB framework

The difficulties which have arisen so far may be overcome if we accept the different solution advanced in the GB theory. There, a newly introduced no­ tion of government eliminates the need of stating binding conditions, being by itself sufficient to define the syntactic domain of binding relations. Govern­ ment, which is operative along with Case assignment rules, states that lexical categories or tense govern and assign Case to what they minimally c-com­ mand, where: (12) Minimal c-command a minimally c-commands β, iff a c-commands |3 and there is no y such that a) a c-commands y b) γ c-commands (5 c) and not y c-commands a for a, γ being lexical, and assuming Š and NP as absolute boundaries for go­ vernment. Government allows the definition of two relevant syntactic do­ mains: (13) a. Every category a containing a governor y and a governed ele­ ment β is a governing category for (3. b. A governing category not properly containing governing catego­ ries is a minimal governing category (mgc). On this basis, a Binding theory can be stated in terms of binding princi­ ples such as 4: (14) a. If a is an anaphor then it is bound in every governing category (including its mgc). b. If a is a pronominal then it is free in its minimal governing cate­ gory. where bound is to be intended as argument-bound, i.e. coindexed with an NP in argument position 5. Referring back to the earlier Latin examples we see that GB theory makes the correct prediction:

ANAPHORIC RELATIONS

33

(15) a. Milites se defendunt. b. Italici Adherbali suadent [uti (ille) se Iugurthae tradat] Se here is bound to an antecedent NP in argument position within its mgc, where it is governed and casemarked: the matrix clause in (15a), the em­ bedded clause in (15b). According to (14a), se has then to be understood as an anaphor. An is occurring in these same positions could never receive a coreferential interpretation (cf. (9) above), being thus free in its mgc, according to the principle (14b) for pronominals. The same pattern is exhibited by accusative subjects of AcI construc­ tions: (16) a. Ariovistusi respondit [non sese- Gallis bellum intulisse] b. Ariovistusi respondit [non eumi Gallis bellum intulisse] where both anaphor and pronominal fall under the appropriate characteriza­ tion, satisfying (14a) and (14b) respectively. The complementary distribution of se and is in subject position of infinitival clauses may be accounted for on purely syntactic grounds in analogy with the preceding examples. Subjects of Acl constructions are ungoverned and cannot receive Case within the infiniti­ val clause, being the infinitive not provided with the feature 'tense'. But, since they do actually receive Case, the suggestion is that, by means of a rule of Š deletion, the matrix verb may govern and assign accusative Case to these em­ bedded subjects. In this way, the mgc becomes the clause containing the gov­ ernor, i.e. the matrix clause. Consequently, se in (16a) is an anaphor, for it is bound to the matrix subject within its mgc. For the same reason, eum in (16b) cannot be coreferential to the main subject since it has to be free in its mgc, ac­ cording to (14b). Let us now turn to the main problem, the status of the indirect reflexive se in examples such as: (17) Decima legioi gratias egit [quod (ille) • de sei optimum iudicium fecisset] We see that, by morphological identity with the reflexive forms in (15) and (16a), se here should be an anaphor, but since, as we have seen, it is free in its mgc, it follows from Binding principles that it has the structural properties of a pronominal. So far the advance offered by the GB theory is to make explicit the contradiction raised inside the OB framework. A feasible solution might be attained on the ground of the new proposals about pronominals advanced in GB theory. In OB, according to its anaphoric

34

A. BERTOCCHI & C. CASADIO

index, a pronoun is free in reference outside its clause, where free implies that it can, but need not necessarily, be bound to an antecedent in the same sent­ ence . Since in Latin se excludes a free interpretation, it emerges that se cannot fall under the characterization for pronouns as given in the OB theory. Com­ pare (18) and (19): (18) John thinks [that Mary loves him] (19) Ariovistus conclamavit [quid ad se venirent] In the case of (19), we cannot say that se is freely understood as referring to some arbitrary person as in (18), where him may freely refer to John or to someone else. The new perspective opened by the GB theory follows from a theory of indexing that gives up the whole system of indexing of OB and simply states that indexing applies freely, in the mapping from SS to LF, assigning to a pro­ noun an index that is either the same as or different from that of a c-commanding argument outside its mgc. In the first case the pronoun will be marked proximate, in the latter it will be marked obviative. Accordingly, two different logical forms will correspond to (20), either (20a) or (20b): (20) John thinks [that Mary loves him] (20) a. for every person x, x = John, x thinks that Mary loves x. b. for every person x, x = John, x thinks that Mary loves him. where in (20a) the pronoun is proximate and we interpret it as a variable bound by the NP John, while in (20b) the pronoun is obviative and is therefore understood as a name, referring to some person whose identity is fixed else­ where. In other words, a pronoun marked obviative cannot have an antece­ dent, while a pronoun marked proximate must have an antecedent. In English the same form is available to express both proximation and obviation and the distinction is found at the level of LF. We will suggest that in Latin, instead, the distinction is morphologically marked, being expressed by the use of two different forms, se and is, which are mutually exclusive. More explicitly, our tentative proposal is to consider the occurrences of se bound to a main subject, which is understood as a logical subject, not just as an anaphor, but as a proximate pronoun, occurring in com­ plementary distribution to the occurrences, in these same contexts, of is which can only express obviation. From the fact that Latin is provided with two distinct morphological forms in the domain of a logical subject to express obviation and proximation,

35

ANAPHORIC RELATIONS

it follows that, in these cases, LF does not play any role in disambiguating the interpretation of pronouns, surface structures being non-ambiguous by them­ selves. We could schematize the dissimilarity existing between Latin and mod­ ern languages such as English in this respect, as follows: LF 1 he obviative DS

→SS

he proximate

LF2 Fig. 2

DS 1

→SS1

LF1

is

DS2

SS2

LF2

se Fig. 3

In (fig.2), corresponding to the English situation, two distinct logical forms are obtained by free application of indexing in the course of the mapping from a unique SS, while in (fig.3), corresponding to the Latin situation, each LF is directly associated with a different SS. Is marks disjunction and is interpreted as a referring expression; that is, it is like a name, while se marks coreference and is interpreted as a bound variable, as appears in (21): (21) a. Caesari mandavit legatis suis ut quae diceret Ariovistus ad sei referrent. b. Caesari mandavit legatis suis ut quae diceret Ariovistus ad eumj referrent. The situation is different when no logical subject is involved, i.e. in con­ texts expressed as a matter of fact, not as the projection of the 'thought' of the matrix subj ect. In these cases, the use of is reflects the same ambiguity exhibit­ ed by modern languages. See for example:

36

A. BERTOCCHI & C. CASADIO

(22) Epaminondas fuit etiam disertus, ut nemo ei Thebanus par esset eloquentia. Ei may be coindexed with the NP Epaminondas, but not necessarily, as in the case of se. Consequently, we propose to assign to is a two-fold characteriza­ tion on the surface: either coreferential to some c-commanding NP outside the mgc or disjoint, which corresponds to the distinction proximate/obviative. Disambiguation is possible at the level of LF. To this extent we can apply to is in direct contexts the same analysis stated for modern languages (cf. (fig.2)) that we restate in (fig.4):

DS

LF1

is obviative

LF2

is proximate

SS

Fig. 4 The morphological identity between anaphors and proximate pronouns in Latin may give rise to ambiguous readings, where the ambiguity hinges up­ on the question which antecedent is relevant in the statement of binding rela­ tions. The alternatives are whether se is bound within its mgc (= an anaphor) or outside its mgc (=a proximate pronoun). In comparison, Romance lan­ guages do not present this kind of ambiguity, being provided with two distinct forms to express reflexiveness or proximation. However, they are ambiguous as to the choice of a referent for the pronominal form, which is either within or outside a sentence, as shown above. Compare Latin and Italian 6 : (23) Dux hortatur milites ut se defendant

t where

= anaphor,

V = proximate pronoun. (24) Il comandante esortò i soldáti a difendersi difenderlo J

ANAPHORIC RELATIONS

37

While in (23) the interpretation may be either that the soldiers must defend themselves or that the soldiers must defend their commander, in (24) ambi­ guity arises as to whether the clitic pronoun lo refers to the commander or to someone else 7 . 3.

Conclusions

Now, we can try to reformulate the binding principles (14a) and (14b) for Latin, according to the proposals advanced before about the different occur­ rences of se and is: (25) a. If a is an anaphor, then it is bound in every governing category, b. If a is a pronominal, then it is free in its mgc and 1) a — se when proximate to a logical subject 2) otherwise a = is That is to say: se must have, in any case, the index of some c-commanding NP in argument position. If the argument to which se is bound is within its mgc, then se is an anaphor. If it is outside the mgc then se is a proximate pronoun. In our opinion, the characterization of se as aproximate pronoun, bound outside its mgc, may be a possible explanation why, in Latin, a morphological form identical to an anaphor may occur free within its mgc.

NOTES 1) Since the content of the rule is "Move a" everywhere (for a read any phrase or category) this is really a rule schema including all the previously distinguished transformations; cf. Chomsky (1980:4). 2) The index 1 is used for arbitrary reference; cf. Chomsky (1980:39). 3) The category Š is assigned to sentences along with their complementizer system according to the base rule Š → COMP S; cf. Chomsky (1980:5). 4) For a full statement of the Binding theory, cf. Chomsky (1981). 5) Argument-bound, where 'argument' means an NP position in S, has to be distinguished from standard binding by a quantifier, which never fills an argument position in S. 6) This example is freely derived from Tekavčić (1972) to whom we owe the characterization of the ambiguity here considered with reference both to Latin and Italian. 7) The differences so emerging between Latin and Romance languages are directly tied to the role played by Latin logical subjects in the distinction between direct and indirect speech. In this perspective, a recent analysis by J.C. Milner fits in with our own results in a significant way. Argu­ ing that indirect speech consists of a set of clauses all related to the same referent, Milner comes to the conclusion that se is a pronoun, sharing the properties of first and second person pronouns and analogously fixing reference to one and the same given referent inside indirect speech.

38

A. BERTOCCHI & C. CASADIO

BIBLIOGRAPHY Banfield, A. (1973), 'Narrative style and the grammar of direct and indirect speech.' Foundations of Language 10. 1-39. Bertocchi, A. and C. Casadio. (1980), 'Conditions on anaphora: an analysis of reflexive in Latin.' In: G. Calboli (ed.), Papers on grammar I,1-46. Bo­ logna: CLUEB. Chomsky, N. (1970), 'Remarks on nominalization.' In: R. Jacobs and P. Ro­ senbaum (eds.), Readings in English Transformational Grammar, 184-221. Waltham: Ginn-Blaisdell. . (1980), 'On binding.' Linguistic Inquiry 11. 1-46. . (1981), Lectures on Government and Binding. Dordrecht: Foris. Ernout, A. and F. Thomas. (1953), Syntaxe latine2. Paris: Librairie C. Klincksieck. Hofmann, J.B. and A. Szantyr. (1965), Lateinische Syntax und Stilistik. Mün­ chen: Beck. Jackendoff, R. (1977), 'Constraints on phrase structure rules.' In: P.S. Culicover, T. Wasow, A. Akmajian (eds.), Formal Syntax, 249-83. New York: Academic Press. Keenan, E.L. (1976), 'Towards a universal definition of 'subject'.' In: C. Li (ed.), Subject and Topic, 303-333. New York: Academic Press. Kühner, R. and C. Stegmann. (1955), Ausführliche Grammatik der lateinischen Sprache. 1. Teil. 3. Auflage durchgesehen von A. Thierfelder. Le­ verkusen: Gottschalk. Milner, J.C. (1978), 'Le système du réfléchi en latin.' Langages 50: 73-85. Tekavčić, P. (1972), Grammatica storica dell'italiano. II: Morfosintassi. Bo­ logna: Il Mulino. Traina, A. and T. Bertotti. (1965), Sintassi normativa delia lingua latina. Bo­ logna: Cappelli.

Latin references [Alexander] cum interemisset Clitumfamiliarem suum, vix a se manus abstinuit (Cic.Tusc.4,79). [Pausanias in agro Troade] consilia cum patriae tum sibi inimica capiebat (Nep.Paus.3,3).

ANAPHORIC RELATIONS

39

[Italici] Adherbali suadent uti seque et oppidum lugurthae tradat (Sall.lug. 26,1). [Ariovistus] respondit non sese Gallis sed Gallos sibi bellum intulisse (Caes. Gall. 1,44,3). [Antonius] non commisisset ut ei senatus tamquam Hannibali... denuntiaret ne oppugnaret Saguntum (Cic. Phil. 6,6). petiit ab oppidanis ut ei liceret legatum ad Caesarem proficisci (Bell.Hisp. 22,5). [Ariovistus] exercitu suo praesente conclamavit quid ad se venirent (Caes. Gall. 1,47,6). [Caesar legatis suis] mandavit ut quae diceret Ariovistus cognoscerent et ad se referrent (Caes. Gall. 1,47,5). Decima legio per tribunos militum ei gratias egit quod de se optimum iudicium fecisset (Caes. Gall. 1,41,2). [Epaminondas] fuit etiam disertus ut nemo ei Thebanus par esset eloquentia (Nep. Epam.5,1).

THE DEVELOPMENT OF LATIN (CASES A N D INFINITIVE)

G U A L T I E R O CALBOLI University of Bologna, Istituto di Filologia Latina e Medioevale SUMMARY In an initial framework the Indo-European subordination is summarily presented in its different stages. The Accusative with Infinitive (AcI) is con­ nected with the other constructions of the nominal forms of verb such as Par­ ticiples and Gerundives which are directly tied with the verb of the main sen­ tence. Some frozen forms such as praesente multis are mentioned in order to demonstrate the instability of these constructions. Then four items are dealt with: (1) the development of Latin subordination, (2) the progressive reduc­ tion of Latin cases, (3) the explanation of the AcI in relation to the recent solu­ tions proposed by Chomsky (in the AcI the Accusative is determined by three conditions: (a) the sentence is lacking in INFLection, (b) the infinitive sen­ tence is without a Š node, (c) the Verb of the main sentence can hold the ac­ cusative subject of the infinitive sentence, and condition (c) can be missing since the system of the cases remains vital), (4) the relationship between the Acl and the quantification of nouns, articles and the word order. Framework Indo-European subordination can come about in two ways: (1) by adding one sentence to another (the subordinate before the principal "as might be ex­ pected of an OV language") 1 , first through an introductory particle, then by reducing this particle to a relative or an interrogative pronoun, and then, in a third phase, by creating new particles and subordinating conjunctions from frozen casual forms of relative and interrogative pronouns, or (2) by connect­ ing with the main sentence a nominal form of the verb such as Infinitive (Ac­ cusative cum Infinitivo, henceforth Acl, and Infinitive without subj ect), Par­ ticiple (Ablative Absolute in Latin, Genitive Absolute in Greek and Sanskrit, Locative Absolute in Sanskrit and Avestan, Dative Absolute in Gothic and

42

GUALTIERO CALBOLI

Balto-Slavic, and Accusative Absolute in Greek, Late Latin and Gothic; cf. K. Brugmann 1916: 961-968; W.P. Lehmann 1972: 987 f.; 1974: 22; 210 f.; Participles in adjectival and predicative use such as Cic. Off.2,25 Dionysius cultros metuens tonsorios ('because he feared the barber's shears') candente carbone sibi adurebat capillum), Gerund, Gerundive and Absolutive fre­ quently found in Sanskrit (cf. A.Thumb-R.Hauschild 1959: 378-385). In the first manner of subordination, in the opinion of Carol Justus (1976; 1980) and W.P. Lehmann (1980), we would have the connection of two sen­ tences, the first corresponding to that which will then become the relative sen­ tence (see the excellent work done on this aspect by Ch. Lehmann 1979: 436442). This subordinate sentence was preposed because of the SOV word order and was in 'focus' as a comment to a term of the main sentence. Such sen­ tences, as has also been shown by Françoise Bader (1973) and others, are intro­ duced by particles (mu, nam, ta, etc.) that are then substituted by particles fur­ nished with a case such as the particle ku- ("ku- functions primarily as a topicmarking, or more exactly, as a focus-marking particle", Justus 1976: 235). Therefore, from the original particle ku we have in Hittite the relative-in­ terrogative kuiš (which in Latin is the relative qui and the interrogative-indefi­ nite quis) and the indefinite kuiški (Latin quisquis). Ch. Lehmann (1979: 436442) has very clearly demonstrated the ways of connection between the sen­ tence introduced by the pronoun of theme ku- and the main sentence, but what is interesting at this point is the fact that the non-inflected particles are substituted by inflected particles (nu is substituted by kuit)2 so that the intro­ ductory formulas take a casual inflection and they agree in gender and number with the noun of the main sentence for which they present the comment. At this point it is necessary to place the development supposed from 'topic promi­ nence' (this allows subordinate sentence as a comment which is not formally tied to the term commented on by the main sentence) to the second phase, known as 'subject-prominent phase'. Only at this time can we place the sec­ ond type of subordination, the type of inflected verbal nouns according to the case requested by the predicate of the main sentence. Until this point, the order of Indo-European languages must have been of the SOV type, as it is proven by the anteposition of the complement sentences in subordination of type (1), and by the casual agreement of the nominal forms of the verb in the subordination of type (2) (cf. W.P. Lehmann 1972). Furthermore, if we ac­ cept the idea that Indo-European languages went from an ergative phase to a Nominative-Accusative one, 3 or at least that these languages developed from

THE DEVELOPMENT OF LATIN

43

the second ergative type (type B) of R.L. Trask 1979,4 with an SOV order (this order is typical of ergative languages (see Trask 1979: 393-404)), and with great use of nominal forms of the verb, we can think that the construct of the Nominative with Infinitive (such as Gaius bonus esse dicitur, with dicitur at the end as required by word order SOV) 5 was born first and then, when these languages became of type Nominative-Accusative, it passed to the true AcI with the subject of the Infinitive interpreted as the object of the main verb such as imperator iussit (or vetuit) milites venire. Even in Hittite there is the Nominativus cum Infinitivo (cf. J. Friedrich 1960: 143: "erscheint das Nomen oder Pronomen, das nach unserer Auffassung Objekt des Infinitivs ist, im Hethitischen als Subjekt des übergeordneten Verbums") and the AcI where the Accusative is interpreted as the--object of the main verb (Friedrich 1960:143). Thus, what is necessary for the subordination of the second type are nominal forms of the verb inflected in accordance with the case required by the sentence itself (main sentence) in which these nominal forms are in­ serted. The second passage is the development of absolutive constructs (Ab­ lative, Genitive, Dative, Locative, Accusative Absolutes, the Absolutive in Sanskrit, etc.). In this manner, such constructions become more and more like autonomous sentences with respect to the sentence in which they find themselves (main sentence). But in the Infinitive arises a modification which not only invests these sentences, but all the uses of the Infinitive: it loses the inflection according to the cases and remains a non-inflected form which loses its nominal character more and more. 6 In fact, only in Vedic Sanskrit we can still find the inflected infinitive that then becomes frozen in the formation -turn of Classical Sanskrit (see L. Renou 1937). With this, the infinitive be­ comes more and more a verb, and less and less a noun as was already shown for Sanskrit by Renou (1937) and for Latin by Perrochat (1932). The Greek infinitive confirms this development (cf. H. Kurzová 1968). It is interesting to note that this freezing of absolutive constructions achieves formations which are generally detached from the verb of the main sentence (this is the case with the Accusative and Nominative Absolute in Latin) 7 and that this freezing achieves the loss of the notion which tied the nominal forms of the verb to the predicate of the main sentence. In fact, we find in Latin types such as absente nobis (Ter.Eun. 649), praesente multis (Rhet.Her.4,16). Hence, we can verify that with the infinitive the loss of the casual inflection brings the infini­ tive to resemble a verb, and the infinitive sentence to a sort of independent sentence, while in the Ablative Absolute, in the presence of a case which is very frequently linked to prepositions such as the ablative, the participle (pres-

44

GUALTIERO CALBOLI

ent in Classical Latin, perfect in Late Latin) becomes nearly a preposition8 (praesente, absente multis, nobis, etc., excepto oppida, about this construct see E. Flinck-Linkomies 1929: 64 f.; and A.L. Castelli 1976). This is a sign of the instability of the constructions with nominal forms of the verb which undergo assimilation to the uses that are closer and more similar, i.e. to the subordinate sentence with finite verb the infinitive constructs, and to the prepositional groups the Absolute Ablatives. What has come into the pic­ ture here is the literary and scholastic Roman development which determined and set the entire system and even enlarged the use of the AcI in subordina­ tion. However, by extending the use of the participle in a system where cases are still in existence, it provoked the re-absorption of types such as praesente multis which have always had very little life and a reduced one at that. Fur­ thermore , the casual system demonstrates a notable vitality in Classical Latin, so much so that the Absolute Ablative with a participle in -tus in deponent verbs is very rare (an isolated example in Sall Jug. 103,7 dein Sulla omnia pollicito docti, quo modo apud Marium, item apud senatum verba facer ent eqs.). The most Absolute Ablatives with passive participles in -tus (not active ones) made it difficult to use these participles with an active value, as is correctly noted by Pierre Flobert 1975: 493, but it makes us think that the Absolute Ablative was interpreted something as a nominal group accompanying the verb of the main sentence, and therefore it was difficult to interpret it with the active verbal meaning of the deponent verbs. Finally, two phenomena are historically ascertained: (1) the develop­ ment of the verbal value of the infinitive in the passage from Plautus to the classic period is responsible for the birth of the -urum fuisse type of unreal in­ finitive, and (2) the AcI expanded despite the Nominative with Infinitive and a completely new type is born in cases as Ter. Andr. 796 in hac habitasse platea dictumst Chrysidem and Nep. Paus.5,3 dicitur eo tempore matrem Pausaniae vixisse.9 In this framework, not everything is clear, nor it is sure. Many points should be looked at more closely. I would like to limit myself to dealing with just a few of these, in particular four items: (1) the development of Latin sub­ ordination, (2) the development of Latin cases, (3) the explanation of the AcI and of similar constructions in relation to the more recent solutions proposed by Chomsky, (4) the relationship between the AcI and the quantification of nouns, or the relationship between AcI, word order, and articles. (1)

It is quite evident from the facts themselves that the Ac I is substituted

THE DEVELOPMENT OF LATIN

45

more and more by constructions with finite verb forms which are introduced by quod, quia. This process can be witnessed already in old Latin, is encour­ aged by Greek and Grecism, is extended in vulgar Latin, and it reaches its full development in Romance languages. In the final phase of the process the facts are clear: the AcI is almost completely abandoned and also in the medieval Latin it is used if the author wants to follow a more literary and refined style but it is given up in favour of quod + finite verb if the author uses common and usual speech as it was demonstrated by Lore Wirth-Poelchau 1977.10 Besides what has already been indicated in Calboli 1978: 209-220, much more material may be alleged. In fact, there exists a series of uses in which the use of the AcI is reduced to the advantage of sentences with a finite verb intro­ duced by quod, quia and the like, for example, in sentences such as percommode factum est, quod els de rebus quae maxime metuuntur, de morte et de do­ lore [...] disputatum est (Cic. Tuse. 4,64) and in others introduced by the de­ clarative quod or quia (see Kühner-Stegmann 1955: II 270-279, and J. Her­ man 1 963: 32-66). Mixtures such as patrem novisti quam sit perspicax (present also in Greek, e.g. Horn.//.3, 192 "tell me then, dear daughter, who this is", Priam asks Helen upon see­ ing Odysseus), mixtures where we have the Accusative + COMP + S (INFL) — therefore we have Š — favored the evolution of the construct of the AcI towards the finite mode. Regarding this construct studied by Claes Lindskog 1896 and by Charles Hyart 1954: 54, I have recently done a study in Calboli 1980. We must also keep in mind mixed types such as nesciendo quae petere, even if these cases are simple infinitives and not Accusative with In­ finitive, and types such as nesciendo quae petere are extended, when the sim­ ple infinitive is expanded and the AcI reduced, cf. Calboli 1981: 149-51. However, the AcI remains for all of Latinity and one must abandon Latin, that is, enter into a language almost without cases, as are the Romance languages, in order to find the complete abandonment of the AcI. Further­ more, one must not ignore the wealth of subordinate sentences in Classical Latin and in Latin literary language in general, and therefore the need to have many types of subordinate sentence, AcI, sentences introduced by conjunc­ tions and with the finite verb (ut, cum, dum, si, quod, quia, quoniam, antequam, priusquam etc. + finite verb), Participles which agree with nouns of the main sentence and Absolute Ablatives, Gerundives and Gerunds, Rela­ tive Sentences and Indirect Interrogatives. It would be necessary to do a vas­ ter study of this aspect, something more than this brief reference. I would however like to advance one of the many examples that could be discussed, in

46

GUALTIERO CALBOLI

which a Roman, Cicero, translates a Greek text. From the translation which is literal with regard to the concepts and the terms, a typical sort of behavior in Latin emerges, one which determines the structure itself of the Latin sen­ tence . It is the first of the of Epicurus translated by Cicero in De natura deorum: Diog.L. 10,139 Cicero puts the sententia in the indirect form and this already denounces the Latin use of carrying in an indirect form what is asserted by others for reasons of scrupulous documentary translation of juridic and official language, as has been proven and documented by Charles Hyart (1954: 137-150): Cic.Nat.Deor.1,45 vere exposita illa sententia est ab Epicuro, quod beatum aeternumque sit ( ), id nec habere ipsum negotii quicquam (we must note the exact translation of orTò with ipsum) nec exhibere alteri, itaque neque ira neque gratia teneri; quod, quae talia essent (= ) imbecilla essent omnia (the Latin translates with this quod + finite form of verb and therefore with a subordi­ nate sentence the coordinated Greek sentence introduced by ).11It would be opportune to think of this passage also with regard to the relationship be­ tween the article and the relative sentence recently studied by Christian Touratier (1980: 363-386) who faces this problem very well in the relationship between le sage and celui qui est sage, but what interests us here is the corre­ spondence between Cicero and Epicurus and the difference in structure be­ tween Latin and Greek: in Cicero, who certainly translates, the entire state­ ment is in the indirect form, two articles with generalizing indications are given with two necessary relative sentences ("relatives déterminatives") and a coordinated sentence, introduced by , is given with a subordinate; there­ fore the Latin author uses subordinate sentences much more than does the Greek author. The use of the infinitive sentence in the indirect Latin style is discussed by Charles Hyart (1954: 30-55). I would only invite more caution be­ fore affirming as Hyart does (1954: 55) that "la construction dico quod, scio quod, est couramment employée dans la langue familière au lieu de l'infinitive". Although it is true that the first example of quod + finite verb can be traced back to Plautus (Asin.52 f. equidem scio iam, filius quod amet meus / istanc meretricem e pruxumo Philaenium) and this construction appears here and there in the familiar tongue, it remains that this construction, as József Herman (1963: 36) states, was unable to find a place but "très sporadiquement dans la langue écrite". I also agree with Herman's explanation that the exis­ tence of sentences with causal quod and quia created confusion, but above all

THE DEVELOPMENT OF LATIN

47

I believe that the written language, which was too rich in subordinate sen­ tences, and therefore needed many types of subordination, could not be de­ prived of a subordinate type such as the AcI. (2) The cases which directly interest us are those of the nominative and the accusative. At any rate, from what has been observed previously with regard to the construction of the nominal forms of the verb, amongst which are also cases such as the ablative, the dative, the genitive, the locative, it is quite evi­ dent that the entire system of cases is involved in this problem. 12 Now, in Latin cases we can see a progressive and slow reduction in Archaic Latin, to a com­ plete stop of the reduction in the literary language used in Classical Latin, and a great reduction in Vulgar and Late Latin. L. Prat's book (1975) dem­ onstrates, leaving out of consideration the value of this work, that about 200 B.C. there is a reduction of old separative ablatives in -ōd, -ād which merge with the endings of the instrumental -o, -ă (this fact is confirmed by Maria Luisa Porzio Gernia 1974 in her more cautious memoir, and the Pompeian in­ scriptions clearly show the "évolution generale du latin vulgaire vers l'élimina­ tion des cas obliques autres que l'accusatif" (V. Vaanänen 1966: 121). In fact, just how much the accusative is used in the place of the ablative dependent on prepositions which hold the ablative in Pompeian inscriptions, is quite impres­ sive: e.g. a pulvinar, cum iumentum, cum basim, cum sodales, si(ne) dulcissimam Philoth..., aurum pro fiirrum diidica(t) (cf. V. Väanänen 1966: 121 f.). (3) Chomsky's explanation (1980 and Lectures 1981a) must be considered the most advanced even if it is an explanation of the infinitive sentences in En­ glish, not in Latin, while Latin is much richer in infinitive sentences than En­ glish. The two types singled out by Chomsky and applied to Latin by Mirka Maraldi (1980) are easily recognized: the type with [-Control] verbs, for example (Ego) puto [S eum indoctum esse] with the object governed by the V of the main sentence across the boundaries of the sentence, and the type with [+Control] verbs, for example lube hunc [SPRO abire hine aliquo], where hunc is the object of iube and controls the PRO of the infinitive sentence. In the first case, it is necessary to think, as Chomsky does, of the deletion of the Š (or S') node as determined by the properties of the verb puto, exactly in the same way as happens with believe in English (cf. Chomsky 1981b:52, who writes: "This property of deleting Š is in part an idiosyncratic lexical property of certain verbs"). The same, but not necessarily, can be said of the second case. Really in Latin the oldest phenomenon seems to be the absence of the

48

GUALTIERO CALBOLI

node Š in AcI (not its deletion, since the infinitive is a verbal noun), and the presence of S will come about only with the verbs and the constructs with which the sentence is accompanied by a complementizer although it is near the infinitive, as, for example nosti patrem [Squam[sPRO(drop) sit perspicax]]. A sentence such as dicitur matrem Pausaniae vixisse (probably marked) gives us an example of the historical evolution towards the absolutive constructs and towards the development of real sentences instead of the infini­ tive. Chomsky's model reveals its wellknown defect: it aims to be universal, but in truth it is built on the English language and Latin can only be explained if we deviate something from this model. But if we do not go too far from Chomsky's model, we can explain the Accusative of the AcI in the following way. This Accusative (of the subject) could be determined by three condi­ tions. (a) That the sentence is lacking in INFLection (as is the infinitive sen­ tence). But this condition is not enough because even the Historical Infinitive is without INFL and just the same its subject is in the Nominative, and we find the Historical Infinitive in main and complement sentences, although the use of Historical Infinitive in complement sentences is not frequent, but see Kühner-Stegmann 1955: I 137, and Hofmann-Szantyr 1965: 368, e.g. Tac. Ann.2,4 ubi minitari Artabanus et parum subsidii in Armenis [..] erat, rec­ tor Syriae [..] circumdat. (b) That the infinitive sentence is without a Š node (and this happens only to infinitive sentences), and (c) that the V of the main sentence can hold the Accusative either as a direct object or as a direct object accompanied by a predicative complement (such is the case verified with the Verba Sentiendi and Dicendi + an infinitive sentence whose predicate is a nominal one, e.g. puto/credo/dico [S te bonum esse]). Conditions (b) and (c) are strictly linked to one another. Just the same, condition (c) can be missing and is missing with types such as constat, apparet, dicitur + AcI, because con­ ditions (b) and (c) are just strictly tied to one another. Originally, condition (b) was possible only because of the presence of condition (c) and until the sys­ tem of the cases remains vital with an order of words fluctuating between SOV and SVO. In this case condition (b) alone guarantees the efficient action even if condition (c) is absent. This is something in the spirit of A-over-A Condition (cf. N.Chomsky-H.Lasnik 1977: 463). In fact, condition (c) works only if con­ dition (b) holds, i.e. we have a dependence on the V of the main sentence iff (if and only if) the complement sentence (S) is without a node Š. Therefore, the condition (c) seems to be a type of subcondition of condition (b) and the pres­ ence of condition (b) can be enough even if condition (c) is missing. The same happens with A-over-A Condition. For A/A condition helps to avoid a filter

THE DEVELOPMENT OF LATIN

49

just as condition (b) is enough to avoid subcondition (c), provided that a vital system of cases still holds as is the case with Latin. This phenomenon is further illustrated in Mirka Maraldi's paper, although she does not share all my views. As confirmation I can quote the following constructions we find in Classical Latin, (l') Cic. de Orat. 2,299 quidam doctus homo atque in primis eruditus accessisse dicitur eique artem memoriae [...] pollicitus esse se traditurum; cum ille quaesisset, quidnam illa ars efficere posset, dixisse ilium doctorem, ut omnia meminisset; et ei Themistoclem respondisse eqs., other examples by Kirk 1938:182. In this case, if we have two or more sentences which depend on a Verbum Dicendi in Passiv (dicitur), the first sentence holds the personal construct (doctus homo accessisse dicitur), while the others take the impersonal construct (dixisse ilium doctorem, Themistoclem respondisse). This shows that infin. without (c) is enough to cause the accusa­ tive.13 (2') We find AcI in exclamations with or without -ne, e.g. Plaut. Men. 1005 erum meum [...] luci deripier in via? Ter. Phorm. 339 tene asymbolum ve­ nire unctum! 497 adeon ingenio esse duro te! I think that this infinitive depends on a Verbum Dicendi or Sentiendi or Cogitandi which is not expressed (cf. G.Calboli 1981:148f). In my opinion, this is the best example of absence of condition (c). The verb of the main sentence can be missing, but the case Ac­ cusative of the NP subject in infinitive sentence is enough to show that this subject depends on the Verb of the main sentence. At any rate, you must exp­ lain why we find Nominative with Infinitive in the construction of the In­ finitum Historicum and Accusative with Infinitive in exclamations. This shows that Infinitive (AcI) in exclamations is an indirect statement (discus­ sion in Calboli 1981) and in this case condition (c) is missing. However, the phenomenon that traditional grammar singles out in the nearing of the infinitive sentence to a real sentence corresponds in Chomsky's theory to the presence or lack of the node Š which at least in an early phase seems to have been present only when AcI is substituted by quod, quia etc. (which are real COMP) + finite verb. To summarize, if we say that the Accusative in the construction of the AcI is due to the Infinitive, that is, only to the absence of the INFL, it is not possi­ ble to explain the construction of the Historical Infinitive where the subject is in the Nominative. And it must be noted that the Historical Infinitive is also in a complement sentence and therefore in a sentence which is certainly fur­ nished with a node Š (condition (b)). I think that the Historical Infinitive, both independent and dependent, can always be a Š, and that this violation of con­ dition (b) would explain the use of the Nominative instead of the Accusative,

50

GUALTIERO CALBOLI

so that we could say that condition (a) alone, that is, the absence of INFL, is not enough to provoke the AcI. With this, we arrive at our explanation where we play with conditions (b) and (c). Furthermore, it must be said that in Latin INFL corresponds in substance to flexion according to the category of the per­ son (1st, 2nd, and 3rd) and it is precisely this that is missing in the Latin infini­ tive, which is instead furnished with tense (but even the tense is in part a newformation certainly in the expression of the future in -türum, -türam esse) and with agreement in the nominal forms (e.g. hominem dicturum esse, mulierem dicturam esse, hominem locutum esse, mulierem locutam esse). An infinitive that has flexion according to the person as in Portuguese, or a verb which is not sensitive to the person, such as the Russian past indicative, would take us away from our conditions and would not be comparable to Latin infinitive.14 The construct of "infinitivo pessoal" in Portuguese can be found also in the Neapolitan dialect of Italy (see P.Tekavčić 1972:532 f.), but it is quite iso­ lated. But besides being founded on the Historical Infinitive, a construction which is present in all periods of the Latin language (see Hofmann-Szantyr 1965:367 f.), my explanation is founded on two elements: (I) The Exclamative Infinitive offers an example of a construction where the subject in the accusa­ tive demonstrates the indirect statement, i.e. it demonstrates the dependence on an unexpressed or deleted verb of saying or believing. I can distinguish be­ tween two sentences such as tu currere and te currere, the first a Historical In­ finitive and the second an Exclamative Infinitive, besides through intonation, essentially for the difference between tu (Nominative) and te (Accusative). (II) When the cases fall in Romance languages, the AcI falls, leaving out of consideration the elements which we talked about in point (1). The forms of the Exclamative Infinitive in Italian confirm all of this. In fact, the Exclama­ tive Infinitive with the subject in the Nominative is found e.g. in Goldoni, Cur. Ace. 1,4 impiegar io medesima le parole e i mezziper trattenerlo! But even that which corresponds to the AcI of Latin is today in Italian a Nominative with Infinitive, for example, Non presumo di essere io (not me) buono e gli altri cattivi; Tutti sanno che tu sei un uomo onesto e non avere mai tu (not te) ingannato alcuno (both Schwendener 1923 and Rohlfs 1969: 87-89, do not give examples with the first and the second person of the singular personal pronoun that alone allows for the case to be distinguished). Naturally I do not accept the explanation of the Exclamative Infinitive given by Professor Vairel-Carron 1975 for the reasons which I have put forth in Calboli 1981: 145-8. In fact, if everything depended on the "intonation as­ cendente", as Professor Vairel-Carron states, we could not, in my opinion,

THE DEVELOPMENT OF LATIN

51

explain why with the same "intonation ascendente" we can have two different constructs such as Sicine hoc te mihi facere! and Tune istuc credis! Nor can we adduce as objection the fact that the enclitic -ne gives the sense of this intona­ tion because in many cases the enclitic -ne is missing. Nevertheless, thanks go to Vairel-Carron for having brought out the problem of intonation, a problem which must be studied further. The conclusion is simple: in order to give a satisfactory explanation of the Latin AcI, all infinitive constructions in relation to subordination and to the various cases of their subjects must be kept in mind. What, I think, must be more closely examined is the relation between our condition (b) and condition (c). (4) Topicalization can be regarded as an old phase pertaining to Indo-Euro­ pean languages, then overcome, but according to W.P.Lehmann 1976:451f., it seems to have left a trace of itself in the construct of the predicative comple­ ment. This can also be said for the ergative phase: it is well known that Latin avoids taking inanimate and abstract things as subject of a sentence (cf. Nagelsbach-Müller 1905:613 f., where we find the following statement: "Es läßt sich [..] nicht leugen, daB der Lateiner, im allgemeinen, wenn er die Wahl hat, persönliche Subjekte den sachlichen vorzieht"; and regarding abstracts see Mikkola 1964:96f.). But above all is in the word order fluctuating between SOV and SVO (and other figures) that comes off the preservation in Latin of the old structure of the sentence up to the imposition of a fixed order SVO, which only if it is realy fixed can substitute the casual distinction of subj ect-object. A logician such as Quine raised the problem of the possibility of export­ ing a noun (in particular, the subject) from a dependent sentence in opaque contexts. This has been resolved by Quine with the use of the infinitive as a construction which alone allows this exportation. I have tried to demonstrate (Calboli 1978) that the birth of the article in Later Latin and in Romance and the fall of the AcI are phenomena to be related. The relational nature of the definite article corresponds to the relational nature of demonstrative pronoun from which the article comes in many Indo-European languages (Romance, German, Greek). I have tried however to demonstrate that in reality it is the SOV word order with infinitive sentence (AcI) which was originally placed before the verb of the main sentence and interpreted as a sort of attributive ad­ jective, that determines the possibility of saving the article. And this effect is obtained, e.g., in a sentence such as amicum reversum esse puto by making agree amicum with reversum or at least by interpreting amicum as the object

52

GUALTIERO CALBOLI

of puto. When the infinitive sentence is substituted by a sentence of relative origin such as the Romance subordinates with che, que, the link (again of the type of comment of a topical element) is no longer given by the case accusative of amicum, but by the demonstrative pronoun ille amicu(s) qui and the use of this pronoun as an article becomes natural. The cause of all this modification is the change from the SOV word order to the SVO. This change makes absolutive the constructions with nominal forms of the verb (cf. W.P.Lehmann 1972:989) and therefore contributes to the nearing of the infinitive sentence to the subordinate of a finite verb type. The persistence of the SOV word order, or rather, the lack of a rigid fixation of the SVO word order sustains the dis­ tinction between subject and object. When this rigid fixation is achieved falls the distinction between subject and object through different cases and falls the system which makes possible the AcI in the extended form that it has in Latin. Constructs such as ital. appare essere tu buono (with nominative tu) are only Latinisms as has been demonstrated by U.Schwendener 1923. In conclusion, the well developed system of cases is tied to the AcI in the first stage of birth and of the expansion of this construction, and is also tied to the expansion of the accusative in the slow evolution of the AcI towards the construction of a real sentence. Here, the accusative is found even when only two of the three conditions necessary are given in order to have the accusative, i.e. lack of INFL (condition a), and lack of node § (condition b), provided that the casual system is vital. In this case a construct such as constat te bonum esse or dicitur matrem Pausaniae vixisse where the accusative is not dependent on the verb of the main sentence behaves, in my opinion, almost in the same way as an absolute construction. As W.P.Lehmann (1972:989) pointed out, such languages that are ambivalent and undecided between word order SOV and SVO, as was the case with Latin, developed absolute constructions (in Latin the Ablative Absolute, in Late Latin the Accusative and Nominative Abso­ lute) . I think the same happened in these cases of the AcI construction. In this sense, M.Bolkestein's explanation (1979) that the Accusative in the AcI de­ pends on the Infinitive, is more right than I thought in Calboli 1980:204 f., al­ though the Infinitive is not enough as is demonstrated by the construct of the Historical Infinitive. The fluctuation between SOV and SVO word order sus­ tains the AcI for as long as this construction remains vital: when the SVO order is fixed, the process that at first froze the constructions containing nomi­ nal forms of the verb in absolutives, arrives at its completion. But it provokes the fall of the cases (at least certainly for the Subject-Object opposition) and it also provokes the fall of the mechanism that sustains AcI as well and, in turn,

THE DEVELOPMENT OF LATIN

53

the fall of the AcI helps in developing the Romance article.15

NOTES 1) W.P. Lehmann 1980:138; for all this section cf. F. Bader 1973; Ch. Lehmann 1979:436-458; W.P.Lehmann 1980:117-140. 2) In fact, the particle nu is more vital in later Hittite than in old Hittite, but we find sentences with nu and kuit and sentences only with kuit (an inflected form such as quod, quia in Latin), cf. J. Friedrich 1960:159. 3) Cf. at last K.H. Schmidt 1980:102 f. 4) In order to understand what R.L. Trask 1979 means with the Type B of ergative languages see Trask 1979: 397: "I propose that this is the origin of Type B ergativity: the incorporation into the in­ flectional paradigm of a nominalized deverbal form with stative force". I would call attention to the following good Trask's remark (1979:399 f.): "The best that I can suggest is that Type B ergativity depends on the existence of a verbal adjective, a participle, and SOV languages, I have the strong impression, are particularly rich in participial forms, and inclined to use them in many cir­ cumstances in which SVO and VSO languages would use finite verb forms1*. 5) I would ask: is it accidental that the explanation through 'subject-raising' is easier in this pas­ sive construction? Or are we dealing in Latin with what is left of an old stage where the infinitive or the Adjective of the complement clause were more closely related to the verb of the main sentence? I would rather agree with this second opinion, since this opinion is more in accordance with the facts pointed out by R.L. Trask 1979. 6) On the inflection of Infinitive cf. W. Boeder 1980. 7) Cf. J.B. Hofmann-A. Szantyr 1965:143 f.; 847. 8) The nature of preposition, or rather, similar to preposition is developed in time above all in cases such as astante civibus (CIL V 895), excepto oppida (lord. Get. 1,6) and in Romance preposi­ tions such as durante, nonostante, medianie, cf. E. Flinck-Linkomies 1929:64; A.L. Castelli 1976:62. 9) Cf. P. Perrochat 1932:75-82; G. Calboli 1962:93-107; J.B. Hofmann-A.Szantyr 1965:665. 10) See in particular pp.70 f.; for example, Gerbert de Reims (Pope Silvester II, 940/45-1003) who had a good education in Aquitaine uses AcI more (97%) than any other of the authors consid­ ered by Dr. Wirth-Poelchau, but notwithstanding this the same writer makes use of quod + finite verb "da, wo eine Anpassung an den gängigen, eher volkstümlichen lateinischen Stil erfolgte". 11 ) For the complex question of Epicurean doctrine present in this text see the rich comment of A.S. Pease, M. Tulli Ciceronis, De natura deorum I, Wissenschaftliche Buchgesellschaft, Darmstadt 1968, pp.300-302, and the edition of G. Arrighetti, Epicuro, Opere 2 , Einaudi, Torino 1973,pp.l21;545-548. 12) But we must not forget that in early Indo-European the case system was quite simple and consisted of two cases, i.e. of Nominative and Objective, the first with ending *-ø, the second with *-N as is believed by K. Shields 1978:194. Therefore, the opposition between Nominative and Ob­ jective (Accusative) was the background of the case system later developed.

54

GUALTIERO CALBOLI

13) Here I must repeat that what I wrote in Calboli 1980:203: "Wenn mehrere koordinierte Nebensätze den Nom.c.Inf. annehmen sollten, wird nur der erste Nebensatz mit dem Nom.c.Inf. konstruiert, während die übrigen den A.c.I. annehmen [...]. Dasselbe Phănomen findet sich auch beim proleptischen Akkusativ, und zwar mit dem alleinigen Unterschied, daB in den anderen beigeordneten Nebensätzen die proleptische NP nach dem ersten Nebensatz mit dem Nominativ, und nicht mit dem Akkusativ ausgedrückt wird: Cael.Cic.epist.8,10,3 nosti Marcellum quam tar­ dus sit etparumefficax, itemque Servius quam cunctator". This confirms my view that in the pres­ ence of a node Š (i.e. in a finite verb clause) the Nominative holds, while without a node Š (i.e. in infinitive clause) the Accusative is introduced. 14) In Russian the use of both the present and the past participles, active and passive, instead of relative or subordinate sentences must be noted. This detail, with the lack of person in the past tense, helps, I think, in explaining the situation of Russian where both Article and AcI are missing, while I believe that both the use of AcI and the presence of Article are related in Latin and Ro­ mance languages. This explanation is already partially given by G. Calboli 1978:252 f. 15) As I thought already in Calboli 1978.

BIBLIOGRAPHY Bader, Françoise, (1973), 'Lat. NEMPE, PORCEO et les fonctions des par­ ticules pronominales', Bulletin de la Société de Linguistique de Paris 68: 2775. Boeder, Winfried, (1980), 'Zur Rekonstruktion von Infinitivkonstruktionen im Indogermanischen', in: P. Ramat (Ed.) 1980:207-224. Bolkestein, A.M., (1979), 'Subject-to-Object Raising in Latin?'. Lingua 48: 15-34. Brugmann, Karl, (1916), Grundriss der vergleichenden Grammatik der in­ dogermanischen Sprachen, 2.Band, 3.Teil, Zweite Bearbeitung, Strassburg: K.J. Trübner. Calboli, Gualtiero, (1962), Studi Grammatically Bologna: Zanichelli. , (1978), 'Die Entwicklung der klassischen Sprachen und die Beziehung zwischen Satzbau, Wortstellung und Artikel', Indogermanische Forschungen 83:197-261. , 'Bemerkungen zum Akk.c.Inf. und zu verwandten Konstruktionen im Lateinischen', in: G. Calboli (Ed.) 1980: 189-208. , (Ed.), (1980), Papers on Grammar I, Bologna: CLUEB. , (1981), 'Le frasi interrogativo-esclamative latine e l'infinito', in: Ch. Rohrer (ed.) (1981 Logos Semantikos. Studia Linguistica in honoreur E. Coseriu, IV, Madrid-Berlin: Gredos-de Gruyter: 133-153. Castelli, Anna Luisa, (1976), Tl tipo praesente nobis', Studi italiani di lingui­ stica teorica ed applicata 5: 57-78.

THE DEVELOPMENT OF LATIN

55

Chomsky, Noam, (1980), 'On Binding', Linguistic Inquiry 11: 1-46. , (1981a), Lectures on Government and Binding. Dordrecht: Foris. , (1981b), 'Principles and Parameters in Syntactic Theory', in: Hornstein, N. & W. Lightfoot (eds.) Explanation in Linguistics, London, Longmans, 32-75. Chomsky, Noam & Howard Lasnik, (1977), 'Filters and Control', Linguistic Inquiry 8:425-504. Flinck-Linkomies, Edwin, (1929), 'De ablativo absoluto quaestiones', An­ nales Academiae Scientiarum Fennicae, Ser.B., Tom.XX, No:l, Helsinki: Suomalainen Tiedeakatemia. Flobert, Pierre, (1975), Les verbes déponents latins, des Origines à Char­ lemagne, Paris: "Les Belles Lettres". Friedrich, Johannes, (1960), Hethitisches Elementarbuch, l.Teil, 2.Auflage, Heidelberg: C. Winter. Herman, József, (1963), La formation du système roman des conjonctions de subordination, Deutsche Akademie der Wiss., Institut für Romanische Sprachwiss. Nr. 18, Berlin: Akademie-Verlag. Hofmann, Johannes Baptista & Anton Szantyr, (1965), Lateinische Syntax und Stilistik, München: C.H. Beck (= 1972). Hyart, Charles, (1954), 'Les origines du style indirect latin et son emploi jusqu'à l'époque de César', Mémoire de l'Académie Royale de Belgique, Cl. des Lettres, Tome XLVIII, Fasc.2, Bruxelles: J. Duculot. Justus, Carol, (1976), 'Relativization and Topicalization in Hittite', in: Ch.N.Li (Ed.) 1976:213-245. Kirk, W.H., (1938), 'Passive verba sentiendi et dicendi with Declarative In­ finitive'. Classical Philology 33:182-187. Kühner Raphael-Carl Stegmann, (1955), Ausführliche Grammatik der lateinischen Sprache, Satzlehre, 3.Auflage durchgesehen von Andreas Thierfelder, Leverkusen: Gottschalk. Kurzová, Helena, (1968), Zur syntaktischen Struktur des Griechischen I In­ finitiv und Nebensatzl, Amsterdam-Prag: A.M. Hakkert-Academia. Lehmann, Christian, (1979), Der Relativsatz, Typologie seiner Strukturen, Theorie seiner Funktionen, Kompendium seiner Grammatik, Arbeiten des Kölner Universalien-Projekts, Nr.36. Lehmann, Winfred Philipp, (1972), 'Contemporary Linguistics and IndoEuropean Studies', Publications of the Modern Language Association of America 81: 976-993. , (1974), Proto-Indo-European Syntax, Austin and London: University of Texas Press.

56

GUALTIERO CALBOLI

, (1976), Trom Topic to Subject in Indo-European', in: Ch.N.Li (Ed.) 1976: 447-456. , (1980), T h e Reconstruction of Non-Simple Sentences in Proto-IndoEuropean', in: P.Ramat (Ed.) 1980: 113-144. Li, Charles N. (Ed.), (1976), Subject and Topic, New York-San FranciscoLondon: Academic Press. Lindskog, Claes, (1896), 'Zur Erklärung der Accusativ-mit-Infin. Konstruk­ tion im Latein', Eranos 1: 121-135. Maraldi, Mirka, (1980), T h e Complement Structure of Perception Verbs in Latin', in: G. Calboli (Ed.) 1980: 47-79. Mikkola, Eino, (1964), Die Abstraktion, Begriff und Struktur, 'Sprache und Denken', Finnische Beiträge zur Philosophie und Sprachwissenschaft, Bd.I, Helsinki: Suomalainen Kirjakauppa. von Nägelsbach, Karl Friedrich-Iwan Müller, (1905), Lateinische Stilistik, 9. Auflage, Nürnberg: K. Geiger. Perrochat, Paul, (1932), Recherches sur la valeur et l'emploi de l'infinitif sub­ ordonné en latin, Paris: "Les Belles Lettres 1 '. Plank, Frans, (1979), Ergativity, Towards A Theory of Grammatical Rela­ tions, London-New York: Academic Press. Porzio Gernia, Maria Luisa, (1974), 'Contributi metodologici allo studio del latino arcaico. La sorte di M e D finali', Memorie delia Accademia Nazionale dei Eincei, CL di Sc. morali, storiche e filologicke, Ser. VIII, Vol. XVII,4: 113-337. Prat, Louis C , (1975), Morphosyntaxe de l'ablatif en latin archaïque, Paris: "Les Belles Lettres". Ramat, Paolo (Ed.), (1980), Einguistic Reconstruction and Indo-European Syntax, Proceedings of the Colloquium of the Tndogermanische Gesellschaft', University of Pavia, 6-7 September 1979, Amsterdam: J.Benjamins. Renou, Louis, (1937), Monographies Sanskrites: Ee suffixe -tu- et la construc­ tion des infinitifs, Paris: Maisonneuve. Rohlfs, Gerhard, (1969), Grammatica storica della lingua italiana e dei suoi dialetti, Sintassi e formazione delle parole, Trad.ital. di T. Franceschi e M. Caciagli Fancelli, Torino: G. Einaudi. Schmidt, Karl Horst, (1980), 'Zur Typologie des Vorindogermanischen', in: P. Ramat (Ed.) 1980:91-112. Schwendener, Ulrich, (1923), Der Accusativus cum Infinitiv o im Italienischen, Inaugural-Dissertation Bern, Säckingen am Rhein: G. Mehr.

THE DEVELOPMENT OF LATIN

57

Shields, Kenneth, Jr., (1978), 'Some Remarks Concerning Early Indo-Euro­ pean Nominal Inflection', Indo-European Studies 6:185-210. Tekavčić, Pavao, (1972), Grammatica storica dell'italiano, Volume II: Morfosintassi, Bologna: Il Mulino. Thumb, Albert & Richard Hauschild, (1959), Handbuch des Sanskrit, II.Teil: Formenlehre, 3.Auflage, Heidelberg: C. Winter. Touratier, Christian, (1980), Larelative, Essai de théorie syntaxique, Paris: C. Klincksieck. Trask, Robert L., (1979), 'On the Origins of Ergativity', in: F. Plank (Ed.) 1979:385-404. Vairel-Carron, Hélène, (1975), Exclamation, ordre et défense, Analyse de deux systèmes syntaxiques en latin, Paris: "Les Belles Lettres". Vaananen, Veikko, (1966), 'Le latin vulgaire des inscriptions pompéiennes, 3ème édition', Abhandlungen der deutschen Akademie der Wissenschaften zu Berlin, KL Sprachen, Literatur u. Kunst 1958, Nr.3, Berlin: Akademie. Wirth-Poelchau, Lore, (1977), AcI und quod-Satz im lateinischen Sprachgebrauch mittelalterlicher und humanistis cher Autoren, Inaugural-Disserta­ tion Erlangen-Nürnberg: J. Hogl Offset-Fotodruck.

LE SYSTEME DES CAS LATINS Aperçu d'une théorie morpho - sémantique. 1

P.de CARVALHO Université de Bordeaux-III

I. Le modèle d'explication du système des cas latins dont j'ai l'honneur de vous présenter ici les grandes lignes repose sur un postulat, avec les deux refus qui en découlent, et sur une exigence. I l: Le postulat est celui de la linguistique psycho-mécanique (G. Guillaume), à savoir: le signifié d'une forme linguistique ne se laisse point réduire aux "valeurs de sens" à la production desquelles cette forme peut être appelée à participer dans le discours. Il se définit plutôt comme un cas général de représentation (de "concevabilité", disait G. Guillaume) engendré par un "psycho-mécanisme" et formant, avec d'autres cas généraux de représentation engendrés par le même psycho-mécanisme, un système rigoureusement cohérent, cette cohérence étant la condition même de son fonctionnement. Ce système doit donc être conçu comme le programme ordonné d'un trajet virtuel, l'activité du locuteur consistant, dans cette pers­ pective, à accomplir effectivement ce trajet, et à actualiser telle ou telle des positions successives qu'il comporte. I2: En conséquence de quoi ne peuvent être retenues, même si l'on est prêt à reconnaître leurs apports respectifs: 1° la théorie historico-comparativ'e classique2, avec tout ce qu'elle a pu entraîner dans son sillage, et notamment par ses échecs, à savoir: l'idée d'un syncrétisme casuel (ablatif latin, génitif et datif grecs); l'opposition cas gram­ maticaux, non-sémantiques, formels vs. cas concrets, sémantiques, imaginée pour répondre à la résistance de certains cas (nominatif, génitif, accusatif) à une description en termes de "valeurs sémantiques"; l'opposition fonction sémantique vs. fonction grammaticale d'un cas donné (J. Kurylowicz, 1949: 135-140); de même, toute analyse inspirée de la "grammaire des cas", dont l'articulation centrale, l'opposition cas ou rôles casuels profonds vs. formes

60

P. DE CARVALHO

casuelles de surface s'enracine elle aussi, pour s'en nourrir, dans l'impossibi­ lité avérée de reconnaître à ces dernières des "sens particuliers clairs" (Ch.J. Fillmore, 1968: 8-9). Enfin, il me faut refuser aussi, comme issue de la même veine et comme aussi peu explicatrice, la proposition, plus récente, tendant à distinguer morphèmes casuels et formes casuelles, ressortissant, celles-ci, à un niveau "morphologique" stricto sensu, purement "formel" (Chr. Touratier, 1978; 1979); 2° les descriptions strictement "syntaxiques", consistant à définir un cas donné tantôt par son aptitude à une ou à plusieurs modalités de relations syn­ taxiques (J. Haudry, 1977: 10: "la valeur d'un cas ne peut se définir qu'à l'intérieur d'une fonction"), tantôt par la vertu, qu'on lui prête, de signaler certains processus transformationnels (Kuryłowicz, 1949: 145; 1964: 180; Benveniste, 1962; J. Haudry, 1968; 1970). Outres qu'elles sont notoirement partielles, incapables de rendre compte d'une quantité non négligeable de faits dans le domaine qu'elles prétendent recouvrir3, ces explications n'en sont pas à vrai dire, car elles s'en tiennent à l'écume des choses. Si, par exemple, le génitif latin assume le plus souvent, et plus volontiers, sans doute, que tout autre cas, la fonction de détermination adnominale, ou s'il se prête effective­ ment à ces exercices transformationnels, c'est en vertu de la représentation qui s'attache, dans le système de la langue latine, à l'unité morphologique qui porte ce nom. C'est cette représentation qu'il s'agit de découvrir, en resti­ tuant, par les voies d'une spéculation abstraite suffisamment hardie -mais tou­ jours soumise à l'observation la plus scrupuleuse des faits d'emploi -, le pro­ cessus psycho-mécanique qui permet de l'engendrer. I 3: Une exigence, enfin, commande la démarche que je propose à votre attention: celle de ne pas mépriser, comme on n'est parfois que trop porté à le faire, les formes apparentes du langage. Dans son effort pour se constituer en science , la linguistique moderne a tendu, de plus en plus, à ne pas s'en tenir aux apparences, et même à s'en défier, ce qui était et reste son droit le plus légitime, je dirais même: son devoir. Elle a su, en procédant ainsi, montrer que les formes directement observables du langage, les énoncés aux­ quels aboutit toute activité langagière, se laissent ramener à des unités plus petites et surtout plus "abstraites" (phonèmes, monèmes, morphèmes, lexè­ mes, etc.) qui, pour n'être plus immédiatement accessibles, n'ont pas pour autant moins de réalité. Cela dit, le langage étant en définitive affaire de forme, ses formes visibles sont elles aussi à considérer, à expliquer: pour­ quoi sont-elles ce qu'on les voit être? C'est dire qu' avec quelques autres je m'accommode mal de certaine attitude, trop répandue, qui conduit à glisser,

LE SYSTEME DES CAS LATINS

61

presque insensiblement, de cette salutaire défiance de principe à une franche et hostile méfiance à l'endroit de la face visible du langage. On agit là "comme si le vrai du langage avait un jour été diaboliquement enterré, comme si un malin génie avait jété sur lui un camouflage farceur et sournois ; comme si, par conséquence, (le) premier soin (de la linguistique) devait être de déblayer, de gratter et, au plus vite, de jeter à la décharge ce mauvais dépôt." (J.-Cl. Che­ valier, 1978: 95). Dans l'affaire qui nous occupe, le respect, au sens étymologique du terme, de ces formes apparentes m'impose de partir d'un fait inévitable: il existe en latin une déclinaison nominale comprenant six postes, ni plus ni moins. Même si l'expression matérielle, physique, "sémiologique"4 de ces postes n'est pas constante, même s'il arrive le plus souvent que deux ou plu­ sieurs d'entre eux se trouvent réunis sous un même signe, selon des affinités dont il faut, d'ailleurs, rendre raison, le fait est là: chacun de ces six cas est compétent pour remplir, seul ou en concurrence avec un autre, un certain nombre de fonctions syntaxiques dans certains contextes, et parfaitement incompétent pour tout le reste : on dit eo Romam comme Romam uidi, mais on ne saurait dire *Romam memini; pour désigner l'actant complémentaire impliqué dans potior, on utilisera tantôt l'ablatif -potior urbe, comme gladiis pugnatum est -, tantôt le génitif - rerum potitus est, comme fuga Pompei, ou comme uiuorum memini -, mais jamais le nominatif, ou le datif: *urbs potiri, *urbipotiri. Pourquoi? Bref, c'est le principe de ces compétences et de ces incompétences qu'il faut s'attacher à découvrir, quelque surprenantes que puissent paraître, du point de vue de nos propres schémas linguistiques, les identités et les nonidentités qui s'expriment à travers elles. A perdre de vue cet impératif, on ris­ que à tout moment de s'égarer dans les méandres d'une spéculation purement "logicienne", sans prise sur la réalité observable et, de ce fait, incontrôlable. II.

La théorie

II 1. Quelques définitions élémentaires: (a) phrase: être linguistique de discours livrant la représentation (la "re­ présentation") momentanée, evanescente, d'un moment singulier de l'expé­ rience d'un locuteur, d'un "cas d'événement". Toute phrase institue un "présent-dont-on-parle", un présent délocuté; (b) nom: être linguistique de langue apportant, partout où il existe -c'est le cas, notamment, dans les langues indo-européennes, la representation d'une substance sémantique conçue sous forme "adynamique" ďespace, par

62

P. DE CARVALHO

opposition à la représentation sous forme "dynamique" de temps qui est le fait du verbe dans ces mêmes langues. Une propriété morpho-sémantique du nom est qu'il ressortit à la personne troisième, délocutée, absente de l'interlocution. Parler de quelqu'un ou de quelque chose en le nommant c'est ipso facto de déclarer "non-moi'7"non-toi"; c'est aussi en faire un objet (délocuté); (c) cas: représentation, affectant toute notion nominale, du rapport de l'objet délocuté (= 3ème. personne nominale) à l'ensemble de l'édifice mental représentatif d'un cas d'événement (= phrase). On opposera le cas unique de la plupart des langues romanes modernes, traduisant le refus de prévoir ce rapport (selon une certaine intuition: l'espace existe a priori, avant le temps, donc "en-dehors" du temps, et ce n'est que secondairement qu'il participe au temps), au cas multiple des langues indo-européennes anciennes, traduisant la prévision de ce rapport avant la déroulement linéaire de la phrase (conformé­ ment à une autre intuition, moins "abstraite" que la première: aucun être nommable n'existe ailleurs que dans le temps qui l'a apporté et qui l'emportera). De cette donnée morphologique première découle une différence importante au niveau proprement "syntaxique". Le cas unique renvoie la responsabilité de la structuration du cas d'événement à la "phrase" proprement dite, plus précisément à la concaténation d'un terme (désormais) absolument "nonprédicatif" - le nom - et d'un terme (désormais) absolument "prédicatif" - le verbe -, avec cette conséquence que la structure ainsi engendrée se confond désormais avec ce que prescrit la définition sémantique d'un verbe particu­ lier; la phrase, alors, n'existe et ne tient que par le verbe. Au contraire, le cas multiple des langues indoeuropéennes anciennes implique que la structure du cas d'événement pré-existe à l'apparition de tout mot verbal. Celui-ci n'a plus, dès lors, de fonction structurante, il n'a plus qu'une fonction dénominative. On comprend dans ces conditions qu'il puisse être réservé pour la clôture de la phrase et même, à la limite, ne pas se présenter (P. de Carvalho, 1978: 241). Voilà qui nous interdit de décrire les cas latins par référence aux propriétés sémantiques du verbe, comme s'ils dépendaient de ces propriétés. Dans une langue telle que le latin, la forme casuelle du nom exprime la relation de l'être nommé à l'ensemble d'un ''présent délocuté" (identifié, ensuite, le cas échéant, par un mot verbal); elle n'exprime jamais une quelconque "dépen­ dance" du nom par rapport à un verbe particulier. 5 II 2. Le système casuel latin est constitué par un ensemble ordonné de positions représentatives des différents rangs concevables, dans cette lan­ gue, de participation de la notion nonviable à la structure d'un présent délo­ cuté. Son ressort profond n'est rien d'autre que la problématique de la per-

LE SYSTEME DES CAS LATINS

63

sonne. Les oppositions qu'il comprend s'expliquent comme le résultat d'une projection, sur la matière à ériger en phrase, de l'affrontement basial moi vs. non-moi. Le système casuel commence à exister en pensée dès l'instant où l'on "réalise" que tel "non-moi" à nommer (objet délocuté) dans le présent de locution présente certaines propriétés qui en font, en réalité, l'homologue du moi. Ainsi, le premier temps de l'édification mentale du système casuel consiste en la discrimination d'une personne délocutée que le moi locuteur voit participer au même titre que lui à l'événement réellement présent qu'est la production de la phrase, celle qui lui apparaît comme le pôle conclusif de cette parole dont il se voit, lui, être le pôle fondateur. Tel est le signifié du cas VOCATIF, représentation de la première personne délocutée, du premier non-moi rencontré par le moi, celui qui est le plus proche de lui, soit: une "personne-dont-je-parle-comme-1'objet-visé-par-ma-parole." Le vocatif consti­ tue à lui seul toute une dimension du système des cas latins. A ce titre, il s'oppose à l'ensemble des cas restants, qui ont en commun de situer le nonmoi nommé à l'intérieur de l'événement que la phrase représente, soit: comme un élément, de rang variable, d'une situation qui n'est plus, rigoureusement parlant, une expérience "présente", mais une expérience "passée" par rap­ port à la phrase qui la "reconstruit." Le deuxième moment du psycho-mécanisme casuel consiste à discerner et à poser un non-moi qui apparaît investi, au sein du présent délocuté, des propriétés de centralité et autonomie que le moi s'attribue dans le temps réel de son existence. Il en résulte le cas NOMINATIF, qui livre la représentation du protagoniste (actif, inactif ou passif) du "petit drame" (L. Tesnière, 1966: 102) : l'être ou chose que l'on voit affronter, en toute autonomie, le devenir qui lui échoit, sous la forme d'une "action", d'une "inaction" ou d'une "passion." Les moments subséquents livreront, par opposition, des images pleine­ ment "objectives" du non-moi délocuté, chacun d'eux représentant un nonprotagoniste de plus en plus éloigné de la position première. A savoir: -ACCUSATIF: non-protagoniste "privilégié", parce qu'appartenant au plan de l'existence du protagoniste, plan dont il constitue le pôle terminal. L'opposition NOMINATIF/ACCUSATIF reproduit, on le voit, dans le "re­ présenté", l'affrontement moi/toi constitutif du présent réel de la parole. Et tout comme le toi n'est rien d'autre qu'un moi "virtuel", "subordonné", le non-protagoniste "présent", "essentiel", "constitutif signifié par l'ACCUSATIF se laisse lui aussi caractériser comme un "sujet virtuel", "subordonné", voire comme un "quasi-sujet", à condition, bien entendu d'appartenir

64

P. DE CARVALHO

à la classe sémantique des "êtres" (animés ou inanimés); -ABLATIF: non-protagoniste non "présent" (non "constitutif du présent délocuté") mais actuel, contemporain du déroulement opératij du présent délocuté: "relación de contiguidad externa" (L.Rubio, 1966:155). Relève de la compétence du cas ABLATIF toute réalité spatiale aperçue dans le champ du présent délocuté, une fois que l'on y a élu un protagoniste et un sous-protago­ niste, ou que l'on a renoncé à le faire; -DATIF: non-protagoniste inactuel ultérieur, supposant l'existence d'un présent délocuté, et en recueillant durablement les apports positifs ou néga­ tifs. Le DATIF est, en quelque sorte, l'homologue du FUTUR; -GENITIF: non-protagoniste inactuel antérieur. C'est le cas de toute unité d'espace exclue du présent délocuté, de ses "marges" et de sa subsé­ quence résultative. Relève de la compétence du GÉNITIF tout ce qui reste à mentionner comme indispensable à la construction du cas d'événement délo­ cuté, une fois que l'on en a décrit et le déroulement opératif et les résultats. C'est, de tous, le cas le plus "négatif", dans la mesure où il traduit le refus de toutes les autres positions systématiques. En tant que tel, il renvoie le nonmoi délocuté à l'arrière-plan, au PASSÉ du présent délocuté.6 La figure 1 résume schématiquement l'ensemble de cette description:

"Moi"

avant

AVANT son "actualité

"DEDANS" large" après

I voc

III ACCUS

après

IV ABL.

le "dehors" du "DEDANS"

VDAT.

=le "dedans" du "DEHORS"

FIG. I: LE SYSTEME DES CAS LATINS: une série ordonnée d'actes de position livrant successivement différents "cas" de l'unité d'espace que "MOI" trouve devant soi comme OBJET de son discours ( = objet délocuté)

NOM.

avant

du "DEDANS"

arnere-plan vu "après coup"

son

après

VI GEN.

= le "dehors" du "DEHORS"

son "prolongement"

avant

APRES ce qu'il comporte d'inactuel

"DEHORS"

CAS D'EVENEMENT REPRESENTE avec ses "tenants et aboutissantes"

APRES

l'actualité stricte du présent délocuté= le "dedans" ses"marges" =

PRESENT REEL DE LA PAROLE

AVANT

LE SYSTEME DES CAS LATINS

66

P. DECARVALHO

Tel est le programme on parcours virtuel de positions en quoi consiste le système casuel latin. Cela dit, et au risque de vous paraître paradoxal, il doit être clair qu'à mes yeux le cas "central" de ce système, le pivot en fonction duquel il apparaît construit et autour duquel il "tourne", c'est-à-dire son terme non-marqué, par rapport auquel tous les autres, nominatif compris, se définissent comme comportant des spécifications particulières, n'est autre que l'ACCUSATIF. Je rejoins ici tout à fait une tradition respectable, repré­ sentée par exemple dans la Grammaire Latine de Madvig (1876: 226). Pour préciser ma pensée,, si vous me le permettez, j'utiliserai ici une métaphore. Représentons-nous le chemin de fer qui a conduit certains d'entre nous de Paris à Amsterdam. Voilà en effet un "programme de positions" ou "parcours virtuel" que l'on peut concevoir comme venant de Paris à Amsterdam, ou allant d'Amsterdam à Paris. Et cependant, pour l'habitant, par exemple, de Rotter­ dam, la première position de ce programme, celle qui lui est la plus immédia­ tement accessible, celle par laquelle il commence, lui, son utilisation person­ nelle de ce parcours virtuel, est bien celle qui se trouve dans sa ville; c'est bien par Rotterdam qu'il a accès au parcours, soit pour "remonter" vers Amster­ dam, soit pour "descendre" vers Paris. Telle est la position de 1'ACCUSATIF dans le système des cas latins. Il livre la forme la plus immédiate du nom. Non pas, à vrai dire, une forme "générale et indéterminée" (Madvig, 1876: 227); celle, plutôt, qui ne comporte qu'une "détermination minimale" du nom en tant que tel, c'est-à-dire en tant que représentation d'un non-moi délocuté "présent". En d'autres termes: on met à I'ACCUSATIF le nom de toute réa­ lité spatiale dont on constate qu'elle constitue, pour une part, le présent délo­ cuté, sans que, pour autant, l'on soit en mesure d'ajouter qu'elle en est le pro­ tagoniste. Si bien que l'on peut représenter schématiquement comme suit (fig. 2) le système des cas latins, vu, cette fois, sous l'angle du locuteur-actualisateur: III.

Conclusion

Le temps me manque ici pour entrer dans le détail de la justification des emplois casuels à la lumière du modèle proposé. Je me bornerai donc, en guise de conclusion, à quelques observations, en me reposant, pour le reste, sur votre propre expérience du latin, et en me permettant de vous renvoyer à des publications ultérieures: 1° Au niveau des emplois, voici quelques faits dont ce modèle rend compte immédiatement: Romam uidi ~ Romam eo~decem annos regnauit ~ante castra considunt: l'objet d'un verbe transitif, le lieu d'aboutissement

SORTIE I VOCATIF

SORTIE III ACCUSATIF

"EN

AVAL

FIG. II: Schéma d'actualisation du systéme des cas latins.

SORTIE II NOMINATIF

"EN AMONT"

SORTIE IV ABLATIF "

SORTIEV DATIF

SORTIE VI GENITIF

LE SYSTEME DES CAS LATINS

68

P. DE CARVALHO

d'un mouvement, la durée d'accomplissement d'une opération, le lieu vers lequel est orientée une existence sont des cas particuliers de la représentation du "non-moi présent "sans plus", donc subordonné (=ACCUS.); mep udet, paenitet, etc.: accus. du non-moi présent face à un état de fait qu'il est aussi impuissant à révoquer qu'à assumer (P. de Carvalho,1980 b : l l ) ; uiuorum memini. me pudet stultitiae meae: génitif de l'objet situé dans le "passé" de l'opération relatée. Devient ausssi immédiatement intelligible l'alternance ACCUS./ABL./GEN . dans la désignation de l'actant 2 de certains verbes transitifs: urbem capio~urbe potior~rerum potior. Dans urbem capio, le verbe désigne une opération allant d'un commencement (y) à une fin (x), avec x = ACCUS. con­ formément à notre définition. Dans urbe potior, la forme passive ("déponente") du verbe est à prendre à la lettre: elle dit clairement que tout se passe dans les limites d'une même et seule personne, y (C → F) ; l'actant 2 ne peut plus appa­ raître comme le rôle conclusif du présent délocuté; comme il est néanmoins "présent", il lui reste la position marginale signifiée par l'ABLATIF. Enfin, dans rerum potior, on a un "actant 2" qui n'est pas strictement actuel, qui, littérale­ ment, ne tient pas dans les limites étroites du présent délocuté 7 ; comme le verbe, de son côté, interdit toute vision d'un résultat extérieur au sujet, il ne reste à l'actant 2 que la position d'arrière-plan. Aussi bien, potiri rerum ne signifie pas tant "prendre effectivement possession d'un objet actuel" (urbe potiri) que "s'installer en personne dominante au centre d'un espace, de façon à exercer un contrôle virtuel sur tout ce qui peut s'y passer." 2. On notera que le modèle proposé élimine purement et simplement quelques (faux) problèmes traditionnels: priorité du génitif "adnominal" sur l'"adverbal", ou vice-versa; accusatif "sujet d'infinitif", qui n'est pas autre chose, on le voit, qu'un sujet privé de son autonomie syntaxique et sémanti­ que, puisque soumis à la rection d'un autre verbe. 8 3. La description "positive" du NOMINATIF fournie par notre modèle s'accorde avec la marque positive dont le signifiant de ce cas est souvent affecté (J. Perret, 1957:154) - dans des conditions qui restent, d'ailleurs, à décrire - et, surtout, permet de saisir et justifier certaines contraintes limitant la formation d'un sujet en latin, cf. notamment: l'impossibilité d'un sujet "pur bénéficiaire", d'où mihi filius est opposé à filium habeo, expression de l'emprise effectivement exercée par un "sujet" sur un "objet" (P. de Carvalho, 1980 b:9); la résistance, confinant à l'impossibilité, à traiter des "noms de lieux" comme sujet d'un verbe opératif : *Roma Carthagini bellum indixit; etc. 4. Au plan théorique, notre modèle rend intelligibles les "synapses" qui caractérisent la "sémiologie" de la déclinaison latine: (a) ablat.dat. = cas

LE SYSTEME DES CAS LATINS

69

de l'au-delà du présent délocuté stricto sensu; (b) dat./gén. = cas de l'inactuel, (c) les regroupements sémiologiques qui, au pluriel, réduisent souvent la déclinaison à 3 signes: duces = "prés. délocuté"; ducihus: son "après" ; ducum; son "avant", etc. D'un point de vue comparatif, on aperçoit la singularité de l'ablatif latin, et l'on s'explique pourquoi le grec, par exemple, a fait l'écono­ mie de ce cas "contradictoire" ( = le dehors du dedans!). Le modèle fait aussi pleinement apparaître la position "axiale" de l'opposition ACCUS./ABL., que confirme, en quelque sorte, l'évolution romane du latin (réduction de la décli­ naison à un "accusatif-ablatif" envahissant, devenu cas-régime universel: G. Serbat, 1975:69). Il permet, enfin, de résoudre le problème des rapports forme casuelle/préposition, en termes non pas d'exclusion réciproque, mais de complémentarité fonctionnelle (P. de Carvalho, 1980 a:48).

NOTES 1) Résumé de la deuxième partie d'une thèse en voie d'achèvement, portant sur le mode de représentation du nom en latin. Voir aussi P. de Carvalho, 1980 a; 1980 b; 1980 c. 2) Sur la confusion, qui caractérise cette conception, entre "valeurs résultant de l'emploi du cas" et "signifié casuel", cf. la critique tout à fait pertinente de L. Rubio, 1966: 156-157. 3) Cf., pour le génitif, la critique formulée par J. Perret (1966:471). 4) "La visée de la psycho-sémiologie est partout et toujours de rendre la psycho-systématique par des moyens qui soient propres à en dénoncer la cohérence intrinsèque. Mais si c'est là un but visé à la continue, il s'en faut de beaucoup qu'il soit régulièrement atteint. Ce but représente l'idéal du mieux, mais la réalité suffisante du bien tolère des en-deçà. La raison en est que, s'il est histori­ quement pressant que ce qui a été construit en pensée soit accroché à une sémiologie, il est moins pressant que cette sémiologie soit parfaite. Aussi, pratiquement, ne lui est-il demandé que d'être suffisante". (G.Guillaume, 1964:243). 5) Cela rejoint certaines analyses récentes tendant à affirmer la primauté syntaxique du cas sur le verbe et invitant à interpréter le signifié d'une forme casuelle donnée en termes deposition plutôt que de fonction (par rapport à un terme régissant): H. Vairel, 1975:63 et 153). 6) On rapprochera cette analyse du concept de "détermination lâche" utilisé par J. Perret à pro­ pos du génitif (1966: 474) et, surtout, de la définition de ce cas par J. Veyrenc: "génitif = nominalisation = désactualisation, virtualisation." (1974: 256). 7) De même, dans les exemples suivants, le complément au génitif a beau être constitué par le nom d'un objet singulier, celui-ci n'apparaît pas moins pour autant porteur d'une virtualité, et c'est celle-ci, visiblement, qui est en cause dans les contextes considérés: deiera te mi argentum daturum ... uiduli ubi sis potitus (Plaute, Rud. 1337) ; se tertium esse cui fatum foret urbis potiri (Sall. Cat. 47, 2). 8) Il me semble que la discussion autour de ce problème a été considérablement obscurcie par la tendance persistante à raisonner comme si le verbe avait, en latin, le même rôle structurant qu'on lui voit exercer dans la syntaxe indo-européenne moderne. Evidemment, si l'on commence par

70

P. DE CARVALHO

réduire l'ACCUSATIF à la fonction d'objet d'une classe de verbes (quelle que soit l'étendue de cette classe), on est mal placé pour s'expliquer l'emploi de cette forme casuelle dans des énoncés où, de toute évidence, la réalité spatiale nommée n'entretient aucune relation objectale avec le verbe à forme personelle de la phrase. Voir à ce sujet les arguments avancés, à juste titre, par A. Machtelt Bolkestein (1979: 15-34), contre toute analyse réduisant l'accusatif "sujet d'infinitif" austatut d'objet.

BIBLIOGRAPHIE Benveniste, E., (1962), 'Pour l'analyse des fonctions casuelles: le génitif latin.' Lingua, 11. Repr.: Benveniste, E., 1966. Problèmes de linguistique générale. Paris, Gallimard: 140-148. Bolkestein, A.M., (1979), 'Subject-to-object raising in Latin?' Lingua, 48, 15-34. de Carvalho, P. (1978), 'Syntaxe et sémantique: verbe et phrase en latin.' Revue des Etudes Anciennes, 80, 239-247. . (1980a), 'Cas et préposition en latin' (à propos de Virg. En. 1, 1-3). Orphea Voce (Université de Bordeaux III), 1, 23-105. . (1980b), 'Réflexions sur le cas: vers une théorie des cas latins.' L'infor­ mation grammaticale (Paris, Heck), 7, 3-11. . (1980c), 'Propositions pour une théorie morpho-sémantique des cas latins.' Revue des Etudes Anciennes. Chevalier, J.-Cl., (1978), Verbe et phrase. Les problèmes de la voix en espa­ gnol et en français. Paris, Editions Hispaniques. Fillmore, Ch.J., (1968), 'The case for case.' In: E. Bach and R.T. Harms (Eds.), Universals in linguistic theory. New York, Holt. 1-88. Guillaume, G., (1964), Langage et science du langage. Paris-Québec, NizetPresses de l'Université Laval. Haudry, J., (1968), 'Les emplois doubles du datif et la fonction du datif indoeuropéen.' BSLP, 63, 141-159. . (1970), 'L'instrumental et la structure de la phrase simple en indo-euro­ péen.' BSLP, 65, 44-84. . (1977), L'emploi des cas en védique. Introduction à l'etude des cas en indo-européen. Lyon, Editions L'Hermès. Kurylowicz, J., (1949), 'Le problème du classement des cas.' Biuletyn P. T.J. 9, 20-43. Repr.: Kuriłowicz, L, (1960) Esquisses linguistiques. WroclawKrakow. 131-150. . (1964), The inflectional categories of indo-european. Heidelberg, C. Winter.

LE SYSTEME DES CAS LATINS

71

Madvig, J.R., (1976), Grammaire Eatine. Traduite de l'allemand par N. Theil (sur la quatrième édition). 3ème. édition. Paris, Firmin-Didot. Perret, J., (1957), 'Sur l'accusatif du latin.' Revue des Etudes Eatines, 35, 152163. . (1966), 'La signification du génitif adnominal.' Revue des Etudes Lati­ nes, 43, 467-482. Rubio, L., (1966), Introducción a la sintaxis estructural del latin. Vol. I, Casos y preposiciones. Barcelona, Ariel. Serbat, G., (1975), Ees structures du latin. Ee système de la langue classique; son évolution jusqu'aux langues romanes. Paris, Picard. Tesnière, L., (1966), Eléments de syntaxe structurale. Deuxième édition revue et corrigée (3ème. tirage, 1976). Paris, Klincksieck. Touratier, Chr., (1978)., 'Quelques principes pour l'étude des cas (avec appli­ cation à l'ablatif latin).' Langages, 50,98-116. . (1979), 'Accusatif et analyse en morphèmes.' BSEP, 74, 43-92. Vairel, H., (1975), Exclamation, ordre et défense. Analyse de deux systèmes syntaxiques en latin. Paris, Les Belles Lettres. Veyrenc, J., (1974), 'Cas et verbe.' Critique, 322 ("Roman Jakobson"), 246259.

THE STRUCTURE OF LATIN COMPLEX SENTENCES 1

ROBERT COLEMAN University of Cambridge SUMMARY Complexity is defined by the incidence of verbals in a sentence. A form of constituent analysis appropriate for the description of Latin complex sen­ tences is set out and illustrated. It is then applied to nine sentences taken from Caesar, Sallust, Cicero, Livy and Tacitus. A method of representing the struc­ tures diagrammatically is exhibited and procedures for quantifying the most significant peculiarities of complex sentences are expounded. 1.1 The simple DECLARATIVE sentence in Latin will usually be found to have one of the following constituencies: (1)

{i:

a VERBAL in the form of an infinitive (the 'historic infinitive')

or a finite verb in the indicative or subjunctive mood, the modal distinction corresponding to factual and hypothetical announcements, e.g. festinare, re­ diit, deserantur. (2) Nnom + {I V : Nnom is a NOMINAL, viz. noun, pronoun or ad­ jective used nominally, marked morphologically as the subject of the verb, e.g. tempus est, liber est, illi festinare, lux rediit, honesti deserantur. (Here and throughout this paper constituencies are presented in what is generally as­ sumed to be their unmarked order in literary Latin of the classical period.) (3) Nacc + {I V : Nacc is a nominal morphologically marked as object. e.g. exercitum inuaserunt, nihil remittere, honestos deserat. (4) Nnom 4- Nacc + {I V : e.g. legiones exercitum inuaserunt, barbari nihil remittere, ille honestos desereret. (5) Pr ( ( Nnom \ (5) Pr ( { )+ Vcop : a

PREDICATE

. . m the nominative, con-

74

ROBERT COLEMAN

sisting of either a nominal or an adjective, with a finite verb having copulative or passive factitive meaning; e.g. ualidus fies, consul eligeretur, eidem sunt. (6) Nnom + Pr 4 Vcop : the initial nominal is the v y { Enom subject. The same lexical restrictions as in (5) apply to the verb, which is often omitted when it is est, rarely in other moods and tenses of esse and hardly ever in other verbs; e.g. uita breuis [est], haec uetera [sunt], consules ualidi fient, Remus rex eligeretur. 1.2 Each of the constituents in (l)-(6) and the sentence as a whole may have adjuncts: the possibilities range from Adj Σ 4 V or AdjV 4 V to AdjΣ + Nnom 4 AdjN + Nacc 4 AdjN 4 AdjV + V or AdjΣ + Nnom + Pr ({Nnom + AdjN) + AdjV + Vcop; Enom + AdjE where AdjΣ = SENTENCE ADJUNCT, AdjV - VERB ADJUNCT etc. The simplest structures may be represented respectively by (7) at rediit = AdjΣ + V, (8) tandem rediit = AdjV 4 V. The NOMINAL ADJUNCT may be either a concordant adjective, as in (9) serena lux rediit = Nnom 4 AdjN (E) 4 V, or a dependent nomi­ nal in an oblique case, as in (10) auctoritas eorum pollebat — Nnom + AdjN (Ngen) + V ; signum projectionis dedit — Nacc 4 AdjN (Ngen) 4 V; mulier erat eximia pietate — Pr Nnom + AdjN (Nabl + AdjN (E)) + Vcop, viz. a predicative nominal in the nominative, with an adjunct consisting of a noun in the ablative with its own adjunct, an adjective, and a finite verb. The VERBAL ADJUNCT may be an invariant adverbial particle, as in

(11) pauorem tandem sedauit = Nacc 4 AdjV (ptc) 4 V, or a nominal in an oblique case with or without a preposition, as in (12) (i) tertio anno pergit = AdjV (Nabl 4 AdjN (E)) 4 V, ad Caesarem uenerunt = AdjV (prep 4 Nacc) 4 V, ab omnibus deserebatur = AdjV (prep 4 Nabl) 4 Vpass. This last example represents with (1) desererentur and (2) honesti desererentur the standard form of the passive trans­ formation. (12) (ii) hospitibus munera misere = Nacc 4 AdjV (Ndat) 4 V; classico signum dedit = Nacc 4 AdjV (Nabl) 4 V. (12) (iii) illud usui est = Nnom 4 AdjV (Ndat) 4 Vcop.

LATIN COMPLEX SENTENCES

75

1.3 Any of the constituents N, E, V, ptc may be enlarged in one or more of the following ways: (a) by attaching a further adjunct to a constituent that is itself already an adjunct, as in (13) aduentu nostri exercitus perterriti sunt = AdjV (Nabl + AdjN (Ngen + AdjN (E))) + Vpass; dies tarn turbidus fuerat = Nnom + Pr (Enom + AdjE (ptc)) + Vcop. (b) by conjoining two or more parallel constituents, usually with a conjunctive or disjunctive particle, as in (14) serena et tranquilla lux rediit = Nnom 4- AdjN (E + ptc + E) + V; munera dederunt et acceperunt = NAcc + V + ptc + V; te tot amicorum causae, tot coloniarum et municipiorum clientelae in forum uocant = Nnom + AdjN (Ngen + AdjN (E)) + Nnom + AdjN (Ngen + ptc + Ngen + AdjN (E)) + AdjV (prep + Nacc) + V. (c) by conjoining two or more adjuncts that occupy the same struc­ tural slot but are otherwise semantically or morphologically dissimilar, as in (15) forte in statione Marcus fuit = Nnom + AdjV (Nabl) + AdjV (prep + Nabl) + Vcop; eodem tempore ex oppido eruperunt = AdjV (Nabl + AdjN (E)) + AdjV (prep + Nabl) + V. 1.4

The negative particles non, haud may come before any constituent ; as in (16) non hospitibus munera misere, hospitibus non munera misere, hospitibus munera non misere. 1.5 INTERROGATIVE sentences are constituted like declaratives but with the following modifications: (a) the addition of an interrogative particle usually in first position or of the enclitic -ne to one of the constituents, as in (17) num classico signum dediti (cf. (12) (ii)), cur ex oppido eruperunt? (cf. (15)), nonne hospitibus munera misere? (cf. (16)), illudne usui est? (cf. (12) (iii)). (b) the replacement of one nominal by an interrogative pronoun or one deictic adjective or particle (talis, tarn etc.) by its lexically related inter­ rogative form, as in (18) quid usui est? (cf. (12) (iii)), quorum auctoritas polleb at? (cf. (10)) a quibus deserebatur? (cf. (12) (i)) quam turbidus dies fuerat? (cf. (13)), quanta conmutatio incessit? (cf. tanta c. i.).

76

ROBERT COLEMAN

1.6 VOLUNTATIVE sentences are constituted like declaratives but with the fol­ lowing modifications: (a) the verb is in the subjunctive, sometimes with the particle utinam, signalling weak volition. The negative is ne and is frequently the only distin­ guishing feature from hypothetical declarations. Thus (19) consul Volscis arma inferat — Nnom + Nacc + AdjV (Ndat) + Vsub; utinam munera misissent — Nacc 4- AdjV (Ptc) + Vsub; honestos ne de­ seras; cf. honestos non deseras ( ← (3)) (b) the verb is in the imperative and the negative, rarely used, is ne, as in (20) Volscis arma inferto = Nacc 4- AdjV (Ndat) + Vimp; honestos ne desere = Nacc + neg + Vimp. 2.1

A sentence is COMPLEX if at least one of its adjuncts contains an explicit VERBAL, though if the verbal is a part of esse, finite or non-finite, it may be omitted (cf. (6), (34), (40)). The class VERBAL is defined not by morphologi­ cal but by syntactic properties, viz. the capacity for verbal rather than nomi­ nal complementation and adjunction. Thus obsidens, which is morphologi­ cally marked for tense against obsessurus and for aspect and voice against obsessus has full nominal inflection; but in the phrase urbem obsidens it has a di­ rect object; cf. urbis obsessor, urbis appetens. Likewise exercitari, which is marked for voice against exercitare and for tense against exercitatum esse, has no other verbal categories; but again its adjuncts are verbal. exercitari in dicendo v. exercitatio dicendi. Similarly urbem dedere v. deditio urbis, apte dicendo v. apta dictione. It follows from the above definition that Romam uenere et munera misere is not a complex sentence, since, although it has two verbals, neither belongs to an adjunct of the other; the constituent structure is AdjV (Nacc) + V + ptc + Nacc 4- V. Contrast the constituencies in (21) below. 2.2 A VERBAL ADJUNCT in any of the structures (11)-(16) may be replaced by one of the following structures: (a) an ADVERBIAL CLAUSE: any of the structures (l)-(6) containing a finite verb is introduced by an 'adverbial' conjunction A, replacing the sen­ tence adjunct if there is one, to produce a clause signalling temporal location, causation, concession, condition or result (real or intended); e.g.

LATIN COMPLEX SENTENCES

77

(21) A + Nnom 4- V in 'postquam prinicipes uenerunt,' hoc ad Caesarem rettulerunt; A + Adj V (Ndat) + neg 4- V in de suis rebus petebant 'quoniam ciuitati non consultuissent'; A 4- Nnom + AdjV (Prep 4- Nacc + AdjN (Ngen)) 4- V in tanta conmutatio incessit 'ut in gratiam nobilitatis lugurtha ueniret'. (b) a NOMINAL PHRASE, consisting of a head nominal with concordant PARTICIPIAL (viz. participle or gerundive), in an oblique case with or without a preposition; e.g. (22) Nabl + AdjN (P) in 'remissa contione' iterpronuntiari iussit; prep + Nacc + AdjN-(P + AdjV (prep + Nabl)) in tertio anno 'post Satricum restitutum a Volscis' iterum consul factus est; prep + Nacc + AdjN (P) in legati circumibant 'ad concitandum bellum'. (c) a GERUNDIAL PHRASE consisting of a head gerund in an oblique case with or without verbal complementation; e.g. (23) Gdat in 'circumeundo1 studebamus; prep + AdjV (Eadv) + Gacc in 'ad celeriter erumpendum' parati sunt; Nacc 4- AdjN (E) + AdjV (Ndat 4AdjN (E)) + Gabl in negotium tibi importasti 'Romana nomina his fabulis adgregando '. (d) a VERBAL-NOUN PHRASE, consisting of the accusative of the verbal derivative in -tu- (the 'supine') or rarely the present active infinitive as head with or without verbal complementation; as in (24) legati uenerunt 'indutias petitum = Nnom + V + AdjV (Nacc 4Sacc) ; tempus est 'ad urbem progredi' = Nnom 4- V + AdjV (AdjV (prep + Nacc) 4- I). Generally speaking, there is in practice a decline in the length and com­ plexity of the subord stuctures from (21) to (24). 2.3 A NOMINAL ADJUNCT in any of the structures (9)-(16) may be replaced by one of the following: (a) a RELATIVE CLAUSE: in any of these structures containing a finite verb any one of the nominals is replaced by a relative pronoun, which usually occupies first position; as in (25) id credebant patribus 'qui proximi steterant' — Nacc + AdjV (Ndat + AdjN (Rnom + AdjN (E) + V)) + V; aliis 'quorum auctoritas polleba' munera misere = Nacc + AdjV (Ndat + AdjN (AdjN (Rgen) + Nnom + V)) + V.

78

ROBERT COLEMAN

(b) a PARTICIPIAL PHRASE, consisting of a head participle, concordant with a nominal in the superordinate structure, with or without verbal com­ plementation; as in (26) exercitum 'Coriolos obsidentem' legiones subito inuaserunt = Nnom + Nacc + AdjN (Nacc + Pact) 4- AdjV (A) + V. principes 'aduentu exercitus perterrit' ad Caesarem uenerunt = Nnom + AdjN (AdjV (Nabl + AdjN (Ngen)) + Ppass + AdjV (prep + Nacc) + V; eos'intus clausos' habebat = Nacc + AdjN (AdjV (ptc) + Ppass) + V, which is the origin of the Romance periphrastic perfect, re-analysed as Nacc + AdjV (ptc) + V (Ppass 4Vaux); hos Haeduis 'custodiendoš* tradit = Nacc + AdjN (Ppass) 4- AdjV (Ndat) 4- V; ea urbs 'rcircumsessa"1 a Celtiberis auxilia arcessiuit — Nnom 4AdjN (E) 4- AdjN (Ppass) 4- Nacc 4- AdjV (prep + Nabl) 4- V. The last example raises a problem. Latin participles are commonly employed as adjectives, so that, where the participle has no verbal adjunct or accusative object, it is not always easy to determine whether it is syntactically verbal or adjectival. In a sentence like iter in insequentem diem pronuntiatum est there seems to us little discernible difference semantically or syntactically between insequentem and posterum. Nevertheless it is probable that the dis­ tinction between qui posterus erat = Rnom + Pr (Enom) 4- Vcop and qui insequebatur — Rnom 4- V, which underlies posterum and insequentem respec­ tively, would have been reinforced for the native speaker by the frequent use of finite forms of insequi [4- Nacc] in the meaning 'follow chronologically'. Even more difficult is the parsing of inflati in a sentence like profectos domo inexplicabiles continuis imbribus uiae et inflati amnes tenebant. For whereas domo guarantees the verbal status of profectos, AdjV (Nabl) + P, the semantic similarity of inflati to tumidi and the coordination with in­ explicabiles point to an adjectival interpretation. Had the author written in­ explicabiles uiae et continuis imbribus inflati amnes, there would have been a strong case for regarding inflati as verbal and analysing the last four words as Nnom + AdjN (AdjV (Nabl 4- AdjN (E)) 4- Ppass). By contrast in­ explicabiles continuis imbribus, like difficiles c. i. for instance, must be analysed as AdjN (E) + AdjE (Nabl 4- AdjN (E)). In fact inexplicabiles is more remote from its verbal base quas explicare non poterant, than inflati from quos imbres inflauerant. Although purists might object to the introduction of such an argument into what is otherwise a structural description, it is nevertheless accepted here (after much hesitation) as grounds for treating in­ flati, like insequentem and circumsessa, as a verbal. In contrast to Greek,

LATIN COMPLEX SENTENCES

79

where the status of participles is unequivocally verbal, this is very much a grey area of Latin morpho-syntax. (c) a NOMINAL PHRASE, in which the adjunct, consisting of an obliquecase nominal (as in(10)), has a dependent passive participial, e.g. (27) Nacc + AdjN (Ngen + AdjN (Ppass)) + Vin 'capti oppidi' signum dedit and belli concitandi studium dissimulabant. (d) a GERUNDIAL PHRASE, in which the adjunct is an oblique-case gerund with or without verbal complementation, as in (28) aliqua in me exercitatio 'dicendi' est = Nnom + AdjN (E) + AdjN (Ggen) + Adj V (prep + Nabl) + Vcop; 'senes eludend' occasio est = Nnom + AdjN (Ggen + Nacc) + Vcop. Although in some instances the semantic difference between the gerund and other verbal nouns is hard to discern (cf. exercitatio orationis), the capac­ ity for complementation seems decisive; e.g. exercitatio apte dicendi. Con­ versely signum profectionis dedit is not regarded as complex, in contrast to e.g. signum proficiscendi ab urbe dedit. 2.4 ADJECTIVAL ADJUNCTS (§1.3) may be replaced by structures parallel to those in §2.3. Thus alongside consulatu digitus erat = Pred (Enom + Adj E (Nabl)) + Vcop we find (29) dignus erat 'qui consul fiere' Pr (Enom + AdjE (Rnom + Pr (Nnom) + Vcop)) + Vcop; cf. (25). Alongside cupidi erant pacis = Pr (Enom + AdjE (Ngen)) 4- Vcop. (30) cupidi erant 'pacis petendae' = Pr (Enom + AdjE (Ngen + AdjN (P))) + Vcop;cf.(27). Alongside equitibus satis latebrosum locum perlustrauit = Nacc + AdjN (E + AdjE (ptc) + AdjE (Ndat)) + V (31) equites tegendo satis latebrosum etc. = Nacc + AdjN (E + AdjE (ptc) + AdjE (Gdat + AdjV (Nacc))) + V; cf. (28). 2.5 The structures specified in (21)-(31) are all OPTIONAL additions to al­ ready well-formed sentences, being AdjN, AdjE and AdjV constituents. A second group comprise structures replacing Nnom and Nacc and are therefore more integral to the basic sentence pattern. It follows that their occurrence in a complex sentence owes less to authorial selection; indeed their excision would in some instances leave an ill-formed sentence.

80

ROBERT COLEMAN

2.6 The SUBJECT NOMINAL , Nnom, may be replaced by one of the following (S denotes the replacement for Nnom in subject position): (a) a NOMINAL PHRASE consisting of a head nominal in the nominative case and a dependent passive participle, as in (32) 'Hannibal uictus' Romanos metu liberauit = S (Nnom + AdjN (Ppass)) + Nacc + Adj V (Nabl) + V. This is semantically comparable to re~ missa contione etc. in (22), not to urbs circwnsessa etc. in (26), though there seems no overt syntactic signal for the distinction. (b) an INFINITIVE usually unmarked for tense, without a subject and with or without complementation; e.g. (33) miserum est 'deturbari fortunis omnibus' = S (AdjV (Nabl 4AdjN (E)) + Ipass) 4- Pr (E) + Vcop; facinus est 'uincire ciuemRomanum'— S (Nacc 4- AdjN(E) + I) 4- Pr(N) 4- Vcop; 'temperantem esse' oportet = S(Pr (Eacc)) + I) + V. In the last example the case of the adjective is typical; the verb oportet is usually classed as 'impersonal'. (c) an INFINITIVE marked for tense with subject nominal in the accusa­ tive, with or without complementation (the accusative + infinitive construc­ tion); as in (34) est inusitatum 'regem reum capitis esse' = S (Nacc + Pr (Nacc 4AdjN (Ngen)) + Icop) + Pr (E) + Vcop;'leges breuis esse' oportet = S (Nacc + Pr (E) + Icop) + V; nuntiatum est legatos ciuitates circumiisse = S (Nacc + Nacc + Iperf) + Vpass. That the infinitival structure is not the complement to an impersonal third person is clear by comparison with nuntiata est haec res (v. *nuntiatum est hanc rem). (d) an INFINITIVE marked for tense with subj ect nominal in the nomina­ tive, with or without complementation. The superordinate verb is passive. Thus (35)'legati' nuntiabantur 'circumiisse' — S (Nnom + I) + Vpass; 'Homerus' traditur 'caecus fuisse"* = S (Nnom + Pr (Enom) + Icop 4- Vpass. 2.7 The OBJECT NOMINAL, Nacc, may be replaced by one of the following (O denotes the replacement for Nacc in object position): (a) a NOMINAL PHRASE consisting of a head nominal in the accusative and a dependent passive participle, as in (36) 'auctorem senatus extinctum' laete tulit — O (Nacc 4- AdjN (Ngen) 4- AdjN (Ppass)) + AdjV (Eadv) + V.

LATIN COMPLEX SENTENCES

81

The note on (32) also applies here. (b) an INFINITIVE, usually marked for tense, with or without com­ plementation. There is no explicit subject but predicative nominals and adjec­ tives are in concord with the subject of the superordinate verb. Thus (37) 'de suis rebus petere' coeperunt = O (AdjV (prep + Nabl + AdjN (E)) + I) 4- V; 'temperans esse' uolebat = O (Pr (Enom) + I) + V; 'Volscis arma inferre' iussus est = O (Nacc + AdjV (Ndat) 4- I) 4 Vpass. The latter could also be analysed as AdjV (Nacc 4- ... 4 I) + Vpass; cf. (39),of which it is the passive transformation. (c) an INFINITIVE, usually unmarked for tense, with or without com­ plementation and with subject nbminal in the accusative, as in (38) uolebat 'legatos circumire' — O (Nacc + I) + V. In some examples the subject to the infinitive can equally be parsed as object to the superordi­ nate verb, as in (39) iussit 'legatos circumire'. If legatos is object to iussit, then the analysis becomes either Nacc +I+V, with a double object implying illudeos iussit (cf. illudeos rogauit), or more likely Nacc 4- Adj V(I) + V, which would be related to (37). (d) an INFINITIVE marked for tense with subj ect nominal in the accusa­ tive, with or without complementation, as in (34); e.g. (40) putabat 'se ab omnibus desertum [esse] = O (Nacc 4- AdjV (prep 4- Nabl) + 1) + V; id exiguum fieri1 sentio; nuntiauerunt 'uos Volscis arma în­ laturos [esse]'. This with (45) and the last example in (34) represents the stan­ dard REPORTED DECLARATION transformation in Latin; cf. ab omnibus desertus sum; id exiguum fit; Volscis arma inferetis. The frequent omission of esse makes parsing difficult in some cases. uacuam sedem regiam uidit could be analysed as O (Nacc (N + AdjN (E)) 4- AdjN (E)) 4 V, 'he saw the empty throne'; alternatively as O (Nacc + AdjN (E)) + Pr (Eacc) + V, with uacuam [esse]. (e) a CLAUSE containing a subjunctive verb with or without com­ plementation, and with the particles ut or ne, usually in first position. The superordinate verb signals either EXHORTATION or AVERSION and its tense gov­ erns the tense of the subordinate. (41) rex praecepit 'ut munera hospitibus mitterent' = Nnom 4 V + O (ptc + Nacc 4- AdjV (Ndat) + Vsub); impero 'ne iter pronuntietur"1 ; uerebatur 'ne ab omnibus desereretur' = V + O (neg + AdjV (prep + Nabl) 4 Vsub).

82

ROBERT COLEMAN

(f) a CLAUSE consisting of one of the structures in (17) and (18) with its verb in the subjunctive and in a tense partly governed by the tense of the superordinate verb, which signals INTERROGATION, (Q denotes a word that is lexically interrogative), e.g. (42) rogaui hospitibusne munera misissent = V + O (Nacc + AdjV (Ndat) + ptc + Vsub); scietis 'quorum auctoritas polleat'= V + O (adjN (Qgen) + Nnom + Vsub) ; sensi 'quam exiguum esset' = V + O (Pr (Enom) + AdjE(Qptc) + Vsub). 2.8 A few of the analyses offered in §§1-2 are controversial, as has been re­ marked from time to time in the text; some may turn out to be incorrect and require revision. Moreover the principles on whichthe presence of a given structure confers complexity upon the sentence may in some cases be strongly disputed. Both the analyses and principles are, however, employed without further comment in §3. 3.1 In this section the foregoing analysis is applied to nine Latin sentences, taken from Sallust lugurtha 13.7, Caesar De Bello Gallico 5.3.5, Cicero Pro Archia 1, Livy Ab Vrbe Condita 1.16.2 (=I), 2.33.6 (=II), 2.59.6 (=III), 7.27.5 ( = IV), 40.33.2 (=V), Tacitus Dialogus de Oratoribus 3.4. Each sen­ tence is divided into its constituents, as defined in §2, and these are numbered consecutively down the page. 3.2 The organization of each sentence is then represented diagrammatically. The numbers, which correspond to those in the decomposed sentence, are placed sequentially along the horizontal axis except for constituents that are enclosed by their superordinates. These are placed before their superordinates. Thus in LIVY II © is enclosed by © and the whole group © - ® by © . The vertical axis indicates chronological progression, with the constituents expounding the earliest events at the top, those expounding the latest at the bottom. Thus in LIVY I 3, 6 and 7 precede © and 4, which precede 5 and ® , which precede © . This enables us to see how far in shaping his material the author has disrupted the chronological sequence: a stylistic consideration of some importance in historical narrative writing. The arrows in the diagram show the direction of dependency:

LATIN COMPLEX SENTENCES

83

©→2→3 means that 1 is subordinate to 2 which is subor­ dinate to © . Thus in TACITUS there is a chain of subordination from © to @ to 5 to 4 to 2 to 1. © © means that 2 is subordinate to © a n d enclosed within it. Thus in LIVY II © is enclosed by @; © , © and © are enclosed by © ; and © , 4, 6, 3 together with © and © are enclosed by © . © (2) 3 means that © and © are coordinate (either conjoined, e. g. by et, or disjoined, e. g. by aut) and as a group subordinate to © . Thus in CAESAR © and @ are coordinate in their dependency on © , which in turn is coordinate with © in its dependency on © . In CICERO © , @ and ® are coordinate in their dependency on 13. ©—2 3 means that © and © are independently subordi­ nate to © . Thus in LIVY I © and © are subordinate to © , which with © , © and © is subordinate to © . © 2 3 means that © and ® are subordinate to © . Thus in SALLUST the coordinate group of © and © is together with © dependent on 1. The letters beside each number indicate the type of constituent: A C G I N

= = = = =

ADVERBIAL CLAUSE: cf. (21). COMPLEMENT CLAUSE : cf. (41), (42). GERUNDIAL PHRASE : cf. (23), (28), (31). INFINITIVAL PHRASE : cf. (33)-(35), (37)-(40). NOMINAL + PARTICIPIAL PHRASE : cf. (22), (27), (30), (32), (36). P = PARTICIPIAL PHRASE : cf. (26). R = RELATIVE CLAUSE : cf. (25), (29).

3.3 After each diagram the CORE of the sentence is defined, consisting of the principal clause and those subordinates which are directly dependent on it. In two instances (SALLUST and LIVY IV) a set of simple sentences is also in­ cluded, corresponding approximately in meaning, one by one, to the con-

84

ROBERT COLEMAN

stituents of the complex. The two sets are transformationally related in that either the constituents can be derived from the simple structures or more often both can be derived from a common underlying structure. Their purpose here, however, is purely stylistic comparison: to show how the same basic information can be processed into simple or complex form. This enables us to examine in detail the distinction between the continuous (eiroméne) and com­ plex (katesîramménë) styles made by Aristotle in Rhet. 3.9.1409A., and within the complex style to discern the authorial choices governing on the one hand which items of information are to be principal, which subordinate, which sub-subordinate (decisions that are largely language-non-specific) and on the other which of the available morpho-syntactic procedures are to be applied to the exposition of those items (decisions that are of course, by contrast, lan­ guage-specific). 3.4 In the following table the complexity of the nine sentences is quantified according to the following criteria: (i) (ii) (iii)

(IV)

(V)

(VI)

(VII)

MAGNITUDE

:

the number of words in the sen­ tence. CONSTITUENCY : the number of constituents. CONSTITUENT WEIGHT : the average size of constituent as determined by the quotient of (i) by(ii) ELABORATION : the quotient of the number of constituents by the size of the core (as defined in §3.3) VARIETY : the quotient of the number of constituents-types (A, C, G, I, N, P, R, as defined in §3.2 fin.) by the number of subordinate con­ stituents. VERBALITY : the quotient of the number of explicit finite verbs by the number of constituents CONTINUITY : the proportion of the sentence that precedes the first possible terminal point, viz. the first con­ stituent to receive both a rightward and a leftward arrow.

85

LATIN COMPLEX SENTENCES

(VIII)

DYNAMIC

(IX)

DISPOSITION

(X)

COMPLEXITY

the proportion of rightwardpointing arrows. the amount of disruption in the order of constituents resulting from enclosing procedures, as measured by the quotient of the number of positional moves by the number of constituents; e.g. in S ALLUST the sequence is 2 4 © © © ; © has moved three places, © has moved one place; hence the disposition =

SALL CAES MAGNITUDE

3 + 1 =0.8. 5 the quotient of the sum of ar­ CIC LIV I rows LIVIIseparating LIV III LIV IV initial LIVV T AC each con­ stituent (viz. having no incoming 24 62 63 33 39 43 38 from 13 7 the principal 9 9 9 clause 8 arrow) 8 by the number of initial con­ 3.4 3.7 4.2 4.8 4.9 7.8 4.8 e.g. 3.0 4.0 2.25 2.6 1.75 in SALLUST 1.6stituents; 1.8 again linked0.65 to © by 0 0.4 0.55 0.5 0.7© is0.4 0.5one arrow, to © by two arrows, © to © by 0.4 0.45 0.9 0.45 0.35 0.6 0.45 the complexity 0.1 1.0 1.0 0.6one;1.0 1.0 index1.0 1+2+1 1.0 = 1.3. 0.75 0.75 0.25 0.35 0.55 1.0

36

45

CONSTITUENCY

5

10

CONST. WEIGHT

7.2

4.5

ELABORATION

1.25

2.5

VARIETY

0.5

0.45

VERBALITY

1.0

0.5

CONTINUITY

0.8

1.0

DYNAMIC

0.75

0.8

DISPOSITION

0.8

1.3

0

=3 0.85 1.1

COMPLEXITY

1.3

2.2

2.3

2.2

2.4

0.15

0.9

1.1

0.25

2.0

2.25

1.8

4.0

86

ROBERT COLEMAN

(If the argument (§2.3) for treating insequentem (constituent 5) in LIVY III and inflati (constituent 8) in LIVY V as verbals is rejected, then these two constituents disappear and the columns will be revised thus: LIV III: 24, 6, 4, 1.5, 0.8, 0.5, 1.0, 1.0, 0,1.55; LIV V: 33, 8, 4.1, 1.6, 0.45, 0.5, 1.0,0.7, 1.1, 1.6.) 3.5

SALLUST(lug. 13.7):

1 2 3 4 3 1 5

sed, ubi Romam legati uenere et ex praecepto regis hospitibus aliisque quorum ea tempestate in senatu auctoritas pollebat magna munera misere, tanta conmutatio incessit ut ex maxuma inuidia in gratiam et fauorem nobilitatis Iugurtha ueniret.

The core is remarkably large, consisting of everything except quorum ea tempestate in senatu auctoritas pollebat. Less surprisingly in a sentence with re­ latively few constituents the index of complexity is very low. The unusually high average constituent-weight (7.2, with limits of 3 and 11) is due to the amount of nominalization: ex praecepto regis, ex maxuma inuidia, in gratiam et fauorem nobilitatis ueniret. {tanta conmutatio incessit does not affect the weighting). Somewhat paradoxically these nominalizations within the con­ stituents go with an unusually high incidence of finite verbs, all placed at the end of their clauses. The fact that © is the recipient of both a rightward and a leftward arrow reveals a potential discontinuity, tanta conmutatio incessit is a well formed simple sentence and in an appropriate context could be a plausi­ ble principal clause to a sentence containing © - © without © . A correspond­ ing set of simple sentences might run as follows: (2) legati Romam uenere. (3) hospitibus aliisque magna munera misere. (4) illorum in senatu auctoritas ea tempestate pollebat. (3) hoc praeceperat rex. (1) itaque res magnopere con-

LATIN COMPLEX SENTENCES

87

mutabatur. (5) antea enim nobiles Iugurthae maxume inuiderant. nunc autem eum gratiorem habebant et fauebant. CAESAR (Gal. 5.3.5): 1 2 3 4 2 5 6 5 7 8 7 9 10 1

sed postea quam non nulli principes ex ea ciuitate, et familiaritate Cingetorigis adducti et aduentu nostri exercitus perterriti, ad Caesarem uenerunt et de suis priuatim rebus ab eo petere coeperunt, quoniam ciuitati consulere non possent, ueritus ne ab omnibus desereretur, Indutiomarus legatos ad Caesarem mittit.

The core consists of four constituents: sed postea quam non nulli prin­ cipes ex ea ciuitate ad Caesarem uenerunt et coeperunt, ueritus Indutiomarus legatos ad Caesarem mittit; which entails a moderate degree of elaboration. The dynamic is high, with only two leftward arrows. The high disposition index (even if © is not divided from © nor ® from ® as in (37) above), is largely due to the enclosing nature of sed postea .. (or sed alone if posteaquam is treated as a single word) .. Indutiomarus legatos ad Caesarem mittit. The chronological sequence of events is reproduced overall, but the coordination of © and © leaves the priority of the former unmarked; cf. © and © in SALLUST.

88

ROBERT COLEMAN

CICERO (Arch. 1.1): 1 2 3 4 5 6 7 6 7 8 9 10 11 10 11 12 13

si quid in me ingeni, iudices, quod sentio quam sit exiguum, aut si quae exercitatio dicendi, in qua me non infitior mediocriter esse uersatum, aut si huiusce rei ratio aliqua, ab optimarum artium studiis ac disciplina profecta, a qua ego nullum confiteor aetatis meae tempus abhorruisse, earum rerum omnium uel in primishicAulis Licinius fructum a me repetere prope suo iure debet.

PJ®

ns

Along with TACITUS this is the largest of the complexes chosen for analysis; it also has the highest constituency, even if (5) is not divided from 4 nor © from ® . The core, si quid in me ingeni, iudices, aut si quae exercitatio aut si huiusce rei ratio aliqua, earum rerum omnium in primis hic Aulus Licinius fructum a me repetere prope suo iure debet, represents more than half the total words but only five of the thirteen constituents, a significant distribu­ tion of weight, combined with high elaboration. The relatively low total of fi­ nite verbs — especially for Cicero — is due largely to the remarkable omission of est in all three of the core subordinate clauses, © , 4, (8). There is a strong preponderance of leftward arrows, but the cumulative rightward movement of the three coordinated dependent clauses of the core (a tricolon of 11, 12 and 22 words respectively) sustains the dynamic to the very end of the period. Somewhat remarkably in a sentence of this kind there is no enclosing in the strict sense of the word; for in the interlacing order of 6 and (7), 10 and 11, the

LATIN COMPLEX SENTENCES

89

first member of each pair is complete before the second. LIVY I ( = 1.16.2): 1

Romana pubes

2 3 4 5 6 7 1 8 1

sedato tandem pauore, postquam ex tarn turbido die serena et tranquilla lux rediit, ubi uacuam sedem regiam uidit, etsi satis credebat patribus qui proximi steterant sublimem raptum procella, tamen uelut orbitatis metu icta maestum aliquamdiu silentium obtinuit.

There is a low degree of elaboration, with a core consisting of five of the eight constituents: Romana pubes sedato tandem pauore, ubi uacuam sedem regiam uidit, etsi satis credebat patribus, tamen uelut orbitatis metu icta maes­ tum aliquamdiu silentium obtinuit. As the core itself indicates, however, there is unusually great variety and, for Livy, a high proportion of finite verbs. Of particular stylistic interest is the counterpoint between the narrative progres­ sion from 2 to 1 and the presentation of earlier events in ® , © , © , all with contrasting leftward arrows.

90

ROBERT COLEMAN

LIVY II (= 2.33.6): 1 2 3 4 5 4 3 6 3 1 7 1 8 9

cum subito exercitum Romanum, Coriolos obsidentem atque in oppidanos, quos intus clausos habebat, intentum sine ullo metu extrinsecus imminentis belli, Volscae legiones, profectae ab Antio, inuasissent eodemque tempore ex oppido erupissent hostes, forte in statione Marcus fuit.

The core has only three constituents: cum subito exercitum Romanum Volscae legiones inuasissent eodemque tempore ex oppido erupissent hostes, forte in statione Marcus fuit. Although this entails high elaboration, the semantic im­ portance of the core is indicated by the fact that it contains nearly half the total words in the sentence. Although the proportion of finite verbs is unexceptional, they are, apart from habebat, all concentrated in the last three constituents. The contrast with the preponderance of participles in © to © is intensified by the intricate dis­ positions of this group, which are not only enclosed in © but, by an uninter­ rupted series of rightward-arrows, also point the reader steadily forward to its verb inuasissent.

LATIN COMPLEX SENTENCES

91

LIVY III (= 2.59.6): 1 2 3 4 5 4 6 7

uictus tandem, quando quidem nihil praeter tempus noxae lucrarentur, remissa contione iter in insequentem diem pronuntiari cum iussisset, prima luce classico signum profectionis dedit.

Here the structural pattern reflects the chronological sequence of events. The core is large — uictus tandem, remissa contione, cum iussisset, prima luce classico signum profectionis dedit — and the degree of elaboration therefore low. So is the average constituent weight. As in LIVY I the variety is great; but the index of complexity is relatively low. As in LIVY II the dynamic is wholly rightward. LIVY IV (=7.27.5): 1 2 1 3 4 5 6 7 8 9 1

tertio anno post Satricum restitutum a Volscis M. Valerius Coruus, iterum consul cum C. Poetelio factus, cum ex Latio nuntiatum esset legatos ab Antio circumire populos Latinorum ad concitandum bellum, priusquam plus hostrum fieret Volscis arma inferre iussus, ad Satricum exercitu infesto pergit.

92

ROBERT COLEMAN

The core is tertio anno post Satricum restitutum a Volscis M. Valerius Coruus, iterum consul cum C. Poetilio factus, iussus ad Satricum exercitu infesto pergit. The relatively small number of finite verbs highlights the chief ac­ tion points as Livy chooses to present them nuntiatum esset, fieret, pergit. The rightward dynamic is interrupted only by the leftward movement of 4 —(5) 6, in chiasmus with ® — 8 — 9 . A corresponding set of simple sentences might run as follows: (2) Satricum a Volscis restitutum est. (3) tertio post anno Coruus iterum consul cum Poetilio factus est. (5) legati ab Antio circumibant populos Latinorum. (6) bellum enim eis concitandum erat. (4) ex Latio hoc nuntiatum est. (7) inde plus hostium futurum erat. (9) itaque Coruus iussus est. (8) 'Volscis arma prius inferto'. (1) ad Satricum exercitu infesto per­ git. LIVY V ( = 40.33.2): 1 2 3 4 5 6 7 6 8 6

ea urbs circumsessa, cum a Celtiberis auxilia arcessisset, morantibus eis, non quia ipsi cunctati sunt sed quia profectos domo inexplicabiles continuis imbribus uiae et inflati amnes tenebant,

LATIN COMPLEX SENTENCES

9 1

93

desperato auxilio suorum in deditionem uenit.

The core is relatively large, ea urbs circumsessa, cum a Celtiberis auxilia arcessisset, morantibus eis, desperato auxilio suorum in deditionem uenit; and the degree of elaboration therefore low. As in LIVY III, the average con­ stituent weight is light, and as in LIVY II there is relatively little structural va­ riety. For Livy the amount of complexity is abnormally small and there is very little enclosing apart from that o f © — ® within 1, which gives a misleading boost to the total index. TACITUS (Dial. 3.4): 1 2 3 2 4 5 6 7 8 7

adeo te tragoediae istae non satiant (inquit Aper) quo minus omissis orationum et causarum studiis omne tempus modo circa Medeam, ecce nunc circa Thyesten consumas, cum te tot amicorum causae, tot coloniarum et municipiorum clientelae in forum uocent, quibus uix suffeceris, etiam si non nouum tibi ipse negotium importasses, ut Domitium et Catonem, id est nostras quoque historias et Romana nomina, Graeculorum fabulis adgregares.

94

ROBERT COLEMAN

The core consists of two clauses: adeo te tragoediae istae non satiant quo minus omne tempus modo circa Medeam, ecce nunc circa Thyesten consumas. The average constituent-weight is abnormally heavy (7.8 with limits of 3 and 13). The high indices of elaboration and complexity result from the remark­ able but not uncharacteristic linear pattern of the constituents, which with its predominantly leftward dynamic is in turn responsible for the prevalent dis­ continuity. There are possible terminal points after 1, 2, 4, 5 and 6. (This and the SALLUST are, significantly, the only two non-periodic structures in the selection). Somewhat un-Tacitean is the high proportion of finite verbs. 3.6 It is now clear that this method of analysis or something like it can yield very precise information about the diversity of Latin complex structure in gen­ eral and about idiolectal characteristics in particular. Even in this brief selec­ tion the contrast between the balanced and padded Ciceronian period, Livian uarietas and the Tacitean 'falling' pattern has shown up clearly. There are no doubt improvements and corrections to be made to the analytical model, the diagrammatic representation and the criteria for quan­ tification. It is perhaps worth noting in conclusion that, though the three parts of the method have been used in concert, they are all essentially independent of one another and any one of them could be employed with a radically differ­ ent version of the other two. 2

NOTES 1) What follows in §§1.1-2.7 is obviously not an exhaustive account of Latin sentence-structure; but it is the author's belief that only a minute number of actual or possible sentences are not ac­ counted for explicitly or by implication and that the inclusion of these would require merely minor additions to the list of structures, not a fundamental revision of them. The method of analysis employed, though devised in a somewhat eclectic way for the immediate purpose, owes much ulti­ mately to the influence of Zellig Harris. 2) This is a revised and expanded version of the paper read in Amsterdam. The author is grateful to members of the Colloquium and especially to Prof. Chr. Lehmann for criticisms, comments and suggestions.

ORGANISATION SYNTAXIQUE DES COMPLEMENTS D E LIEU ET TEMPS EN LATIN ARCHAÏQUE*

H. F U G I E R Université de Strasbourg RESUME Dans l'ensemble des circonstants de lieu-temps, le fonctionnement syn­ taxique distingue deux sous-ensembles. 1 ) Le syntagme prépositionnel à 3 ter­ mes illic | intus | in Mac hirnea. Applicable au lieu dans ses 3 variétés (ques­ tions ubi, quo et unde), et ses 2 fonctions S.P.v. et S.P.p., ainsi qu'au temps, cette formule donne lieu à des combinaisons théoriques, dont seule une sélec­ tion non aléatoire se trouve réalisée. En position de préverbe, la préposition, changeant le sens et la construction du verbe, intervient dans l'"ordre de pro­ cès". Suivie d'un nom verbal ou d'une phrase qu'elle subordonne, elle entre en système avec le verbe principal et les particules de type iam. 2) La série pro­ gressive des ad-verbes invariables numquam, semel..., saepe, semper. Quan­ tificateurs verbaux correspondant aux quantificateurs nominaux, ces adver­ bes ont des rapports indirects avec l'aspect mais surtout des relations avec cer tains auxiliaires de modalité. Conclusion: temporalisé par 1) et quantifié par 2), le procès peut recevoir aussi des qualifications, temporelles quant à leur sens; mais les adverbes-adjectifs qualifiants tels que continuo, prédicats de phrase, ne sont pas syntaxiquement des circonstants de temps. 0. Ce qui fait les compléments "de lieu et temps" n'est pas la forme morpholo­ gique, car cette catégorie de circonstants embrasse aussi bien des adverbes étrangers à toute flexion comme semper que des noms déclinés comme (apud) cenam; aussi bien les adverbes à thème démonstratif hic,istic,illic ... que ceux formés sur des thèmes aussi variés que ante,post,intus-intro ... (Cupaiuolo, 1967 : 84,115-123,129 ; Knight, 1921: 65,69). Ce n'est pas non plus le sens, car si quelque chose nous autorise à parler de compléments de "lieu" ou "temps", c'est moins la dénotation spatiale ou temporelle que l'existence d'une corréla­ tion entre une telle valeur sémantique et un fonctionnement syntaxique

96

H. FUGIER

défini. De même est-ce encore une différence syntaxique qui nous invite à reconnaître, dans la masse globale des compléments " de lieu et temps", ces deux sous-ensembles que sont respectivement le S.P. et l'ad-verbe : sousensembles dont les descriptions successives constitueront cet exposé. 1. Un S.P. : Le type illic intus in illac hirnea 1.1. Description distributionnelle Un type très attesté de complément de lieu ou temps est celui qui s'ordonne autour d'un groupe préposition + N ou G.N., ainsi ad cenam (Men., 124), a principio {Bacch., 1006) ; in illac hirnea (Amph., 432) ; in hunc diem (Ps. , 128) ; intra muros ciuicos {Bacch., 4) ; media nocte (préposition 0 : Amph., 514) ; ad nos {Ps., 650). Ace noyau peuvent s'adjoindre, en séquence syntagmatique, deux autres éléments : - d'une part un adverbe pronominal hic {istic,illic) ou huc {istuc,illue) ou hinc {istinc,illinc) ou hac (istac,illac) ; - d'autre part un adverbe choisi parmi les deux termes d'une paire, elle-même prise dans la liste close intus {intro) / f oris (foras),ad dexteram ou dextra I ab laeua,sursum ou de supero I deorsum, ante ou prius / post. La plus longue chaîne possible comporte donc trois éléments, que nous sym­ boliserons respectivement par : Ll L2 L3 parex. illic in tus in illac hirnea (Amph.,432). Sur des chaînes analogues en Allemand, cf. R.Steinitz (1969:chap. 8). Une telle séquence répond, en matière de sens et de référence, à l'exigence du repérage d'un objet dans l'espace ou le temps, puisque l'objet n'est bien situé que par l'indication simultanée d'une position relative au "je" observateur : hic = "ici près de moi, à l'instant où nous sommes", d'une position dans l'espace-temps -orienté lui-même par rapport à "moi maintenant" : ad dexteram ≠ ah laeua,ante ≠ post, d'une position absolue,c.à.d. définie par référence à une réalité indé­ pendante de "je" : in hirnea,noctu. Commun aux compléments de lieu et de temps, ce schéma souffre cependant quelques variations, qui nous obligent à considérer séparément chacun des deux cas. 1.1.1. Lieu L'existence d'un "lieu où l'on va" et d'un autre "d'où l'on vient" à côté de

COMPLEMENTS DE LIEU ET TEMPS

97

celui "où l'on est" — suivant les termes de nos vieilles grammaires—multiplie par 3 les points d'application du schéma L1 + L2 + L3.I1 faut entendre par là que les trois formules telles que par exemple, pour "ubi" : foris illic extra scaenam (Capt. ,60) pour "quo" : Aliquo ad cenam foras {Men., 124) pour "unde" : hic intus 0 (Ps.,1121) hinc ø a nobis (Ps.,617), sont, non seulement possibles chacune pour son compte, mais aussi compossibles. On calculera aisément le nombre de combinaisons théoriquement offer­ tes, pour l'ensemble des séquences L 1 ; L 2 ; L 3 ; L1 + L2; L 1 + L 3 ; L 2 + L 3 ; L 1 + L 2 + L 3 , quand interviennent la série ubi uniquement „ la série quo „ (la série unde ou l'ensemble de 2 séries ubi-quo, ubi-unde,quo-unde \ ou l'ensemble des 3 séries ubi-quo-unde. Il suffira de penser, avant de consulter le triangle de Pascal, - qu'on peut se dispenser de prendre en compte, outre ubi, quo et unde, la série qua. Les réalisations de celle-ci dans le corpus sont en effet numériquement négligeables, puisque de rares L 1 {hac en Bacch., 1166), L 2 {recta en Hec., 372) ou L 3 {alia uia en Heau. ,789) forment tout au plus d'exceptionnel­ les séquences ou correspondances à deux termes {hac vs. ab laeua = L1 vs. L 2 en Cist. ,641); - que l'ordre respectif des divers éléments intervenant dans une même séquence n'est pas pertinent. Moyennant ces précisions, on verra que la plus longue séquence possible L 1 {ubi) + L 2 (ubi) + L 3 (ubi) + L 1 (quo) + L 2 (quo) + L 3 (quo)+ L 1 (unde) + L 2 (unde) + L 3 (unde) autorise : 9 combinaisons de 1 élément 4- 36 combi­ naisons de 2 éléments + 84 combinaisons de 3 éléments + 126 combinaisons de 4 éléments + 126 combinaisons de 5 éléments + 84 combinaisons de 6 élé­ ments + 36 combinaisons de 7 éléments + 9 combinaisons de 8 éléments + 1 combinaison de 9 éléments, soit 507 séquences différentes. Il va de soi cependant que la connaissance du corpus gagne moins au calcul des combinaisons théoriques qu'à l'observation des séquences effectivement réa­ lisées. A ce point, on est invité à tenir compte du fait que le complément de lieu peut remplir deux fonctions syntaxiques distinctes, suivant qu'il comple­ mente soit la P. entière soit le seul G.V. . Dans le premier cas nous trouvons, outre des masses quantitativement comparables -dont l'inventaire exhaustif

98

H. FUGIER

occuperait trop de place pour l'intérêt qu'il présente- de L 1 (ubi), L 1 (quo) et L 1 (unde), de L 2 ..., de L 3 ..., des séquences associant entre eux 3 termes de série ubi (hic intus in fano Fidei, Aul. ,617), d'autres associant 3 termes de série quo (aliquo ad cenam foras, Men.,124), d'autres beaucoup plus rares asso­ ciant divers termes de série unde (hine ab dextera en And., 734: hinc ab ostio en Cist.,659). Mais nous trouvons aussi des séquences hétérogènes, qui conjoi­ gnent des éléments issus de séries différentes -beaucoup plus intéressantes car elles introduisent le problème de la compatibilité entre les séries. De fait, il s'en faut de beaucoup que les séries ubi,quo et unde se combinent entre elles de façon aléatoire ; il s'en faut aussi qu'elles se distribuent au hasard entre les positions L 1, L 2 et L 3. Ubi ne coexiste ni avec quo ni avec unde, alors que ces derniers apparaissent au contraire en association régulière : c'est-à-dire, en interprétation sémantique, que la position statique dans l'espace est une chose, et le mouvement défini par son origine et son terme, une autre chose, non conciliable avec la première dans le même énoncé. Quant aux positions occupées respectivement par quo et unde, le corpus présente, pour de très rares séquences où l'ordre des positions décroît de unde à quo, soit L 3 (unde)→ L 1 (quo) (Quid si adduco tuum cognatum huc ab naui Naucratem, Amph.,849), une masse de séquences orientées en sens croissant, soit L 1 (unde)→ L 2 (quo) (Tollam ego ted in collum atque intro hine auferam, Bacch,.571) ou beaucoup plus souvent L 1 (unde)→ L 3 (quo) (Hanc iam oportet in cruciatum hinc abripi,And. ,786) : comme si (soit dit pour le sens) le mouvement se déterminait mieux à partir de ego vers un terme objectif que par la démarche inverse. Les données du problème changent quand la série considérée ne com­ plemente plus la P. mais le seul G.V. — sur les fonctions grammaticales des compléments de lieu-temps,v. H.Pinkster (1972:chap.6,3), et sur le S.P.v. sous-catégorisant le verbe, H.J.Seiler (1968). Cette situation propre au lieu puisque le temps n'accède qu'à la fonction syntaxique S.P.p.- entraîne quel­ ques conséquences. Certes la structure séquencielle comporte toujours les mêmes éléments -nommons-les cette fois 1 1, 1 2, 1 3 —, figurant soit à l'état isolé 11 (huc ueni en Amph. ,50 ; hic habito en Most. ,498), ou 12 (abi intro en Bacch. ,227) ou 13 (adambulabo ad ostium en Bacch. ,768 ; uno adsto in loco en Mén.,56), soit en association partielle 1 1 + 1 2 (te...hoc intro ire nolo,Hec. ,348) ou 12 + 13 (intro ire in aedis,Amph. ,617), ou 11 + 13 (Dum Uli hue ad uos exeant, Cist., 782), soit en chaîne complète 11 + 12 + 13 (ire hue intro ad filium, Bacch., 906). L'effectif des séries, en revanche, se réduit à deux. En effet les verbes admettant un complément de lieu ne sauraient

COMPLEMENTS DE LIEU ET TEMPS

99

appartenir qu'à l'un des deux groupes lexico-syntaxiques : - habitare,adstare .... nécessairement complemente en ubi; - {ad)uenire, (red)ire, commigrare, currere, concedere, reuerti..., nécessaire­ ment complemente en quo. Les verbes de chaque groupe admettant chacun, en chaque énoncé où ils figu­ rent, un seul S.P.v. à la fois, les 7 séquences énumérées ci-dessus ne peuvent se réaliser que : en série ubi ou en série quo, soit 14 combinaisons possibles. A l'unique S.P.v. s'ajoute éventuellement un S.P.p. de lieu c'est-à-dire une des séquences L décrites ci-dessus. Il n'est guère utile, quoi-qu'aisé, de calculer le nombre des combinaisons théoriques autorisées par les 7 séquen­ ces 1 ajoutées à l'ensemble des séquences L : tant l'accroissement des possibili­ tés théoriques change peu les réalisations effectives. Comme tout à l'heure nous avons vu prédominer, dans le corpus, la succession orientée L 1 (unde) → L 3 {quo) sur l'ordre inverse L 3 {unde)→ L 1 {quo), ici encore et à plus forte raison (en l'absence de 1 unde), trouvons-nous une forte majorité de L 1 {unde)→ 1 3 {quo) : in Asiam hinc abii {Heau. , 111) ou LI {unde)→ 1 2 {quo) : Cesso hinc ire intro ? {Hec. ,324) contre de rares cas de L 3 {unde)→ 1 1 {quo) : Ex Sicyone huc peruenisti {Ps. ,1174) ou L 3 {unde)→ 1 2 {quo) : Quisnam a nobis egreditur foras ? {Heau.,561). Cet état de choses confirme que le prélèvement des combinaisons effectives sur les combinaisons théoriques n'est pas aléatoire; il montre en sus que 1 et L,c.à.d. le S.P.v. et le S.P.p. de lieu, loin de se trouver juxtaposés au hasard s'ajustent proprement l'un à l'autre. A cet égard, le S.P.p. de temps se com­ porte tout autrement. 1.1.2. Temps A première vue, le temps semble jouir de moindres possibilités que le lieu. Syntaxiquement, il n'exerce que la fonction S.P.p. En matière d'organi­ sation structurelle, il ne réalise que la série ubi. A côté du "temps où l'on est", il n'existe en effet ni de "temps où l'on va" ni de "temps d'où l'on vient", pour la raison évidente que ego ne peut quitter son propre présent pour se déplacer vers son avenir ou son passé. Le "temps où l'on est", en revanche, fait l'objet d'une analyse plus fine que le lieu de même série. La temporalité s'ordonne en "avant" vs. "après", autour d'un repère fourni par le point "présent". Ce "present" est le plus naturellement celui de moi-ici-maintenant; il peut toute­ fois être constitué aussi par une date objective indépendante de ego, ainsi "les Ides de mars de l'année 700 après la fondation de Rome". Puis autour de l'un

100

H. FUGIER

ou l'autre point central se distribuent 4 postes (I,II,III,IV et I',II',III',IV), le long de l'axe qui court du repère vers l'"avant" ou l'"apres", correspondant à l'origine ou au terme d'un mouvement soit centripète soit centrifuge. On obtient ainsi les deux schémas suivants: post hoc momentum

ante hoc momentum I

ab... usque adhuc

II abhinc (tres annos)

-

ab hoc momento usque ad... III ego nunc An hoc momento) ← →

ante Idus

a futuro tempore usque ad hoc momentum IV post Idus

I' ab... usque ad Idus

ab Idibus ad...

III'

a futuro tempore usque ad Idus

IV'

Idus J IF ab Idibus usque ad (Idus antecedentes)

Tous ces postes sont pourvus dans le corpus, à l'exception de IV et IV' - et aussi, sauf erreur, de IF: les estimations temporelles dans le sens de la rétro­ version , malaisées et rarement utiles, ne sauraient être fort usitées. Abhinc se trouve en revanche, cf. Interea mulier quaedam abhinc triennium ex Andro commigrauit huc, And.,69-70 (Lundström,1961). Certes, le "lieu où l'on est" pouvait déjà faire l'objet d'une pareille sché­ matisation, avec des flèches centripètes et centrifuges, autour d'un "moi-ici" ou d'un objet pris pour repère spatial. Le parallélisme ainsi institué aurait per­ mis de comparer les réalisations observées respectivement dans les schémas "temps" et "lieu" -et de remarquer par exemple: -qu'au signifiant de L 1 hic, istic,illic,ne correspond en aucun cas, pour T 1 , un signifiant istic ou illic: puisque le temps "où tu es", "où il est", sont de nul intérêt en regard du lieu "où je suis" ; et correspond très rarement un signifiant hic {ubi...hic iam = "à ce moment", Most. ,107): pour ce dernier, son absence s'explique assez bien par le fait que hodie "ce jour où nous sommes" et nunc "présentement", tous deux très usités, comme ils impliquent hic "en cet ins­ tant où je suis, dont je parle", rendent celui-ci superflu. -Qu'aux nombreux signifiants de L 2 intus/foris, sursum/deorsum ... corres-

COMPLEMENTS DE LIEU ET TEMPS

101

pond pour T 2 l'unique paire ante(prius)/post. Mais cette micro-analyse mérite à peine d'être poursuivie. Théoriquement possible en effet, l'analyse interne du "lieu où l'on est" s'exprime beaucoup moins dans la langue que celle du temps -bien que quelques sous-systèmes res­ treints méritent description, ainsi celui de ante-prae-pro, ob, post, étudié par Ph.Marcq (1971): de sorte qu'on ne saurait mettre "lieu" et "temps" en équi­ valence systématique. Il est frappant à cet égard que les combinaisons des séquences T avec L ou l, à l'inverse de ce que nous avons constaté supra pour celles de L à L ou de L à 1, semblent purement aléatoires. Mise à part une liai­ son hodie-hic "ici aujourd'hui", dont la fréquence passe la moyenne (Cist.,16,etc...), aucune contrainte significative ne fait, des enchaînements observés: soit divers L (quo) + T (Amph. ,778; Hec. ,36), soit divers L (ubi) + T (Amph., 120), soit divers L (unde) + T (Amph. ,758), soit une séquence complexe (L unde + L quo) + T (Most. ,440), soit 1 + T (Men. ,287; Men. ,692) - une organisation plutôt qu'une simple suite juxtaposée. Surtout, certaines particularités du "temps" ont des implications qui leur sont propres. Par exemple, le huc de usque adhuc au poste I, le -hac de antehac ou posthac qui peuvent figurer aux postes I et II, impliquent un verbe d'énonciation -on s'en expliquera infra en 1.2.2.; ou encore, c'est un nom verbal, voire une P., qui peut tenir la place occupée par Idus au centre de notre second schéma: par où il apparaît que le c.c. de temps s'intègre en certains cas au système verbal. C'est là un fait syntaxique de telle conséquence qu'il nous convainc de porter maintenant au niveau syntaxique la description de nos S.P.p. de lieu et temps. 1.2. Description Syntaxique Les circonstants de lieu et temps sont affaire de syntaxe, non point par leur seule fonction de S.P.p., mais aussi par le fait qu'ils entrent en relation systématique avec le verbe en tant que tel. La chose se produit en deux domai­ nes, dont l'un intéresse "lieu" et "temps", l'autre le "temps" seulement. 1.2.1. Groupes Prépositionnels et Préverbation Sous sa forme la plus habituelle, le problème soulevé par les verbes pré­ fixés consiste à élucider la relation qu'ils entretiennent avec le complément nominal prépositionnel qui les accompagne. Les préverbes et prépositions de lieu-temps, entre autres, donnent lieu à quelques questions de ce type : - Le rapport au cas nominal de la préposition (Perrot,1966) ou du préverbe (Lehmann, 1981); - Le type de complémentarité observable entre le préverbe et le groupe pré­ positionnel. Le préverbe, en effet,

102

H. FUGIER

. tantôt confirme la préposition, soit par répétition littérale, notamment quand le groupe préposition­ nel relève de la position 3, ainsi abeunti a milite {Hec., 91), ad me adit (ft., 966), soit du moins par équivalence, notamment quand le groupe préposi­ tionnel relève de la position 1 ou 2, ainsi qui hinc abiit modo (Men.,336), eius iussu nunc huc me adfero (Amph.,989), ou intro adspexi (And. ,365) ; . tantôt lui fournit son terme opposé, en réalisant, soit l'élément unde, tandis que le groupe prépositionnel réalise l'élé­ ment quo, ainsi abii...ad hortum nostrum {Most.,1045), uicinam nostram huc euoca (Hec., 720), intro abi {Ps.,890), soit-beaucoup plus rarement- l'élément quo, tandis que le groupe pré­ positionnel réalise l'élément unde ou ubi, ainsi tuam qui amicam hinc accersebat {Ps. ,719), Hi sciunt qui hic adfuerunt {Ps. ,721). - Le degré d'intégration du préverbe au lexème verbal. Entre l'un et l'autre, 4 dispositions relatives sont attestées: ante signa pedes ire, ante ire agmen, ex castris copias educere, anteceder ø, que distinguent à la fois leur ordre diachronique d'apparition -même si la coexistence de préverbe + lexème verbal avec V + groupe prépositionnel remonte très haut dans l'histoire des langues indo-européennes (Zuntz,1936) - et leur valeur systématique (Marcq, 1971, 83-84). Des trois problèmes ainsi énumérés pour mémoire seul le dernier touche à notre propos, en ceci qu'il montre à la fois l'identité de nature entre la préposi­ tion et le préverbe par-delà les variations d'arrangements syntagmatiques, et l'unité qui s'instaure cependant entre préverbe et lexeme verbal. Or ce pro­ cessus d'univerbation n'est pas sans entraîner des conséquences, à la fois pour le sens et pour la construction des verbes considérés. Le sens peut être abordé à partir du préverbe, de façon à attacher à chaque préverbe tous ses effets sémantiquement modificateurs, ainsi pour ad- (F.Thomas, 1938:29 sq. et 39 sq.); ou, mieux, à partir du lexème verbal, pour mesurer la différence entre le verbe simple et ses diverses formes préfixées: ainsi uideo, muni du préfixe local in- "en direction de, en se portant contre...", donne-t-il "détester, envier" {Ep.,109), mais avec le préfixe local-temporel pro- "devant soi, d'avance", devient "pourvoir à, s'occuper de" {Most. ,526) (Guiraud, 1964:910). De telles oppositions sémantiques impliquent en certains cas -pensons aux dérivés en per-, spatiaux et de là perfectifs- la distinction entre deux "ordres de procès", dont maints linguistes font une affaire d'aspect

COMPLEMENTS DE LIEU ET TEMPS

103

(Fuchs, 1978:4-5; Leonard-Fuchs, 1978). Mais puisqu'en tout état de cause le sens nous importe moins ici que la syntaxe, les préfixés nous présentent surtout de frappantes innovations constructionnelles -bien décrites par M.Lemaire (1979). Ainsi, s'écartant du tran­ sitif uideo, inuideo s'emploie-t-il soit absolument, soit avec le datif (Pers. ,777; Ps., 1135); de mêmeprouideo avec le datif (Most.,526). 1.2.2. Groupes Prépositionnels et Temps Verbaux Pour la bonne part d'entre eux qu'introduit une préposition à valeur rela­ tive telle que simul cum X, ante X, post X, les circonstanciels de temps impli­ quent un terme qui leur est extérieur, soit: le V dénotant le procès qui a lieu précisément "en même temps que X, avant X, après X" ; ils forment avec ce V un système lié par la corrélation X ↔ V. Quelques précisions sur X et V feront mieux saisir cette importante fonction syntaxique des c.c . temporels. Dans le schéma général de P.: "Un tel a fait telle chose avant X"

(S.N.I + S.V.) + (prép. + X) X admet diverses réalisations; mais V, lui -c'est-à-dire plus précisément le procès auquel V renvoie- peut se situer à deux niveaux différents du discours, à savoir, celui de l'énoncé ou celui de l'énonciation. Comme cette distinction prime l'autre en intérêt linguistique, nous choisirons pour clé d'analyse V plu­ tôt que X, pour décrire le jeu complet des corrélations possibles entre X et V. a) Quand V se trouve être un verbe d'énoncé, il apparaît en l'état effective­ ment réalisé, c.à.d. comme une forme de conjugaison. Quant au X qui l'accompagne, il laisse varier à la fois (en structure de surface, s'entend) son identité morphologique et son mode de rattachement syntaxique à V. On trouve en effet: a) Le type Obseruabo quid agat hominem: post adibo atque loquar (Men. ,465). Le lieu où s'opère la corrélation est la P. post adibo atque loquar. L'élément V, adibo atque loquar, y est accompagné, en fait de X, par le circonstant post ø. Si l' ø était rempli il le serait par la P. qui précède immédiate­ ment, obseruabo quid agat hominem. Post est remplaçable, dans une organi­ sation identique, par simul (Ps. ,863), prius (Most. ,539), ou ante (Most., 199). (3) Le type Earn (sc. pateram) huic dedisti, posthac rursum obsignasti clanculum (Amph.,196-191). Il ne diffère en rien de a) sinon par le -hac occupant l ' 0 , qui renvoie anaphoriquement à la P. antérieure Earn huic dedisti. Antehac se rencontre aussi (And.,186). y) Le type Postquam aspexi, ilico cognoui (Heau. ,656-657). La P. aspexi, qui selon le modèle a) aurait précédé post cognoui comme la première de deux

104

H.FUGIER

indépendantes, se trouve ici introduite sous post par le subordonnant quant. Prius quant fournit des insertions du même genre (Most., 326-327). Enfin le N verbal prius tua opinione (hic adero, Amph. ,545) équivaut structurellement à prius quant opinaris. En résumé, V et X se situent respectivement ainsi: a) Obseruabo ... ; post 0 adibo

X

V

(3) Earn huic dedisti, posthac rursum obsignasti X

V

y) postquam aspexi, ilico cognoui prius tua opinione hic adero X V. b) Mais il arrive aussi que X ne se montre corrélé à aucun V présent dans l'énoncé. Il n'est pas pour autant dépourvu de tout lien avec quelque procès que ce soit; mais celui auquel il se réfère n'est autre que le procès d'énoncia­ tion: post X, ante X mesurent alors la distance qui sépare, en avant ou en arrière, le temps de X de celui où se produit l'acte énonciateur. Soient par exemple: a) Nupta est illa, obsecro? |.-Scies haud multo post (Bacch. ,852-853). Nul constituant de l'énoncé ne se prête cette fois à occuper la place consécutive à post. Aussi bien cette place est-elle vide par principe, puisque post articule entre eux non pas deux éléments de l'énoncé, mais un V de l'énoncé (scies) avec le procès d'énonciation. Scies haud multo post = littéralement "Tu vas le savoir peu après (sc.l'instant où je parle)". Prius joue le même rôle que post, par exemple en Cist. ,546-547 : Hinc ex hisce aedibus paulo prius uidi exeuntem mulierem. Cf. ante ti prius dans And.,239. (3) Et caue posthac, si me amas, umquam istuc uerbum ex te audiam (Heau., 1031). Ici anaphorique, -hac renvoie au présent de l'énonciateur: "Et veille, de l'instant présent, à faire que ...". De même font postea, postilla, antehac, antidhac. Connecteur de deux temporalités, soit toutes deux énoncées, soit une seule énoncée mais mise en relation avec le présent de l'énonciateur, post (prius, simul) collabore dans cette fonction avec une série de particules qui, toutes différentes qu'elles soient par leur point et leur mode d'intervention, parfont le système de l'intertemporalité: iam, (iam) diu, (iam) dudum, (iam) nunc, (iam) pridem ... Pour qui part d'un instant-repère (ego- ou allocentrique, peu

COMPLEMENTS DE LIEU ET TEMPS

105

importe ici), iam indique, soit une avancée du procès vers l'avenir: → soit une récapitulation du procès, considéré rétrospectivement à partir de l'instant-repère: ← cf. D.Haller-Anstett (1970:54 et 58). Nunc fournit seulement un signifiant supplémentaire au signifié '"avancée", et diu, dudum, pridem, au signifié "retrospection". Mais n'oublions pas que notre enquête présente porte sur l'insertion de iam dans des P. à deux termes temporels V 1 et V 2, articulés par postquam et antequam (simul atque se trouvant évidemment exclu de tels sys­ tèmes de décalages chroniques): et notons que postquam indique le début de la portion de temps qui s'écoule avant l'instant-repère; priusquam le début de la portion de temps qui s'écoule après l'instant-repère. Compte tenu, enfin, du fait que iam peut porter aussi bien sur V 1 que sur V 2, de ces diverses sortes de données résultent dès lors les réalisations suivantes: N.B. Pour la clarté du schéma, on a modifié l'ordre des mots dans les textes utilisés, pour amener iam au plus près de la barre verticale "instant-repère". Aussi bien l'ordre des mots n'est-il pas pertinent pour notre question actuelle. -postquam iam V2 | V 1 : postquam pueri septuennes iam sunt \ pater onerauit nauim (Men. ,24-25); -postquam V2 | iam V1 : Ubi illo immigrat nequam homo... \ iam hic aedibus uitium additur (Most. ,105-107); -iam V 1 | priusquam V2 : Dicampatriiam \priusquam malum istoc addis → (Bacch., 382-383); - V 1 | priusquam iam V2 : ne se trouve pas, sauf erreur. Iam semble incom­ patible avec priusquam; -postquamV 2 \ iam (diu) V 1 : Postquam bibimus iam diu factum est (Pers., 822). Seul cas où la flèche franchisse la barrerepère. Le "depuis longtemps" s'étend vers l'arrière jusqu'au début du procès bibere: le fran­ chissement est inclus dans la situation "depuis que -iam V 1 | ante V 2 :

(Si) cui in re tali subuenisti iam antidhac (Aul., 396). Le système créé par la collaboration de post (ante) et de iam (diu...) dis­ tribue et organise des niveaux temporels, dont le signifiant normal est le temps verbal: la relation entre les temps de V 1 et V 2 constitue donc, lui aussi, un sous-système corrélé avec post-iam. L'ensemble des adverbes temporels + V., susceptible d'une description complète en Latin comme en d'autres lan-

106

H.FUGIER

gues (pour le Français, Klum,1961), confirme notre constatation antérieure, sur le changement de construction verbale par la préverbation. Dans les deux cas, les prépositions de lieu-temps interviennent sur le verbe, et par là, concernent la P. entière; dans les deux cas, l'ensemble signifiant lieu-temps correspond à un fonctionnement grammatical. Il doit en aller de même pour l'autre groupe majeur des circonstants, qu'il nous reste à décrire. 2.

Des Ad-Verbes: La Série Numquam

...Semper

2.1. Description Lexicologique Il existe un groupe d'adverbes, exclusivement temporels, qui à l'inverse des précédents n'entretiennent aucun lien formel avec la flexion nominale et qui, pour le sens, indiquent tous avec quelle fréquence se répète le procès dénoté par le verbe. Comme la fréquence est affaire de "plus ou moins", ces adverbes tendent à s'ordonner en série à la façon d'une gradation. Chaque terme de la série requiert une courte description explicative: - NUMQUAM, "ne pas ... à quelque moment que ce soit", nie la réalisation du procès en un quelconque instant du passé, du futur ou d'un présent assez largement défini pour comporter plusieurs points temporels distincts: c'est une négation "à dimension temporelle" (Bacri, 1976,chap. IV). La valeur négative vient-elle à éclipser la temporalité, numquam fonctionne alors comme une pure et simple négation, "absolument pas! jamais de la vie!" {Numquam hercle factum est, Men.,533; Id quidem edepol numquam erit, Men.,665; numquam ullo modo, Amph.,560). Le fait est avéré en particulier lorsque, associé à ullus, numquam nie l'existence plutôt que l'assignation temporelle du procès: Scelestiorem ego annum argento fenori \numquam ullum uidi, quam hic ... — scelestiorem annum nullum uidi (Most.,532); lorsqu'il nie la vérité de l'assertion plutôt que la réalité du procès: Ubi illae armillae sunt quas una dedil \ -Numquam dedisti = "Mais tu ne me les a pas donnés! Il n'est pas vrai que tu me les aies ...!" (Men.,536-537); ou lorsqu'enfin l'emploi indifférencié de numquam-nusquam atteste la neutrali­ sation des sèmes temporels-locaux au profit de la seule négation, cf. la pré­ sence simultanée de umquam usquam dans Amph., 1036, et la liaison nusquam-ubi dans Ego nusquam dicam, nisi ubi factum dicitur (Men., 10). Cf. aussi Amph., 620; Ps.,845. La valeur temporelle, en revanche, s'impose-t-elle fortement, alors num­ quam "à nul moment", et par implication "pas une seule fois", devient l'éche­ lon inférieur d'une gradation "zéro fois, une fois, deux fois...". Le sens impli-

COMPLEMENTS DE LIEU ET TEMPS

107

que "pas une seule fois" ressort auprès d'un parfait aoristique à valeur ponc­ tuelle: O fortunate, nescis quid mali praeterieris, qui numquam es ingressus marel {Hec.,418-419) = "Tu ne sais pas à quels maux tu as échappé, toi qui pas une seule fois n'as pris la mer!". Il autorise numquam à figurer en parallèle avec d'autres termes de la même gradation tels que saepe ou semper. Ainsi dans I' Hec.,486-487, Pamphile reconnaît-il que Philoumène n'a "pas une seule fois" {numquam) manqué à ses devoirs envers son époux, qu'elle s'est au contraire "souvent" {saepe) comportée comme il le souhaitait; tandis que numquam s'oppose à semper comme deux termes extrêmes de la même échelle, en Cist. ,48-49 ou Most. ,306-307. L'appartenance de numquam à la série se mani­ feste enfin syntaxiquement par le fait qu'il admet le même type de complé­ ments que les autres termes. Il se fait suivre en effet d'un circonstant lui-même temporel, qui relève: - soit du type ante ou post + N, ainsi ante hunc diem {Hec.,863), ante hesternum diem (Ps.,731), antidhac {Ps. ,620-621), post hunc diem {Men., 477), pos­ tilla {Men., 1117), postquam natussum {Men., 399-400); - soit du type ablatif-locatif sans préposition, ainsi hodie numquam (Amph., 398), ou (n) umquam in uita mea (Heau., 1006). A quoi on comparera semper ante ou antehac {As., 861), a paruolo semper {And., 35), decies die uno (Aul., 70), saepe aetate in sua {Men., 839). -SEMPER clôt vers le haut la gradation dont numquam fournit le degré infé­ rieur - paire contrastée souvent mise à profit pour rythmer le dialogue {Hec., 860-861). Semper, littéralement "une fois", mot à double entente, dont une seule des virtualités sémantiques correspond aux exigences de notre grada­ tion. En tant qu'il vaut "une fois pour toutes", d'où "durablement, pour tout l'avenir", semper accompagne perpetuus {Most., 306) ou perennis (Amph.,14), et rien ne distingue de usque "continument" (usque adero domi, Cas.,530) le semper de As.,654: Di te seruassint semper. Ce n'est pas un tel adverbe, avec son signifié "durée", qui trouve sa place dans la série "zéro fois, parfois ...". Mais "une fois" peut s'entendre aussi bien "une fois renouvela­ ble" , d'où un semper "fois après fois, à tous les coups", notamment lorsqu'un autre lexème adjoint vient préciser la périodicité de cette répétition, tel cottidie dans Aul. ,23-24: Huic filia una est; ea mihi cottidie aut ture aut uino aut aliqui semper supplicat. Et ce semper-ci répond bien à numquam.. - ALIQUANDO, SEMEL, BIS ... Entre "zéro fois" et "à toutes les fois" s'étagent naturellement les termes servant à nombrer les occurrences du pro­ cès. Aliquando "un certain nombre de fois" recouvre un certain nombre de possibilités numériques, dont chacune s'explicite dans l'un des adverbes de la

108

H. FUGIER

série arithmétique: semel "une fois", bis "deux fois", ter ..., decies (AuL., 70), miliens (Eun., 422)... Iterum, rursum, denuo viennent en équivalence avec bis. - SAEPE estime, sans les chiffrer, le nombre d'occurrences affectant le pro­ cès. Ce thème adjectival, encore muni d'un comparatif et d'un superlatif, à l'origine devait se comporter syntaxiquement comme continuo (infra, 3.1.): il prédiquait la P., de sorte que le procès se trouvait qualifié de "fréquent", lit­ téralement "dense". Mais dès nos plus anciens textes, son intégration à la série des adverbes nombrants ne fait aucun doute; chez Plaute, ses relations contextuelles avec semel (Cist.,504), iterum (Hec., 186-187), miliens (Eun., 422 et 429) ... la rendent incontestable. Mais ce qui fait, de l'ensemble numquam; aliquando, semel, bis ..., miliens; saepe; semper, une série homogène, ce sont surtout les traits de fonc­ tionnement syntaxiques communs à tous ces éléments. 2.2.

Description Syntaxique

2.2.1. La Quantification Verbale Les termes de notre série sont des ad-verbes en ce sens précis qu'ils por­ tent sur le procès et non sur le lexème. Certes plusieurs d'entre eux coexistant dans la même P. peuvent se situer à des niveaux différents de la structure phrastique, ainsi dans As., 420-422: Abscede ac sine me hunc perdere, qui sem­ per me ira incendii, \ cui numquam unam rem me licet semel praecipere furi, \ quin centiens eadem imperem atque ogganiam. Numquam me licet domine syntaxiquement semel praecipere; or tout se passe comme si l'adverbe de la complétive insérée entretenait avec le verbe un rapport plus étroit que ne fait celui de la principale insérande: "je ne puis jamais me contenter de direles-choses-une-bonne-fois" (trad. A.Ernout). Cf. Aul., 69-70: ... ita me miseram ad hunc modum \ decies die uno saepe extrudit aedibus = " et souvent, dix-fois-par-jour-il-me-met-à-la-porte" (trad. A.Ernout). Est-ce une illu­ sion, un effet de syntaxe? Peut-on même légitimement parler de relation "plus ou moins étroite" entre l'adverbe et le verbe? En serait-il ainsi, cependant, que même dans leur état le plus "serré" la plupart de nos adverbes n'en vien­ draient jamais à porter sur rien d'autre que le procès. Procès et lexème verbal, il s'agit bien en effet de deux choses différentes: pas plus qu'en Français "sif­ fler deux fois, trois fois ..." = "reproduire deux, trois fois l'acte simple de ..." ne se confond avec "siffloter" = "se livrer, sans interruption mais sans réitéra­ tion non plus, à l'activité de ...", pas davantage centiens eadem imperare n'équivaut à imperitare. L'activité continue ou habituelle, centrée sur sa pro­ pre réalisation, qu'exprime -itare, n'admet pas pour signifiant, en règle géné-

COMPLEMENTS DE LIEU ET TEMPS

109

rale, les adverbes de répétition. S'excluent notamment de ce rôle: - numquam, pour des raisons évidentes; - les adverbes correspondant à un nombre cardinal précis: bis, ...milieus. Car seul le procès se nombre; l'activité, elle, se prolonge. A cet égard semper, "à chaque coup", nombrant distributif, ne peut guère affecter que le procès; - saepe, dans la plupart de ses emplois. Admis que cet ancien adjectif soit presque partout nombrant = "plus d'une fois, de nombreuses fois" (cf. Fac semel periculum. -Feci saepe, en Cist.,504), et parfois encore quelque peu qualitatif = "de façon serrée, répétitive" ; admis qu'il voisine avec soleo (Solet iocari saepe mecum illoc modo, Men., 317), et accompagne parfois des formes en -itare (uisitatus saepe, Ps.,727): il reste qu'en l'immense majorité des cas, saepe + V s'entend comme "il arrive souvent que V". Cf.Capt., 108-109: Sed aperitur ostium, \ unde saturitate saepe ego exii ebrius = "d'où il m'est arrivé plus d'une fois de sortir ivre" ; ou de façon évidente dans la citation des adages, ainsi Capt. ,165: Ut saepe summa ingenia in occulto latent = '"Comme on con­ state souvent que les grands esprits ...". Malgré l'imperfection de cette trop brève analyse, nous voyons assez bien désormais en quel sens les ternes de la série numquam,... semper méritent le nom d'ad-verbes: ils se rapportent au procès, pour en dénombrer les occur­ rences. Quantificateurs du procès, les adverbes nombrants gagnent à cette défi­ nition un statut logique et linguistique original. Le logicien discerne dans le rapport de semper à aliquando celui de l'universel à l'existentiel; la négation "à dimension temporelle" lui pose des problèmes particulièrement complexes de formalisation; l'afférence des adverbes nombrants au procès n'est pas sans incidence sur les problèmes de l'assertion et du posé/présupposé. Quant au linguiste descripteur du Latin, laissant pour aujourd'hui des questions qui dépassent en tous sens le cas particulier de cette langue, il observera la corres­ pondance qui associe terme à terme nos quantificateurs verbaux aux quantifi­ cateurs nominaux: numquam fortunam commutauit ↔ nulla fortunae commutatio aliquando „ „ ↔ aliquot fortunae commuta­ tion es semel, bis... „ „ ↔ una, duae „ saepe „ „ ↔ plures semper „ „ ↔ omnes Faute d'études suffisantes sur les aspects tant théoriques qu'empiriques de la question (toutefois Lewis, 1975). on ne saurait présenter catégoriquement les

110

H. FUGIER

quantificateurs nominaux comme produits des quantificateurs verbaux par l'effet de la transformation nominalisante. Pour le moins, la série nullus, unus,...omnes fournit-elle une paraphrase convenable de la série numquam, semel,...semper. Pour un éventuel travail d'approfondissement, des maté­ riaux seraient fournis par quelques configurations syntaxiques intéressantes, telles que feriet saepius vs. una plaga dans Amph. ,704-705; ou Hec, 180-181: Neque lites ullae inter eas, postulatio \ numquam. La quantification verbale inclut-elle, ou implique-t-elle, des valeurs aspectuelles? Mettra-t-on en relation, comme le fait W.Dressler (1968: § 1318 et 26-28), la "pluralité verbale" avec les notions de perfectif/imperfectif? Peut-être, si l'on entend, avec cet auteur, par pluralité verbale des réalisations telles que l'itératif. Mais si l'itératif, comme nous le croyons, a pour signifiant en Latin des suffixes verbaux ou des circonstants de "manière", et pour domaine d'intervention sémantique le lexème plutôt que le procès; si la cor­ respondance de la pluralité verbale à la pluralité nominale doit s'établir, comme il nous semble, du signifiant-signifié "itération" au pluriel interne, c.à.d. qui analyse le signifié du singulier en plusieurs éléments: exercitare vs. exercitationes uirtutum = "l'exercice (aux formes multiples) de la vertu", mais des adverbes quantificateurs verbaux aux seuls pluriels nombrants: "changer deux fois" ↔ "deux changements", alors nos adverbes nombrants n'ont pas de relation directe avec la perfectivité. Tout au plus parlerons-nous à leur sujet de compatibilité avec certaines indications lexématiques, en ce sens que tel lexème verbal impliquant "ponc­ tualité" exclut la co-présence d'un nombrant {*decies mori). 2.2.2. Quantificateurs Verbaux et Modalisation Quantificateurs du procès, nos adverbes touchent par là à une catégorie à première vue très différente: les auxiliaires de modalité. "Auxiliaire" par exem­ ple licet parce que dans licet feli regem spectare (proverbe) les rapports de sous-catégorisation jouent de regem vers spectare et non vers licet. "De moda­ lité" parce qu'ils correspondent aux diverses modalisations possibles d'un procès que l'énonciateur estime soit possible, soit nécessaire; si, omettant les modalités aléthiques hors de propos ici, nous ajoutons avec une partie des phi­ losophes antiques que la temporalité également vaut pour une sorte de moda­ lisation (Gardies,1979:65), nous retiendrons donc trois auxiliaires, potest, debet, solet. Quels sont donc les adverbes nombrants qui correspondent à cha­ cun d'entre eux?

COMPLEMENTS DE LIEU ET TEMPS

111

- Potest correspond parfaitement à aliquando, au besoin à saepe. On trouve aussi fere = "généralement" (Capt.,232) ou fit (Capt., 25); et sous certaines contraintes sémantiques, licet. - Debet correspond exclusivement à semper. - Solet couvre l'espace qui s'étend de aliquando à semper; entre les deux, saepe lui fournit son plus juste équivalent. Cf. solet iocari saepe en Men., 317; et Men.,302-305: Non scis quis ego sim, qui tibi saepissume \ cyathisso apud nos, quando potasl ... |.- Tun cyathissare mihi soles, qui..? Il reste beaucoup à trouver sur ces adverbes nombrants -même si leur relation à l'auxiliaire cor­ respondant ne mérite pas proprement le qualificatif de symétrique: puisqu'en effet le "toujours" n'est pas le "nécessaire" mais seulement son conséquent, puisque tout debet ne se reformule pas en semper, ni, encore moins, tout sem­ per en debet (Capt.,800: Semper memini = debeo meminissel)', ni, pour des raisons analogues, tout potest en aliquando ou inversement: ni tout solet en saepe ou inversement. La zone de non-correspondance peut d'ailleurs se définir en termes syntaxiques. Pour "pouvoir" notamment, admettent la reformulation par aliquando l'impersonnel potest = "il est possible que", et les emplois du verbe personnel possum liés transformationnellement à l'impersonnel: pos­ sum uincere = potest fieri ut uincam, d'où aliquando uinco; la refuse, le verbe possum centre sur le sujet et exclusivement personnel = "je suis capable de": possum uincere = "je suis bien assez fort pour vaincre" = potest fieri ut... et donc = aliquando uinco. Les contraintes d'horaire imposées aujourd'hui, et de façon plus géné­ rale les insuffisances dont souffre encore la description linguistique des quan­ tificateurs en Latin, nous arrêtent ici. Du moins les correspondances établies, les paraphrases suggérées, montrent-elles déjà que cette classe d'adverbes joue sans conteste un rôle syntaxique effectif et original: au sens "X fois" cor­ respond bien un fonctionnement propre. Après le S.P.p. Ulic intus ... et l'adverbe nombrant, existe-t-il encore, parmi les circonstants de lieu et de temps, un autre groupe qui se définisse par la coïncidence d'un sens et d'une syntaxe? 3. Les C.C. de lieu-temps, un ensemble fini 3.1. Un type "continuo''? Si seinei, ter ... relèvent à l'évidence des quantificateurs verbaux, n'en pourrait-on dire autant de primo, tertio ...? Puis de postremo, symétrique de primo par le sens et la morphologie? Faut-il, dans la foulée, prendre en compte le groupe entier des thèmes adjectivaux à désinence ablative qui

112

H. FUGIER

caractérisent le procès par sa fréquence (crebro, raro), sa durée (assiduo, perpetuo), ou le décrivent comme rapide, subit, consécutif au procès précédent, ou tardif (cito; subito, repente; continuo; sero)? Faut-il enfin, à primo-postremo, joindre les formes d'accusatif primům et postremuml La syntaxe en décidera. Tous ces ablatifs — sans doute aussi les accusatifs — sont des prédicatifs de phrase, c'est-à-dire des adjectifs prédiquant un sujet qui n'est autre que la P. entière. Relativement à ce sujet-phrase, continuo joue le même rôle que l'adjectif prédicatif par rapport au nom sujet ou complément d'objet direct -à ceci près que la question de l'accord casuel du prédicat se pose diffé­ remment quand le sujet n'est pas lui-même un nom fléchi: Equités parent citi (Amph. ,244) Ahtiquam rem nouam ad nos proferam (Amph. AIS) Capimus consilium continuo (Bacch.,300). Quant au sens, il est celui qu'autorise la construction predicative: capimus consilium capimus consilium continuo ia

continuum

= "Nous prenons une décision, (prise de décision) immédiate". Sens et struc­ ture syntaxique, ici encore, semblent donc bien s'accorder. Seulement, est-ce autre chose qu'une apparence? Car la construction, cette fois, ne caractérise pas en propre le groupe continuo ... : puisque tout adverbe dit "de manière" tel que certo, mais aussi tel que sancte ou sapienter, prédique également la phrase. Aussi bien les adverbes de type continuo ne sont-ils pas non plus pour le sens vraiment des temporels. Ils qualifient le procès en lui attribuant le caractère de "fréquent, rapide ..." signifié par leurs lexemes respectifs: mais il y a loin de là à situer le procès dans l'espace ou le temps, ou à le quantifier, comme faisaient les types précédemment décrits, en un fonctionnement spé­ cifique et cohérent. 3.2. Situer et Quantifier Bref, il n'existe que deux sous-classes d'adverbes de lieu et temps: a) Un type de séquence pronom deictique 4- adverbe d'orientation + (préposition + N), jouant le rôle syntaxique de circonstanciel de P. pour "lieu" et "temps" circonstanciel de V. pour "lieu", dételle sorte que: - la préposition (de lieu surtout), fonctionnant aussi bien comme préverbe,

COMPLEMENTS DE LIEU ET TEMPS

113

institue des oppositions lexicologiques entre le verbe préfixé et le simple, non sans implications quant à 1'"ordre de procès"; - le N., s'il est de type verbal, entre en système avec le temps principal, celui de l'énoncé ou celui de renonciation, contribuant ainsi à construire la tem­ poralité de la phrase entière, b) Un type d'ad-verbe étranger à la flexion nominale, formant une échelle de quantificateurs verbaux parallèles aux quantificateurs nominaux, qui nombrent les occurrences du procès, et sans toucher vraiment à l'aspect entraînent cependant quelques contraintes de compatibilité avec cer­ tains thèmes verbaux; et qui correspondent en un certain sens à des auxi­ liaires de modalité. Le premier type participe à la temporalité de la phrase, le second assure la quantification du procès. Quant au type continuo, il ne contribue que par son sens lexématique, éventuellement temporel, à la qualification du procès. *NOTE DE LECTURE: ABREVIATIONS. N - nom; G.N. = groupe nominal; V = verbe; G.V. = groupe verbal; c e . = complément circonstan­ ciel; S.P. = syntagme prépositionnel; S.P.v. = syntagme prépositionnel de verbe; S.P.p. = syntagme prépositionnel de phrase; P. = phrase.

BIBLIOGRAPHIE

Bacri, N., (1976), Fonctionnement de la négation, Paris-La Haye, Mouton. Cupaiuolo, F., (1967), La formazione degli avverbi in Latino, Napoli, Libreria scientifica. Dressler, W., (1968), Studien zur verbalen Pluralităt. Iterativum, Distributivum, Durativum, Intensivum in der allgemeinen Grammatik, im Lateinischen und Hethitischen, Wien-Köln-Graz, Böhlaus. Fuchs, C., (1978), 'L'aspect, un problème de Linguistique générale: éléments de réponse dans une perspective énonciative, D.R.L.A. V., 16, 1-30. Gardies, J.L., (1979), Essai sur la logique des modalités, Paris, P.U.F. Guiraud, Ch., (1964), Les verbes signifiant "voir" en Latin. Etude ďaspect, Paris, Klincksieck. Haller-Anstett, D., (1970), Iam, mémoire dactyl, Université Strasbourg IL Klum, A., (1961), Verbe et adverbe. Etude sur le système verbal indicatif et sur le système de certains adverbes de temps à la lumière des relations verbo-

114

H.FUGIER

adverbiales dans la prose du Français contemporain, Stockholm-Göteborg-Uppsala, Almqvist et Wiksell. Knight, C.M., (1921), Greek and latin adverbs and their value in the recon­ struction of the prehistoric declensions, Cambridge, University Press. Lehmann, C , (1981), 'Case roles and preverbs', This volume. Lemaire, M., (1979), Recherches sur la suffixation et la préfixation des verbes latins, Paris, thèse 111° cycle Université Paris VII, dactylographiée. Leonard, A.M., et Fuchs, C , (1978), Vers une théorie des aspects, Paris-La Haye, Mouton. Lewis, D., (1975), in Keenan, E.L.(ed.) Formal semantics of natural lan­ guage, Colloquium of the King's College Research Centre, Cambridge, University Press, 3-15. Lundström, S., (1961), Abhinc und ante, Studien zur Geschichte der lateinischen Zeitdifferenzbestimmungen, Lund, Gleerup. Marcq, Ph., (1971), 'Structure d'un point particulier du système des préposi­ tions spatiales en Latin classique', Linguistique, 7, 81-92. Perrot, J., (1966), 'Le fonctionnement du système des cas en Latin', Revue de Philologie,40, 217-227. Pinkster, H., (1972), On Latin Adverbs, Amsterdam, North Holland publ. Seiler, H.J., (1968), 'Probleme der Subkategorisierung mit Bezug auf Bestimmungen des Ortes und der Zeit', Lingua, 20, 337-367. Steinitz, R., (1969), Adverbial Syntax, Berlin, Akademie-Verlag. Thomas, F., (1938), Recherches sur le développement du préverbe latin ad-, Paris, Klincksieck. Zuntz, L., (1936), Die hethitischen Ortsadverbien arha, pară, piran als selbststăndigen Adverbien und in ihrer Verbindung mit Nomina und Verba, diss., München, Pilger Drückerei.

Le corpus de textes utilisé est: Plaute et Terence, éd. Collection des Universi­ tés de France, Paris, Belles-Lettres.

115 REACTION A LA LECTURE DE H. FUGIER

CH. GUIRAUD M. Guiraud remercie H. Fugier pour son exposé courageux et, dans l'ensemble, convaincant. Il ne parlera pas de la première partie, distributionnelle, avec laquelle il est d'accord. Avec quelques réserves cependant, par exemple: qu'est-ce qui distingue L 1 (Ulic) et L 2 (intus) sinon la morphologie? Pour H. Fugier, les adverbes temporels-locaux peuvent être auxiliai­ res - soit de temps. Ex. : post aduentum eri = postquam erus aduenit -soit de mode. Ex.: aliquando fit ut équivalant à potest fieri ut, c'est-à-dire éventuel. - soit ďaspect. Ex. : les quantificateurs verbaux qui marquent la répétition du procès. Mais alors: il y a vraiment toutes les caractéristiques verbales dans les adverbes de temps et de lieu et comment s'articule leur syntaxe avec celle du verbe? H. Fugier dit que des adverbes comme perpetuo, cito, etc. prédiquent l'ensemble. Ex.: capimus consilium continuo serait à lire: capimus consilium et id continuum est. "Nous prenons cette décision et cette décision est immé­ diate". Mais ne peut-on réinterpréter ainsi d'autres niveaux du système? Ex. : "Il a fait cela dans cet endroit" peut devenir "Il a fait cela et c'était dans cet endroit". Pour finir, une question très générale: si H. Fugier a pris comme cor­ pus Plaute et Terence, c'est sans doute en raison de l'abondance du matériel. Mais cette analyse serait-elle valable à toutes les époques et pour tous les niveaux de la langue latine?

L'IMPORTANCE D U VERBE DANS LA PHRASE LATINE

CH. G U I R A U D Université Paris X RESUME L'importance du verbe n'a pas été établie par les grammairiens anciens, malgré la terminologie. Tout simplement sans doute parce que les grammai­ riens anciens n'ont pas exploité l'opposition philosophique entre sujet et pré­ dicat. Cependant, dès Priscien, le verbe apparaît chargé d'une valeur particu­ lière dans la mesure où il devient le centre de la notion de transitivité. La linguistique contemporaine est, sur ce point, beaucoup plus systéma­ tique que les Anciens puisque le verbe est maintenant présenté comme l'élé­ ment de base (doctrine de L.Tesnière, reprise par H.Happ, F.Charpin). Nous ne pouvons accepter ces positions, ni sur le plan du discours (où n'importe quel élément peut être emphatisé ou focalisé), ni sur le plan de la langue. En effet, il existe en latin des phrases sans verbe ou phrases nomina­ les, que l'on peut définir de manière positive. De plus, le latin peut contribuer à rétrécir l'importance du verbe person­ nel dans la phrase, avec l'existence et le développement de l'infinitif dit "de narration". L'importance du verbe n'a pas été établie par les grammairiens anciens, et uerbum). Tout simplement sans doute parce malgré la terminologie ( que les grammairiens anciens n'ont pas exploité l'opposition philosophique entre sujet et prédicat. Pourtant l'analyse verbe/nom se trouve déjà dans la Cratyle1 et il est probable que Platon n'en est pas l'inventeur. Mais, s'il a été facile de distinguer le verbe dans la variété des éléments d'une phrase, il n'a pas été facile de le définir. Aristote semble être le premier à s'être posé le pro­ blème de la valeur du verbe dans la phrase. S'il ne l'a pas résolu, du moins a-til laissé son empreinte dans une définition du verbe et dans quelques remar-

118

CH. GUIRAUD

ques qu'on peut en rapprocher. La définition d'Aristote a été souvent reprise, mais en général assez mal comprise, parce qu'on ne l'a pas reproduite dans son intégralité. On doit citer la phrase suivante: "Le verbe est un mot qui signifie en outre le temps, dont aucune partie ne signifie séparément du tout, et qui est toujours le signe de ce qu'on dit d'une chose" .2 Cela est très positif et très moderne: on croirait lire un chapitre de la Grammaire générale et raisonnée de Port-Royal. Mais Aristote tourne court lorsque, dans les explications qui suivent ces lignes du De interpretatione, il refuse le nom de verbe à toute ), ainsi qu'à toute forme qui exprime un expression négative (ex. : temps autre que le présent. Aristote raisonne ici en logicien car, pour un grammairien, la négation ne réagit pas sur la nature du verbe. Chez les grammairiens latins, la notion de "ce qui est affirmé" ( chez Aristote) n'apparaît qu'assez tard. Le mot praedicamentum se lit au 4è. siècle, chez Marius Victorinus3, et encore le sens n'apparaît-il pas tout de suite: praedicamentum ne devient l'équivalent de qu'au 5è.siè­ cle (traduction des textes aristotéliciens par Boèce). Priscien ne semble pas avoir toujours très bien vu l'importance du verbe, si l'on se réfère à des affirmations comme celles-ci: à propos de la construction d'un terme à un cas oblique, il note tantôt que ce terme se rapporte au verbe 4 , donc en tant que celui-ci est transitif, tantôt que ce terme se rapporte au nomi­ natif5, c'est-à-dire à la substance initiale. Cependant, dès Priscien, le verbe apparaît chargé d'une valeur particu­ lière dans la mesure où il devient le centre de la notion de transitivité. "C'est à partir de cette constatation du rôle tenu par le sens des verbes dans la constitu­ tion de la transition que les verbes ont fini par être choisis, à un degré ultime de cette évolution, comme centre de la notion, en tant qu'ils constituent la médiation de la transition". 6 . Pour retrouver des analyses syntaxiques utilisables sur le rôle du verbe, il faut aller jusqu'au 17è.siècle7, avec la Grammaire générale de Port-Royal: "Le verbe qui n'est rien autre qu'un mot dont le principal usage est de signifier l'affirmation, c'est-à-dire de marquer que le discours où ce mot est employé est le discours d'un homme qui ne conçoit pas seulement les choses, mais qui en juge et qui les affirme".8 On sait que la linguistique contemporaine a systématisé le rôle du verbe dans la phrase. Mentionnons à ce propos la doctrine de L.Tesnière 9 , reprise pour le latin par H.Happ 10 sur le verbe "centre de phrase". L'analyse d'une phrase commence par le verbe et tout s'ordonne à partir de lui. Certains lin­ guistes ne discutent même pas ces positions, qu'ils établissent comme des

L'IMPORTANCE DU VERBE

119

axiomes. Ainsi F.Charpin: "Sur le plan syntaxique la forme verbale apparaît essentiellement comme le noeud de l'énoncé, c'est-a-dire comme le régissant qui commande un ou plusieurs subordonnés. Toutes les fonctions s'organi­ sent par rapport au verbe. Il constitue le centre syntaxique de la phrase". 11 On ne saurait accepter de pareilles propositions. Elles ne sont déjà pas valables au plan du discours. Il est bien évident en effet que l'utilisation de la langue en discours amène à "emphatiser" ou à "focaliser" tel ou tel élément d'une phrase; ce peut être le verbe, ce peut être aussi un quelconque autre élément. L'on voudrait maintenant établir qu'au plan du fonctionnement syn­ taxique l'importance du verbe doit être restreinte. Il convient d'examiner les problèmes posés par l'existence de phrases sans verbe, moins nombreuses sans doute en latin que dans une langue comme le grec ancien, mais qui apparaissent tout de même. La fréquence de ces phra­ ses sans verbe varie suivant les auteurs: très élevée chez Tacite 12, elle est moyenne chez Salluste et Tite-Live. Des sondages font apparaître que, dans les Métamorphoses d'Ovide, il y a à peu près une phrase sans verbe tous les cent vers, alors que, chez Virgile, on trouverait, pour le même corpus, quatre ou cinq phrases sans verbe. 13 Les phrases sans verbe doivent-elles être considérées comme des phrases à verbe sous-entendu? Dans cette optique, un verbe, exprimé ou sousentendu, est toujours présent dans la phrase. Telle était la position des gram­ mairiens latins: Charisius14 parle d'ellipse, Priscien15 de sous-entendu. Cer­ tains grammairiens modernes, tout en ayant du mal à admettre l'explication par l'ellipse, prennent une position intermédiaire et essaient d'expliquer le phénomène non-verbal en latin par le lointain souvenir d'un passé indo-euro­ péen où la phrase nominale régnait sans partage. Telle est la thèse de J.Per­ ret16 que l'on peut résumer ainsi: si l'on compare l'usage de la phrase sans verbe en latin et en français, on constate qu'en français le phénomène est sty­ listique, alors qu'il appartient en latin à la structure de la langue. D'où l'expli­ cation historique: "On admettra volontiers que cette différence tient à ce que le latin est encore relativement proche de l'état indo-européen. Quoique son système de la phrase se soit entièrement regroupé autour du verbe, quoiqu'il interprète comme phrases verbales incomplètes les phrases qui prennent la suite des phrases nominales d'autrefois, il subsiste dans le latin un assez grand nombre de ces structures pour qu'elles n'y produisent pas cet effet d'anomalie violente qu'elles ont en français".17 Or, une autre explication est possible. On peut considérer qu'une phrase dite "sans verbe" doit être définie comme une phrase à prédicat nominal. Il y a

120

CH. GUIRAUD

donc deux types de phrases, la phrase verbale et la phrase nominale, pourvues l'une et l'autre de la même fonction d'assertion. E.Benveniste 18 est sans doute le premier à avoir établi la chose avec autant de netteté. On reprendra ici quel­ ques formules de son article: "Un énoncé est ou nominal ou verbal". 19 "La phrase nominale ne saurait être considérée comme privée de verbe. Elle est aussi complète que n'importe quel énoncé verbal". 20 Mais, sur un point, la doctrine de Benveniste flotte un peu: il s'agit de la fonction assertive. Tantôt il la considère comme dévolue au verbe: "Nous définirons le verbe comme l'élément indispensable à la constitution d'un énoncé assertif fini".21 Tantôt il l'attribue aussi à la phrase nominale: "Nous dirons que la phrase nominale en indo-européen constitue un énoncé assertif fini, pareil dans sa structure à n'importe quel autre de même définition syntaxique". 22 C'est la deuxième position qu'il convient de retenir. En effet des phrases comme omnia praeclara rara ou chauds les marrons, douze métiers treize misères sont aussi com­ plètes qu'elles peuvent l'être. Il n'y manque rien. Il faut donc se débarrasser des positions flottantes comme celles de Benveniste ou d'autres linguistes.23 Autrement dit, la fonction-phrase, bien que très souvent exprimée statisti­ quement par un verbe, n'est pas liée à la présence d'une forme verbale à un mode personnel. On ajoutera pour le latin un dernier argument: l'existence et le dévelop­ pement dans cette langue de l'infinitif dit "de narration". Les Anciens propo­ saient déjà de l'expliquer par une ellipse: celle du verbe coepi. Tel est le cas pour Quintilien24 et Priscien.25 Chez les modernes nous retiendrons de l'arti­ cle de P.Kretschmer 26 deux tentatives d'explication qui vont dans le même sens, puisqu'elles cherchent pour l'infinitif "historique" ou "de narration" une origine grammaticale rattachant cette tournure au contexte: Plaute, Rud., 393-394: O facinus inpudicum! Quam liberam esse oporteat, seruire postulare. "O le crime éhonté: une fille qui devrait être libre, vouloir la tenir en escla­ vage". La deuxième phrase aurait pu se détacher du contexte et glisser vers la valeur: "On veut tenir en esclavage une fille qui devrait être libre". Salluste, Jug.,101,11: Tum spectaculum horribile in campis patentibus: sequi, fugere, occidi, capi. Il faudrait sous-entendre le verbe "être" aussi bien dans la première

L'IMPORTANCE DU VERBE

121

phrase (Tum...patentions) que dans la séquence d'infinitifs. Nous avons dans tous ces cas des tentatives visant à refuser l'autorisation à un infinitif de consti­ tuer un prédicat. Or, c'est bien de cela qu'il s'agit: dans des phrases du type de Homo clamare, qui sont complètes, il y a un sujet (homo) et un prédicat (cla­ mare), bien que ce dernier ne relève pas de la catégorie des formes verbales à un mode personnel. On peut s'interroger pour finir sur les raisons qui ont poussé les linguistes à accorder au verbe une telle importance dans la phrase. Il est certain qu'en latin la plupart des phrases sont verbales. Par ailleurs le verbe porte synthétiquement les catégories de temps, de mode, de personne. 27 Mais, de même que, stylistiquement, n'importe quel mot peut être mis en valeur dans une phrase, de même on doit affirmer que la présence d'un verbe à un mode per­ sonnel n'est pas indispensable pour qu'il y ait phrase: le prédicat peut être nominal (au sens large) et il peut, particulièrement en latin, être constitué par un infinitif.

NOTES 1) On connaît les grandes lignes du Cratyle. Deux thèses sont en présence: celle de Cratyle (les noms sont justes par nature), celle d'Hermogène (la nature n'est pour rien dans cette justesse, qui est affaire de convention). L'analyse verbe/nom apparaît en 424 c- 425 a. Socrate distingue d'abord ); puis les syllabes qui servent à composer les les voyelles; les muettes ( ). noms et les verbes ( 2) De l'interpret., chap.3. On ne retrouve dans la Poétique, chap.20 que deux parties sur trois de cette définition: "Le verbe est un composé de sons significatif, avec idée de temps, et dont aucune partie n'est significative par elle-même". Il manque ici la valeur assertive. 3) Cf.J. Perrot, 1961, Les dérivés latins en -men et -mentum, Paris, Klincksieck,p.74. 4) Priscien, Institutions grammaticales, livre 17, cité d'après H.Keil, Grammatici latini, en abrégé G. L. K., III,154: "Obliqui casus pronominum omni modo ad uerba feruntur". 5) G.L.K., III, 116-117: "Inest igitur intellectu nominatiuus in ipsis uerbis, quo sine substantia significari non poterat". 6) M.Baratin, 1977, Priscien et la constitution d'une syntaxe latine, Paris, p.272. 7) En ce qui concerne la Renaissance, on pourra retenir la formule de L.Kukenheim, 1951, Con­ tribution à l'histoire de la grammaire grecque, latine et hébraïque à l'époque de la Renaissance, Lei­ den, Brill, p.76: "Dans la grande majorité des ouvrages, la syntaxe occupe peu de place dans l'ensemble du livre". 8) 1830, Grammaire générale et raisonnée contenant les fondements de l'art de parler..., avec les remarques de Duclos, nouvelle édition suivie de la logique ou l'art de penser, Paris, A.Delalain, p.290. 9) 1965, Eléments de syntaxe structurale, 2è.édit., Paris,(nouveau tirage: 1969).

122

CH. GUIRAUD

10) 1976, Grundfragen einer Dependenz - Grammatik des Lateinischen, Vandenhoeck & Ruprecht, Göttingen. On se reportera aussi avec profit à l'article de R.Hiersche, 1979, Les "précurseurs" allemands de L.Tesnière, L'information grammaticale, Paris, n°.3, pp.24-28. 11) 1977, L'idée de phrase grammaticale et son expression en latin, Lille, p.471. 12) F.G.Moore, 1903, Studies in Tacitean ellipsis. Descriptive passages, TA Ph A, 34, pp.5-26. 13) Voir Th.Winter, 1907, De ellipsi verbi esse apud Catullum, Vergilium, Ovidium, Statium, Iuvenalem obvia, Dissert. Marburg. 14) G.L.K. I, 268. 15) G.L.K. III,152-3. 16) 1956, Le verbe latin, Paris, Centre de Documentation universitaire, pp.25-26. 17) o.c, p.26. 18) 1950, La phrase nominale, Bulletin de la Société de linguistique de Paris, tome 46, pp. 19-36, repris dans E.Benveniste, 1966, Problèmes de linguistique générale, Paris, Gallimard, tome I, pp.151-167. 19) Problèmes ...I, p.157 20) Problèmes ...I, p.159 21) Problèmes ...I, p.154 22) Problèmes ... I,p.l58 23) Ainsi L.Hjelmslev, 1948, Le verbe et la phrase nominale, Mélanges Marouzeau, Paris, Bel­ les-Lettres: "Toute phrase nominale comporte dans son contenu certains éléments que la gram­ maire traditionnelle devrait logiquement reconnaître comme des éléments verbaux, sans qu'elle comporte nécessairement de verbe".(p.258). 24) Inst. orat., 9, 3, 58: "Quarum (figurarum) una est quam libro proximo in figuras ex distuli, cum substractum uerbum aliquod satis ex ceteris intellegitur, ut Caelius in Antonium "stupere gaudio Graecus" : simul enim auditur "coepit" ". 25) G. L. K., III, 228: "per ellipsin uerbi "coepi" soient auctores proferre infinita, ut Terentius in Andria: ego illud sedulo Negare factum deest enim "coepi"." 26) 1910, Zur Erklärung des sogenannten Infinitivus historicus, Giotta, II, pp.270-287. 27) Ces catégories peuvent être exprimées aussi par des mots indépendants. Egalement en fran­ çais: Vérité aujourd'hui, erreur demain.

INTERROGATION, INTERROGATION RHÉTORIQUE ET EXCLAMATION EN LATIN

FRANÇOIS HOFF Strasbourg RESUME Des frontières nettes séparent l'interrogation, l'interrogation rhétorique et l'exclamation en qu-. Dans l'interrogation proprement dite, le mot en qu­ est "vide" ; dans l'interrogation rhétorique, quand elle a une réponse possible, il représente une donnée objective, existante ou non, connue du locuteur et de l'interlocuteur; dans l'exclamation, il vaut pour le "haut degré" du consti­ tuant marqué par qu-. En outre, les noms et adjectifs "de qualité" ne peuvent être utilisés que dans l'exclamation. 0. Introduction Les énoncés indépendants contenant des morphèmes qu- sont a priori susceptibles de recevoir plusieurs interprétations: qui consul peut être une interrogation (attendant une réponse: Cicero; nullus consul) ou une exclama­ tion (qui équivaut à optimus consul par exemple). Notre projet est de montrer que des frontières nettes séparent a) les interrogations proprement dites et les interrogations rhétoriques. On considère traditionnellement, à tort, que ces dernières sont une sorte de mélange d'interrogation et d'exclamation, b) les interrogations et les exclamations. 1. L'interrogation en quL'interrogation est un énoncé construit pour mettre l'interlocuteur dans la situation d'être obligé de répondre. 1 Cette condition remplie, il y a quatre possibilités, selon que le locuteur connaît d'avance ou non la réponse, et pré­ suppose ou non que l'interlocuteur la connaît.

124

FRANÇOIS HOFF

1.1. Le locuteur ignore la réponse et présuppose que l'interlocuteur la sait. C'est la percontatio des rhéteurs, la question proprement dite, par la­ quelle le locuteur complète ou vérifie un savoir. 1.2. Le locuteur ignore la réponse et présuppose que l'interlocuteur l'ignore aussi: la question est posée pour embarrasser l'interlocuteur, ou dans un but argumentatif. Dans cette catégorie entrent les questions sans réponse possible: par exemple la figure de rhétorique de la dubitatio {quid dicam?), et les questions à présuppositions fausses: "Si p est vrai, pourquoi non q?" (p étant évidemment faux): cf. Cic.Mil.51. 1.3. Le locuteur sait la réponse et présuppose que l'interlocuteur l'ignore: autre cas de question pour embarrasser. Les questions d'examen font partie de cette classe. (Les cas 1.2 et 1.3 sont les seuls admis par Searle (trad.fr. 1972 108-109) comme définissant la question.) 1.4. Le locuteur sait la réponse et présuppose que l'interlocuteur la sait aussi: c'est l'interrogation rhétorique, appelée (comme les deux précédentes) interrogatio par les rhéteurs. 1.5. Dans ce dernier cas, l'énoncé marqué comme interrogatif ne garde de l'interrogation que sa valeur de mise en demeure conventionnelle de répondre. Elle prétend contraindre l'interlocuteur à exprimer par une énonciation, et à prendre à son compte un contenu qui "va sans dire" dans la situa­ tion, mais qui doit être dit, pour la poursuite de l'argumentation du locuteur. 2 L'interrogation rhétorique est donc un acte illocutoire dérivé3: une déclaration réalisée par une interrogation. De cette déclaration peut être dérivé un troisième acte illocutoire: par exemple un ordre, avec quin. Dans certains cas, le latin oppose deux adverbes: l'un, neutre, servant à la percontatio et à l'interrogatio, l'autre réservé à l'interrogatio: cur : quid cur non : quidni; quin quomodo : qui 1.6. La déclaration ainsi produite peut avoir deux formes, selon la situa­ tion: négative ou positive. En général, les interrogations rhétoriques avec quvalent pour une déclaration négative: Cic.Phil.2,87: Ubi enim tu in pace consistes? L'interrogation vaut pour nusquam, réponse imposée à Antoine. Mais l'autre cas existe aussi: Cic.Mil.7: In qua tandem urbe hoc homines stultissimi disputant? (= Romae).

INTERROGATION ET EXCLAMATION

125

Cic.Lig.7: Apud quem ... hoc dico? (= Caesarem) Cic.Mil.54: Videte nunc ilium egredientem e uilla subito — Cur? — ... ( = quia Milonem interficere uult) 2. L'exclamation en quElle se distingue de l'interrogation par le fait qu'elle n'appelle pas de réponse. Il n'est pas possible de la décrire comme un cas particulier de l'interrogation rhétorique à réponse positive: comme les exemples ci-dessus le mon­ trent, le contenu de cette dernière est une réalité objective (une personne, un lieu, un mobile ...), alors que l'exclamation se définit par sa valeur sémanti­ que 'intensive": elle est "une structure qui exprime le haut degré en l'absence des marqueurs spécifiques de cette catégorie sémantique" (Milner 1977:111). Le "haut degré" se répartit en deux sous-catégories: 2.1. Les quart titatifs4 quot; quantus,a,um: adjectifs, portent sur le nombre et la mesure. quotiens: circonstant. quantum: "adverbe de quantité"; quanti: "adverbe de degré" 5 (quanto, "ablatif de mesure", portant sur un adjectif ou un adverbe, n'est pas réellement un quantitatif). 2.2. Les non-quantitatifs Ils portent sur une qualité manifestée par un nom, ou sur le haut degré d'un adjectif ou d'un adverbe. - avec un adjectif ou un adverbe: quam, quanto. Quam, qui n'apparaît qu'avec le degré positif, est en distribution com­ plémentaire avec quanto, réservé au comparatif. - avec un nom: qui,ae,od; qualis.6 Seul des deux, qualis peut être attribut; il est aussi beaucoup plus rare (5 occurrences contre environ 130 de qui... ) et n'est utilisé qu'avec des noms ani­ més dans des phrases verbales, alors que qui... est principalement (83%) accolé à des noms non-animés, dans des phrases nominales. - avec un verbe (exclamatifs à valeur adverbiale): quam; ut. Ils apparaissent avec certains verbes susceptibles d'une spécification de degré 7 : Cic. Phil.2,20: Quam id te, di boni, non decebat! 3. Interrogation et exclamation 3.1 Certain mots en qu- ne peuvent être exclamatifs: - ceux qui portent sur une identité objective: uter; quis,ae,id.6 - ceux qui portent sur une circonstance objective: ubi, quo, unde, qua:

126

FRANÇOIS HOFF

quando; quomodo. - ceux qui portent sur la cause ou le but: cur, quare, et les interrogatifs "rhétoriques" qui, quin, quid, quidni. L'exclamation est une appréciation subjective d'une qualité ou d'une quantité portées à leur plus haut degré: ce n'est jamais un contenu objectif en soi (l'identité d'une personne, un lieu, un moment) qui fait l'objet d'une exclamation/ De même, les adverbes interrogatifs qui ne commutent pas avec un adverbe de degré ou de quantité, c'est-à-dire ceux qui n'ont pas de corrélatif en t- (plus qui... et ut) sont limités à l'emploi interrogatif. 3.2. Pour les quantitatifs, il peut y avoir doute entre une lecture interro­ gative et une lecture exclamative. L'interprétation est liée à un choix du lec­ teur, en fonction du contexte. Le fait important est que cette interprétation est soit interrogative, soit exclamative, et qu'elle ne saurait être un mélange des deux. Cic. Cat. 1,16: Quotiens tibi iam extorta est ista sica de manibus! Cic.Phil.2,45: Quotiens te pater eius domu sua eiecit? (ponctuation de l'édition des Universités de France) Phil.2,45 est ici une percontatio dont le sens est: "Dis-moi donc combien de fois..."; Cat. 1,16 n'a pas cette valeur interrogative: le sens est: "J'ai échappé très souvent à tes tentatives de meurtre." 3.3. Pour les non-quantitatifs, nous reprenons le principe posé par Milner (1977:113) "Sont interrogeables les noms et adjectifs qui définissent une classe d'êtres ou une propriété objectivement définie, en-dehors de l'énoncé particulier où ils apparaissent; ne sont pas interrogeables (et sont donc susceptibles, par eux-mêmes, d'être exclamatifs) les noms et adjectifs qui ne définissent leur référence que par un acte d'énonciation individuel." 8 Avec un mot en qu-, les premiers noms et adjectifs (dits "classifiants") peuvent être exclamatifs, à condition d'être marqués (dans une théorie gene­ rative) par un spécificateur spécial, sinon ils sont interrogatifs. Les seconds (dits "de qualité") sont réservés à l'exclamation. Sans discuter le problème de la validité de ce principe à propos du latin, nous faisons les constations suivantes: - qui... et qualis avec un classifiant sont interrogatifs ou exclamatifs {qui liber), qui... portant dans ce cas sur l'identité en interrogation, sur la qualité en exclamation. - les mêmes avec un nom de qualité sont exclamatifs: Cic.Sest.26: qua... superbia...!

INTERROGATION ET EXCLAMATION

127

- mais l'énoncé est toujours interrogatif lorsque qualis se trouve dans une subordonnée infinitive (avec putamus, existimatis), quel que soit le statut du nom qu'il détermine. - pour les adjectifs, la classification de Milner doit être nuancée; nous pro­ posons, pour quam9: (l)adjectifs non susceptibles de degré: auxiliaris, romanus (2)adjectifs de qualité: honestus, turpis, praeclarus (3)adjectifs mixtes: magnus, breuis, caecus Les adjectifs de la classe (1) ne peuvent se trouver ni dans une interroga­ tion ni dans une exclamation. Les adjectifs de la classe (2) sont réservés à l'exclamation, sauf dans une subordonnée infinitive: Cic.Phil.5,22: Quam auidum in pecuniis locupletium (hune futurum fuisse censetis) ? Les adjectifs de la classe (3) peuvent se trouver dans une interrogation ou une exclamation. Pour quam + adverbe et quanto, la répartition est la même, la classe (1) étant évidemment impossible. 4. Conclusions On peut objecter que les classes des noms et des adjectifs proposées ont des frontières imprécises, et que très souvent, seule l'interprétation du contexte permet de choisir entre interrogation et exclamation. Il n'en reste pas moins vrai que les énoncés étudiés ne se chevauchent pas; des frontières nettes, définies par la pragmatique, les séparent. Dans la percontatio, le mot en qu- est "vide" ; dans l'interrogation rhétori­ que, il représente une donnée objective (nulle ou positive); dans l'exclama­ tion, il vaut pour le "haut degré" du constituant qu'il représente ou déter­ mine. Outre le problème de la description fonctionnelle de certains mots en qu-, il resterait à tenter d'intégrer l'ensemble des énoncés interrogatifs et exclamatifs dans un seul système. Enfin, une telle approche ouvre la porte à une étude plus fine de l'interro­ gation rhétorique, et au-delà, à une relecture des théories anciennes sur les figures.

128

FRANÇOIS HOFF

NOTES 1) Ducrot 1972:92; Hoff 1979:46-53, pour le latin. 2) Hoff 1979:141-145; Calboli 1981:7 la définit par la notion de défi. 3) Schmidt-Radefeldt 1977:390; sur l'acte illocutoire dérivé, cf. en dernier lieu Communications 32, 1980, les articles de Anscombre et Zuber, avec la bibliographie antérieure. 4) Le corpus est constitué par les discours et les oeuvres rhétoriques de Cicéron. 5) Définitions de la syntaxe d'Ernout-Thomas 1964; la description fonctionnelle de ces mots semble n'avoir pas encore été faite. 6) Le pronom quis,ae,id peut être exclamatif dans des tours comme quae dixit! (Cic.Phil.2,80; 11,33). Il est alors synonyme de qualis. 7) Milner 1974:92sq. 8) Ces derniers sont des noms et des adjectifs qui donnent à un énoncé "tu es (un) X" une valeur d'insulte ou de louange (Milner 1974:114). 9) Cette classification est celle que H. Vairel (1975: 78 sqq.) utilise pour décrire les adjectifs qui entrent dans les énoncés du type lepidum homineml Ces adjectifs sont également ceux qui sont réservés à l'exclamation avec quam.

BIBLIOGRAPHIE Calboli, G. (1981), 'Le frasi interrogative-esclamative latin e l'infinito' , in Logos Semantikos (Festschrift Coseriu), Berlin, De Gruyter, 133-53. Culioli, A. (1974), A propos des énoncés exclamatifs' , Langue française 22, 6-15. Ducrot, O. (1972), Dire et ne pas dire, Paris, Hermann. Elliott, D. (1974), Toward a grammar of exclamations' , Foundations of Lan­ guage , 11, 231-246. Ernout, A., Thomas, F., (1964), Syntaxe latine, Paris, Klincksieck. Hoff, F. (1979), Le système linguistique de l'interrogation en latin classique, Thèse de 3 è cycle, Strasbourg. Milner, J.C. (1974), 'Les exclamatives et le Complementizer' , Actes du Col­ loque franco-allemand de grammaire transformationnelle, I, Tübingen, Max Niemeyer, 78-138. .(1977), 'De l'interprétation exclamative comme valeur sémantique ré­ siduelle' , in M.Ronat, Langue, théorie générative étendue, Paris, Her­ mann, 109-122. Schmidt-Radefeldt, J. (1977), 'On so-called "rhetorical" questions' Journal of Pragmatics, l, 375-392.

INTERROGATION ET EXCLAMATION

129

Searle, J.R. (1972) Les actes de langage, Paris, Hermann, trad.fr. de Speech Acts, Cambridge University Press, 1969. Vairel-Carron, H (1975), Exclamation, ordre et défense. Etude de deux sys­ tèmes syntaxiques du Latin. Paris, Belles Lettres.

WORD ORDER WITHIN LATIN N O U N PHRASES

JAN R. D E J O N G University of Amsterdam SUMMARY In this paper it is argued that in the Latin noun phrase, in the unmarked case, modifiers designating persons follow their head nouns. In this case both head and modifier make their own contribution for establishing the reference of the phrase. Preposition may occur when the modifier has a special pragma­ tic function (contrast or topic) and when head and modifier are not indepen­ dently referring items. 0. Introduction In this paper I want to discuss the factors that determine the position of nominal and pronominal modifiers relative to their head nouns. I have re­ stricted my subject matter to modifiers designating persons. This was done in order to bring about a certain degree of semantic uniformity in the material. The material I have investigated is comprised of noun phrases containing head nouns accompanied by different categories of modifiers: possessive pro­ nouns, and genitives of nouns, proper nouns, and demonstrative and anaphorical pronouns. l Research was based on a corpus consisting of the fol­ lowing texts: Caes. Gall. 2-4; Sall. Cat., Cic. Cat. 1-4; Q. Curt. 3; Livy, parts amounting to ± 60 capita. Roughly speaking, modifiers designating persons precede their head nouns as often as they follow them. There are some variations in frequency between categories and authors, but these are not really significant. It has long been known that the distribution of pre- and postposition is not ran­ dom. Postposition has been recognized as the neutral order. As far as the reversed, marked, word order is concerned, Marouzeau (1922), Adams (1976) and others have shown that contrast is the chief factor causing a

JAN R. DE JONG

132

personal modifier to be preposed. I agree that among the pragmatic condi­ tions that determine word order in these noun phrases, contrast is the most important one, but, I think, there are others as well. I shall come back to these below. First in ( 1 ) below I give a rather tentative approximation of what the func­ tion of postposition is. ( 1 ) By postposition the speaker instructs the hearer to evaluate the reference of the phase as a whole on the basis of the references of the head noun and the modifier (and do nothing more). However, since not all heads and modifiers are of a referential nature we should modify this to what is shown under number (2). (2) The speaker instructs the hearer to evaluate the reference and/or sense of the phrase as a whole on the basis of the references and/or senses of the head and the modifier. I shall proceed now to the pragmatic conditions that cause a modifier to be preposed. 1.

Contrast

The first condition to be distinguished of course is contrast. Contrast oc­ curs in a number of different varieties. I will give some examples of them below. In general, one could say that contrast is present when at least three conditions, stated below, are fulfilled. (3)

(a) two terms, for example noun phrases, are brought into connection with each other (b) both terms share some property (c) both terms differ in some property

I will show how these conditions work on two examples (4) and (5). (4) (5)

non hostem auctorem sed rem spectare ('he had regard, not to the suggestion of the enemy, but to fact', Caes. Gall. 5,29,3) ne propter bonitatem agrorum Germani qui trans Rhenum incolunt e suis finibus in Helvetiorum fines transirent ('lest the excellence of the farmlands might tempt the Germans who dwell across the Rhine to cross from their own into the Helvetian borders', Caes. Gall. 1,28,4)

In (4), the two terms hostem auctorem and rem are brought into connection by the adversative conjunction sed. The property they share is that they have the same function (obj ect) in the state of affairs, while both terms differ in the lexi­ cal items they contain.

WORD ORDER

133

In (5), the two terms Helvetiorum fines and suis finibus are brought into con­ nection just by playing a part in the same state of affairs. The property they share is that they have the same head nouns. The property in which they differ is the specification given by the modifier. The parts of the surface manifestations of these terms which represent the points of difference we may call the carriers of contrast. In (5) they are suis and Helvetiorum. In English the carriers of contrast are marked by contrastive stress, while in Latin, in the case of personal modifiers, they are marked by the preposing of the modifier to its head noun. Example (6) shows that the common property of the two terms may be seman­ tic rather than lexical similarity as it was in (5): (6)

cum his navibus nostrae classi eiusmodi congressus erat, ut ... ('when our own fleet encountered these ships it proved ...',Caes. Gall. 3,13,7)

The following example illustrates that both terms in an opposition are not al­ ways expressed explicitly. In this case the other term is given in a presupposi­ tion which accompanies the statement: (7)

(non) auxilia ex sua civitate in Treveros missa ('that no auxiliaries had been despatched from their state to the territory of the Treveri', Caes. Gall 6,9,6)

Numerous kinds of contrast can be distinguished. I shall give examples of a few of them below. (8)

sed nescio quo pacto omnium scelerum ac veteris furoris et audaciae maturitas in nostri consulatus tempus erupit ('but in some strange way all these crimes and this long-standing madness and audacity have come to a head in the time of my consulship', Cic. Cat. 1,31)

(9)

Alexander non ducis magis quam militis munia exsequebatur ('Alexan­ der performed the duties not more of a commander than of a soldier', Q. Curt. 3,11,7) quae (sica) quibus abs te initiata sacris ac devota sit nescio, quod earn necesse putas in consulis corpore defigere (T do not know what sacrifi­ ces you made to hallow and consecrate it because you thought that you must plunge it into the body of a consul', Cic. Cat. 1,16) (occupaverat) animos tanta omnium oblivio ut multo maior pars (Romanorum) Veios in hostium urbem ... quam recto itinere Romam ... fugerent ('such complete forgetfulness of everything that a much greater number fled to Veii, a hostile city,... , than by the straight road to Rome', Liv. 5,38,5)

(10)

(11)

134

JAN R. DE JONG

(8) is a case in which there is no contrast with some specific other person, but rather with anybody: the identity of the "possessor" has a special significance in the given context: Cicero wonders why all the evils mentioned have come to a head in the time of his (of all people) consulate. 2 In (8), as well as in (4) - (7), there was contrast between the references of the personal modifiers. Contrast may also occur between the senses of different personal modiers, expecially of course when, such as in (9), the modifiers do not have any reference: "Alexander performed the duties not more of a com­ mander than of a soldier". Similarly in (10) and (11) the statements are not about the identity of persons, but rather about their properties: "the body of, not just somebody, but a con­ sul" and "Veii, a hostile city". Possessive pronouns can be used in a way similar to the cases just men­ tioned. (12) and (13) are examples. (12)

at Dareus ... se vero tantum facinus negat facturum, ut... suos milites iubeat trucidari ('But Dareus declared that he certainly would not com­ mit such a crime as to order his own soldiers ... to be butchered', Q. Curt., 3,8,4)

(13)

deinde ... concessumque (est) in condiciones ut plebi sui magistratus essent ('then ..., and a compromise was effected on these terms: the plebeians were to have magistrates of their own', Liv. 2,33,1)

Here contrast is not between two 'possessors' but rather between 'ownership' and 'not-ownership'. It is the kind of contrast that is expressed in English by the adjective "own": Darius declares that he would not butcher "his own sol­ diers". There is no contrast here between Darius (referred to by suos) and some other possesssor. In other words Darius does not imply that he would butcher someone elses soldiers. Rather, the contrast here is between the treatment by Darius of his own property and things outside his property. A similar case is (13): "the plebejans were to have magistrates of their own". What is implied is not that in the past the plebejans had other people's magis­ trates, but rather that previously there were no magistrates which they could call their property whereas later there were. Examples (14) and (15) are in­ stances of what I call 'intrinsic contrast'. (14) (15)

quot ego tuas petitiones... effugi? ('how many of your thrusts... did I escape', Cic. Cat. 1,15) si suas copias Haedui in fines Bellovacorum introduxerint et eorum agros populari coeperint ('this could be done if the Aedui led their own

WORD ORDER

135

forces into the borders of the Bellovaci and began to waste their lands', Caes. Gall. 2,5,3) ut (Belgae) agros Remorum popularentur qui magno nobis usui ad bel­ lum gerundum erant ('to lay waste the lands of the Remi which were of great service to us for the conduct of the campaign', Caes. Gall. 2,9,5)

(16)

The nature of the states of affairs is such that there may be said to be a direct interaction between the agent and the person referred to by the modifier. In (14) it is thus implied that by escaping Catiline's attack Cicero also escaped Catiline himself. And in (15) not only the lands of the Bellovaci were affected by the violence of the Haedui but the Bellovaci themselves as well. In (16), on the other hand, where the order is reserved, it is explicitly stated that the ac­ tion of the Belgae is directed against the Romans rather than against the Remi. In (17), we again have a case where the person designated by the personal modifier is affected directly by the action. (17)

ea commemorando militum animos accendebat ('by mentioning these he fired the spirits of the men', Sall. Cat. 59,6)

One could characterize all these cases (14 -16 and 17) as follows: semantically the modifier is more or less detached from its head and brought on one level with the major constituents. It is interesting to see then that Latin has a con­ struction by which, though it is more restricted in its application, the 'posses­ sor' is actually detached from the 'possessed' on the syntactic level as well: the sympathetic dative construction. (18) is an example which closely parallels (17): (18)

ea vero res Romanis auxit animos ('this circumstance increased the ar­ dour of the Romans', Liv. 37,33,5) 2.

Topic

Another factor which causes a personal modifier to be placed before its head noun is the Topic status of the modifier. I follow Dik (1978) in defining Topic as the function of a constituent that pre­ sents the entity 'about' which the predication predicates something in the given setting. Numbers (19) and (20) are examples of personal modifiers having Topic status: (19)

feminarum praecipue et gaudia insignia erant et luctus ('the women especially exhibited extremes of joy and grief', Liv. 22,7,12)

136 (20)

J A N R . D E JONG

Gallbrum alacer ac promptus est animus ('the temper of the Gauls is eager and ready', Caes. Gall. 3,19,6)

In most of cases like these the modifier not only precedes the head noun, but also tends to be placed in a more forward position in the sentence. The same tendency to detach modifiers with Topic function from their head nouns can be observed in languages with much heavier restrictions on word order than Latin. For example, one can translate (19) into Dutch: "vooral van de vrouwen waren vreugde zowel als verdriet opvallend" (litt. : "especially of the women both joy and grief were conspicuous"). I will now deal with some details connected with this type of preposed modifier, which are illustrated by the following examples: (21)

(22)

(23)

(quibus rebus permota civitas atque immutata urbis facies erat...) at Catilinae crudelis animus eadem illa movebat ('(these precautions struck the community with terror, and the aspect of the city was changed...) But Catiline's pitiless spirit persisted in the same attempts', Sail. Cat. 31, 4) inter quos Alexandri femur leviter mucrone perstrictum est ('among those wounded Alexander himself was slightly grazed in the right thigh by a sword', Q. Curt. 3,11,10) militibus longe dispar animus est ('but the spirit of the soldiers was far from being matched', Liv. 23,29,7)

Example (21) shows that a personal modifier may have Topic function and be used contrastively at the same time, when it marks the transition from one topic (civitas) to another (Catilind).3 Sentence (22) is introduced by the phrase inter quos. Such phrases serve to re­ strict the range of possible Topics in the predication that follows them 4 : a pre­ dication following inter quos has to be 'about' someone included in the class referred to by quos. In (22) therefore, the Topic of the predication has to be Alexandri and cannot be Alexandri femur. Since, in these cases, assignment of the Topic function to the modifier is almost obligatory, word order patterns as the following are hardly to be exspected: (24)

?Inter quos femur Alexandri leviter mucrone perstrictum est

Example (23), when compared to (20), shows that, in these cases as well, a da­ tive sometimes has the same function as a genitive detached from its head noun.

WORD ORDER

3.

137

Referential unity

A third condition which causes a modifier to be preposed is what I call re­ ferential unity. This occurs when a personal modifier refers to the same class of persons as the phrase as a whole does, or (perhaps better) where the headnoun refers to a collective and where the modifier specifies the parts of which it is made up. This condition is fulfilled with the phrases "the people of the Suebi" and the "band of slingers" in (25) and (26) :5 (25)

(26)

Sueborum gens est longe maxima et bellicosissima Germanorum om­ nium ('the Suebi are by far the largest and most warlike nation among the Germans', Caes. Gall. 4,1,3) ante hanc aciem posuerat funditorum manum ('before this battle-line he had placed a band of slingers', Q. Curt. 3,9,9)

It seems likely that this condition of referential unity also determines the fixed order of modifier + head when the noun numerus is used in the idiomatic way shown in: (27)

(28)

qui ex his secuti non sunt in desertorum ac proditorum numero ducuntur ('any of them who have not followed ... are reckoned as deserters and traitors', Caes. Gall. 6,23,8) C. Manlius ex suo numero legatos ad Marcium Regem mittit ('C. Manlius sent a delegation from his army to Marcius Rex', Sall. Cat. 32,3)

(28) is somewhat problematic. As it stands it does not fulfill the condition just mentioned, since suo refers to Manlius: 'C. Manlius sends a delegation from his midst'. Perhaps it should be explained as showing the order ex suorum numero would have had according to the rule and which is after all what it means. Genitives with nouns such as gens and manus may be considered as partitive genitives.6 There is another group of head nouns which are often accompanied by a partitive genitive, for example pars, quisque, plerique and similar expres­ sions. With these the situation is just the other way around. Here the modifier (the genitive) can not be coextensive with the phrase as a whole. And indeed what we find is that (except where it is used as a Topic or has contrastive func­ tion) the modifier is postposed to the noun. (29) is an example of this. (29)

magnam partem eorum interfecerunt ('they slew a great part of them', Caes. Gall. 2,23,1)

138

JAN R . D E JONG

One could add the genitives vestrum and nostrum which are used almost ex­ clusively in this function are always postposed, as vestrum is in (30) (30)

singulis vestrum milia denarium dari iussi (T have ordered a thousand drachmes to be given to each of you', Q. Curt. 5,5,22) 4.

Context reference

The fourth condition for the preposing of a personal modifier that I have distinguished is context reference. When a noun phrase containing a state-ofaffairs noun refers back to a fact known to the hearer or reader by previous mention in the context, the modifier is usually preposed. (31) and (32) are typ­ ical examples. (31)

(32)

(Caes. Gall. 4,21: cum iis (legatis) una Commium... mittit. Huic imperat... adeat civitates horteturque ut populi Romani fidem sequantur) Commius Atrebas venit quern supra demonstravimus a Caesare praemissum. Hunc illi (Britanni) cum ad eos oratoris modo Caesaris mandata deferret, comprehenderant ('(along with them he sent Com­ mius... Him he commanded to visit what states he could, to exhort them to seek the protection of Rome). Commius the Atrebatian came, who, as shown above, had been sent by Caesar into Britain. When he delivered Caesars messages, they had seized him', Caes. Gall. 4,27,4) illi (Germani)... reverti se in suas sedes simulaverunt et... reverterunt... atque... inopinantesque Menapios qui de Germanorum discessu per exploratores certiores facti... ('they pretended to retire to their own houses and districts... and they returned... and caught the Menapii uninformed and unawares, for, having learnt through their scouts of the departure of the Germans...', Caes. Gall. 4,4,6)

In chapter 4,21 Caesar tells us Commius was sent to Britain to deliver de­ mands of Caesar. In 4,27 he explicitly ("quern demonstravimus") refers back to that passage. Thus the phrase Caesaris mandata in (31) may be considered as an instance of context reference. In (32), it is first stated that the Germans return home and later on an action noun refers back to the same event: Germanorum discessu. Compare with these also the following cases: (33)

(Gall. 3,17,5: Sabinus... castris se tenebat) superiorum dierum Sabini cunctatio ('(Sabinius confined himself from camp) the hesitation of Sabinius during the previous days ...', Gall. 3,18,6)

WORD ORDER

(34)

139

qui (Annius) una cum hoc Furio semper erat in hac Allobrogum sollicitatione versatus ('Annius, who had always been associated with this Furius in tampering with the Allobroges', Cic. Cat. 3,14)

Although in general I do not take into account cases in which a personal mod­ ifier comes between the head noun and another modifier, since it is not sure that these cases could be treated just like other cases of pre- or postposition, I nevertheless cite (33) and (34) here since the other modifiers themselves {superiorum dierum and hoc) clearly indicate that the phrase refers to some earlier mention of the event in the context. Therefore postposition of the modifiers in these cases would be least expected. A similar case in (35). (35)

(Ianum ... indicem pacis bellique fecit (Numa)) bis deinde post Numae regnum clausus fuit ... ('(and built the temple of Ianus as an index of peace and war). Twice since Numa's reign has it been closed', Liv. 1,19,3)

In the preceding paragraph the actions of Numa were mentioned, the last which was the building of the temple of Janus. It is hard to interpret post Numae regnum as an independent reference to the reign of Numa because of the adverb deinde in the same clause which clearly refers back to the preceding passage. 5.

Conclusion

The function of preposition of personal modifiers could (again very ten­ tatively) be summed up as follows: Preposition serves as a signal to the hearer that he should either change his strategy in establishing the reference of the phrase or that he should do some­ thing more than just establish the reference of the phrase: he should also take into account that the modifier is used contrastively or has topic function. In the foregoing I have not said much about differences in behaviour be­ tween the various categories of personal modifiers: genitives of nouns, proper nouns and pronouns and possessive pronouns. I believe that in principle they obey the same rules. Some uses are more frequent with some categories than with others: possessive pronouns cannot occur as independent noun phrases and therefore can hardly be detached from their head nouns to serve as the topic of the clause. Genitives of demonstrative pronouns, on the other hand, very often (by virtue of their anophorical function) serve as a Topic. A typical example is (36):

140 (36)

JAN R. DE JONG

(mortales) quibus profecto contra naturam corpus voluptati, anima oneri fuit. eorum ego vitam mortemque iuxta aestumo ('in these men we see, contrary to Nature's intent, the body a source of pleasure, the soul a burden. For my own part I consider the lives and death of these men about alike', Sall. Cat. 2,8)

Proper nouns generally behave like other nouns with the exception of fixed patterns like 'Publius Marci films', where a proper noun is preposed to nouns like filius, frater etc. These are probably best regarded, as Adams (1976) does, as inherited from an older stage of the language. I have tried to explain word order patterns of a class of complex noun phrases in Latin solely on the basis of the pragmatic status of the constituents. Sometimes word order rules have been claimed to exist that are of phonologi­ cal or syntactic/semantic nature: for example the rule that words are arranged in such a way as to avoid hiatus or the rule that the subjective genitive precedes the noun and the objective genitive follows it. Since I have found that these rules at least within the corpus that I studied have a rather limited explanatory power I have proposed a set of rules of solely pragmatic nature. I must, however, emphasize that neither do these rules account for all in­ stances. Obvious counter-examples remain, as well as a number of doubtful cases. Also one should not be surprised to find different word orders in quite similar contexts. This however is to be expected. There are situations in which a speaker has considerable freedom to mark a particular constituent for con­ trast- or topic status. I would like to conclude my paper by pointing to one particularly striking example of this in the following pair of sentences: (37)

(38)

sed ita censeo: publicandas eorum pecunias, ipsos in vinculis habendas per municipia ('this, rather, is my advice: that their goods be confis­ cated and that they themselves kept imprisoned in the free towns', Sall. Cat. 51,43) (words of Caesar) itaque censuit (C. Caesar) pecunias eorum publicandas, ipsos per municipia in custodiis habendos ('therefore he recommanded that the goods of the prisoners be confiscated and that they themselves be im­ prisoned in the free towns', Sall. Cat. 52,14) (words of Cato).

NOTES 1) I left out of consideration genitive modifiers containing a relative pronoun, since the position of these modifiers is subject to purely syntactic restrictions.

WORD ORDER

141

2) As a special instance of this kind of contrast I regard cases in which a speaker, by means of a preposed second person possessive pronoun, makes a direct appeal on the audience, e.g.: (39)

versatur mihi ante oculos aspectus Cethegi et furor in vestra caede bacchantis ('before my eyes there rises the countenance of Cethegus and his madness as he revels in your death', Cic. Cat. 4,11)

3) Panhuis, in his comments, has some doubts concerning my analysis of this passage. The point that I wished to make with this example was that constituents can serve as Topic in a sentence, and at the same time show a contrast relation with some other constituent, for example the Topic of some other sentence. If we simplify the state of affairs described in Sall. Cat. 31,1-4 to the following pair of sentences: (40)

urbs trepidabat. at Catilina in inceptis perseverabat ('the city was panic-striken. But Catiline persisted in his attempts')

we would have such a situation: there is a contrast between urbs and Catilina (the two terms are brought into connection with each other by the 'change-of-topic'-conjunction at, they share the property of being subject/topic of their sentences, and they are different in their respective refer­ ences) . Whether or not one analyses the actual passage Sall. Cat. 31,1-4 in the same way depends on the degree of parallelism one assumes to exist between Catilinae crudelis animus and a presupposed civium animus. It seemed to me that the presence of the adjective crudelis rather disturbs this paral­ lelism . If one claims, as Panhuis seems to do, that the modifiers have a contrastive value by virtue of having the same head (actually present or presupposed), one would somehow imply that the citi­ zens' animus was crudelis as well, which is obviously not meant here. 4) Within the Functional Grammar framework of Dik (1978) inter quos would probably be said to have Theme function. Theme is defined there as "the function which presents a domain or uni­ verse of discourse with respect to which it is relevant to pronounce the predication which follows". Contrary to Topic constituents Themes do not hold a direct relation to the predicate. 5) The factors "referential unity" and "discourse reference" are the two least well-established among the ones mentioned in this paper. Nevertheless, among the cases of personal modifiers which are otherwise difficult to explain, instances of these two types seemed to be rather frequent. I therefore came to the conclusion that these two factors should be considered sufficient conditions for preposition. I considered (25) to be problematic because a contrastive Sueborum gens im­ mediately following a non-contrastive Suebis is rater awkward, Perhaps Panhuis is right in calling funditorum in (26) contrastive, although it still bothers me that an acies should by implication be called a manus. 6) The noun phrases mentioned consist of a head noun referring to some kind of collective, and a modifier which specifies the "substance" of which the collective is made up. These genitives, there­ fore, have more in common with partitive genitives as in pars hostium or poculum vini than with explicative genitives such as in nomen servitutis ox poena mortis.

BIBLIOGRAPHY Adams, J.N. (1976), 'A typological approach to Latin word order', Indogermanische Forschungen 81, 70-99. Dik, S.C. (1978), Functional Grammar, Amsterdam. Marouzeau, J. (1922), L'ordre des mots dans la phrase latine, vol I, Paris.

142 COMMENTS ON J.R. D E JONG's PAPER

DIRK P A N H U I S Leuven The most convincing part of de Jong's paper is the section on the pre­ posed modifier which occurs in contrast to some other term. This section pre­ sents, with some additional refinements, the traditional view that the order Noun-Modifier is the unmarked order, and that the marked Modifier-Noun order occurs for contrast. De Jong also correctly points out that in English the carriers of contrast are marked by means of contrastive stress. I understand that, by implication, he wants to say that such contrast in Latin is expressed by the marked word order and not by stress. It seems important to underline that for Latin there is no evidence to assume such a contrastive stress, as do many German scholars (e.g., Leumann-Hofmann-Szantyr, 1965-1977, I:236, II:397-404), when they talk about "Betonung," "Haupttonstelle," "Tonstärkere Ausdrücke," "Tonschwache," etc. See also Panhuis (1982:152-156). The section on the preposed modifier as topic is the most interesting one for me, although, to a certain extent, I disagree with the Jong's analysis of sen­ tence (21). (21)

At Catilinae crudelis animus eadem illa movebat. (Sall., Cat. 31, 4)

This sentence cannot adequately be discussed out of context. Chapter 31, 1-3, describes the state of mind of the citizens of Rome, and particularly of the women. Then follows the sentence under discussion. In this sentence, some comment is made, not about Catiline, but about Catilinae crudelis animus. The comment is that he is moving on with his same projects {eadem illa movebat). The entire noun phrase Catilinae crudelis animus is the topic (or the theme, in the theory of Functional Sentence Perspective) of the sentence, just as the state of mind of the citizens and the women is the topic (theme) of the preceding sentences (even if the word animus does not occur explicitly in the text). Within this topic noun phrase, one can also observe a pragmatic (communicative) dimension in the order of the constituents, viz. the contras­ tive dimension discussed earlier. Within this noun phrase animus is the topic (theme), while the genitive Catilinae is a comment about animus. The pre­ posed Catilinae, then, is not the topic (theme), but the comment (rheme) in the noun phrase, because the genitive Catilinae stands in contrast to the citi­ zens and the women. The preposed genitive in sentence (21) has thus to be

143 classified under Section 1. The reason for this different analysis seems to be that de Jong does not look at the two different levels in the grammatical hierarchy (sentence and noun phrase), but only looks at the modifier in the noun phrase, and takes part of the topic noun phrase as the topic, whereas I would consider that part as the comment (rheme) within the topic noun phrase. For the organization of com­ municative units on various levels, see Svoboda (1968). I agree with de Jong's analysis of sentence (36). The modifier eorum oc­ curs in the first part (the thematic part) of the sentence, while the head nouns vitam mortemque occur in the second or rhematic part. The different dis­ course function, or more precisely, in the terms of the theory of Functional Sentence Perspective, the difference in degree of Communicative Dynamism, is the reason for the disjunction (hyperbaton). See also Panhuis (1982:72-80). Still, there seems to be a problem which I cannot solve, but which I want to raise nevertheless. In (36), the preposed eorum is the topic (theme); it re­ fers to mortales in the preceding sentence. However, in general, the marked preposing seems to be associated with comment (rhematicity), rather than with topic (theme). Maybe the solution has to be sought in the fact that the preposed modifier can be topic (theme), if it is also disj oined from its rhematic head noun and occurs in the beginning and thematic segment of the sentence. In such a case, preposing is not a noun phrase phenomenon, but a sentence phenomenon. I am sceptical about the explanations given in Sections 3 "referential unity" and 4 "context reference". With respect to the "referential unity", de Jong seems to consider Sueborum in (25) as a partitive genitive. First, I think it is a genitivus explicativus, and second, I do not see when a partitive genitive is preposed and when it is not. With respect to "context reference", it is not clear to me when some previous mention in the context leads to a noun phrase with a preposed modifier and when not. I would rather explain the preposed modifiers in sentences (25), (26), and possibly (31) in Sections 3 and 4, as contrasted modifiers in the sense discussed earlier. In (25), Sueborum gens is the topic (theme) of the sentence, but Sueborum is contrasted to the Usipetes and Tenctheri, who are the topic (theme) of the preceding sentences. (The Suebi are mentioned in the preced­ ing sentence, but not as topic.) In (26), funditorum is preposed because it is contrasted to the phalanx and the equites, mentioned earlier in the text. In (31), there is no clear proof that the preposed Caesaris is contrasted with something else, but neither is this interpretation impossible. Caesaris could be

144 contrasted with Commius, or it could simply be highlighted without contrast.

BIBLIOGRAPHY Leumann, M., Hofmann, J., and Szantyr, A. (1965-1977), Lateinische Gram­ matik. 2. vol. München: Beck. Panhuis, D. (1982), The Communicative Perspective in the Sentence: A Study of Latin Word Order. Amsterdam: Benjamins. Svoboda, A. (1968), "The Hierarchy of Communicative Units and Fields as il­ lustrated by English Attributive Constructions." Brno Studies in English 7:49-101.

LATIN PREVERBS A N D CASES

CHRISTIAN LEHMANN Cologne

SUMMARY The syntax of verba composita, especially the argument structure of their preverbs, is investigated under the general hypothesis that it will be regular to the extent that the constructional behavior of the verbum compositum can be explained on the basis of the constructional behavior of its parts, namely the verbum simplex and the preverb. A preverb is — like a preposition — a local relator, which takes two arguments, the locatum and the relatum. In preverbation, the locatum is identified with the subject of the intransitive and the ob­ ject of the transitive simplex. The relatum very often remains implicit. If it is expressed, it may be couched in an adverbial phrase, in particular a preposi­ tional phrase, in particular one introduced by the preposition identical to the preverb. Or it may be taken directly in a mere case, the dative, ablative or ac­ cusative. The latter two cases are subject to certain restrictions: it is not com­ mon for preverbs to have rection, as prepositions have. Preverb and preposi­ tion are not functionally equivalent; the paradigmatic relation between preverbial and prepositional constructions is irregular because the former belong essentially to word-formation, while the latter belong to syntax. 1. Introduction The subject-matter of the present contribution is the syntax of Latin pre­ verbs. The central question is: How is the manner in which verba composita take complements to be described? And does it have a regular connection with the case-rection of the verbum simplex, on the one hand, and that of the preposition corresponding to the preverb, on the other? To my knowledge, this question has not yet been posed in Latin linguistics. The investigation re-

146

CHRISTIAN LEHMANN

ported upon here has had to start from scratch. I have attained little more than a classification of the relevant phenomena. All farther-reaching hypotheses are to be regarded with extreme caution. The initial assumption is that the argument structure of verba composita will be regular just to the degree to which their meaning and constructional behavior can be explained on the basis of the meaning and constructional be­ havior of their parts, namely the verbum simplex and the preverb. This search for syntactic regularity entails, therefore, the limitation to such verba com­ posita whose meaning has a regular relationship to the meaning of the simplex. I shall concentrate on such verbs as aufero or distraho, whose mean­ ing can be largely predicted from the meanings of few and ab-, or traho and dis-, respectively, and I will not treat such verbs as incipio or decerno, which do not allow this. This restriction has the natural consequence of a concentration on preverbs with local meanings, leaving other kinds of meanings (as, e.g., in pergaudeo "rejoice greatly") out of consideration. 2. Basic concepts Preverbs and prepositions are (together with adverbs, which I leave out of consideration here) coordinate categories and jointly subordinate to the notion of 'local relator' (LR), which I have to introduce here. 'Local' must be taken to cover 'spatial' and 'temporal', but it correctly emphasizes that spatial relationships are at the heart of the matter. Some examples will show in which sense prepositions and preverbs are relators: (1)

(a) ille qui stillantem prae se pugionem tulit (Cic.Ph.2,30) (b) Pontico triumpho trium verborum praetulit titulum: veni, vidi, vici (Suet. Caes. 37) (c) a portu me praemisisti domum (PI. Am.602)

In each of the examples, the semantic relation of 'x (being) before y' is inhe­ rent by virtue of the presence of prae, though it is not always made explicit. In (a), x is the dagger and y is he; in (b), x is the inscription and y is the proces­ sion; in (c), x is the speaker and y is the hearer. In semantically different, but structurally analogous ways, two terms are correlated by all the LRs. We call x the locatum and y the relatum. These two may be characterized as follows: The locatum is something whose position in space or time is described by the construction. This may be a person or a thing, as in the examples given; or it may be an event. We may leave this latter possibility open at the moment (see §5.2. ) and confine ourselves to the observation that if the LR is a preverb,

LATIN PREVERBS AND CASES

147

then its locatum is a nominal concept and thus an argument of the verbum compositum. The relatum is the item with respect to which the locatum is localized. It is an animate, lifeless or abstract object, a nominal concept which serves as a point of reference for the localization. Syntactically speaking, if the LR is a preposition, the relatum is its complement or rectum, as is se in (l)(a). If the LR is a preverb, then the relatum may be one of the arguments of the verb, as is in (b) Pontico triumpho. This is normally distinct from the argument which represents the locatum - as it is in (l)(b) -, but this need not be so, as we shall see later. Finally, the relatum may be implicit. Thus the fact that in (l)(c) it is the subject of the verb is not wholly determined by the syntax of the sentence; we infer it, partly, from its meaning and context. Whereas the distinction between locatum and relatum is not always straightforward with verba composita, as we shall see in §4.4.2., they behave quite differently when an LR takes the position of the predicate of a sentence (or clause). The locatum is then always the subject, as (2) shows. (2) nec spes ulla super (Val.F1.8,435) The relatum, however, if it is not implicit as here, will occupy an oblique case role, as can be seen in (3). (3) (in fossam) quae erat ante oppidum (Caes. B. G. 2,32,4) Thus, whenever a local construction is intimately tied up with a complex syn­ tactic environment, the reduction to its embryonal form, as in (2) or (3), can serve as a heuristic criterion to determine locatum and relatum. Almost all of the Latin preverbs have a basic local meaning. This would have to be defined as in the following examples, where x is again the locatum and y the relatum. x ex- y: x moves out of y x sub- y: x is or moves under y There are peculiarities associated with some of the preverbs which need not be dealt with here. It is self-understood that such definitions of local relations are not based on extra-linguistic reality; on the contrary, these relations are defined with respect to the speech-situation and would, in their complete form, incorporate the speaker's deixis. A preverb, contrary to other LRs, does not take arguments indepen­ dently from the verb to which it is prefixed. That is to say, both its locatum and

148

CHRISTIAN LEHMANN

its relatum have to be arguments of that verb. We may therefore approach the analysis with the following general hypothesis: We start from a preverb and a verbum simplex. Each has its argument structure associated with itself: the preverb has a locatum and a relatum; the verbum simplex has a subject and possibly a direct object and further complements. In the formation of a ver­ bum compositum, the argument structure of the preverb is superimposed onto that of the verbum simplex. One of the following may then happen to either of the arguments of the preverb: either it may be added to the argu­ ments of the verbum simplex, thus enriching the argument structure of the verb; or it may be lost in preverbation; or it may be identified with one of the arguments of the verbum simplex. With these assumptions in mind, we may now specify our initial question: Among the syntactic functions associated with a verbum compositum, which are those typically occupied by the locatum and the relatum of its preverb? 3. The syntactic function of the locatum We will begin with the syntactic function of the locatum. Empirical inves­ tigation turns up a general rule here which is valid for all of the productive types of preverbation and most of the unproductive types. Consider the fol­ lowing pairs of sentences, whose first member is intransitive and the second transitive; the locatum has been italicized. (4) (5) (6)

(a) (b) (a) (b) (a) (b)

Caesar ... ad cohortandos milites ... decucurrit (Caes.B.G.2,21,l) hic postquam in aedes me ad se deducit domum (Pl.Mf.121) si nemo hac praeteriit (Pl. Ci.683) ut te hodie quasi pompam illā praeterducerem (Pl.Mz.67) simul ac primum nubes succedere soli coepere (Lucr.5,286f) sues ... subigunt in umbrosum locum (Var.R. R. 2,4,6)

The rule is: I II.

If a verbum simplex is intransitive, then its subject is the locatum of the preverb in composition. If a verbum simplex is transitive, then its direct object is the locatum of the preverb in composition.

It is understood that what applies to the object of a transitive verb, applies to its subject if the verb is passivized. The rule could be simplified if the subject and object of a verbum simplex were automatically also the subject and object of the compositum; however, as we shall see later on, this is not always so.

149

LATIN PREVERBS AND CASES

The fact that this rule relies not on the notion of the subject, but on that of the subject of the intransitive and the object of the transitive verb seems to be an ergative trait in Latin. There has been much quarreling among scholars in recent times about whether a certain language is ergative, and another one is accusative. It would seem sensible to recognize that every language shares some ergative and some accusative features. Even the most accusative lan­ guage such as Latin is ergative to a certain extent. When we ask for the functional basis of our rule of the syntactic function of the locatum, we have to consider the way in which the argument structures of the preverb and the verbum simplex are superimposed. We have seen above that if an LR takes the position of the predicate of a sentence, its locatum must be the (intransitive) subject, whereas the relatum takes an oblique position. We therefore face the following parallelism: verb

absolutive

local relator

other arguments

locatum relatum

Now, if an LR is combined with a verb to form a unit with a single argument structure, this analogy becomes effective, and the locatum of the preverb is identified with the absolutive of the verb. The fact that every verb has an ab­ solutive argument and that every LR has a relatum accounts for the near ab­ sence of exceptions to this rule. 4. The syntactic function of the relatum As for the syntactic function of the relatum of a preverb with respect to the verbum compositum, the situation is much more complex. The following alternatives present themselves: 1. 2.

3.

The relatum may either be expressed by a nominal constituent of the clause or it may be implicit, i.e. not so expressed. If the relatum is expressed, it may be adjoined to the verbum compositum either in an adverbial phrase, especially a prepositional phrase, or in a simple case. If the relatum is adjoined to the verbum compositum in an adverbial phrase, it may either be governed by the preposition corresponding (i.e.

150

4.

CHRISTIAN LEHMANN

identical) to the preverb, or it may be adjoined otherwise. If the relatum is adjoined to the verbum compositum in a simple case, this may either be the one governed by the preposition corresponding to the preverb or another one.

The regularities determining these choices are to a great extent still unclear to me. I will illustrate them by some examples and, in discussing these, point out some tendencies that seem to be candidates for rules. 4.1. The relatum is implicit If the relatum is not expressed by a nominal constituent of the clause, it is implicit and must be inferred from the context. We may then also say that the preverb is deictic as to its relatum. The examples (4) and (6)(b) above illus­ trate what is meant. The following comments seem to be in order: 1. Many examples might be adduced of relata which are not only not ex­ pressed but which are not at all adjoinable to the verbum compositum in ques­ tion. Thus consider: (7) (8)

(tun redimes me,) si me hostes interceperint? (Pl. As. 106) locum nacti ... quern domestici belli ... causa iam ante praeparaverant (Caes.B.G.5,9,4)

The verbs intercipio, praeparo and many others forbid the expression of the relatum of their preverb even though used in a concrete spatial or temporal sense. The total absence of a relatum is, of course, much more frequent if the preverb does not have such a concrete meaning - a case not dealt with here. 2. One form of the implicit relatum requires special attention. Not in­ frequently does a preverb refer back to the subject of the verbum com­ positum. This occurs mainly with transitive verbs such as praemitto in (l)(c). Other such examples are: (9) Si te forte meae gravis uret sarcina chartae, abicito ... (Hor.Ep.l,13,6f) (10) libenter extremum spiritum vitae ediderim (Cic.Ph.11,9,22) The identification of the relatum with the subject might be made explicit in each of these examples by adding a reflexive phrase like prae te in (l)(c), a te in (9) and e me in (10). The fact that this sort of reflexivity is actually inherent in such verbs can be proven by examples from the texts: (11) abige abs te lassitudinem (Pl. Mer.113)

LATIN PREVERBS AND CASES

151

(12) oves nullum fructum edere ex se sine cultu hominum et curatione possent(Cic.n.d.2,63,158) This — reflexive or implicit — identification of the relatum of the preverb with the subject of the verb depends on the following constellation: the simplex must be a transitive verb of motion, i.e. of transport (maybe also in a figura­ tive sense), and the dynamic relationship between the transporting subject and the transported object must be such that the dynamic relationship be­ tween the relatum and the locatum of the preverb is already implicit in the former relationship or at least its most natural specification. Deixis evidently plays a role here. In concluding this section, it must be stated that, apart from the obliga­ tory transitivity of a few verba composita (cf. (24) and (25) below), it is quite normal for a verbum compositum to occur without an expressed relatum. This is essential in that it indicates that most preverbs are like adverbs, and unlike prepositions, in that they need not be rectional, but may be deictic as to their relatum. 4.2. The relatum is in an adverbial phrase We will defer consideration of relata adjoined with the prepositional counterpart of the preverb to §4.3. and will consider only distinct prepositions and adverbs. Examples of adverbs functioning as relata have been seen above in (5) (hāc and illa). Examples of prepositional phrases are: (13) Abin e conspectu meo? (PI.Am.518) (14) Itaque paulatim ex castris discedere ac suos clam ex agris deducere coeperunt. (Caes.B.G.4,30,3) (15) is me ad illam inlexit. (Pl.Au.737) With certain preverbs, this way of adjoining the relatum is quite regular. In fact, examples such as the above testify to the partial synonymy of a) ab, de and ex, b) ad and in, c) prae with ante, d) pro with ab, de, ex and ante. Whereas the semantic relationships in a) and b) are reciprocal, those in c) and d) are not: ante is less specific than either prae or pro ; and ab, de and ex may in­ troduce relata of pro, but not vice versa. On the other hand, there are perfectly clear cases of adverbial phrases

152

CHRISTIAN LEHMANN

which are not the relata of the verba composita on which they depend. Such is the case, e.g., in (4) and (6)(b) above, where the phrases ad cohortandos mi­ lites, in aedes ... ad se ... domum and in umbrosum locum are not the relata of the preverbs. It appears that a number of preverbs whose relatum does not play the role of a goal nevertheless imply one, in addition to their relatum. This is especially true for the preverbs whose relatum is a source: ab-,de-, ex-, and pro-. But it also occurs with others like per- and sub-. Additional exam­ ples are: (16) ecfunde hoc cito in barathrum (Pl.Cw.121) (17) incolumem legionem in Nantuates, inde in Allobroges perduxit (Caes.B.G.3,6,5,) The implication of a goal is so strong with some such verbs that this is by far more frequently specified in the clause than the relatum, which is mostly left implicit. It is safe to assume that this depends not only on the preverb but also on the verb. 4.3. The relatum is adjoined with the corresponding preposition For each of the preverbs we are dealing with, except dis-, re- and se-, there is a preposition that corresponds to it in that it is formally and semantically (almost) identical to the preverb. For all of them, except these latter ones, we might expect that their relatum can be adj oined by means of this cor­ responding preposition. In fact, this construction, which we will call duplica­ tion , is rather unevenly distributed among the different preverbs. We will first illustrate duplication with some typical examples: (18) de via decedite (Pl.Am.984) (19) quom ex alto puteo susum ad summum escenderis (P1.Mi.1150) (20) Hunc in collum ... impone. (Pl.Pe.691f) The overall frequency distribution of duplication is roughly as follows: frequent with ab-, de-, ex-; ad-, in-; com-, inter-, sub; rare with ob-, per-, pro-, trans-; never with ante-, circum-, post-, prae-, praeter-, super-. I cannot present an analysis of these facts here. A diachronic consideration to be taken into account would be that duplication is rare or impossible with the recent prepositions like trans and praeter, whereas it is common with the old­ est group, which ex-, in-, etc. belong to. However, one question that should be asked is: Why does duplication occur at all? Is it not redundant to say decedo

LATIN PREVERBS AND CASES

153

de, escendo e, impono in and the like? What if we eliminated one occurrence of the LR in such constructions? We shall return to this question at the end of this report. 4.4 The relatum is governed by the preverb When an LR is a preposition, it governs the case of its relatum; we say that it has rection or is rectional. Accordingly, we say that a preverb has rection and governs its relatum when this depends on the verbum compositum in the case which it would have to take if the preverb were a preposition. Since every Latin preposition takes its complement either in the accusative or in the ablative, we will subdivide the discussion according to these two cases. 4.4.1. The relatum is in the ablative The prepositions taking the ablative are ab, cum, de, ex, in, prae, pro, sub, super (and se). Here are examples of their rectional use as pre verbs : (21) Sceledre, manibus amisisti praedam. (P1. Mi.457) (22) ipse omnes copias castris eduxit (Caes.B.G.4,13,6) (23) pacis nomine bellum involutum reformido. (Cic.Ph.7,6,19) The frequency distribution of this construction is as follows: frequent with ab-, de-, ex-, pro-; possible with in-; almost never with com-,sub- (only with transitive verbs), super-; never with prae-. From this we may conclude that the only ablative regularly governed by preverbs is the separativus, the locativus being much more restricted. This corres­ ponds to the general distribution of the ablative outside preverb syntax. One may therefore ask whether the ablative is really strictly governed by the pre­ verb, or whether the adjunction of the relatum of the preverb to the verbum compositum is not rather based, in part at least, on the meaning of the abla­ tive. 4.4.2. The relatum is in the accusative The prepositions taking the accusative are: ad, ante, circum, in, inter, ob, per, post, praeter, sub, super and trans. The preverbs inter- and post- can gov­ ern the accusative only if the verbum simplex is intransitive, normally a verb of the lexical field "go". Here are some examples:

154

CHRISTIAN LEHMANN

(24) Aliquot me adierunt (Ter.And.534) (25) Quid me circumsistitis? (Pl.Men.998) (26) nonnulli ... fossam transire et macerkzm transcendere conantur. (Caes.B.G.7,70,5) Preverbs governing the accusative on a transitive verbum simplex may be seen in the following examples: (27) Eho istum, puere, circumduce hasce aedis et conclave! (Pl. Mo. 843) (28) flumen Axonam ... exercitum traducere maturavit (Caes.B.G.2,5,4) This construction is, however, extremely rare. This restriction suppresses the double accusative which will appear in such constructions, and this seems to be the ratio essendi of the restriction. Support for this hypothesis comes from the fact that in none of the other cases where a preverb on a transitive simplex is permitted to govern the accusative does a double accusative appear. At least two types of transitive verbs are relevant here: 4.4.2.1. Transitive verba composita with perConsider the verba composita derived from intransitive simplicia, as in (24) - (26). Their subject is identical to the locatum, and their object identical to the relatum of their preverb. This same situation holds true also with at least one compositum derived from a transitive verb, namely perquiro: if x perquirit y, then x per- y (though it is not true that x quaerit y). The situation is mini­ mally different in xperlegity and x perfïcit y: here the same situation holds true for the simplex (x legit y, x facit y), but there is no locatum for per-: ? per- y. Now if we start from the fact that the object of the simplex is identical to the re­ latum of per-, we have a model for further derivation of verba composita such as perfringo, perrumpo. For these, the same relations hold as for perlego above; however, in addition to these, an instrument may be used with the lat­ ter. This leads us to a last subtype where the instrument which can be used functions as the locatum of per-. Consider perforo: if z perforaty with x, then z forat y (with x), and x per- y. The same relations hold for pertundo and percutio (if we disregard the lack of a simplex for the latter). This is the first type of derivation from transitive simplicia with preverbs governing the accusative without a double accusative emerging from this. 4.4.2.2. Verba composita of affection Affection here means that the subject of the verbum compositum affects

LATIN PREVERBS AND CASES

155

the object with something (cf. afficio in the first example below). This type is semantically closely related to the last subtype of the preceding paragraph, but not dependent on a particular preverb. Observe the following examples: (29) (30) (31) (32)

Quaecumque afficiet tali medicamine vultum (Ov.Med.67) [postes] sunt inducti pice (Pl. Mo. 827) aedïs venalïs hasce inscribit litteris. (P1.Tri.168) sagittā Cupido cor meum transfixit. (Pl. Pe. 27)

All of these constructions conform to the formula: 'z affects object y with ob­ ject (instrument) x, and x LR y' ; i.e. the locatum of the preverb appears in the ablativus instrumentalis, and the relatum becomes the object of the verbum compositum. Even in this case, part II of our rule for the syntactic function of the locatum of a preverb, which says that the locatum is the direct object of a transitive simplex, remains true. This can be shown when the simplex can be used, together with the corresponding preposition, in a paraphrase of the con­ struction ; then at least some of the simplicia take the locatum of the preverb as their direct object. Thus: ("facio" medicamen ad vultum), duco picem in postes, scribo litteras in aedes and figo sagittam trans cor. We will see in the next section that it is possible to construe the locatum as the direct object even with some of the composita; but then the relatum is not in the accusative. If the locatum were the direct object of the verbum compositum here, and if, at the same time, the relatum were construed rectionally, we would have double accusative with these verbs. The hypothesis introduced above says that this should be avoided. This can be achieved in two ways: Either the preverb must not take its relatum rectionally. Then this is adjoined with a pre­ position or in the dative (see §4.5.). Or we must ignore the fact that the locatum is the object of the simplex and rather pay attention to the fact that it is an instrument in the activity designated by the compositum. Then the locatum is adjoined in the ablative. Each of the verbs in (29) - (32) yields to one of the alternatives; some even admit of both. The evidence of the transitive verba composita derived from intransitive simplicia (type adeo) and of the verba composita of affection shows that the rule for the syntatic function of the locatum has to refer to the verbum simplex and not to the compositum. This has been taken account of in the above for­ mulation (§3.). We need only the following supplement to part II of the rule:

156

CHRISTIAN LEHMANN

If the relatum of the preverb is the object of the verbum compositum, then the locatum is either in the ablative or not adjoinable. The first of the alternatives refers to the type afficio, the second to the type perlego. It is understood that in the construction with double accusative (type traduco), the relatum is not the direct object of the compositum. More than half of the Latin prepositions take the accusative. If rectional use of prepositions as preverbs were the rule, conflicts such as those analyzed above would arise with most of the transitive verbs in preverbation. Instead, the verba composita of affection are the only type where the rection of the pre­ verb prevails over that of the simplex. What is immeasurably more common is that the rection of the verb prevails and that the preverb has no rection at all. This will also be seen in the next section. The typical role of a preverb does not consist in changing the argument structure or even the transitivity of a verb, but in bringing the local specifications expressed by certain LRs nearer to the verb. This is also true when a preverb is allowed to govern the ablative. Change of the transitivity occurs in a moderately regular way only in one case, namely when a preverb governing the accusative is attached to an intransitive simplex (type adeo). In contradistinction to the type traduco (transitive simplex), here the relatum of the preverb really becomes the direct object of the verb, as is proven by passivization: (33) cum neque praetores diebus aliquot adiri possent (Cic.Q.fr. 1,2,15) Without playing down the transitivity-altering force of the preverb in this type, we may state that it occurs almost exclusively with verbs of the lexical field "go" and that this use of the accusative can be naturally explained as a grammaticalization of the accusativus directionis commonly taken by such verbs. 4.5. The relatum is in the dative There are some preverbs whose relatum is freely adjoined in the dative. These are ante-, prae- and post; in- and sub-; inter- and ob-. Examples are: (34) (a) (b) (35) (a) (b)

virtute regi Agathocli antecesseris (Pl.Ps.532) bonum anteponam prandium pransoribus (Pl.Men.274) Aiax in silva ... gladio incubuit. (Auct.Her.1,11,18) Hannibal ... quemque iussit ... foribus nomen suum inscribere (Liv.25,10,8)

For some verbs, e.g. succedo, this is practically the only construction.

LATIN PREVERBS AND CASES

157

For some preverbs, there is a restriction on dative-adjunction of the re­ latum to intransitive verbs. Thus, the relatum of ad-, com- and super- is freely taken in the dative by transitive verbs, but seldom or not at all by intransitive ones. Similarly, there are rare examples of transitive verba composita with circum- governing the dative but no such intransitive verb. We might suppose that this is connected with the avoidance of the double accusative discussed in the preceding section. Ad-, circum- and super- freely govern an accusative re­ latum if prefixed to intransitive simplicia, as we have seen. Their taking the dative with transitive simplicia seems to be interpretable as one of the alterna­ tives to avoid the double accusative. This is further corroborated by the fol­ lowing evidence: Some preverbs never or almost never have their relatum adjoined in the dative: ab-, de-, ex-,pro-;per-,praeter-, and trans-. Here we note again the ten­ dency of ab-, de-, ex- and pro- to behave alike. It is remarkable that just those preverbs most intimately associated with the ablative most bluntly reject the dative. Again, on the basis of the double accusative hypothesis, we do not ex­ pect these to take the dative. Per-, praeter- and trans-, on the other hand, either (and relatively seldom) take their relatum in the accusative (per- may also be duplicated), or the adjunction of the relatum is not grammaticalized at all, which means that it remains implicit or is couched in an adverbial phrase. It must be emphasized that the construction with the relatum of the pre­ verb in the dative does not stand apart from the others, but is tightly integrated into the overall preverb syntax. Many verba composita take the relatum either in the dative or by means of duplication (e.g. inscribo). Special attention should be drawn to the quasi-transformational relation, to be found with sev­ eral verbs, which exists between the construction 'locatum in the accusative - relatum in the dative', as evidenced by (34)(b) and (35)(b), and the construction 'locatum in the ablative - relatum in the accusative', as evidenced by the verba composita of affection in (29)-(32). Not all of the verbs of affection allow the transformation; observe, however, the close parallelism in the pair (31)/(35)(b). The frequency and even regularity of the dative expressing the relatum of certain preverbs is difficult to account for. One might be inclined to suppose that if the relatum of a preverb is adjoined to the verbum compositum by means of a mere case, i.e. without an intervening preposition, then it would be

CHRISTIAN LEHMANN

158

most natural for it to appear in the case which the prepositional counterpart of the preverb governs. This is evidently not so. Not only does the rection of the preposition not prevail; what is more, it is totally irrelevant for the dative-rection of the relatum whether the accusative or the ablative is associated with the preverb (as a preposition). The dative is equally frequent with ante- (acc.) and prae- (abl. ), and it accompanies illabor and succedo as well as incubo and subiaceo, although the prepositional counterpart with the simplicia of the former two would govern the accusative and with the latter two, the ablative. From this it must be concluded that the rection of the preposition is, with a large group of preverbs at least, irrelevant for the verbum compositum. It is not the preposition that is prefixed to the verb, but rather an LR which has two phenotypes: one as a preposition, which has a rection, and one as a pre­ verb, which mostly has no rection, namely in every case except the one discus­ sed in §4.4. We arrive here at a conclusion already touched upon at the end of §4.4. The normal situation in preverbation is the following: There is a verb which contracts relations with a subject and one or more complements. And there is an LR which contracts one relation with a locatum and one with a relatum. If the LR is prefixed to the verb, the locatum is identified with one of the argu­ ments of the verb according to the rule of the syntactic function of the locatum. But the other open slot of the preverb, its relatum, is normally added (with the exceptions noted in the preceding sections) to the case-frame of the verb. The preverb is not a preposition; so the argument occupying this slot is not normally subject to case-rection. The open slot is semantically there, but syntactically underspecified. This is why there are so many different possibilities to adjoin the relatum to the verbum compositum. The accusative and the ablative are among these possibilities. However, as we have seen in §4.4., their use is subject to certain constraints; they are often not the approp­ riate cases. Exempting the genitive from the adverbal cases, what remains is the dative. Most of the examples show that this use is an extension of its proper function, which is to express an object indirectly participating in an action or process. 5. Conclusion In conclusion, let me briefly comment on the absolutive hypothesis and on the relation between the preverb and the preposition.

LATIN PREVERBS AND CASES

159

5.1. The absolutive hypothesis In §3. we saw that the locatum of a preverb of a verbum compositum is regularly identified with the absolutive argument of the verbum simplex. In §4.2.-4.5. the relatum was seen to emerge as an oblique argument of the com­ positum, sometimes being identified with an oblique argument of the simplex. Furthermore, in §4.1., which dealt with the implicit relatum, we saw that it may relate back to the subj ect of a transitive verb. In an ergative system, this would be an oblique syntactic function, too. In view of this, it might seem more ap­ propriate not to treat the relatum identified with the transitive subject as an instance of the implicit relatum. We could then strengthen the absolutive hypothesis by saying that the locatum of a preverb is identified with the ab­ solutive argument of the verbum simplex, and its relatum becomes one of the arguments which in an ergative system are oblique. Though this tends to be empirically true, attention must be drawn to two sorts of exceptions: 1. The relatum may perhaps be identified with the sub­ ject of an intransitive simplex in some cases such as procumbo, lit. "to fall in front of oneself". 2. The relatum is identified with the object of a transitive simplex in several verba composita with per- as discussed in §4.4.2.1. As long as such exceptions are not explained, we can only maintain the first half of the absolutive hypothesis, which says that the intransitive subject and the direct object are treated alike, and leave its second half, which says the transitive subject is treated alike with other oblique syntactic functions, for further in­ vestigation. More evidence for a syntactic function 'absolutive' in Latin could proba­ bly be found in nominalization: both the subject of the intransitive and the ob­ ject of the transitive verb appear in the genitive — subjectivus and objectivus, respectively —, if the verb is nominalized. However, a genitivus subjectivus related to the subject of a transitive verb seems to be subject to special con­ straints (pace Kühner/Stegmann 1962, I:413); this occurs with some fre­ quency only if a genitivus objectivus is also there. Thus, occisio Caesaris would almost always mean that someone killed Caesar, not that Caesar killed someone. Evidence of a more semantic nature might be found in such lexical pairs as egeo —privo or abundo — dono (as suggested to me by Harm Pink­ ster), whose second members might be described as having an ergative in ad­ dition to the absolutive (and the other oblique case) present in both members. However, this evidence can be appreciated only when we have more precise knowledge about causative constructions in ergative and accusative systems.

160

CHRISTIAN LEHMANN

5.2. The function of preverbs Preverb and preposition are not functionally equivalent; x adit y and x it ad y are not synonymous; there is no regular transformational relationship be­ tween the two constructions. Preverbation belongs primarily to word-forma­ tion and only secondarily to syntax. Prepositions are rectional, preverbs are generally not, apart from the cases discussed in §4.4. Therefore, preverbs do not affect the valency of a verb in a regular way, apart from the subregularities analyzed. If we regard adjuncts joined to the verb by a preposition as belonging to the case-frame in a wider sense, then preverbation does not even enrich the case-frame of a verb. At least, it does not make local relations syntactically ac­ cessible that would have been inaccessible before. On the contrary, if we con­ sider the verbs of motion and transport centrally involved in preverbation, e.g. duco, a source and a goal is often already implicit in the meaning of the simplex. What the preverb does, e.g. in educo or adduco, is to make this local relation an expressed part of the meaning of the verb. Thereby, the relata of the local relations involved move nearer to the verb and become adjoinable with a mere case, whereas if the preverb were not there, they would have to be adjoined by a prepositioin. We might mention here the hypothesis that relata of preverbs are always verbal complements, whereas relata of prepositions may be complements or adjuncts (see Horrocks 1980: 201-204). But this syntactic consequence is epiphenomenal. We have seen that it is very common (after Cicero with de­ creasing frequency) that the relatum of a preverb is adjoined by a preposition, in particular by duplication. The preverb does not make the preposition superfluous, neither paradigmatically nor even syntagmatically. We might rather take recourse to a semantic hypothesis and conceive of the argument structure or case-frame of a verb as composed of various con­ centric ranges (cf. Pinkster 1972:91-101). The innermost would be the verb with its absolutive argument; the next would include the nearest oblique argu­ ments and might be called the nucleus. Adding to this other oblique argu­ ments which might be said to belong to the periphery, we come to the next range which is the center; and adding to this certain elements such as sentence modals, we arrive at the sentence level. We might then say that the relata of prepositions tend to belong to the periphery, but those of preverbs tend to be­ long to the nucleus. Similar considerations apply to the locatum. Additional syntactic justification of this semantically-based hypothesis would be called for.

LATIN PREVERBS AND CASES

161

BIBLIOGRAPHY Horrocks, Geoffrey C. (1980), 'Verb compounds in Greek: the elimination of a transformational rule'. Traugott, E . C et al. (eds.), Papers from the 4th International Conference on Historical Linguistics. Amsterdam: J. Benjamins; pp. 199-209 Kühner, Raphael/Stegmann, Carl (1962), Ausfuhrliche Grammatik der lateinischen Sprache. Teil 2: Satzlehre. 2 vols. Leverkusen : Gottschalk Pinkster, Harm (1972), On Latin adverbs. Amsterdam/London : North-Hol­ land

REACTION À LA LECTURE DE CHR. LEHMANN P . D E CARVALHO Sur votre idée initiale, je suis tout à fait d'accord: ce qui est en jeu avec le mécanisme de la préverbation est bien la localisation d'un être ou chose dont le nom correspond à un des arguments compris dans la structure casuelle, ou valence, du verbe. Je suis néanmoins plus réservé quant à l'usage que vous fai­ tes de votre concept principal, celui de locatum, avec les analyses syntaxiques qui en découlent. Vous commencez, en effet, par poser que "le locatum d'un verbe com­ posé est son sujet, si le verbe est intransitif, et son objet direct, si le verbe est transitif." En d'autres termes, plus explicitement "sémantiques": est locatum le "sujet" ou l'"objet" affecté par le résultat de l'opération signifiée par le verbe. Ou encore, pour reprendre vos propres expressions: 'le rôle syntaxi­ que et sémantique le plus intimement inhérent à la signification du verbe." L'ennui, me semble-t-il, est que vous ne tenez pas assez fermement à ce principe, car vous ne tardez guère à évoquer soit des expressions comportant, selon vous, un locatum à l'ablatif — p.ex. les "verbes d'affection", cf. eum [animum] [deus] circumdedit corpore et uestiuit extrinsecus : Cic. Tim. 20 , où, à vous en croire, ce serait corpore, ablatif, le locatum — soit des expressions où vous n'arrivez pas à identifier de locatum, cf. certains composés en per-,

162 évoqués en 4.4.2.1., p.ex. tuas litteras perlegi. Or je crains qu'il n'y ait là un malentendu. Apparemment, pour vous, ce qui définit le locatum est la position matérielle que se trouve occuper, "dans la réalité des choses", tel objet ou telle personne soumise à un mouvement non moins réel. Cf. d'ailleurs vos définitions en 2, elles sont toutes de la forme "x fait mouvement vers y", "x se trouve placé dans telle ou telle position (devant, derrière, etc.) par rapport à y". Ainsi, dans deus animum uestiuit corpore, ce serait corpore le locatum, parce que c'est la chose nommée "le corps" qui est vue déplacée. Dans litteras perlegi, comme évidemment "la lettre" ne bouge pas, vous refusez d'en faire votre locatum. C'est d'ailleurs pourquoi vous aviez, dès le départ, situé au centre de vos réflexions des verbes tels que aufero, distraho, etc., dès lors que, plus que d'autres, ils évoquent l'image d'un mouvement réel imprimé à une personne ou à une chose. Ce qui entraîne l'inconvénient majeur de vous obliger à ignorer superbement celui de tous les verbes latins qui est, peut-être, le plus ouvert à la préverbation, je veux dire sum. Vous m'accorderez peut-être que c'est assez surprenant, d'autant que vous êtes le premier à invoquer, dans l'avant-dernier paragraphe de votre ex­ posé, l'influence des composés de ce verbe comme un facteur possible de l'ex­ tension de la construction dative des composés en général! Dans ces conditions, je suis amené à vous poser la question suivante: ne voudrait-il pas mieux définir le locatum d'une façon plus "abstraite", permet­ tant de cerner, par-delà le contenu lexical de chaque préverbe, quel que soit ce contenu, l'apport constant de la catégorie Préverbe en tant que telle? Dans cette perspective, et pour ne pas m'en tenir à une critique purement négative, voici que je me risquerais à proposer dans le prolongement de votre analyse, qui, rigoureuse et fine, assurément, reste néanmoins trop asservie à la particularité lexicale de chacun de ces préverbes et manque, de ce fait, à faire voir la forme générale qui les soustend. A mon avis, il faudrait définir le locatum comme faisant référence, dans tous les cas, non à la localisation "réelle", dans l'expace ou dans le temps, de l'argument "objet", mais au fait même que cet "objet" se présente, aux yeux du locuteur, comme "positionné" de telle ou telle manière dans l'espace ou dans le temps — et ce du fait d'un certain "cas d'événement" (représenté par la notion verbale), mais par-delà le durée stricte de celui-ci. A charge, ensuite, pour la notion verbale, pour la notion préverbale et pour le "complément" éventuellement appelé par la pre­ mière, ou par la combinaison de la première et de la seconde, de spécifier: 1°

163 de quelle modalité d'événement il s'agit; 2° quelle position particulière appa­ raît occuper le locatum. 3° le cas échéant, quel est le repère extérieur qu'impli­ que ce positionnement. Ce serait, autrement dit, la représentation résultative de ce "positionnement" l'apport général de tout préverbe; c'est par cet apport qu'un verbe préverbe, quel qu'il soit, s'oppose à la version non préverbée de la même notion verbale, dont le propre est de ne pas "faire voir" ce position­ nement. Deux exemples rapides suffiront à préciser ma pensée. Dans deus animum circumdedit corpore, le locatum serait bel et bein l'objet anïmum. Cela ne signifie pas, bien entendu, que cet animus soit, en l'occurrence, soumis à une forme quelconque de mouvement. Cela veut dire simplement: "il y a là désormais, consécutivement à l'opération dedit, dans la situation telle que la voit le locuteur — ce que j'ai nommé ailleurs le "présent délocuté" — un ani­ mus situé dans l'espace d'une certaine façon, celle que désigne circum, autre­ ment dit: "un animus entouré de ...", avec, naturellement, un "complément" pour désigner l'argument nouveau suscité par cette modification apportée par le préverbe à la notion verbale. De même, dans Tuas litteras per legi, le locatum est, encore et toujours, l'"objet" litteras : "il y a là, posée d'une certaine manière dans le présent délocuté — soit: traversée de part en part par une cer­ taine opération — une lettre". Ici, bien entendu, point de repère extérieur: "la lettre" apparaît, en quelque sorte, posée par rapport à elle-même : comme un entier implicitement opposable à chacune et à la série de ses parties. Vous devinez, je pense, les conséquences de la généralisation proposée. Il me semble qu'elle permettrait de renforcer la cohérence de votre descrip­ tion et de surmonter certaines impasses. Parmi ces conséquences j'en retien­ drai une seule, pour finir. D'ou vient-elle donc, cette tendance, si puissam­ ment accusée à partir d'un certain moment, à donner un complément "datif" à un grande nombre de verbes composés? Mais elle est inscrite, en puissance, dans la forme générale instituée par la préverbation! En thèse générale, on peut avancer que tout verbe composé porte, virtuellement inscrite en lui, l'image d'un résultat, même si la modalité particulière de ce résultat n'a pas, à première vue, de rapport avec la représentation associée à la forme casuelle DATIF. En effet, qui dit résultat dit, inévitablement: existence possible d'un être affecté plus ou moins durablement, d'une manière ou d'une autre, par la présence "en lui" de ce résultat, "contenant" en lui, dans son existence pro­ pre, ce résultat. Ainsi, quand je lis chez Horace (Sat. 1,3, 36-37) namquel

164 neglectis urenda filix innascitur agris, je ne puis me résoudre à considérer, comme vous semblez le suggérer à la fin de la version élargie de votre exposé, qu'il y aurait là un usage "arbitraire", vaguement "suspect." Certes, dans la "réalité des choses", le résultat particulier aperçu consiste en "la présence de fougères dans le lieu, dans l'espace inanimé désigné par agris". Et cependant ces "champs non cultivés" ne se laissent pas concevoir uniquement comme un espace inerte servant, "hic et nunc", de cadre à une présence. On peut aussi y voir un Être intégrant durablement, dans sa propre existence ultérieure, la présence de ces fougères. Bref, Horace veut peut-être dire que des champs quel'on oublie de cultiver ''ont en eux, portent en eux, des fougères!" On com­ prendra ce que je veux dire si l'on veut bien se reporter au texte d'Horace (de l'importance du contexte dans l'analyse syntaxique!): on l'apercevra alors que ces neglecti agri ne sont rien d'autre que la métaphore d'une âme humaine qui n'a pas su se perfectionner. Si l'on admet cette façon de voir, si d'autre part l'on veut bien consentir à examiner sans préjugé la théorie morpho-sémantique des cas latins proposée ici même, avec la description qu'elle comporte du signifié du cas Datif—une sorte de "futur" casuel! —, l'extension de la construction "dative" de la plu­ part des verbes composés apparaîtra, finalement, assez prévisible. L'histoire de cette extension serait, alors, sans doute, à décrire en termes d'une résis­ tance plus ou moins nette, et finalement surmontée dans bon nombre de cas— opposée par chaque préverbe à l'orientation générale inscrite dans le méca­ nisme de la préverbation. Dans le cadre général de cette hypothèse, une attention particulière devrait être portée à la relation "logique" existant entre l'item lexical investi du statut de repère extérieur ou relatum et l'"objet" posé comme locatum, p. ex.: 1. relatum "animé" intégrant, "possédant" un locatum "inanimé", comme dans l'exemple d'Horace; 2. relatum "inanimé conte­ nant" à l'endroit d'un locatum "inanimé contenu", comme dans votre exem­ ple (35 b) Hannibal...quemque iussit... foribus nomen suum inscribere (Liv. 25,10,8). Une toute dernière observation. J'ai été particulièrement frappé par votre suggestion, en 3, à propos de l'existence de traits "ergatifs" en latin. Le terme même d'"ergatif' est sans soute à manier avec prudence. Il me serait difficile, d' autre part, de poser le problème comme vous venez de faire, en ter­ mes de simple "coexistence" de traits ou éléments. Il reste qu'au fond ce que vous avancez là rejoint tout à fait certaines conclusions auxquelles j'ai pu par-

165 venir dans ma réflexion sur les cas latins. A ceci près, toutefois, que je serais pour ma part enclin à aller un peu plus loin et à me demander si le latin est, si peu que ce soit, une langue ... accusative! Quoi qu'il en soit, il est assez réconfortant de constater que des recher­ ches tout à fait indépendantes, et fondées sur des postulats radicalement opposés, peuvent parfois conduire à des résultats convergents.

NEW APPROACHES TO ACCUSATIVE SUBJECTS: CASE THEORY VS. RAISING

MIRKA MARALDI University of Bologna, Istituto di Filologia Latina e Medioevale SUMMARY This paper is the resultof a reconsideration of some proposals concerning infinitival clauses advanced by Chomsky's more recent linguistic theory. In this framework Case Theory, which interacts with other subtheories of gram­ mar such as Binding Theory and Theory of Control, offers a means of arriving at an analysis of infinitive constructions as an alternative to Raising analysis suggested in particular by Postal. Working within the perspective of these re­ cent results, I propose a derivation of Latin Accusative with Infinitive comple­ ments which can explain the status of these constructions, without resorting to a Subject-to-Object Raising transformation, a rule considered redundant and unmotivated in the general theory of Transformational Grammar and insuffi­ ciently explicative in Latin grammar. As has been emphasized by Bolkestein (1979), Miller (1974), Pillinger (1980), in a lot of cases the analysis of the Accusative in the Accusativus cum Infinitivo (henceforth AcI) complement as a Raised Object gives rise to sev­ eral difficulties.1 The following sentence types support this statement: (1) (2) (3) (4)

dicitur eos venisse 'It is said that they have come' manifestum est eum abisse Tt is clear that he has gone' constat te bonum esse Tt is certain that you are good' earn admoneo eos profectos esse T warn her that they have left'

168

MIRKA MARALDI

In these examples, any attempt to analyse the Accusative as part of the main sentence seems to fail inevitably: in (1) because of the passive form of the main verb, in (2) because the main verb is not subcategorized for an NP com­ plement (see the ungrammatical *constat te, *dico te, *scio te) or because of its being subcategorized for both an NP complement as direct object and an AcI clause, as in (4). 2 All these data, which support the view that it is often impossible to characterize the Accusative subject in AcI as a Raised Object, also substan­ tiate the claim that AcI functions as a single constituent of category S at all points in the derivation, as does the English AcI construction according to Chomsky's theoretical model. Chomsky assumes, as the simplest theory, that there are two types of clauses, tensed clauses and infinitival clauses, which are both constituents of category S generated in the base. As a consequence, there is no need to hypothesize Raising to Object as part of the derivation of infinitival clauses. The choice between a tensed and an infinitival clause as a complement is, on the contrary, determined by language-specific lexical con­ straints indicating what form of complement clause a verb may select.3 In this paper, I will first give some support to the basic idea of analysing AcI as a single S-constituent on a par with a finite clause.4 Secondly, as a con­ sequence of the rejection of the proposal that the case form of the Accusative may be due to a Raised Object status, I will try to explain, by means of Case theory, why a subject-constituent is marked Accusative and not Nominative as the subject-constituent of a finite clause must be. In traditional grammars the term AcI generally indicates any construc­ tion in which an Accusative noun is present together with an infinitive. Thus the following sentence types are both considered examples of AcI clauses: (5) (6)

dico te venisse 'I say that you have come' cogo te venire T force you to come'

As a matter of fact, there appears to be no relevant difference between these surface sentences, but an analysis of their structures shows that they are quite distinct from each other. In particular, they differ at the level of thematic role representation, i.e. in the representation of grammatical relations (such as 'subject', 'object', 'indirect object', 'clausal complement', etc.) that enter into determining thematic roles (like 'agent', 'patient', 'goal', 'theme', etc.), 5 as the corresponding representations emphasize:

ACCUSATIVE SUBJECTS

(7)

169

[ s [ NPj e ][ VP [v j - dico ][ s [ NP te ][ VP venisse ]]]

(8) [ s [ N P j e ] [ V P [ v j c o g o ] [ N P i t e ] [ s [ s [ N P i PRO][ VP venire]]]] The well-formedness of these structures is predicted if we assume, in agreement with Chomsky (1981: 31 ff.) that much of the information expressed in them is, effectively, determined by the lexicon. These structures, in fact, meet the so-called '0-criterion', that is the condition on the thematic role rep­ resentation, since they assign the grammatical relations which determine the thematic roles, as specified in the lexical entries for the two verbs examined. Thus, if the lexical entry for cogere specifies that it is strictly subcategorized for an object NP complement and a clausal complement, assigning to each a thematic role ('theme' and 'goal', respectively), whereas the lexical entry for dicere specifies that it is strictly subcategorized only for a clausal complement with the thematic role 'theme', then (7) and (8) meet the 0-criterion and there­ fore are well-formed. If we assign to (5) a representation analogous to (8), the result will be ruled out. See the representation in (9): (9)

[s[NPje][VP[vjdico][Npite][s[NPiPRO][Vpvenisse]]]]

This structure would be a violation of the 9-criterion; the object te could receive no thematic role, since the strict subcategorization frame of dicere specifies that it takes only a clausal complement and only this element plays a thematic role. If we accept that the assumptions made on the lexicon and on D-structures (=deep structures), now considered as a kind of "projection" from lexi­ cal structure via the 9-criterion, are correct, the (7) and (8) present clear evi­ dence in favour of analysing (5) and (6) as different from each other. (5) is an example of a real Acl clause; in it the Accusative neither bears any grammati­ cal relation to the main verb nor does it have any thematic role. It has only the grammatical function of subject that determines its 'agent' role in relation to venisse, within the clausal complement subcategorized by the main verb. The ill-formedness of (9) prevents the analysis of the Accusative as a real object or as a Raised Object; it functions only as the subject of the AcI clause. (6), instead, is a control structure, 6 in which the Accusative is part of the main sentence as a real, underived object. The representation given in (8) al­ lows a correct semantic reading of (6) without the need to assume an unneces­ sary Raising into Object position, as Pepicello (1977: 217) does. The fact that

170

MIRKA MARALDI

in (6) the Accusative is felt to be related to both main verb and infinitive di­ rectly follows from the theory of Control, which determines that the NP cate­ gory PRO, i.e. the pronoun without phonetic matrix that functions as subject of the embedded infinitive, has to be interpreted as coreferential with te. By rules of control, in fact, PRO in (8) is controlled by and coindexed with its an­ tecedent (the object te) from outside of its own S, as required by the third prin­ ciple of Binding theory (cf. Chomsky 1981). This analysis explains, therefore, why te is understood as 'agent' of the embedded infinitive even if it bears the grammatical relation object to the main verb and has, consequently, a different thematic role. We may represent the structures underlying AcI clauses and infinitival clauses of control as follows, respectively: (10) [ [ NP . +phonetic S S matrix

VP

(11) [S[S NP = PRO s s -phonetic matrix

VP

+infin.

]]

]] + infin

While (11) is predicted by Case theory, (10) is not. If we assume, in accor­ dance with Chomsky (1981) that the structures underlying infinitival clauses given in universal grammar (henceforth UG) are like (11),7 then as a conse­ quence, at least two questions arise. How may we account for the occurrence of structures as in (10)? And how may the Accusative marking be explained if, in conformity with the theory of Case, subjects of infinitives do not receive Case? Before giving a plausible answer to these questions, I will briefly outline Case theory. Following some suggestions advanced by Lightfoot (1980) in re­ lation to Chomsky's Case theory, I assume that UG makes two kinds of Case available: those assigned at the level of S-structure and those assigned in the base. While surface Cases are assigned to NPs only on the basis of the struc­ tural notion of 'government' (a general notion which underlies both Case theory and Binding theory), Cases assigned to NPs in the base are determined by idiosyncratic properties of governing lexical items. For example, in Latin, where particular verbs subcategorize a dative or ablative object, these Cases would be assigned in the base. Government is defined as follows:

ACCUSATIVE SUBJECTS

171

(12) a governs |3 iff a minimally c-commands ß and there is no S or NP between a and |3 a = [± N, ± V], or INFL(ection) where minimal c-command is defined as in (13): (13) a minimally c-commands |3 iff 1) a c-commands (3 (i.e. if neither dominates the other and if the first branching node dominating a dominates (3) 2) there is no Γ such that a c-com­ mands γ and Γ c-commands β and not Γ c-commands a Case assignment rules dependent on government are as in (14): (14) a. NP → Nominative iff governed by INFL, where INFL is finite, containing Tense and AG (= agreement, i.e. a collection of features, person, gender, number, index, case) b. NP → Accusative iff governed by V (14a) expresses that Nominative is the Case assigned to subjects of finite verb forms. I have decided to state this rule in terms of INFL government along the lines discussed in Chomsky (1981: 52,263ff.) in view of two facts which characterize Latin. Firstly, Tense cannot be the only element which distinguishes finite clauses from infinitival clauses. In fact, Latin AcI clauses, as Bolkestein (1979: 33) rightly observes, are less nonfinite than their English counterparts, since they not only show agreement in case, gender and number between the subject and the infinitive when the latter consists of a perfect or future tense participle plus esse, but can also express tense, the infinitive having a complete verbal value (cf. Perrochat 1932). Secondly, in Latin there is an independent motivation for INFL government: Latin is a language which has 'missing subjects', i.e. it shows empty Nominative NPs. According to Chomsky (1981: 241f. ; 256ff.) this property can be accounted for by assuming that, like Italian, which shows the same property, Latin has an index on the node AG. By means of this index INFL becomes a possible governor for the empty subject and can be coindexed with it (see above (7) and (8) where the main empty subject has been coindexed with the main verb). (14b) states that Accusative is the Case assigned to object NPs which are subcategorized by verbs, since government, expressing a kind of local control, determines the structural domain of subcategorization.8

172

MIRKA MARALDI

From government and Case assignment as in (14) it follows that UG makes (11) but not (10) available as the structure underlying infinitival clauses. In fact, in (10) the lexical subject is not governed either by INFL, since it is infinitival, or by the main verb, because of the presence of the node Š which is assumed to be an absolute barrier to government and, consequently, cannot be Case-marked. In these conditions a structure like (10) would always be ruled out by the Case Filter that marks as ill-formed any structure in which an NP is lexical and not Case-marked (cf. Chomsky 1981:49;175). But if we assume that particular grammars may delete the node Š under certain condi­ tions, then we have the possibility of preserving the structure in (10). The pur­ pose of the definitions in (12) and in (13) is clear: a lexical category or INFL governs its sisters and also the non-sisters that it c-commands, if there is no in­ termediate Š- or NP- boundary and if there is no other lexical category or INFL that governs the non-sisters in question. The direct consequence of this statement is that a lexical category can govern the subject of its clausal com­ plement if there is no Š- but only an S- boundary and if there is no INFL gov­ ernment in the embedded clause. In AcI clauses the second condition is always met. As far as the first con­ dition is concerned, I think that in Latin there are sufficient reasons to support a rule deleting Š. In fact Š deletion can be applied if particular structural fac­ tors are met, namely: Š (which rewrites as COMP (= complementizer) S) must contain a lexically null COMP and an infinitival S and the main verb must be one which may undergo the rule (i.e. a verb that can occur in the environ­ ment: NP VPr [+infin.]' where NP is lexical and the subject of the infinitive). Since these conditions are always met by Latin AcI clauses, it is reasona­ ble to think that Š deletion is available in Latin.9 As a consequence, the follow­ ing structure can be obtained: (15) α [S β where no node Š is present and a governs (3 By postulating Š deletion we may account for the Accusative subject of the AcI clause: it results from a property of 'exceptional Case-marking' that allows main verbs to assign Accusative case to the subject of the embedded S if Š deletion has been applied. Note that Š deletion is not only required to express this property of 'ex­ ceptional Case-marking', but is independently motivated. Within this theory of grammar it is also required to account for the occurrence of Nominativus cum Infinitivo constructions, as well as the occurrence of se as an anaphor and of is as an obviative pronoun, when they are subjects of AcI clauses. The situa-

173

ACCUSATIVE SUBJECTS

tion which I refer to is well illustrated by the following structures: (16) [ s [ N P i Ille][ V P [ v dicitur[ s [ N P i t][ V P venisse]]] (17)

[S[NPiIlle][VP[vdixit][S[NPiSe][VPVenturum

esse

]]]

(18) [ S [ N P i Ille][ V P [ V dixit][ s [ N P j eum][ V P venturum esse]]] If we apply Š deletion these structures observe the principles of Binding theory concerning trace, reflexive and pronoun (cf. Chomsky 1981: 188fn.)1() and are consequently well-formed. If we assume, therefore, that Latin grammar, like English grammar, has a rule deleting Š, then we can provide a plausible answer to the questions above: the occurrence of AcI clauses can be attributed to the availability in Latin of Š deletion ; this rule makes it possible for the subj ect of the infinitive to be governed and Accusative marked by the main verb across an S-boundary. Since in English not all verbs can assign Case to an embedded subject even if they govern it, the exceptional Case-marking in this language is depen­ dent on idiosyncratic properties of the main verb (see believe vs. seem). In Latin, instead, there is no need to postulate this property in the lexicon; it is sufficient that the main verb is one which may undergo the rule deleting §. An analysis of the Latin AcI clause in these terms might be generalized to the vari­ ous instances of this complement type, in which the Accusative has to be con­ sidered exclusively as the subject of the infinitive, as in the examples (1), (2), (3) and (4), and there is no need to hypothesize an untenable grammatical re­ lation between the main verb and the NP subject of the embedded infinitive. As far as (1) is concerned a further clarification has to be provided. This example seems to contradict the statement that the basic property of passive morphology is that it absorbs Case (cf. Chomsky forthcom.:34). Because of this property we always ought to obtain a Nominativus cum Infinitivo con­ struction and not an AcI clause as a complement of passive verbs of the dicor type. In other words, instead of the type dicitur eos venisse in (1), we only ought to have the following sentence type: (19) (iī) dicuntur venisse with an S-structure which is similar to that in (16), where the trace indicates that the embedded subject has been moved to become the main subject and consequently to receive Nominative case. (16) is well-formed since after S de-

174

MIRKA MARALDI

letion the trace is governed but not Case-marked, the main verb being in the passive form. Thus, if we accept that the property of passive morphology is that of absorbing Case, we may account for the occurrence of examples like (19). This analysis, however, may not be generalized to the occurrence of sen­ tences like (1), where the main verb seems to assign Accusative case to the em­ bedded subject it governs, notwithstanding its passive morphology. Perhaps it could be suggested that in Latin a passive main verb has the property of assigning Accusative case optionally, under the restrictions im­ posed by the Case Filter and by the principles of Binding concerning traces: thus, if there is a trace, then the main verb cannot assign Case, since the trace of an NP movement is never Case-marked; on the contrary, if there is an NP left, the main verb has to assign Case. The optional quality of such a Case assignment corresponds to the alter­ native occurrence of two constructs: the personal one, where the embedded subject becomes the subject of the higher clause and the so-called imper­ sonal one. The possibility of assigning Accusative case is limited to the occur­ rence of an impersonal passive verb. In this case the embedded infinitival clause is felt as a semantic unit, corresponding to a finite clause of the type : It is believed that they have come. As far as I can see, this seems to be a plausible explanation of the cooccur­ rence of sentences such as (1) and (19) in Latin. Nevertheless, in order to ob­ tain a more complete analysis of these types a deeper investigation into the status of Latin personal and impersonal constructs is required, which takes into account diachronic by the side of synchronic considerations. This re­ quirement seems to be confirmed by the fact that the impersonal type dicitur eos venisse is not found in Archaic Latin while, from Cicero onwards, it occurs with increasing frequency.

NOTES 1) On the dubious theoretical importance of Raising-to-Object within the Transformational Grammar framework itself see Lightfoot (1976) where a clear comparison between Chomsky's and Postal's theories is given. 2) For a discussion on the verb admonere see Bolkestein (1976). 3) It is well known that certain verbs select the form of complementizer and even tense of a sen­ tential complement (cf. Bresnan 1970; Grimshaw 1979). 4) On the use of AcI in replacement of quod-constructions and ut-constructions see, in particu­ lar, Perrochat (1932).

ACCUSATIVE SUBJECTS

175

5) See Chomsky (1972, chap. 3); Jackendoff (1972); Wasow (1980). 6) This distinction is proposed in Maraldi (1980). 7) Various languages provide evidence for this statement (see Italian, French, German, for example). Latin too seems to give some support to it: while Classical Latin and in particular the lit­ erary style greatly extends the use of AcI, familiar/colloquial Latin on the other hand tends to re­ strict the occurrence of the infinitive in control structures and to replace the AcI clause with finite verb clauses. This is shown particularly by some concordances of use between Latin of Plautus, Post-Classical Latin and Romance languages. Compare, for instance, the Plautine promitto facere with the Classical promitto me facturam esse and the Italian prometto di fare. On this issue see Perrochat(1932). 8) On the relationship between subcategorization and government see Hendrick (1980). I use the term Accusative which is usual in Latin syntax in place of Objective used by Chomsky. 9) On Š deletion in English grammar see Lightfoot (1980: 279 ff.). 10) For a more extensive analysis of the occurrence of se and is in these structures see the paper by Bertocchi-Casadio in this volume.

BIBLIOGRAPHY Bertocchi, A., & C. Casadio. 'Anaphoric relations, pronouns and Latin com­ plementation.' This volume. Bolkestein, A.M., (1976), 'AcI- and ut-clauses with verba dicendi in Latin'. Giotta 54, 263-291. . (1979), 'Subject-to-Object Raising in Latin?.' Lingua 48, 15-34. Bresnan, J., (1970), 'On complementizers: towards a syntactic theory of com­ plement types.' Foundations of Language 6, 297-321. Chomsky, N., (1972), Studies on semantics in generative grammar. The Hague: Mouton. . (1981), Lectures on Government and Binding. Dordrecht: Foris. . (forthcoming), A naturalistic approach to language and cognition. Grimshaw, J., (1979), 'Complement selection and the lexicon.' Linguistic In­ quiry 10, 279-321. Hendrick, R., (1980), On the form and functioning of subcategorization. Bloomington: Indiana University Linguistics Club. Hofmann, J.B., & A. Szantyr, (1965), Lateinis che Syntax und Stilistik. Mün­ chen: Beck. Jackendoff, R.S., (1972), Semantic interpretation in generative grammar. Cambridge, Mass.: MIT Press.

176

MIRKA MARALDI

Kühner, R., & C. Stegmann, (1955), Ausführliche Grammatik der lateinischen Sprache. Satzlehre. 3. Auflage durchgesehen von Andreas Thierfelder. Leverkusen: Gottschalk. Lightfoot, D., (1976), T h e theoretical implications of Subject Raising.' Foundations of Language 14, 257-285. . (1980), T h e history of NP movement.' In: T. Hoekstra, H. van der Hulst and M. Moortgat (eds.), Lexical grammar. Dordrecht: Foris Publications. Maraldi, M., (1980), T h e complement structure of perception verbs in Latin.' In: G. Calboli (ed.), Papers on Grammar I. Bologna: CLUEB. Miller, D.G., (1974), 'On the history of infinitive complementation in Latin and Greek/ The Journal of Indo-European Studies 2. 223-246. Pepicello, W.J., (1977), 'Raising in Latin'. Lingua 42, 209-218. Perrochat, P., (1932), Recherches sur la valeur et l'emploi de l'infinitif subor­ donné en Latin. Paris: Les Belles Lettres. Pillinger. O.S., (1980), T h e accusative and infinitive in Latin: a refractory complement clause.' Journal of Linguistics 16, 55-83. Postal, P.M., (1974), On Raising: one rule of English grammar and its implica­ tions. Cambridge, Mass.: MIT Press. Wasow, T., (1980), 'Major and minor rules in lexical grammar.' In: T. Hoekstra, H. van der Hulst and M. Moortgat (eds .) Lexical grammar. Dordrecht: Foris Publications.

LES RELATIVES EN LATIN: "RAISING" OU "MATCHING"?

J.P. M A U R E L Université de Toulouse-Le Mirail

1.

"Raising Analysis"

1.1. Dans la grammaire transformationnelle, on a d'abord proposé pour les relatives l'analyse suivante: on suppose que la phrase matrice et la phrase enchâssée contiennent deux NP lexicalement identiques; la deuxième occurrence (celle de la phrase enchâssée) est alors effacée et remplacée par le relatif. C'est ainsi que Kuroda (1968) pose que la forme de base de: (a)

the material which lay on the table was the tissue

est: (a') the material (WH the material lay on the table) was the tissue. De la même manière, pour Chomsky (1965), sous-jacente à: (b) the man who had been fired returned to work, nous avons: (b') the man — (wh — the man — had been fired) returned to work. En fait, Chomsky suppose d'une manière plus précise que les deux oc­ currences du NP plutôt que d'être identiques sont en réalité "non-distinctes". Dans la théorie phonologique, en effet, on dit que deux segments sont dis­ tincts si, pour le même trait, l'un est spécifié positivement et l'autre négative­ ment. En ce qui concerne les entrées lexicales, on considère comme non-dis­ tinctes (et donc effaçables) celles qui possèdent le même ensemble de traits in­ hérents (par opposition aux traits non-inhérents, introduits par transforma­ tion). Ce type d'analyse est traditionnellement désigné par le terme de "match­ ing analysis", abrégé ici en M — A. 1.2. A partir des années 70, on voit apparaître une analyse alternative, connue comme "raising analysis" (ici R — A ) . Elle est développée essentielle-

178

J.P. MAUREL

ment par J.R. Vergnaud (1974), tenant compte des travaux de Brame (1968) et Schachter (1973). Dans cette nouvelle perspective la phrase matrice ne con­ tient pas de NP réalisé mais une place vide dans laquelle est promu ("élevé") le NP de la phrase enchâssée. C'est ainsi que pour Vergnaud: (c) Les syndicalistes que Pinochet a jetés en prison a pour structure profonde (simplifiée ici):

-W H Pinochet a jeté Det syndicalistes en prison. Alors (en négligeant diverses étapes), le NP de la phrase matrice est promu sous N et la relative est extraposée sous S. Nous allons maintenant examiner rapidement les arguments donnés pour cette analyse par Vergnaud et les autres. Ils sont de trois types: 1.2.1. Le premier argument fait intervenir les contraintes pesant sur certaines locutions idiomatiques comme to make headway en anglais, ou pren­ dre part à, tirer parti de, faire cas de en français. C'est ainsi que, alors que (1) est non-grammaticale: (1) *I1 est surpris de la part de Jean, (2) , pourtant très proche, est grammaticale: (2) Il est surpris de la part que Jean a prise à son succès.

LES RELATIVES EN LATIN

179

L'explication serait la suivante: dans (1) part ne peut être employé (dans ce sens) sans être régi par le verbe prendre. Mais la grammaticalité de (2) vient du fait qu'en structure profonde part est, dans la relative, complément de prendre. On oppose de même: (3) * Wang-Hung-Wen a pris aux travaux du 9e congrès la part qu'il décrit dans son livre et (4) il décrit la part que W.H.W. a prise (3) est non-grammatical alors que part est dans la même proposition que prendre! C'est que ces deux éléments étaient séparés en structure profonde, part ayant, selon R — A, sa source dans la relative. On aurait un raisonnement proche de celui-ci avec des paires comme: (5) * La manière a été astucieuse (6) La manière dont elle a résolu le problème a été astucieuse. Nous ne discuterons pas ici longuement ce type d'argumentation; nous le ferons systématiquement dans une autre étude. Disons simplement que le phénomène est beaucoup plus général et concerne toute une série de NP ac­ compagnés de leurs "verbes opérateurs", comme (commettre) une agression, (poursuivre) un but, e t c . . Ce qui est en cause alors, c'est la "structure argumentale" des NP en question, et les contraintes pesant sur l'absence possi­ ble de tel ou tel argument, ainsi que du verbe opérateur. Ce dernier peut être en fait absent (sous certaines conditions), comme le montre: (l') Sa part dans cet enlèvement a été décisive, ce qui enlève beaucoup de poids à l'argumentation de Vergnaud. Il faut d'ailleurs remarquer à propos de (l'), que nos jugements de gram­ maticalité diffèrent nettement de ceux de J.R. Vergnaud, pour qui (l') est non-grammatical! Il en est de même pour (3), pour nous parfaitement gram­ matical. Et si c'est le cas, cela ruine R — A, qui en prédit la non-grammaticalité ( à partir d'une relative contenant: *il décrit la part dans son livre). On voit mal enfin comment mettre en place R — A pour des phrases comme (3') signalées par Carlson (1977): (3') Il a pris ce qui apparaît avoir été une grande part à — 1.2.2. Le deuxième type d'arguments concerne certains problèmes d'ac­ cord. Dans une phrase comme (7): (7) Ces femmes sont plus intelligentes que Jean,

180

J.P. MAUREL

l'effacement de l'adjectif est possible, malgré les différences de genre et de nombre, car les deux unités sont non-distinctes (les traits de genre et de nombre étant pour l'adjectif des traits amenés par les transformations que suppose l'accord). Pourtant (8) est impossible: (8) * Marie n'est pas la comédienne que son père était. Pour Vergnaud, c'est qu'il n'y a pas en fait effacement mais promotion, les traits donnés dans le premier cycle (la relative) étant incompatibles avec ceux impliqués par la phrase matrice. Pour nous cependant l'explication est plutôt que dans (8) nous n'avons pas des adjectifs, mais des NP {le/la comédien (ne)) dans des prédications at­ tributives, pour lesquels les traits de genre et de nombre sont inhérents. 1.2.3. Le dernier type d'arguments concerne certains faits de pro­ nominalisation. Schachter oppose ainsi: (9) the opinion of him that John thinks that Mary has et (10) *the opinion of John that he thinks

:

(les termes coréférents sont soulignés). La pronominalisation se fait obligatoirement "en arrière", ce qui est facilement explicable si him dans (9) a sa source dans la relative. Mais J.R. Vergnaud a détruit lui-même cette argumentation. Dans R — A il faut supposer que le NP est d'abord déplacé en tête de la relative (= "wh — fronting") avant "raising" (la promotion directe dans la phrase matrice vio­ lerait des contraintes sur les règles, telles que la contrainte du sujet spécifié ou celle des phrases à temps fini). Or la pronominalisation doit avoir lieu après wh — fronting. La configuration est donc, dans tous les cas, celle d'une pronominalisa­ tion en arrière, phénomène qui doit d'ailleurs de toute manière recevoir une explication, cf: (11) the portrait of his mother always depresses John (Il faut faire appel à des notions comme "C — command", cf. Guéron 1979). 1.2.4. Nous négligeons une autre série d'arguments tels que ceux propo­ sés par Schachter et Carlson (op. cit.) qui en fait reviennent à proposer une R — A pour d'autres types de structures (les "cleft sentences" de Schachter et ' les "amount relatives" de Carlson) et à noter certaines propriétés communes entre ces structures et les relatives classiques, ce qui n'est qu'une argumenta­ tion indirecte.

LES RELATIVES EN LATIN

181

Nous passons aussi sur les difficultés techniques de R — A: lors de la montée du noeud COMP vers le N (vide) supérieur, Vergnaud est obligé de supposer une "dislocation" de l'élément N et de l'élément wh — qui l'accom­ pagne, une domination "non-exhaustive" de ces éléments par COMP; de poser que la relative a son origine dans le spécifieur du N supérieur, puis est extraposée, alors que ce point est discuté dans la littérature — cf. Jackendoff: 1977 — e t c . . Nous voulons seulement conclure cette analyse de R — A, en disant que pour nous, il n'existe aucun argument décisif en faveur de R — A qu'on puisse tirer des faits considérés. Nous allons confronter cette théorie avec certaines données remarquables du latin. 2.

R — A et Les Faits Latins

2.1. Les relatives non-restrictives (ou appositives: Ra) 2.1.1. Quelle que soit la position que l'on ait par rapport à R — A, il est de tradition de considérer que, de toute manière, cette analyse ne peut s'appli­ quer qu'aux relatives restrictives (ici: Rr) et en aucun cas aux relatives nonrestrictives (cf. en particulier Huot 1978). Les arguments invoqués semblent recevoir une confirmation très nette à la lumière des faits latins. On peut trou­ ver en effet, dans les Ra latines, un NP dans la relative, soit identique au NP de la phrase "matrice": (12) ... in castris Heluetiorum tabulae repertae sunt... quibus in tabulis... ratio confecta erat (Caes, B.G., 1, 29, 1), soit différent de ce NP: (13)... in finibus Carnutum, quae regio totius Galliae media habetur... (Caes, B.G., 6, 13, 10) (14) (Amici) sunt... firmi... eligendi, cuius generis est magna penuria (Cic, Lael, 62). Il semble alors qu'il n'y ait pas, en position d'"antécédent", de place vide, comme l'exige R — A, et qu'en tout état de cause le NP contenu dans la rela­ tive ne puisse être promu. D'autre part certains phénomènes d'accords ou de "syllepse" sont diffici­ lement explicables dans R — A: (15) Sulla cum magno equitatu in castra uenit, quos uti cogeret Romae relictus erat (Sall, I ., 95,1) (16) ... omnes Belgas, quam tertiam esse Galliae partem dixeramus (Caes, B.G., 2, 11, 1).

182

J.P. MAUREL

(17) ... seruili tumultu, quos tarnen aliquid usus ac disciplina subleuarent (Caes, B . G . , l , 4 0 , 5 ) . On voit mal en (15) pourquoi, si equitatus a été promu, quos est masculin et pluriel. Il faudrait supposer en (16) que les règles d'accord (ici partern → quam) sont postcycliques, ce qui va à rencontre de ce que soutient Fauconnier (1974). Et on ne voit pas quel NP a été promu en (17). Enfin, il existe en latin ce que Huot appelle des "appositions" de type II", c'est à dire pouvant se rapporter non à un nom, mais à une proposition: (18) ... imbecilii ualent, et quod difficilius dictu est, mortui uiuunt (Cic, Lael.,23). D'une manière générale, on peut avoir des appositions à un élément quelconque X" (où X = N, V, A, Prep.), ce qui ne fournit plus dans la phrase matrice la place vide nécessaire à R — A. 2.1.2. Cependant, ces propriétés ne nous paraissent pas susceptibles de nous permettre de trancher entre R — A et M — A. On ne saurait parler véri­ tablement en effet de M — A pour les Ra, dans le sens vu plus haut, puisqu'il n'y a pas réellement effacement d'un élément non-distinct (effacement "com­ mandé" ou "autorisé" par la présence de la première occurrence), mais coréférence entre un élément X" et un NP (= N"). Ce NP peut avoir alors, soit la forme qui + N et apparemment une référence autonome (si N est lexicalement identique à un N précédent, comme dans (12), nous avons affaire à un cas particulier, le cas général étant représenté par (13)), soit la forme du sim­ ple relatif qui. On peut cependant considérer qu'il s'agit encore d'un cas parti­ culier où N est réduit à certains traits (dans (15) et (17), par exemple: (+ humain), (+ masculin), (+ pluriel), etc...; dans (16), il faut supposer cepen­ dant que les traits initiaux sont effacés par l'accord avec partem). C'est peutêtre d'ailleurs un faisceau de traits que tente de matérialiser eorum dans: (19) ... quorum eorum unus surrupuit currenti cursori solum. (Pl, Tri., 1023) D'une manière générale se pose alors le problème de la possibilité d'une coréférence entre deux unités autonomes. Celle-ci semble autorisée (cf. Milner 1976) par la présence d'un déictique: Chomsky... ; ce linguiste... ou lorsque le deuxième NP est "générique" par rapport au premier: un cheval... ; le quadrupède

183

LES RELATIVES EN LATIN

Même si nous n'interprétons pas l'adjectif relatif comme participant d'une quelconque deixis, les phrases ci-dessus montrent bien que nous nous trouvons dans le deuxième cas; et en particulier lorsque le N est réduit à des traits généraux comme (+ humain) etc... 2.1.3. Nous pensons donc qu'une telle analyse des Ra n'est pas incompat­ ible avec R — A. Nous pensons au contraire que la description structurale de la relative Ra que nous proposons (c'est à dire la présence de N dans la rela­ tive) est celle qu'il faut supposer à la base de R — A. On peut seulement sup­ poser que les conditions structurales de la montée ne sont pas remplies (pas de place vide A dans la phrase matrice) et que donc le NP reste dans la relative. Une analyse identique pourrait alors être proposée pour des phrases comme: (20) Quem in portum numquam hostis accesserat, in eo piratae nauigauerunt (Cic, Suppl., 138), ou même: (21) uideamus quanta sint quae a philosophia remedia ... adhibeantur (Cis, Tusc., 4, 58) sur lesquelles nous reviendrons largement: ici encore R — A serait structuralement bloquée, par absence d'une place vide supérieure. 2.2. Les systèmes comparatifs 2.2.1. Il existe en latin des comparatives formées sur le système de l'ad­ jectif (comme il en existe formées sur le système du nom quantifié, que nous ne considèrerons pas ici), du type de: (22) ...cum esset talis qualem te esse uideo. On a fait l'hypothèse (Milner 1973 et 1978 a) que leur formation repose sur les mêmes principes généraux que celle des relatives. Milner (1978 a) propose d'ailleurs une configuration générale:

.... x...Y-QU... Ls'

s' (+qu)

COMP

J

J

J

(+qu)

pouvant rendre compte de l'ensemble des systèmes corrélatifs dans le cas où X

184

J.P. MAUREL

= Y: si alors X = Y = N", on obtient des relatives. Si X = Y = A", on obtient les phrases du type de (22), dans laquelle A" s'analyse en (Spec, A') et A', et A' est non-spécifié; ce qui s'épelle morphologiquement en talis (corrélateur supérieur) ou en qualis (corrélateur inférieur, dans COMP). On peut imaginer alors que le A" supérieur est vide et que R — A peut s'appliquer. 2.2.2. Il existe aussi des phrases du type de: (23) ... nihil esse tam detestabile quam uoluptatem si on les analyse comme précédemment en faisant de tam (quam) le (Spec, A'), alors une phrase comme: (24) ... non tarn praeclarum est scire latine quam turpe nescire semble incompatible avec R — A, puisqu'on trouve dans la phrase matrice et dans la phrase enchâssée deux adjectifs (lexicalement différents d'ailleurs). On perdrait alors une généralisation importante et qui rend compte de nombreux phénomènes (Milner 1973). On peut cependant sauver la possibilité de R — A en analysant différem­ ment les A" de (23) et (24) et en supposant une structure: A"

(Spec, A')

A'

Q"

(Spec,Q')

A

Q'

(où Q" représente le noeud "quantité", soit, probablement le degré en ce qui concerne l'adjectif). tam (quam) serait alors l'épel morphologique de (Spec, Q') et d'un Q' non spécifié. Et le processus de "wh — movement" aurait pour cible non pas A", mais Q" (ce qui semble nécessaire par ailleurs). On peut alors toujours supposer une place vide sous le Q" de la phrase matrice et sauver R — A, tout en autorisant deux occurrences d'adjectif (en laissant de côté les difficultés

LES RELATIVES EN LATIN

185

techniques concernant l'application de R — A à un élément de type Q"). 2.3. Les relatives restrictives (= R r) 2.3.1. Les considérations qui précèdent, ne permettent donc pas réelle­ ment d'éliminer R — A. On peut semble-t-il sauver cette analyse même dans le cas de certaines Rr telles que: (25) ... est ex eo numero qui semper sancti sunt habiti (Cic, Arch., 31) Ce genre de phrase fait difficulté aussi bien pour R — A que pour M — A, puisqu'il faut supposer deux NP distincts dans la phrase matrice et la phrase enchâssée. Cependant si nous considérons que l'équivalence: ex illo numero = ex illorum numero, proposée par les grammaires classiques, est plus qu'une équivalence sémanti­ que, on peut poser que (25) est équivalente à: (25') ex eorum numero qui..., ou même peut-être à: (25") ex eo numero (hominum) qui, ..., et retrouver des schémas classiques. 2.3.2. Mais l'existence de tours comme: (26) ... diem dicunt qua die ... omnes conueniant (Caes B. G., 1, 6, 4) (27) ... iter in ea loca facere coepit quibus in locis esse Germanos audiebat (Caes, B.G., 4, 7, 1), nous paraît plus décisive. Si, à la rigueur, on peut imaginer pour (26) une inter-prétation avec Ra, cela nous semble impossible (malgré Touratier 1980, p. 188 sq) de faire de ea dans (27) un anaphorique d'un élément antérieur (et donc de la relative une Ra) : ce n'est j amais le cas dans les nombreux exemples équivalents de César. D'ailleurs le procédé n'est pas propre à César; cf: (28) Est causa qua causa simul mecum ire ueritus est (Pl. Epidic., 41) (29) ... ex ea parte homines cuius partis nos uel principes numerabamur (Cic, Quir., 13) (30) ... ad earn causam referendum est quae causa genuit ciuitatem (Cic, Rep., 1,26), (nous empruntons ces exemples à Touratier 1980). Il s'agit certes d'emplois minoritaires ("marqués"), et, semble-t-il, réser­ vés à un registre de langue juridique. On pourrait alors, comme nous l'avons

186

J.P. MAUREL

fait avec les Ra, considérer qu'il s'agit de cas particuliers ou R — A est blo­ quée (par la présence d'un NP supérieur identique). Mais le phénomène sem­ ble différent: les deux NP sont toujours identiques (à la différence des Ra). Et d'autre part, il ne semble pas raisonnable de supposer pour (27) et (27') (= cas "non-marqué"): (27') ... in ea loca... in quibus... deux structures sous-jacentes différentes, l'une supposant une place vide dans la matrice ((27') avec R — A), l'autre supposant une place remplie par un NP. Nour préférons voir dans ces tours un argument contre R — A invitant à un autre type d'analyse pour les relatives en latin. 3.

M — A en Latin

3.1. Cadre théorique 3.1.1. Ce qui, pour nous, caractérise en propre le système des relatives en latin, ce sont les deux faits suivants: 1) A la différence du français et de l'anglais, par exemple, et même du grec homérique (cf. l'emploi du thème de démonstratif comme relatif), il n'y a jamais en latin de formes différentes du relatif permettant de distinguer les Rr et les Ra. 2) On peut trouver, dans tous les types de relatives le N "pivot" dans la relative. C'est le cas des Ra: cf (12), (13), (14); des Rr: cf (26) à (30) ci-des­ sus; c'est le cas aussi des relatives "sans antécédent" du type (20) ou du type (21). C'est ainsi que la relative dans (27) pourrait se trouver dans toutes les configurations suivantes: (27)

( = R r ). . . in ea loca quibus in locis...

(27)

(= Ra)... in finibus Carnutum, quibus in locis...

(27)

: Quibus in locis..., in iis...

Ces régularités invitent à proposer une description syntaxique unitaire du processus de relativisation en latin, et à chercher dans les conditions (syn­ taxiques et sémantiques) d'apparition des relatives, ce qui permet de les dif­ férencier au point de vue de l'interprétation. 3.1.2. Nous choisirons pour décrire la relativisation, le cadre théorique général proposé par Chomsky (1977 b, 1977 c, 1980). On part des règles de base (optionnelles):

187

LES RELATIVES EN LATIN

S' — COMP S COMP -» ± WH - WH → .... e .... On définit ensuite une règle "wh — movement" qui a la forme générale suivante:

±

S'

WH

[COMP

J wh-phrase

S

[COMP α i

...I

wh-phrase

S ±WH COMP

α

J

J J

... s

e ... αi

J

où αi e] est la trace de l'élément déplacé; coindexée avec lui, à laquelle Chomsky suggère d'assigner le trait + COMP A cette structure on peut appliquer une règle d'interprétation qui donne le statut de quantificateur à l'élément (wh — phrase) dans COMP et le statut de variable liée par ce quantificateur, à un élément à l'intérieur de la position occupée par la trace. En ce qui concerne plus précisément les relatives, il faut bien sûr supposer que COMP a la valeur — WH. En ce qui concerne l'inter­ prétation, Chomsky et Vergnaud suggèrent que la règle donne une forme logique (LF) contenant une variable liée (dans les positions ou apparaît une variable liée dans les questions) et y fournit une construction en "tel que". Soit, pour (27) par exemple: (des lieux x tels que "Germanos esse in x"). Pour nous, cependant, il semble que la variable soit, comme nous le ver­ rons, liée par un élément extérieur à S', et qu'en fait la relative S' contienne une variable libre dans S', avec l'interprétation suivante: la relative définit une classe ďobjets. Nous choisirons alors la notation traditionnelle (Grize 1967), soit:

|

x /

R (x) } ,= les x qui satisfont à R(x).

188

J.P.MAUREL

3.2. Relatives restrictives 3.2.1. Nous partirons par hypothèse de la structure suivante:

(Spec, NT) En fait, nous ne trancherons pas le problème de savoir si une telle struc­ ture est ou non une structure profonde, et si S' a son origine sous (Spec, N') (Vergnaud, parmi d'autres), ou si elle est engendrée dans la base sous N" (Jackendoff 1977). Nous renvoyons à la discussion du problème chez ce der­ nier auteur. Nous n'excluerons même pas la possibilité d'une "Chomsky — NP . Ces débats ne nous semblent pas cruciaux adjunction" du type: / \ NP pour ce qui nous concerne.

S'

Ce qui est essentiel, c'est que S'occupe la place d'un modificateur restric­ tif (Jackendoff), c'est à dire contribuant à la référence du N" en restreignant son extension (par l'adjonction d'un prédicat). On peut alors faire intervenir la "contrainte de Kuroda" (1971) (cité par Fauconnier (op. cit.): dans cette configuration, S' doit contenir au moins une occurrence de N" (il est impossi­ ble d'identifier un objet au moyen d'une description qui ne fasse pas allusion à cet objet). On aura donc (avant wh — movement) quelque chose comme:

(On peut supposer que la forme qu — N' du N" de S' est celle que prend ce N" sous le trait (+ wh) qui se transmet par "percolation" à ses divers compo­ sants).

LES RELATIVES EN LATIN

189

On se trouve apparemment devant le paradoxe d'une coréférence enchâssée (Fauconnier): un terme est identique à la structure totale qui le contient. Mais nous avons dit que les deux N"i sont en fait non-distincts. Nous poserons alors (avec Jackendoff) qu'au point de vue de l'interpréta­ tion sémantique, c'est le spécifieur du N" supérieur qui fonctionne comme un opérateur logique (éventuellement un quantificateur), et lie le "wh —phrase" de COMP. Nous pouvons considérer alors chaque cas particulier: 3.2.2. Si N' SUPest spécifié: nous sommes dans le cas des Rr classiques. 3.2.2.1. Normalement N' INF est alors non-répété. Ce N' n'est pas effacé à strictement parler, puisqu'il reste certains traits (qui se communiquent par accord au relatif). Par exemple, dans (27'), quibus reçoit les traits ( — animé), (+ pluriel), e t c . . Le N' est simplement lexicalement nul. Le résultat est quelque chose comme:

3.2.2.2. Dans le cas marqué, le N' I N F e s t répété. On sait (cf. Chomsky 1977 b) qu'il n'est pas exclus par principe qu'on puisse trouver du matériel répété, à condition qu'il y ait pour cela des motivations stylistiques (Chomsky imagine un dialogue du type: Loc A: Jean est plus courageux que Pierre n'est intelligent LocB: Non, Jean est plus courageux que Pierre n'est courageux). Il semble qu'en latin le N' INF soit répété dans le registre juridique et plus précisément avec des noms comme res, causa, dies, locus, pars... : cf. (26) à (30). Ce qui doit correspondre à une interprétation comme: "dans des lieux justement où..." pour (27) "à la cause précisément qui..." pour (30) e t c . . Cette répétition est rendue possible par l'existence générale de relatives (Ra etc...) contenant un N en latin. 3.2.2.3. En ce qui concerne l'interprétation des Rr en général, elle est déterminée par la nature du Spec, supérieur. Avec is par exemple, on a nor­ malement une description définie, dont l'opérateur est noté . Pour (30), par exemple, on a une forme logique:

J.P. MAUREL

190 (30') \x

(x = CAUSA)

(x genuit ciuitatem)

Pour (26) on a (26') : 3x(x = DIES) (omnes conueniant x) etc... 3.2.3. Si N' SUPest non-spécifié: 3.2.3.1. La conséquence est que le N" supérieur est caractérisé par un ensemble de propriétés interprétatives: (1) pour que ce N" soit interprétable, il faut supposer qu'il puisse rece­ voir l'indice du N" inférieur (Milner 1980); (2) en surface, S' n'a pas la forme d'une restriction, la prédication qu'elle contient constituant à elle seule la référence du N"; (3) La variable est liée par un Spec, généralement nul. Il nous semble alors que l'interprétation conséquente a pour effet d'identifier la référence du N" à la classe d'objets définie par la relative S', ce qu'on peut assimiler à la quantification universelle: Vx (R x) 3.2.3.2. Dans le cas le plus rare S' est dans une position syntaxique quel­ conque (en fait on peut essayer de préciser avec Touratier, les positions syn­ taxiques marginales autorisant une telle construction): cf (21) et: (31)

.. .in Carnutes, Andes, Turnos, quaeque ciuitates propinquae his locis erantubi... (caes, B.G., 2, 35, 3).

On peut avoir une idée de l'interprétation assignable à ce type de relati­ ves dans des phrases comme: (32)

uiuere seligentem quae secundum naturam et si quae contra naturam reiicientem (Cic, Fin., 3, 9, 31) {si quae souligne l' "indétermination").

Dans le cas le plus général, S'est en tête de la phrase. Nous sommes dans le cas de ce que nous appellerons les Relativ es-Thèmes. Le modèle en est donné par (20) répété partiellement ici: (20) quem in portum..., in eo... La relative S' définit une classe d'objets (ici réduite à un), et construit par là une référence nominale qui servira de thème à la suite du discours, le nom y étant alors représenté par un élément anaphorique. On peut remarquer que la référence peut être "actuelle", selon les pro­ priétés de S': c'est le cas de (20). Mais elle peut être aussi "générique": (33)

Quam quisque norit artem, in hac se exerceat (Cic, Tu., 1, 41)

LES RELATIVES EN LATIN

191

ou "virtuelle": (34)

haec... qui uideat, nonne cogatur confiteri deos esse? (Cic, N.D., 2, 12)

D'autre part, l'élément anaphorique peut être un NP ayant la même tête N que celui contenu dans S': (35)

... qua in ciuitate... lis aestimata sit, in eadem ciuitate ... esse concessum ut... (Cic, Verr. II, 3, 184),

ou une tête différente: (36)

... fit, quibus quisque in locis miles inueterauerit, ut multum earum regionum consuetudine moueatur (Caes. B.C., 1, 44, 2) Cet élément peut être un "pronom" cf. (20), (33) et:

(37)

... et quae grauissime adflictae erant naues, earum materia... utebatur (Caes, B.G., 4, 31, 2),

ou une anaphore nulle: cf. (34) et: (38)

Quae ciuitates... existimantur, habent... (Caes, B.G., 6, 20, 1) (généralement alors, l'anaphorique est en position de sujet).

Au point de vue syntaxique, on peut supposer que S' est alors dans la position TOP définie (cf. Chomsky 1977 b) par les règles suivantes: S" → TOP S' S' → COMP

S"

s Il est remarquable aussi que l'anaphorique soit systématiquement en tête de la proposition qui le contient, déplacé peut-être dans une autre position TOP. 3.2.3.3. il faut ajouter que: 1) même dans le cas où N' un Spec. réalisé: (39) (40) (41)

SUP

est non-spécifié, on peut parfois trouver

Hos quos uidetis stare hic captiuos duos, ... istant ambo... (PL, Capt., 1,2) ... centuriones ex eo quo stabant loco recesserunt (Caes, B. G., 5, 43, 6) ... Di ilium infelicent omnes qui... leno... immolant (Pl. Poen., 449-51).

On peut imaginer que l'apparition du spécifieur est due à des contraintes

192

J.P. MAUREL

pesant sur la forme de S', et permet d'éviter par exemple ex/quo dans (40) {quo est un locatif), ou infelicent/qui dans (41). 2) il reste à expliquer le phénomène appelé "attraction inverse" illustré par: (42) (43)

Naucratem quem conuenire uolui, in naui non erat (PL Amp., 1009), ou: tu istos captiuos duos heri emi..., his indito catenas (Pl, Capt. , 110-112).

On pourrait voir dans ce phénomène (propre au latin archaïque) une antéposition secondaire d'un N' contenu dans S', et renvoyer donc (42) et (43) au schéma général des relatives-thèmes: on peut en effet comparer (39) et (43), remarquer la constante solidarité du N' et du relatif qui le suit, noter le déplacement de heri dans (43) etc... Cependant, si l'on considère des phrases comme: (44) (45)

ille qui mandauit, eum exturbasti ex aedibus (Pl, Trin., 137) ego qui in mari prehendi, ... ei darei negatis quicquam (Pl, Rud., 1291-2), il semble difficile de partir de: (44')

*qui ille...

ou (45') *qui ego... On peut considérer donc la structure des tours avec attraction inverse comme basique (variante archaïque des tours où le N' est intégré, avec des adjectifs relatifs enclitiques). Le problème est alors celui de la configuration d'accord donnant son cas au N' préposé. Mais ce problème existe tout aussi bien dans des tours comme: (46)

illos qui dant eos derides (Pl. Asin., 527).

Ce qui est clair c'est que cette configuration a pour effet de rendre impos­ sible une interprétation de la relative comme Ra, malgré le N propre de (42) et ille ou ego dans (44) et (45): le N" est ininterprétable sans la relative (cf. la reprise par ei dans (45)). 3..3. Relatives non-restrictives 3.3.1. Emonds (1979) a proposé une formulation générale du processus de "formation de parenthèses", soit une règle attachant à une S' une autre S' contenant un élément [PRO i] (affecté du trait 4- wh) anaphorique d'un cons­ tituant H" de la première phrase, H" étant la projection maximale d'une catégorie lexicale N, V, A ou Prep. La formation des Ra relèverait de ce pro-

193

LES RELATIVES EN LATIN

cessus général. On peut cependant se demander ce qui déclenche alors le processus de relativisation (wh — placement et wh — movement). La règle d'attachement a beau "subordonner" les deux S' à un même noeud E, elle n'implique pas que la parenthèse ne puisse être tout aussi bien constituée par une S' coordonnée contenant un simple anaphorique. 3.3.2. Nous préférons avec Huot (1978) en appeler à une problématique générale de l'apposition. Quelle que soit la place syntaxique que doivent attri­ buer les règles de base aux diverses appositions (cf. Huot, et Jackendoff 1977: pour ce dernier, les éléments non restrictifs sont engendrés sous N'", soit pour les Ra:

... N" S'

, les restrictives étant engendrées

[N''' sous N" :

...

N'

N"

S']), nous considérons que les Ra et les appositions

nominales relèvent de la même analyse. Et plutôt que d'expliquer ces derniè­ res par la réduction des Ra, il nous semble plus intéressant de faire des Ra des N" d'un type particulier, dont la structure serait:

COMP

e

e qu

S

N

et dont les règles d'interprétation seraient les mêmes que celles proposées cidessus dans le cas où N' est non-spécifié. Plutôt qu'une simple coréférence, la position "apposition" impliquerait ici une relation inclusive du type "êtreun", renvoyant donc à la classe d'objets définie par la relative S'; cf. en fran­ çais: (47)

J'ai appris l'arrivée de ton frère, une nouvelle très agréable pour moi,

qui est l'équivalent du latin:

194 (47')

J.P. MAUREL

Audiui fratrem tuum aduenisse, qui nuntius (quod) mihi gratissimus (-mum) fuit;

cf. de même (48) P. Clodius, uir prudentissimus... Ce qui, en évitant de faire de qui un anaphorique, permettrait de ne pas donner pour les Ra des règles interprétatives et des formes logiques fonda­ mentalement différentes de celles proposées pour les Rr. 4. Conclusion Après l'abandon de R — A, notre effort a été de proposer pour tous les types de relatives en latin une analyse (M — A) unitaire tant au point de vue structural qu'au point de vue interprétatif. Nous pouvons tenter maintenant de définir les propriétés caractéristi­ ques de l'élément qui commun à tous ces types de relatives. Au point de vue diachronique, on a posé (cf. Lehmann 1979, Haudry 1979 etc...) que qui était un indéfini. Les structures du type (20) seraient les plus anciennes, avec quem fonctionnant comme un indéfini spécifique: "il y avait un port où ... ; les pira­ tes y naviguèrent". C'est d'ailleurs ce type de paraphrase que Kuroda (68) propose même pour les restrictives ordinaires: ce que serait ici (20'): (20')

... in eo portu in quem...

D'ailleurs la parenté morphologique indo-européenne entre relatifs, interrogatifs et indéfinis serait à mettre en parallèle avec leur caractère com­ mun, au point de vue logique, d'opérateurs liant des variables. Dans: (49)

spectare quos (homines) amauit (= relative)

et (49') spectare quos (homines) amauerit (= inter-indirecte) la seule différence réside dans la valeur — WH (49) et + WH (49') de COMP et dans celle des opérateurs logiques que cela suppose pour les formes logiques de ces deux phrases. Au point de vue synchronique, on peut affirmer que: 1) qui est la forme que prend, dans COMP, sous le trait + wh (par percolation) l'élément (Spec, N'); 2) une relation interprétative peut lier cet élément à un (Spec, N') supérieur, dont la forme non-marquée est épelée is (ou avec un anaphorique postérieur de même forme); 3)

on peut considérer que qui et is sont des éléments non-distincts

LES RELATIVES EN LATIN

195

(ils n'ont pas de traits inhérents pour lesquels is serait marqué + et qui -, ou l'inverse). La seule différence serait le trait ( + COMP) affecté par la transformation de wh — movement. Si l'on suit Milner (1978 a) et que l'on considère que dans les corrélations les deux termes sont marqués (+ wh), on aura: (Spec, N ' ) ( + w h ) (Spec, N ' ) ( +

→ is

wh) (+ C O M P )

→

qui

(On peut penser à la parenté du thème i- et de celui du relatif -yo- en grec, ancien "anaphorique fort" disent les grammaires). 4) Au point de vue de l'interprétation, is et qui ne sont ni "définis" ni "indéfinis" (cf. pour is les tours du type: is uir qui ueniat). C'est la description qui les contient qui est par ailleurs interprétée comme ± définie. - qui définit une classe d'objet par une proposition - is signale ou condense une telle proposition.

BIBLIOGRAPHIE Brame,M. (1968), A new analysis of the relative clause: evidence for an in­ terpretive theory (M.S.) M.I.T. Carlson, G.N. (1977), 'Amount relatives', Language 53, 520-42. Chomsky,N. (1965), Aspects of the theory of Syntax, Cambridge Mass., M.I.T. . (1977 a), Essays on form and interpretation, Amsterdam, North-Holland . (1977 b), 'On wh — movement, in Formal Syntax (Culicover, Wasow, Akmajian eds.), 71-132. . (1977 c), 'Filters and control', Linguistic Inquiry 8, 425-504. . (1980), 'On binding', Linguistic Inquiry 11, 1-46. Dean, J.P. (1966), Determiners and relative clauses, Miméo M.I.T. Emonds,J. (1979), 'Appositive relatives have no properties, Linguistic In­ quiry 10, 211-43. Ernout,A. et Thomas,F. (1951), Syntaxe latine, Paris, Klincksieck. Fauconnier G. (1974), La coréférence: syntaxe ou sémantique? Paris, Seuil. Guéron,J. (1979), 'Relations de coréférence dans la phrase et le discours, Langue française 44. Grize,J.B. (1967), 'Logique', in Logique et connaissance scientifique, Paris, N.R.F.

196

J.P.MAUREL

Haudry,J. (1979), 'Antéposition de la relative en I.E.', La Linguistique 15, 101-11. Huot,H. (1978), 'Appositions et relatives appositives', Recherches Linguis­ tiques 5 Jackendoff,R. (1977), X Syntax, Cambridge Mass, M.I.T. Kleiber,G. (1980), 'L'opposition relative restrictive/relative appositive', L'information grammaticale 7. Kuroda,S.Y. (1968), 'English relativization and certain related problems', Language 44, 244-66. . (1971), 'Two remarks on pronominalization', Foundations of Language 7, 183-98. Lehmann, C. (1979), 'Der Relativsatz vom Indogermanischen bis zum Italienischen', Die Sprache 25, 1-25. Milner,J.C. (1973), Arguments linguistiques, Mame. . (1976), 'Réflexions sur la référence', Langue française 30. . (1978 a), De la syntaxe à l'interprétation, Paris, Seuil. . (1978 b), 'Cyclicité successive, comparatives et cross-over', Linguistic Inquiry 9, 673-93. . (1980). 'La prolepse en grec ancien', Lalies 1. Schachter,P. (1973), 'Focus and relativization', Language 49, 19-46. Touratier,C. (1980), La relative, Paris, Klincksieck. Vergnaud, J . R . (1974), French relative clauses, Cambridge Mass., M.I.T.

COMMENTS ON MAUREL'S PAPER

CHRISTIAN LEHMANN I agree with you on what the principal problems in the analysis of Latin relative clauses are and on the primary, "superficial" analysis of the examples. I found your argumentation quite clear. Our disagreement centers around your solutions to the descriptive problems. I should like to articulate my criti­ cism on two levels: first within the descriptive model you have chosen, second from outside that model. You say that Raising Analysis fails in some cases such as relative con­ structions with two occurrences of identical or distinct NPs inside and outside the relative clause. However, promotion of the head is a transformation

197 whose structural description is only met when in the higher NP of the relative construction there is an Ň dominating a dummy to which the internal Ň may be promoted. Transformations apply only when their structural description is met. If that external Ň does not dominate a dummy, promotion simply does not apply. It would have to be made optional, anyway, in order to account for internal head relative clauses. Thus, I do not see how this could constitute a counterargument to Raising Analysis. This brings up a question concerning the model of generative grammar you have chosen here. In the most recent literature, surface structure is the basis for semantic interpretation. This means that there is no need to provide a uniform deep structure for different surface structures in order to account for their semantic similarity or identity. This makes one ask why it must be either Matching Analysis or Raising Analysis. Why can't you just generate two sorts of deep structures for relative clauses and apply Raising Analysis or Matching Analysis, as the case may be? This brings me to my second level of criticism, from outside the model. When you compare, within one language, constructions which are functionally loosely (!) equivalent, and similarly when you compare such equivalent constructions across languages, you find that they differ structurally in regular ways. Each one fulfills its function in its own right; they are alternatives, all on a par in this respect. For instance, you may form a relative clause putting it in front of its head, or behind, or you may put the head in the relative clause, and there are further possibilities. There is no indication in the data that one of the constructions is more basic than the others, these latter being merely derived. Generative grammar requires us to explain one structure by another struc­ ture, to reduce structures to each other. For instance, Matching Analysis re­ duces the ordinary postnominal relative clause to the postnominal relative clause with repetition of the head noun, which is, in fact, a rare, marked var­ iant. One may ask what is the initial plausibility, the intuitive appeal of a theory of language which requires us to do such things. Does it not appear more plausible to treat all structures on a par and to explain them by the func­ tions they fulfill?

PRAEDICATIVUM* (QUANTIFYING ADJECTIVES A N D ADJECTIVES DENOTING PHYSICAL OR MENTAL STATE)

H. PINKSTER University of Amsterdam SUMMARY I will try to show that there are semantic arguments against considering Praedicativa as Adjuncts of some sort; this holds both for quantifiers and ad­ jectives denoting physical or mental state, albeit in a different way. Quan­ tifiers show some remarkable parallels with quantifiers in other languages. They occur with constituents which do not allow of attribution; they do not occur in predication-like constructions; they show very few restrictions in their occurrence in Latin sentences, however, in comparison with other lan­ guages. They cannot be regarded as embedded predications. Praedicativa consisting of adjectives of physical or mental state, on the other hand, can be regarded as embedded predications which refer to the same time as the main predication. One of the terms of this embedded predication is coreferential with one of the arguments of the main predication ; they seldom go with satel­ lites. 0. Introduction My main interest in this paper will be in discussing the way in which the use of what I call Praedicativa can be related to that of Adjuncts (Adverbials) on the one hand (section 1) and Attributes and Subject-/Object-Complements in the sense of Quirk e.a. (1972) on the other (section 2). More specifi­ cally, I will confine myself to two types of adjectives that may function as Praedicativum, namely quantifiers and adj ectives denoting physical or mental state, the term 'state' being used in rather a loose sense. It is not my intention to present suggestions of how my observations would fit in technically into particular linguistic theories. Part of my terminology, however, derives from

200

H. PINKSTER

Functional Grammar (Dik 1978). Quantifiers are, roughly, those adjectives that give an answer to the ques­ tion 'how many': exemples are ambo ('both'), omnis ('all'),plerique ('most'), pauci ('few'), nonnulli ('some'). Examples of adjectives denoting physical or mental state are maestus ('sad'), laetus ('glad'), ebrius ('drunk'), salvus ('safe'). The reason for confining myself to these adjectives is not, that I con­ sider them in some way representative of praedicativa in general, but exactly because I assume (and know) that there are quite remarkable differences be­ tween the various types of Praedicativa. Also, I do not consider it impossible that within the types of adjectives that I do discuss there may be individual dif­ ferences. Paying special attention to omnis just because the relevant data are readily accessible may blur certain facts about other quantifiers. The follow­ ing sentences exemplify the Praedicativum under discussion. (1) Homines qui gestant quique auscultant crimina, si meo arbitratu licet, omnes pendeant ('If I could have my way, tale-bearers and their hearers would all be strung up', Plt. Ps. 428) (2) At ea subterfugere potis es pauca, si non omnia ('Well, you have a chance to escape a little of it, if not all, Plt. Cap. 970-1) (3) His nonnulli etiam minabamur ('Against these some of us even ut­ tered their threats', Cic. Lig. 33) (4) Amphitruo uxorem salutat laetus speratam suam ('Gladly does A. greet his darling wife', Plt. Am. 676) (5) Ita me Amor lassum animi ludificat ('Yes, Love mocks me in my weariness of soul', Plt. Ci. 215) Before turning to my main topics I will briefly touch upon some general aspects of praedicativa. Firstly, quite some attention has been given in recent times to the ways by which one might distinguish between praedicativa and expressions with iden­ tical formal properties (agreement) such as Attributes, Subject-/ObjectComplements, Apposition, etc. I refer to Fugier (1978: 124-8)1 and Happ (1976: 286-7) for the problem of Attributes, and to Pfister (1973)2 for Subject/ Object-Complements. I assume that there will be contexts in which a decision can hardly be made. On the other hand the statistically predominant cases are those in which there is no reason for assuming a Subject/Object-Complement and in which praedicativa are in a sense obligatory, viz. when there is no con­ stituent with which the adjective could be interpreted as an Attribute (Proper Nouns, often (ex. (4)), furthermore personal (ex. (5)), anaphoric (ex. (2)), in-

PRAEDICATIVUM

201

terrogative, relative pronouns, and in the case of 'zero-pronominalisation' (ex. (3)). For the remainder of this paper I will take it for granted that there will be instances which cannot be decided upon precisely. Nevertheless, there are also enough clear-cut instances and I will confine myself to those. Secondly, the notion 'Praedicativum' is sometimes used to cover quite different phenomena. I only mention Scherer's application of the notion 'Praedicativum' to the so-called ablativus absolutus-construction as well (1975: 194-5). However, the very criterion of optionality which applies to Praedicativa (Pfister 1973) makes it clear that this is not correct. In (6), e.g. invitis cannot be omitted: (6) Vobis invitis ... una libella liber possum fieri ('Despite you I can get freed for a farthing', Plt. Cas. 315)3 Thirdly, I follow Fugier & Corbin (1977) for Latin and a.o. Quirk e.a. (1972: 258 ff.) for English, who have presented arguments for differentiating between certain types of adjectives on other than semantic grounds. In my opinion the usefulness of distinguishing several types of adjectives appears also from juxtaposition patterns such as the following: (7) Ita me gessi ut salvi omnes conservaremini ('My actions have sec­ ured the salvation of you all', Cic. Cat. 3,25) (8) Non vides ut palantes solae liberae grassentur ('Don't you see how they straggle along aimlessly, alone, untended, Plt. Ba. 1137) (9) Ubi sunt isti scortatores qui soli inviti cubant ('Where are those young blades that hate a lonely couch', Plt. Am. 287)4 I now pass on to my main topics. 1. Praedicativa and Adjuncts The first point concerns the usual explanation of the occurrence of adjec­ tives as Praedicativum. In reference grammars such as K.-St. (:I, 234 ff.) and Sz. (:171 f.) these adjectives are said to occur 'instead of adverbs'. A similar view, albeit in a different framework, can be found in Fugier (1978:127,129)5. A familiar kind of argument for assuming some sort of synonymy be­ tween adjectives and adverbs is based on the occurrence of sentences like (10) and (11), in which adjectives and adverbs are coordinated or are part of a parallel construction (more examples in K.-St. I, 238; see also Fugier (1978: 129)).

202

H. PINKSTER

(10) Dum alii trepidi (v.1. trepide), cedunt, alii segniter subeunt ('While the one part in disorder was yielding ground, and the other was slow in coming up', Liv. 27, 12, 15) (11) Hoc primis diebus ... et frequentes et impigre fecerunt ('In the early days they did this both frequently and stoutly', Liv. 36, 23, 4) (et before impigre is lacking in most mss.)6 Apart from instances of coordination one may also point to cases like (12) and (13), where it is difficult to explain why in one case we find the adjective cupidi, in the other cupide. (12) Libentes cupidi que condicionem acceperunt ('They accepted the condition gladly and eagerly', Bell. Af. 90, 2)7 (13) Cupide accipiat ... faxo ('I'll soon make him eager to treat', Ter. Ad. 209) Against this two types of argument there are several arguments for not identifying Praedicativa with Adjuncts and I think that these arguments are stronger. Firstly, I have found cases of juxtaposition of adjectives and adverbs, be it a few cases only: (14) Refertum praeda, quam effuse avidi atque egentes rapiebant ('(a land) filled with livestock which his greedy and impoverished men gathered in from far and wide', Liv. 22,9,3) (15) Cum densiores manipulatim resistere hostibus instituissent ('When they had started to resist the enemies in companies, closely crowded together', Sisenna frg. inc. 31) Two examples of juxtaposition of an adjective and an adverb in the function Disjunct (sentence adverbial) are: (16) Non recte agis ignominiosus ('You have no right to bring this ac­ tion, because you have been deprived of civil rights', Quint. 3,6,75) (17) Hostili in terra turpiter iacuit insepultus ('In a hostile land he lay shamefully unburied', Cic. Inv. 1,108) Secondly, but more importantly, the fact that there are cases in which it does not make much difference for understanding the utterance whether the state of affairs is defined by additional information about one of the partici­ pants (as is the case with a Praedicativum) or about the action involved (Ad­ junct) is no proof of synonymy. Actually, there are cases in which the differ-

PRAEDICATIVUM

203

ence in meaning is quite clear and there are also cases in which an adjective is excluded. K.-St. seem to be aware of this. Thus, in order to illustrate their opinion that adjectives in a sense replace adverbs, they remark: 'Die lateinische Ausdrucksweise ist in solchen Fällen ohne Frage lebendiger, energischer und anschauliger, indem der nähere Umstand einer Handlung zugleich in die Persönlichkeit des Handelnden aufgenommen wird, als: Socrates venenum laetus et libens hausit' (: I, 234). On the other hand, they say that both types of expression are used 'ohne wesentlichen Unterschied der Bedeutung', (: I, 237). An adjective in the function Praedicativum is excluded in impersonal sentences. This explains why we find (18) by the side of (19). (18) Romam inde frequenter migratum est ('Many persons came to live in Rome', Liv. 1,11,4) (19) Conveniunt frequentes prima luce ('They gathered at daybreak in large numbers', Liv. 1, 50, 2) As for semantic differences between adjectives and morphologically re­ lated adverbs, consider the following examples: (20) Unde vos ire dicam ... tarn maestiter vestitas ('Where can you have come from so sorrowly attired?, Plt. Rud. 265) (21) Cum (Philippides) in certamine poetarum praeter spem vicisset et laetissime gauderet, inter illud gaudium repente mortuus est ('When Ph. had unexpectedly won the first prize in a contest of poets and was rejoycing exceedingly, he died suddenly in the midst ofhisjoy',Gell. 3, 15,2) (22) (Quid velimus dicere) laete an severe, abundanter an presse ('What we wish to express) in luxuriant or austere language, at length or with conciseness', Quint. 8, 3, 40) Laetus and maestus are often found as Praedicativum. The related adverbs are very rare. Still, only the adverb is possible in (20), because it is not the mental state of the girls that is commented upon, but the type of clothes they are wear­ ing (see also the following context). In the same vein, in (22) the adverbs characterize the product of speaking, not the speaker. Similarly, in (21) it is the gaudium that is laetissimum. Cum laetissimus gauderet would be rather silly. I also refer to K.-St. 's distinction (: I, 238)prudens et sciens = 'wissentlich und mit Vorbedacht' vs. scienter et prudenter = 'in kluger und geschickter

204

H. PINKSTER

Weise': one may do something sciens and prudens in a very stupid way. (See also Happ (1976: 292) and Shackleton Baily (Comm. on Cic. Att, 8, 3, 4) on the 'stylistic necessity' for Cicero to use invitius in Cic. De orat. 2, 364).8 As a consequence, any suggestions that such adverbs and adjectives are more or less synonymous are mistaken. It is relevant to this question that the examples on which such suggestions are based are limited to Praedicativa that are related to Subject-constituents in the sentence and to active sentences. Consider (23) and (24). (23) Beluarum modo caecos in foveam missos ('They had been driven blindly, like wild beasts, into a trap', Liv. 9, 5, 7) (24) Eadem nos formido timidas terrore impulit ('We poor women were struck with the same terror', Plt. Am. 1079) Nobody would suggest that these adjectives are used 'instead of' adverbs. 9 Furthermore, there are classes of adj ectives where the identification with adverbs is much less attractive than ex. (10) - (13) might suggest. As far as quantifiers are concerned, it is in many cases quite impossible to say that, when occurring as Praedicativum, they are used 'instead of' adverbs. Con­ sider the following examples. (25) Ipsi ex silvis rari propugnabant ('The enemy came out of the woods to fight in small numbers', Caes. Gal. 5,9,6) (26) Hostes crebri cadunt ('The foes fall in heaps', Plt. Am. 236) In the majority of cases replacing the adjective by an adverb would result in a difference in meaning. So in (25) raro instead of rari and in (26) crebro in­ stead of crebri would specify the frequency of the events mentioned, not the quantity of participants in that event (for rari/raro see also Happ (1979:292)). Generally speaking adverbs related to quantifying adjectives either express the frequency of the action expressed by the (main) predication (e.g. plerumque) or the intensity with which the action is performed. Apart from semantic differences adjectives in the function Praedicativum and adverbs also differ with respect to their position in the sentence. Gener­ ally speaking, both quantifiers and adjectives denoting physical or mental state occur more often after the Predicate and the average number of inter­ vening words is smaller than in the case of adverbs (for adverbs see Pinkster (1972: 106-7)). Summarizing, although a description of Praedicativa as Adjuncts of some sort may have some prima facie attractiveness in the case of adjectives that are morphologically related to manner adverbs, on closer inspection

PRAEDICATIVUM

205

there is enough counterevidence to drop that kind of explanation. This is even more conspicuous in the case of quantifiers occurring as Praedicativum. 2. Praedicativum, Attribute and Subject Complement The second point I want to raise concerns the question of how adjectives in the function Praedicativum can be described if compared with attributive adjectives and adjectives occurring as Subject or Object Complement. I will first briefly state the usual way of looking at it and then discuss quantifiers and adjectives of mental or physical state separately. 2.1. This question has been dealt with most intensively in recent times by Heilmann (1973), Happ (1976) and Fugier (1978). A textbook treatment can be found in Sz. He observes (: 172) 'Die Erklärung fur diesen Gebrauch ist darin zu suchen, dass in alien diesen Fällen das Adj. vorzugsweise in prädikativer, selten oder gar nicht in attributiver Verwendung auftritt, also seine aus der Grundsprache ererbte Autonomie beibehält (z.B. Plt. Amph. 1039 perii miser...)'. The explanation given by Happ and Heilman is that adjectives oc­ curring as Praedicativum can be explained as derived from an embedded pre­ dication in which the adjective is Subject Complement and the noun, or pro­ noun, with which the Predicativum is related is the Subject: (27) Galli laeti in castra pergunt = Galli in castra pergunt. (Galli) laeti sunt ('The Gauls gladly enter the camp = The Gauls enter the camp. They are glad'). See Happ (1976: 284). However, as Happ (1976: 294 ff.), and Heilmann be­ fore him, have demonstrated such paraphrases do not always yield acceptable results. An example is (28): (28) Ego vivo miserrimus (T live miserably', Cic. Att. 3, 5) ≠ *Cum miserrimus sim ego vivo Also, invitus, omnis, totus, and other adjectives pose problems. Heilmann and Happ want to solve this problem by looking for other types of para­ phrases, which roughly have the same meaning, and are acceptable as well. 2.2. As far as quantifiers occuring in the function Praedicativum are con­ cerned, there might be some reason to describe quantifiers as embedded pre­ dications, if they occurred either as Subject Complement in a copulative sen­ tence (which they do not) or in some other typical Predication-like construc­ tion, e.g. in the ablativus absolutus-construction. Now, whereas we do find li­ beris ('greedy') and invitus ('unwilling'), of which to my knowledge no occur-

206

H.PINKSTER

rences as a Subject Complement are attested either, in the abl. abs.-construc­ tion, quantifiers such as omnis are excluded: (29) Ne citerior Gallia nobis invitis alicui detur ... ('That Cisalpina Gaul is not given to anyone against our will', Cic. Prov. 39) (30) *Ne citerior Gallia nobis omnibus alicui detur ...10 In view of this there seems to be much more reason to explain the possi­ bility for quantifiers to function as Praedicativum by referring to their func­ tioning as Attributes. In fact, in general linguistics quantifiers are regarded as typically attributive (or, in terms of Functional Grammar, as 'term operators'). 11 Within a transformational generative framework Carden (1976) and Akmajian & Heny (1975: 224 f.) for English and Link (1974) and Vater (1980) for German have drawn attention to pairs of sentences like (31) - (32).12 (31) All the men have found their way home (32) The men have all found their way home In the studies referred to above a sentence like (32) is transformationally de­ rived from (31) by a rule called Quantifier-Floating (or Moving), which sepa­ rates the quantifier all from its attributive position within the Noun Phrase all the men. The concept of 'Floating' has been developed, of course, to explain dif­ ferences in word order between sentences such as (31) - (32), the assumption being that the expressions are synonymous.13 In this respect, the concept is of little use in Latin. There is, however, a parallel between Quantifier-Floating and let us say 'Praedicativum-formation' in Latin in as far as they are both ob­ ligatory in the case of personal pronouns, etc. (see the introduction): in En­ glish, for example, it is not possible to say all the we have found our way home (cf. ex. (31)). The fact, however, that the phenomenon is obligatory in specific cases might be an argument against assuming a common origin for normal At­ tributes and Floaters as well. Another reason for some linguists (e.g. Dik 1981) to hesitate about the equivalence of (31) and (32) and therefore to reject an identical basic (seman­ tic) structure is that in many languages Floating is restricted to a limited number of syntactic or semantic functions. Thus Floating may be restricted to Subjects only. In the case of Latin there seem to be very few, if any, restric­ tions on the syntactic and/or semantic functions in which quantifiers in the function Praedicativum occur. I give a survey of my findings:

PRAEDICATIVUM

(i)

207

Quantifiers function as Praedicativum in relation with arguments of all types, e.g. (a) Subject: ex. (1), more in Merguet, p. 466, a (b) Object: ex. (2), more in Merguet, p. 466, b (c) Recipient: ex. (33) below (d) Complement: exx. (34) - (37)

(ii) We (e) (f) (g)

also find quantifiers with satellites Agent in passive sentence: exx. (38) - (39)14 'Criterion': ex. (40) Beneficiary: I have not yet found examples, but they must be possible

Apart from these cases in which the quantifier is related to one of the so-called major constituents of the sentences, we also find examples of omnis function­ ing as a Praedicativum on a lower level, in relation to an attributive noun phrase (exx. (41) - (42) and as a Complement within a Praedicativum (ex. 43). (33) Quas ... leges ... fuerit impositurus nobis omnibus ('The law which he intended to inflict upon us all', Cic. Mi. 33) (34) Ille corporis motu tantum amorem sibi conciliaret a nobis omnibus ('Merely by the motions of his body he won all our hearts', Cic. Arch. 17) (35) Quae detrimento, maculae ... nobis omnibus esse possint ('That may cause injury and dishonour to all of us', Cic. Ver. 3,144) (36) Uni potius intus ero odio quam hic sim vobis omnibus ('Better to bore one man inside than stay here and bore all of you', Plt. Poe. 922) (37) Ambo ambobus nobis sint obnoxii ('That the pair of them be under obligations to the pair of us', Plt. As. 284) (38) Non multi mihi ab uno sic diligi videntur ut hic ab his universis ('Few men, I think, are so much beloved by a single individual as my client is by this entire community, Cic. Clu. 198) (39) Id agendum nobis omnibus fuit (This should be done by all of us', Cic. Dom. 11) (40) Quid ego communes leges profero, quibus omnibus es exul ('Why need I quote the universal principles of law, by all of which you are an exile', Cic. Parad. 32) (41) Quorum omnium testimoniis de hac Dionis pecunia confirmatum est ('Everyone of these has testified to the facts about this money that Dio paid', Cic. Ver. 2, 23)

208

H. PINKSTER

(42) De omnibus nobis si qui tantulum de recta regione deflexerit ('Let anyone of us all turn aside from the straight path, be it never so lit­ tle', Cic. Ver. 5, 176) (43) Cum multo nobis (v.l. bonis) omnibus veniret iratior quam illis fuerat ('He was becoming much more enraged against us all than he had been against those', Cic. Phil. 3, 4) It seems then, that in Latin there are very few restrictions, if any, on quan­ tifiers occuring as Praedicativum. Notice that the very occurrence of quantifiers with attributive noun phrases is again an argument for not explaining them as more or less equiva­ lent to adverbs (see above). Of course, it also makes the assumption of a com­ mon basis for Attributes and Praedicativa problematic. 2.3. I now pass on to adjectives of physical and mental state. These adjec­ tives occur in 'Predication-like' constructions, and there is no objection, therefore, to describing Praedicativa of this type as embedded predications15 of which one term is coreferential with one of the terms in the main predica­ tion. The presence of the copula in ex. (43) iratior quam illis fuerat also points in this direction. There are, on the other hand, at least two objections to relating this type of Praedicativa to Attributes. One objection is the circumstance that Praedicativa as a rule denote a property of one of the constituents of the sen­ tence, which property is relevant only as long as the state of affairs in which that constituent plays a role exists. Thus, in sentence (4), Amphitruo is glad at the moment he greets his wife, and it is irrelevant whether he is glad his en­ tire life or sometimes glad, sometimes sad. Also in a different type of Praedicativum (44), the people referred to had lost their consciousness while lying on the floor. In the same line parvolam puellam in (45) was true at the time referred to by olim: (44) Plerique ut fusi sine mente ac sine ullo sensu iacerent ('Most of them would be lying about the place fuddled and unconscious', Cic. Ver. 5,28) (45) Nam ego illanc olim ... parvolam puellam ... sustuli ('A long time ago, when she was only a little tot, I picked her up', Pit. Ci. 123-4) As a consequence, it is unlikely for lexemes denoting properties of a more permanent character to occur as Praedicativum, unless the context con­ tains additional information. In (4), for example, instead oilaetus something

PRAEDICATIVUM

209

like calvus ('bald') might be acceptable if either Amphitruo's wife or the audi­ ence knew that Amphitruo was by no means bald before. In the light of this (46) is difficult to explain. (46) (Cerberus) toto ... ingens extenditur antro ('He stretches his bulk all over the den', Verg. A. 6, 423) Attributive adjectives are not necessarily restricted in such a way.16 Notice also, that this entire argument is irrelevant to the problem of quantifiers. Another objection against relating Praedicativa to Attributes concerns the fact that Galli laeti in castra pergunt in the Praedicativum interpretation necessarily refers to all Gauls, whereas in the attributive interpretation this need not be so. Apart from these semantic objections against treating laetus, etc. in the function Praedicativum as dislocated Attributes of some sort, there is also no need for dealing with quantifiers and the adjectives involved in an identical way on account of the conditions on their occurrence as Praedicativum. These adjectives are not found in the same range of constructions as quantifiers are. In my material I have found no counterparts for (41) - (43) (wordgroup level) and only two (Beneficiary) satellites, viz. (47) and (48). (47) Ludos ego hodie vivo praesenti hic seni faciam quod credo mortuo numquam fore ('It's a royal send-off I'll give the old chap today, while he is alive and with us, which is more than he'll ever get when he dies, I'm thinking (Plt. Mo. All) (48) Vivo fit quod numquam quisquam mortuo faciei mihi ('While alive more is done for me than anyone will do for me when I'm dead', Plt. Am. 459) The material contains, for example, no Agents in passive sentences, which confirms an observation by Fugier (1978: 134). The reason for this could be that from the point of view of information structure it would be too much first to specify a state of affairs by a satellite indicating the Agent and furthermore to specify the Agent by a state of affairs in which that Agent is involved. Be­ neficiaries may be an exception to this, since they seem to take up a higher position in the 'semantic function hierarchy' (Dik 1978: ch. 5).17 3. Remaining Problems One might ask, in view of the discrepancies between quantifiers and physical and mental state adjectives, whether it is feasible at all to say that they both function as Praedicativum, as I have done throughout this paper. My

210

H. PINKSTER

reason for doing so is firstly that both types of adjectives are obligatory in cer­ tain contexts, mentioned in the introduction, and secondly that this also holds for other types of Praedicativa such as consul, iuvenis, preposition phrases (cf. ex. (44)), noun phrases in the ablative (so-called ablativus qualitatis) etc. The range of occurrence as Praedicativum seems to be related to the type of constituent involved. A second problem, not to be solved here, 18 concerns the way in which Praedicativa can be dealt with in a grammatical model. More specifically, how can we distinguish between embedded predications func­ tioning as Attribute and those functioning as Praedicativum?

NOTES *) This paper is an interim-report on a research project which has also been sponsored by ZWO (Netherlands Organization for the Advancement of Pure Research). Rina Knaap and Rodie Risselada assisted in assembling examples and testing hypotheses. Simon Dik has drawn my attention to Quantifier Floating. Machtelt Bolkestein has commented on an earlier version of the text. I also point to my Inleiding Latijnse Taalkunde, ch. 5 (Amsterdam, 1982). Pieter Niewint helped me with my English. I am very grateful for their help. Translations of Latin texts are taken from the Loeb editions, whenever possible. 1) Tests by which to distinguish between attributive constructions and constituents functioning as Praedicativum are pronominalisation, relativisation and question. 2) As for the possible ambiguity between Praedicativum and Subject Complement, Pfister has drawn attention to the following distinction: Subject Complements are obligatory, whereas Praedicativa are optional. Thus with the copula esse two constituents are obligatory, one of them being the Subject Complement, e.g. bonus in: Cicero est bonus. The distinction is quite clear in it­ self, but much depends on what we would like to regard as copulative verbs. K.-St. (: I, 15) count the verb nasci ('to be born') as a 'copula-like1 verb, and as a consequence, one might ask whether virum in virum me natam (not natum) vellem ('I could wish I had been born a man', Ter. Ph. 792), is a Praedicativum or a Subject Complement. Me natam vellem alone would be quite uninformative, of course. Especially in poetry there seems to be a tendency to avoid 'colourless' esse. In a similar way one might discuss the relationship between Praedicativa and an Object Complement as Penelopen difficilem procis in Non te Penelopen difficilem procis Tyrrhenus genuit parens ('No Penelope art thou, unyielding to thy suitors, nor of Tuscan parents born', Hor. C.3, 10, 11) Nor­ mally one would not be tempted to take gignere for a three-place verb (comparable to e.g. reddere). In this context, however, Penelopen etc. could hardly be omitted without making the sentence un­ informative. Again, especially in poetry, there is a tendency to create three place verbs. 3) It is not uncommon either to apply the notion of Praedicativum to expressions like primo vere, in medio horto (e.g. Happ 1977: 49) as well. The reason for doing so probably lies in the typical 'par­ titive' relationship in these expressions. However, we find this relationship also in cases where the attributive relationship seems beyond doubt: Si media nox est sivest prima vespera ... tamen est eundum ... ('be it midnight or be it early eve, still you must go, Plt. Cu. 4) Illustriora enim sunt quae in principiis quam quae in mediis causis dicuntur ... ('For

PRAEDICATIVUM

211

statements made at the beginning stand out clearer than those made in the middle of a case', Cic. de Or. 2, 323) Furthermore, as Sommer (1928: 34-7) has observed, we do find the same partitive relationship in other cases as well, where an explanation as Praedicativum is not very likely. Exemples are frigidus annus and: ad earn orationem venio quae est propria vestrae quaestionis ('I pass on to that part of my argument which bears specifically on the issue before you', Cic. Mi. 1). 4) See Pinkster (1972: 108-33) for rules of coordination. Other instances: Hor. C. 3,4. 29: Verg. A. 4,32; 205. It is not always possible, of course, to decide whether we are dealing with juxtaposition or with zero-coordination (the latter e.g. in Cic. Dom. 139). 5) Fugier's test is not confirmed by my material: I have found no evidence that a Praedicativum may correspond with a question word quomodo, or related expressions. Of course, in view of my description later on, it would have been rather embarrassing as well if such an example could be found. As for the example (11), in answer to a question of Ph. Baldi, I observe that coordination of frequentes and impigri would be awkward as well (cf. (12) - (13)). 6) A few other examples Plt. Men. 1073; Cic. Fam. 7, 22; S. Rosc. 123; Imp. Pomp. 47; Att. 2, 19, 5; Sall. Jug. 31, 2; Liv. 2, 30, 12) 7) See TLL s.v. cupidus 1425, 37: 'Praedicative pro adverbio'. 8) I leave out of account the fact that for certain adjectives there are no morphologically related adverbs (see Menge 1965: § 21) and the fact that metrical considerations may explain turbidus (in e.g. Stat. Th. 7, 669) instead of turbide. See Bednara (1906), Engel (1914), Priess (1909), etc. 9) On manner adverbs an adjectives related to Subject-constituents see now Vester (1983). We also find instances of coordination of an Adjunct and Praedicativum in the accusative case (related to the Object-constituent), e.g.falso atque insontem arguis (Plt. Ba. 474). Falso cannot be replaced by an adjective. 10) Lubentes in Plt. Ci. 692 is exceptional. Invitus and lubens are not used attributively either with nouns denoting human beings until Gellius. They are exceptions to what will be said below on physical and mental state adjectives. 11) Quantifiers are treated as predicates in Lakoff (1971). Criticism in Dowty (1979: 19). In Logic, quantifiers are treated as propositional operators. See Allwood e.a. (1977: 6 ff.). 12) See also Chung (1976) and (1978: 189-96) on Indonesian and Tonga, respectively, and Perlmutter (1976) on Portuguese. 13) I leave this assumption as it is. We will see later that Praedicativa like maestus are not synonymous with attributes. 14) For a discussion on the dispensability of Agents in passive sentences see Stein (1979: 126 ff.). 15) Quirk e.a. (1972: 258-64) pay attention to the question which adjectives may function as At­ tribute, which as Subject- or Object- Complement, and which both ways. As far as I know, the type of adjectives under discussion can be used in both ways in Latin, be it that, for example, laetus and maestus are used attributively much less frequently (see also the quotation from Sz. on p. 205). Some adjectives with a negation-prefix (e.g. indemnis, 'suffering no damage1) do not occur attribu­ tively, but this may have a statistical explanation. Also two-place adjectives (e.g. exsors, 'deprived of') occur seldom, if at all, as Attribute. Examples of adjectives that are not or hardly ever used as Subject Complement, at least in relation with Noun Phrases denoting human beings are diversus ('situated apart') and assiduus ('constantly present'). They fall outside the scope of this paper.

212

H. PINKSTER

16) There are exceptions to the rule that a Praedicativum has the same temporal reference as the state of affairs in the main predication. K.-St. (I. 239-40), among others, refer to the so-called 'proleptic use' of participles and adjectives. Examples are: Incute vim ventis submersasque obrue aut age diversos ('Hurl fury into the winds, sink and overwhelm the ships, or drive the men asunder', Ver. A. 1, 69-70) Agunt eum praecipitem poenae civium Romanorum ('He is being swept into mad­ ness by those executions of Roman citizens', Cic. Ver. 1,7) Cf. Verg. A. 11, 484 ipsum pronum sterne solo. Verg. A.3, 141 turn sterilis exurere Sirius agros may be considered attributive. The type is rather familiar in Dutch with predicates with a resultative (terminative) Aktionsart, e.g. Hij verfde de deur groen ('He painted the door green') (See Honselaar 1980: 50-2). I also pass over two other notable exceptions, viz. past and future participles. In examples like Tac. Ann. 6, 24: (Tiberium callidum olim et tegendis sceleribus obscurum huc confidentiae venisse, ut... ('That, once so astute, so impenetrable in the concealment of his crimes, he had attained such a pitch of confidence that') where olim goes with callidum and obscurum, the adjectival phrase cannot be understood as a Praedicativum since olim explicitly refers to a time differing from venisse. More examples in TLL s.v. olim 561, 60; 563, 4 vv. ; s.v. ante 130, 69. 17) See also Kiefer (1980) on the Focus function of so-called 'adverbial modifiers of State' as in 'Did you buy the car in pieces?' Examples of Praedicativa related to arguments are Plt. Am. 895; Au. 62, Ba. 517; Cas. 63; Mi. 333; St. 597; Am. 890; Cic. Ver. 4,29; TLL s.v invitus 234, 26 ff.; Merguet, Lex. Reden, s.v. invitus 749, b. A doubtful case is Cic. Ver. 2, 155, where ab invitis could be regarded as an Agent. In that case we have to assume zero-pronominalization. 18) Several people have drawn attention to this problem in the discussion. a.o. Christian Lehmann. For a transformational approach compare Williams (1980). Additional note: Recently I found two exceptional instances of an adjective of physical state as Praedicativum on the word group level, viz.. Liv. 31,18,6 eorum vivorum and Cic. de Orat. 2,9 eorum...vivorum et praesentium.

BIBLIOGRAPHY Akmajian, A. & F. Heny (1975) An Introduction to the Principles of Transfor­ mational Syntax. Cambridge, Mass., MIT Press. Allwood, J. e.a. (1977) Logic in Linguistics. Cambridge, University Press. Bednara, E. (1906) 'De sermone dactylicorum Latinorum quaestiones I, IF. ALL 14, 317-60; 532-604. Carden, G. (1976) English Quantifiers. New York, Academic Press. Chung, S. (1976) 'An Object-creating Rule in Bahasa Indonesia'. Linguistic Inquiry 7,41-89. . (1978) Case Marking and Grammatical Relations in Polynesian. Austin, Texas U. P. Dik, S.C. (1978) Functional Grammar. Amsterdam, North-Holland.

PRAEDICATIVUM

213

. (e.a.) (1981) Work in progress on 'Typologie van universele kwantifikatie'. University of Amsterdam. Dowty, D.R. (1979) Word Meaning and Montague Grammar. Dordrecht, Reidel. Engel, A. (1914) De Quinti Horatii Flacci Sermone metro accommodato. diss. Breslau. Fugier, H. (1978) 'Les constructions prédicatives en Latin'. Giotta 56, 122143. Fugier, H. & J.M. Corbin (1977) 'Coördination et classes fonctionelles dans le syntagme nominal latin'. Bull. Soc. Ling., 72, 245-273. Happ, H. (1976) Grundfragen einer Dependenzgrammatik des Lateinischen. Göttingen, Vandenhoeck & Ruprecht. (1977) Zur Erneuerung der lateinischen Schulgrammatiken. Frankfurt a/ M, Diesterweg. Heilmann, W. (1973) 'Generative Transformationsgrammatik im Lateini­ schen'. AU, 16, 46-64. Honselaar, W. (1980) Valenties en Diathesen. Diss. University of Amster­ dam. Kiefer, F. (1980) 'Yes-no questions as Wh-questions', in: J.R. Searle, F. Kiefer, M. Bierwisch (eds.) Speech Act Theory and Pragmatics. Dor­ drecht, Reidel, 97-119. Lakoff, G. (1971) 'On generative semantics', in: Steinberg, D.D. & L.A. Jakobovits (eds.) Semantics. Cambridge U.P. Link, G. (1974) 'Quantoren Floating im Deutschen', in Kiefer, F. & D. Perlmutter Syntax und Generative Grammatik. Frankfurt, Athenaion. Menge, H. (1965) Repertitorium der Lateinischen Syntax und Stilistik. Mün­ chen, Hueber 196514. Merguet, H. (1877) Lexikon zu den Reden des Cicero. Jena (4 vols.) Repr. Hildesheim, Olms 1962. Perlmutter, D. (1980) 'Evidence for Subject Downgrading in Portuguese', in: Schmidt- Radefeldt (ed.) Readings in Portuguese Linguistics. Amsterdam, North-Holland. Pfister, R. (1973) 'Prädikationsbezogene Sprachbetrachtung im Lateinis­ chen'. Münchener Studien zur Sprachwissenschaft 31, 151-67. Pinkster, H. (1972) On Latin Adverbs. Amsterdam, North-Holland. Priess, H. (1909) Usum Adverbii quatenus fugerint poëtae latini quidam dactylici. Diss. Marburg. Quirk, R. e.a. (1972) A Grammar of Contemporary English. London, Longman.

214

H.PINKSTER

Scherer, A. (1975) Handbuch der Lateinischen Syntax. Heidelberg, Winter. Sommer, F. (1928) Zum attributiven Adjektivum. München. (Abh. Acad. Wiss. Bayern). Stein, G. (1979) Studies in the Function of the Passive. Tübingen, Günter Narr. Vater, H. (1980) 'Quantifier Floating in German', in: J. v.d. Auwera (ed.) The Semantics of Determiners. London, Croom Helm. Vester, E. (forthcoming) Instrument and Manner Adjuncts in Latin with spe­ cial attention to the use of the ablativus gerundii. Diss. Amsterdam. Williams, E. (1980) 'Predication'. Linguistic Inquiry 11, 203-38.

ABBREVIATIONS: K.-St.: Kühner, R. & C. Stegmann (19122) Ausführliche Grammatik der lateinischen Sprache II. Satzlehre. Sz.: Szantyr, A (1965) Lateinische Grammatik II. Syntax und Stilistik. München, Beck. T.L.L. Thesaurus Linguae Latinae. Leipzig, Teubner.

III. TEXT STRUCTURE

CONVERSATION OPENINGS IN THE COMEDIES OF PLAUTUS

M A R I A E. H O F F M A N N University of Amsterdam SUMMARY This paper presents a first attempt to apply modern conversation analysis to Plautine comedies. Investigations have been restricted to conversation openings in particular. Some parts of the openings are compulsory: the sum­ mons-answer sequential pair, and the identification. Optional are greetings, exclamations, and vocatives. Within the opening, the summons-answer pair has a special status, characterized by a combination of features. In ordinary speech, the parts of a conversation opening are minimalized; in comedy, however, minimalization is deliberately discarded, in order to get comic ef­ fects by, for instance, elaborate and overdone summonses, identifications, greetings and vocatives. In this paper I try to apply modern conversation analysis to the comedies of Plautus. Most of the research until now was based upon conversations in modern languages. Coulthard (1977: 52-92) gives a good summary of Sacks' work. Schegloff (1968) analysed telephone calls; classroom interaction has been described by, for instance, Barnes (1969), Flanders (1970), Bellack (1966) and others, cited in Coulthard (1977: 93). Ten Have (1980) worked on telephone calls as well, and on doctor-patient conversations. Modern drama texts — not recordings of actual theatre productions, but the texts only — have been analysed by Deirdre Burton (1980) of the Birming­ ham English Research Group. She used, among other plays, Pinter's The Dumb Waiter and Ionesco's The Bald Primadonna, that is the English transla­ tion of La cantatrice chauve. But important differences exist between modern drama and Plautus. Firstly, Latin is a so-called dead language. No native speakers have survived to tell us about the correct intonation or the proper use of certain expressions. Secondly, the relation between the actors and the audi-

218

MARIA E. HOFFMANN

ence in comedy differs fundamentally from that relation in modern drama. Plautus makes actors speak explicitly to the audience, and this is done fre­ quently. Kraus (1934) collected all examples of explicit addresses to the audi­ ence. The differences mentioned, however, do not make the application of conversation analysis techniques impossible. Much more work has to be done, of course. I have not yet looked at all examples of conversation open­ ings in Plautus. I have left aside those conversations where a third person breaks into an existing interaction between two people. I have not yet com­ pared Plautus and Terence, which will be rewarding, I guess. And, for the mo­ ment, I did not try to analyse conversations as a whole; I just focused on one feature: the opening, to see whether in the analysis the opening should be given a special status. I will start the survey of what I've been doing by explaining some notions from the field of conversational analysis that I will use throughout this paper. As a conversation opening, I consider, as usually is done, the whole series of conversation turns, from the summons until the mentioning of the intended main topic of the conversation. Every turn has the length of one utterance. Originally, investigators took the turn or utterance as the basic unit of descrip­ tion. But sometimes a turn clearly shows two or more smaller parts that might be considered as basic units: an accepting comment, followed by a directive, as in (1): (1) Pl. Men. 1129 sqq Men. 2 : ... quod tibi nomen est, fecit mihi. Men. 1 : credo ita esse factum ut dicis. sed mihi hoc responde. Men. 2: roga. (Men. 2: ... your name, he gave to me. Men . 1 : I think it happened quite so. But answer me this question. Men. 2: Ask it.) In such cases, we say that there are two moves in the turn. A speaker who wants to start a conversation must be sure the hearer agrees to listen. This is an absolute precondition for a succesful start of a conversation. Many other abso­ lute and relative preconditions with regard to conversations can be listed, but I will leave this for the moment. The person who wants to start a conversation makes a summons, and the other person must give an answer to let the first speaker know that he is willing and ready to listen. Many times being willing and being ready coincide, but remember the well-known restaurant prob­ lems. Having caught the waiter's eye in a restaurant, you may infer from his

CONVERSATION OPENINGS

219

nodding that he is willing to listen to you, but it may take another five minutes before he is really ready to come to your table. As for summons and answer, both may be either verbal or non-verbal. Summons and answer form a sequential pair, that is, they belong together. The summons must be answered, otherwise no interaction is possible. Con­ versation openings consist of: a summons-answer sequential pair, that can be found at the beginning of any conversation. Then, there is an identification, a greeting, a vocative; the order of these elements is not a fixed one, as we shall see. Only the identification is compulsory. In some cases, preliminary re­ marks about health, the weather, etc. are made, and then, at last, the first speaker is expected to come to the main topic of the conversation. The sum­ mons-answer pair, greeting, identification, vocative, these all seem to be finished as quickly as possible in ordinary conversation, according to, for in­ stance, ten Have (1980). I will call this the minimalizing principle. As I cannot speak about all features of conversation openings, let's have a look now at a few examples that have to do with the summons-answer pair and their being verbal or non-verbal, and with identification. In (2), two old gentlemen, Nicobulus and Philoxenus, are standing at the doorstep of the Bacchis sisters. They want to get their sons out of that house of bad reputa­ tion. (2) Pl. Ba. 1117 sqq Nic. : quid dubitamus pultare atque huc evocare ambos foras? Phil. : hau moror. Nic. : heus Bacchis, iube sis actutum aperiri fores, nisi mavoltis fores et postis comminui securibus. 1120 Ba. : quis sonitu ac tumultu tanto nominat me atque pultat aedis? Nic. : ego atque hic. Ba. : quid hoc est negoti nam, amabo? quis has huc ovis adegit? Nic. : ovis nos vocant pessumae. (Nic. : Why don't we go straight up and knock, and call them both out here? Phil. : I don't tarry. Nic. : Hi! Bacchis! Have your door opened this instant, unless you prefer to have door and doorposts smashed with axes. (enter the two Bacchises into doorway.) Bacc. : Who is raising such a din and uproar, calling me and beating on the house?

220

MARIA E. HOFFMANN

Nic. : This gentleman and I. Bacc. : (to her sister) My dear, what does this mean? Who drove these sheep here? Nic. : (to Phil.) They're calling us sheep, the sluts.) From line 1120, it is clear that they not only said (line 1117) that they would knock at the door, but that they actually did knock. Several cases could be mentioned in which people tell they are going to knock at a door, but then, there is no knocking at all, because, for instance, the man who was expected to be inside, turns up outside, or, somebody prevented the visitor from knock­ ing. (Instances: As. 382, Amph. 449, etc.) See also: Duckworth, The Nature of Roman Comedy (1952). At line 1120 the door must open, either just before, or during the words Quis ... to ... aedes. It seems unlikely, although possible, that line 1120 is said by Bacchis inside the house, and that the door is opened between 1120 and 1121. So the summons (1 move) is partly non-verbal: a knock on the door, and partly verbal : Heus, and a vocative, that reinforces the summons. (lines 1118-19) Bacchis' turn can be taken as consisting of two moves. The first one, a non-verbal reaction, is the answer to the summons: she goes to the door and opens it. The second move is the questioning of the iden­ tity of the visitors. The summons is comicized by an elaborately inflated lan­ guage (1118-19), in which an extra-loud knocking is mentioned, and actually carried out. The answer is not very special, rather minimal, and so is the ques­ tion about the identity. But Nicobulus' reply Ego atque hic is no sufficient reply, so the sisters identify them as sheep, and the scene carries on with the sisters addressing each other, and the gentlemen doing the same, joking and feeling insulted, respectively. Notice that there is no greeting. In example (3), Peniculus the parasite already announced in line 465 that he wanted to go towards Menaechmus. At that moment, Menaechmus Sosicles was speaking with the courtisane Erotium, at her doorstep. Peniculus makes a few remarks aside, and repeats the adibo in 486. (3) Pl. Men. 465, 486 sqq 465 Pen. : opservabo quid agat hominem. post adibo atque adloquar. 486 Pen. : adibo ad hominem, nam turbare gestio. Men.2: quis hic est qui advorsus it mihi? Pen. : quid ais, homo levior quam pluma, pessume et nequissume, flagitium hominis, subdole ac minimi preti?

CONVERSATION OPENINGS

221

490

quid de te merui qua me caussa perderes? ut surripuisti te mihi dudum de foro! fecisti funus med apsenti prandio. qur ausu's facere, quoii ego aeque heres eram? Men.2: adulescens, quaeso, quid tibi mecum est rei 495 qui mihi male dicas homini ignoto insciens? an tibi malam rem vis pro male dictis dari? Pen. : post earn quam edepol te dedisse intellego. Men.2: responde, adulescens, quaeso, quid nomen tibist? Pen. : etiam derides quasi nomen non gnoveris? 500 Men.2: non edepol ego te quod sciam umquam ante hunc diem vidi neque gnovi; ... 465 (Pen.

: I'll observe what the fellow's up to. Then I'll go up and have a word with him.

486 Pen. : I'll up to the fellow, for I'm aching for a row! Men.2: (aside) Who's this, advancing on me? Pen. : See here, you rascal lighter than a feather, you base, villainous scoundrel, you outrage of a man, you tricky good-for-nothing! 490 What have I ever done to you that you should spoil my life? How you sneaked off from me at the forum a while ago! You've interred the luncheon, and I not there! Why did you dare do it, when I was as much its heir as you? Men.2: Sir, what have you to do with me, please, that you 495 abuse me, a stranger? Or do you want to be given a bad time in return for this bad language? Pen. : Oh Lord! Again! You've given me that already, I perceive. Men.2: Please answer me, Sir, what's your name? Pen. : What? Making fun of me, as if you didn't know my name? 500 Men.2: Good Lord, I have never seen you or known you before this day, so far as I know; ...) From the dialogue it is clear that Peniculus goes towards Menaechmus with­ out speaking a word: a non-verbal summons. An identification scene follows: Menaechmus Sosicles does not know Peniculus at all, but on the other hand, Peniculus supposes Menaechmus Sosicles to be his twin brother Menachmus. Identification between people that know each other takes place by recogni-

222

MARIA E. HOFFMANN

tion in a few milliseconds. So Peniculus immediately starts to accuse Menaechmus Sosicles for having had luncheon without his parasite — the in­ tended topic of the conversation. The following re-identification fails com­ pletely, and is interwoven with a lot of information Menachmus Sosicles should have known if he were the twin brother, but that is new to him now. We see that the minimalizing principle of telling each other: I am A, and I am B, has been discarded by Plautus deliberately. This deviance causes the comic ef­ fect. The slave Sceparnio in (4) is digging clay for the making of new roof tiles, because the house has been damaged by a storm. (4) ?\.Rud. 97sqq Dae. : heus, Sceparnio! Scep. : qui nominat me? Dae. : qui pro te argentum dedit. Scep. : quasi me tuom esse servom dicas, Daemones. 100 Dae. : luto usust multo, multam terram confode. (Dae. Scep. Dae. Scep. 100 Dae.

: Hey, Sceparnio! : Who's calling me? : The man that paid out money for you. : That's a good deal like calling me your slave, Daemones. : We need a lot of clay, dig deep.)

Sceparnio's master Daemones makes a verbal summons by means of an excla­ mation, Heus, and a vocative, Sceparnio. Whether he goes towards Sceparnio cannot be guessed from the text. Sceparnio anyway must be in such a position that he doesn't see Daemones. Either bent over his spade while digging, or sit­ ting down in a bent position? Now several possibilities exist: Sceparnio may recognize his master's voice, but is unwilling to react properly, and to say "What do you want, master?", or he really doesn't recognize the voice. Any­ way, apparently Sceparnio does not lift his head when answering the sum­ mons. This answer is a question about the identity of the summoner. Here is a problem in the description of the conversation: what exactly is the answer to the summons? Is it the being attentive itself? Can this be separated from the question about the identity of the caller? After that periphrastic way of iden­ tification, Sceparnio jokes about it, and, by using Daemones' name, makes sure the identification is satisfactorily carried out. Then, Daemones comes to the main topic, and, having the right to command his slave, starts with a direc­ tive.

CONVERSATION OPENINGS

223

From these examples we see that always a summons is needed. This is true for a conversation with a single person, but also for addressing an audi­ ence or a god — and the other person has to give some sign that he is listening. An identification, too, is compulsory. But a greeting is not. When there is a greeting, it may be short, as in Mercator 283-4, where two old gentlemen say: Lysimache, salve, and Euge, Demipho, salveto! But a greeting, too, may be extended with a humorous effect, as in (5): (5) PL Amph. 676 sqq 676 Amph.: Amphitruo uxorem salutat laetus speratam suam, quam omnium Thebis vir unam esse optumam diiudicat, quamque adeo cives Thebani vero rumiferant probam. valuistin usque? exspectatun advenio? Sos. : hau vidi magis. 680 expectatum eum salutat magis hau quicquam quam canem. Amph. : et quom [te] gravidam et quom te pulchre plenam aspicio, gaudeo. Alc. : opsecro ecastor, quid tu me deridiculi gratia sic salutas atque appellas, quasi dudum non videris, ... 676(Amph. : Gladly does Amphitryon greet his darling wife, whom her husband judges to be the one best lady in all Thebes; yea, and justly do the citizens of Thebes bruit her virtue. Have you been well all this time? Are you glad to see me? Sos. Glad? None more so! 680 Welcomes him about as warmly as she would a dog! Amph. : I'm glad to see you pregnant, and far gone with child. Alc. : Good gracious! Why are you making fun of me with all these greetings and salutations, as if you had not seen me a little while ago, ...) Alcumena considers the real Amphitruo's elaborate greeting as a joke, be­ cause she has been visited only a short time ago by the false Amphitruo, that is, by Jove. Getting the attention of a big audience is difficult. By coming to the theatre people show their willingness to listen, but when the play starts, they have to be ready to listen, as well. In some cases, the prologus simply says: advortite animum. But have a look at the sophisticated way of getting the audi­ ence's attention in example (6):

MARIA E. HOFFMANN

224

(6) Pl. Amph. 1 sqq, 15 sq. Merc. : ut vos in vostris voltis mercimoniis emundis vendundisque me laetum lucris adficere atque adiuvare in rebus omnibus, et ut... 15

ita huic facietis fabulae silentium itaque aequi et iusti hic eritis omnes arbitri.

(Merc. According as ye here assembled would have me prosper you and bring you luck in your buyings and in your sellings of goods, yea, and forward you in all things; and according ... 15

then in such degree will ye keep still while we are acting this play and all be fair and square judges of the perfor­ mance).

The first part of the prologue is a rather heavy, complicated description of Mercury's tasks, in a series of subordinate clauses, possibly spoken in a loud, solemn voice. Then, in line 15, in the main clause, the prologue asks the public to be silent. Afterwards similar pompous parts alternate with summonses to listen. We may imagine people gradually becoming quiet, hearing the solemn voice, and saying to each other: "Hush, the play is to begin!" Remember the summons cannot be made by going towards the audience, nor by calling the names of noisy people. Speaking to a god is a very stereotypical activity: you can't go towards a god, nor can you say: Heus, te voco!, for how would a god know that it's not a human being that is called, but a god? And ... which god? Let's have a look at example (7): Pl.Capt. 976 (7) Heg. : serva, Iuppiter supreme, et me et meum gnatum mihi. (Heg. : Jove Almighty, save me and save my boy for me!) The only possible way to summon a god is to call the god's name. Apparently, people know they don't have to wait for the god to answer the summons. Gods are supposed to listen always and everywhere to anybody calling them. There­ fore, people may even put the god's name later in the sentence, and not in the first position.

CONVERSATION OPENINGS

225

I will come now to the conclusions that may be drawn from this analysis of Plautine conversation openings. The first thing we see is that modern ways of approaching a text by means of conversational analysis clearly apply to Latin texts as well. Another conclusion may be that the sequential pair summonsanswer has a special status, that is marked by the following combination of re­ quirements: a)

b)

c)

absolute preconditions concerning speaker, hearer and utterance, are required. The speaker must ask for the hearer's attention, the hearer must agree to start the conversation, the summons must be perceptible, and so must the answer be. the summons-answer-pair is compulsory in any conversation, be­ cause it serves to fulfil the above mentioned absolute precondi­ tions. It is the first thing to be done except when addressing a god. the summons-answer-pair has a stereotypical content, that may be paraphrased by "Hey, I want to say something to you. Are you wil­ ling and ready to listen?", followed by the hearer's: "Yes, I am wil­ ling and ready to listen to you". Just because the content is always the same, non-verbal means are as effective as verbal ones. Going towards somebody and the knock at the door have already been mentioned, and we might add the telephone bell, touching some­ body's shoulder, or catching his eye.

Still other conclusions, specifically connected with Plautine comedy, are : 1)

2)

the going towards somebody, that serves as a summons, many times is mentioned explicitly: Adibo hominem. This way of sum­ moning a hearer until now has not been given sufficient attention in papers on conversation openings. It is kind of a surprise that Latin comedy supplies this information abundantly. the normal pattern seems to be minimalizing the opening proce­ dures of a conversation. Now, as we all know, overdoing activities is a feature of comedy: drunkenness, behaving miserly, beatings, and so on, but we see also that even the smallest parts of conversa­ tion openings are distorted, extended, maximalized, in order to add comic force to the play. Not all parts of the opening, however, seem equally suited for extension to a comic scene. So examples of overdone summonses may easily be found: extremely loud and frequent knocking at a door. But until now I have not found any example in which the answer to the summons is overdone. Greet-

226

MARIA E. HOFFMANN

ings and vocatives are comicized, too, and this is especially the case with identification. Preliminary remarks like "Isn't the weather fine?" etc. are seldom used in the comedies of Plautus. The in­ trigue is so complicated that no time should be lost by the actors saying things irrelevant to the plot.

NOTE:

The translations of the comedy fragments are taken from Paul Nixon's Loeb edition, with a few minor changes.

BIBLIOGRAPHY Burton, Deirdre (1980), Dialogue and Discourse, a sociolinguistic approach to modern drama dialogue and naturally occurring conversation. London, Boston and Henley, Routledge and Kegan Paul. Coulthard, Malcolm (1977), An Introduction to Discourse Analysis. London, Longman. Have, Paul ten (1980), 'Openingssekwenties', in: Fooien, Ad, Jan Hardeveld en Dick Springorum (red.), Lezingen van het Congres ConversatieAnalyse, gehouden op 17, 18 en 19 dec. 1979 in Nijmegen, 63-84. Groningen, Xeno. Kraus, Walther (1934), '"Ad Spectatores" in der römischen Komödie', Wiener Studien 52, 66-83. Schegloff, E.A. (1968), 'Sequencing in Conversational Openings', American Anthropologist 70/6, 1075-95.

IV. SEMANTICS

U N E ANALYSE SEMANTIQUE ET PRAGMATIQUE DES PRONOMS INDEFINIS EN LATIN

A. O R L A N D I N I Université de Bologna, Istituto di Filologia Latina e Medioevale RESUME Les pronoms interrogatifs et relatifs wh- participent d'une même nature sémantique: dans le language courant ils jouent le rôle de ''quantificateurs propositionnels". La formalisation du quantificateur wh- interrogatif (dans un contexte epistémique) et du quantificateur wh- relatif (en dehors de ce con­ texte) est semblable à celle de l'article the dans une description définie (cf. Cushing 1977: 79). Les pronoms indéfinis représentent la partie présupposée tant des interrogatifs que des relatifs. La représentation sémantique des pro­ noms indéfinis singuliers par le quanticateur existentiel Ex tient compte des deux lectures possibles de re ou de dicto (transparente ou opaque), selon le scope différent du quantificateur dans la proposition. Parmi les indéfinis à lec­ ture transparente une ultérieure subdivision est opérée en introduisant un cri­ tère pragmatique pour distinguer le pronom spécifique quidam de l'indéfini non-spécifique aliquis (cf. Kasher-Gabbay 1976). Le pronom quisquis joue le rôle de liaison entre les indéfinis singuliers et les indéfinis relativisés pour lesquels nous proposons la formalisation par le quan­ tificateur universel Vx. Dans des langues très différentes entre elles, indo-européennes ou non indo-européennes, on peut vérifier des affinités morphologiques dans le triple système pronominal: interrogatif, indéfini et relatif. Les pronoms interroga­ tifs et les pronoms indéfinis ont des réalisations bien semblables entre eux en allemand (wer, welcher) et même en grec et en latin, où les pronoms ont la même réalisation lexicale ( quis, quid). En grec aussi bien qu'en latin l'interrogatif est marqué par un accent tonique, alors que l'indéfini est enclitique.

230

A. ORLANDINI

De même les pronoms interrogatifs et relatifs présentent la marque mor­ phologique wh- en anglais en w- en allemand, alors que le latin emploie un thème qui pour ces pronoms. Ces diverses relations morphologiques dépen­ dent de la grammaticalisation d'une propriété sémantique commune. L'ana­ lyse que nous allons développer a pour but de déterminer les liens sémanti­ ques que présente la forme logique de ces pronoms, compte tenu, dans cer­ tains cas, de leur développement historique dans la langue latine. Nous voyons les pronoms interrogatifs et relatifs wh- participer d'une même nature sémantique. Ces pronoms dans le langage courant jouent le rôle de quantifi­ cateurs. 1 Compte tenu de la distribution syntaxique différente de ces pro­ noms, on peut remarquer que les pronoms indéfinis représentent la partie présupposée tant des interrogatifs que des relatifs. Une who- question non rhétorique, par exemple: (1) Who lives here? présuppose en effet la proposition: (2) Someone lives here tandis que la réponse à cette question est représentée par une liste exhaustive des éléments de l'ensemble (les individus qui vivent ici). Grâce à cette réponse le locuteur dira qu'il connaît: (3) The one who lives here. Les relations entre les questions et les énoncés déclaratifs pourraient éclairer la nature sémantique des rapports qui existent entre les pronoms interrogatifs et les pronoms relatifs. On peut remarquer aussi qu'une proposition relative telle que: (4) The one who lives here scrubbed the floor présuppose la même proposition: (2) Someone lives here. Il s'agit dans le cas de la proposition (4) d'une relative determinative, ayant pour antécédent un pronom tel que the one, qui doit nécessairement recevoir une caractérisation (cf. Touratier 1980: 381). Cette relative semble tirer son origine de la juxtaposition de deux propositions indépendantes, dont la pre­ mière (la présupposition où se trouve le pronom indéfini) est une déclarative construite comme un prédicat qui dénote une propriété, et la seconde propo­ sition rappelle anaphoriquement le pronom indéfini:

PRONOMS INDEFINIS

231

(5) Someone lives here, he scrubbed the floor. La grammaticalisation des relations qui existent entre les deux propositions exige la subordination. Dans une phase intermédiaire (cf. Lyons 1977: 760) on aurait la proposition presque-anglaise: (6) The someone lives here scrubbed the floor. Nous croyons donc qu'on peut analyser les pronoms interrogatifs wh- aussi bien que les pronoms relatifs comme "quantificateurs propositionnels". Par ce terme on entend que la représentation sémantique de ces pronoms exige une présupposition formalisée par le quantificateur existentiel ( 3x) et traduite dans le langage courant par le pronom indéfini someone. Cette représentation sémantique exige aussi une assertion formalisée par le quantificateur universel (V x), car la quantification concerne un ensemble. La formalisation du quanti­ ficateur wh- interrogatif (dans un contexte épistémique) et du quantificateur wh- relatif (en dehors de ce contexte) est semblable, à notre avis, à celle que S.Cushing (1977: 79) a proposée pour l'article the dans une description définie telle que: (7) The theorem that Fermat discovered has been proven qu'on peut formaliser ainsi: (8) a)(3x)(Theorem (x); Fermat-discovered(x),-(E),-(E),-(E),-(Ey)(Theorem(y); Fermat-discovered (y), y=x)) b) (Vx)(Theorem (x), Fermat-discovered (x); Proven (x)). où a) représente la présupposition: qu'il y a un théorème découvert par Fermat et qu'il est unique, et où b) représente l'assertion: que tous les théorèmes découverts par Fermat ont été démontrés. Il faut d'abord préciser que l'analyse présuppositionnelle est propre à l'emploi non généralisant de l'article défini the, car on ne pourrait pas l'utili­ ser pour un énoncé qui semble ne pas avoir de présupposition d'existence: (9) The beaver builds dams qu'on formalise par le quantificateur universel: (10) All the beavers build dams c'est-à-dire: (11) If x is a beaver, then it builds dams. On trouve la même distinction pour le pronom relatif, car l'analyse présuppo­ sitionnelle n'est pas propre à un énoncé universel tel que:

232

A. ORLANDINI

(12) He who breaks shall pay. Il faut remarquer aussi, en rapprochant le pronom relatif de l'article défini2, qu'il y a des langues, par exemple le grec ancien et l'allemand, où le der, die, das). relatif a été concurrencé par le thème de l'article ( L'évolution du pronom indéfini en pronom relatif, historiquement attes­ tée en hittite dans le cas du pronom kwis (cf. Hahn 1933) coïncide sémantiquement avec le passage de la quantification existentielle à la quantification dans un ensemble, avec une présupposition d'existence et aussi d'unicité (dans le cas des pronoms singuliers). On va maintenant donner un panorama des principaux pronoms indéfinis sin­ guliers, analysés en rapport à la présupposition d'existence, et au scope diffé­ rent du quantificateur existentiel (3x), grâce auquel ils sont représentables sémantiquement, et enfin en rapport aux préconditions pragmatiques du locuteur. Ensuite on analysera les pronoms indéfinis relativisés, qu'on peut formaliser par le quantificateur universel (Vx). On va d'abord distinguer parmi les pronoms singuliers un pronom indé­ fini spécifique, qu'on trouve lorsque le domaine du quantificateur est repré­ senté par un individu particulier déjà connu par le locuteur (angl. a certain) et un pronom indéfini non-spécifique, lorsque l'individu est choisi dans une liste disjonctive. L'ambiguïté qui existe entre la lecture spécifique et la non-spéci­ fique n'est pas représentable par un scope différent du quantificateur existen­ tiel (3x), parce que de nature pragmatique (cf. Kasher-Gabbay 1976:150). Il s'agit pour le locuteur, dans le cas de la lecture spécifique, d'une connaissance non anaphorique de l'individu dénoté par l'expresion indéfinie, dont on veut donner une description seulement partielle. La lecture spécifique d'une de­ scription indéfinie sera pour le locuteur tout à fait pareille à la lecture referen­ tielle d'une description définie, même si elle restera toujours indéfinie pour l'interlocuteur. Celui-ci connaît l'objet dénoté par le pronom indéfini spécifi­ que en tant qu'il appartient à un certain groupe, mais il n'est pas à même de l'identifier ultérieurement. C'est dans cette situation pragmatique qu'on emploie en latin le pronom indéfini quidam: (13) Cic. Cluent. 63,177 Quidam ex advocatis, homo et honoribus populi ornatus et summa virtute praeditus, intelligere se dixit. Du point de vue de la sémantique il y au contraire une distiction qui peut être représentée par le scope différent du quantificateur existentiel, mais cette distinction ne concerne pas l'ambiguïté qui existe entre le pronom spécifique

PRONOMS INDEFINIS

233

et le pronom non-spécifique. Il y a deux lectures possibles pour une proposition telle que la (2): si le locu­ teur dit de quelqu'un qu'il a la propriété de vivre ici -qu'il le connaisse ou nonon va parler d'une lecture de re; dans ce cas le scope du quantificateur existen­ tiel est plus étendu. Mais si l'on dit seulement que l'ensemble des individus qui vivent ici n'est pas vide, nous appellerons cette lecture de dicto; dans ce cas-ci le scope du quantificateur existentiel est plus réduit. L'interprétation referen­ tielle se trouve dans un contexte "transparent", c'est-à-dire vrai-fonctionnel, alors que pour l'interprétation de dicto on peut parler d'un contexte "opaque" (cf. Quine 1960: 141 ss.). C'est la signification spécifique qui donne au pro­ nom quidam la possibilité de recevoir seulement la lecture de re (c'est-à-dire "transparente"). Une telle lecture exige la présupposition d'existence, qui est déterminée bien souvent par la nature du prédicat et aussi par des modes vraifonctionnels, dans la signification logico-standard du terme, comme l'indica­ tif dans les propositions déclaratives. En effet nous pouvons remarquer que le pronom quidam n'est jamais employé sans la présupposition d'existence; par exemple il ne se trouve pas dans les énoncés universels, dans les propositions suppositives, dans les potentiels, parce qu'une lecture opaque de ce pronom serait incorrecte. 3 La langue anglaise, aussi bien que la française et l'italienne traduisent le pronom indefini spécifique quidam par a certain, un certain, un certo, et ce n'est pas par hasard, car dans le latin tardif quidam a été concur­ rencé par certus. Toutefois on emploie aussi bien souvent someone, quelqu'un, qualcuno, qui recouvrent l'aire du pronom quidam et celle du pro­ nom aliquis, lequel représente l'indéfini latin non-spécifique.4 On peut donner du pronom aliquis une lecture transparente, ou une lecture opaque, selon qu'il exige ou non la présupposition d'esistence. Aliquis pré­ sente une lecture transparente par exemple dans les propositions déclaratives au parfait: (14) Rhet. Her. 4,13 'Ergo aliquid fuisse necessum est. Quid aliud, nisi id, quod dico, potest esse?' Alors que le pronom aliquis reçoit une lecture opaque, il a néanmoins chez le locuteur une précondition pragmatique positive: parce qu'il croit que pourrait exister un individu satisfaisant les conditions posées par le prédicat. Cet emploi du latin aliquis répond à la valeur sémantique du pronom indéfini anglais someone et du français quelqu'un: (15) Ter. Hec. 651 hunc videre saepe optabamus diem / quom ex te esset aliqui qui te appellaret patrem.

234

A. ORLANDINI

Parfois c'est le pronom anglais anyone emphatique qui traduit aliquis lorsque sa signification est opposée à celle de nullus: (16) Cic.Cato 44 si aliquid dandum est voluptati. La même analyse sémantique que nous venons de proposer pour aliquis peut être utilisée pour le pronom quispiam. Cet indéfini possède aussi bien la lecture transparente que la lecture opaque, mais celle-ci est sans doute plus fréquente. L'emploi de quispiam est très rare dans les propositions déclarati­ ves, en particulier au parfait (où on admet seulement la lecture transparente): (17) Cic. Mur. 62 Dixisti quippiam: 'fixum et statutum est'. Au contraire le pronom quispiam a été classiquement plus employé dans les propositions subordonnées de caractère hypothétique: (18) Cic. Tusc. 2,12 si grammaticum se professus quispiam barbare loquatur. Dans ces propositions il n'y a pas la présupposition d'existence, mais la pré­ condition pragmatique du locuteur est orientée vers un doute positif, c'est-àdire qu'il croit possible l'existence d'un individu dénoté par l'indéfini. Quant au pronom quis, il possède presque seulement la lecture opaque. On le trouve couramment dans le même domaine sémantique que le pronom anglais anyone non emphatique. Il s'agit: —d'énoncés conditionnels universels: (19) Cic. Parad. 44 filiam quis habet: pecuniast opus; - d e propositions hypothétiques: (20) Nep. Paus. 4,4 locum fecerunt sub terra, ex quo posset audiri, si quis quid loqueretur. On rencontre également le pronom quis: -après nē final ou complétif : (21) Prop. 1,3,30 neve quis invitam cogeret esse suam; -dans les interrogations rhétoriques négatives: (22) Cic. Lael. 41 Num quid simile populus Romanus audierat? -dans les défenses: (23) Cic. Leg. 3,9 eundem magistratum, ni interfuerint decem anni, ne quis capito, eqs.

PRONOMS INDEFINIS

235

La précondition pragmatique du locuteur est en général neutre ou bien néga­ tive en ce qui concerne l'existence de l'individu dénoté par cet indéfini.5 Parmi les indéfinis latins indiquant l'existence, l'equivalent direct négatif du pronom aliquis est quisquam. Au contraire de quis, qui reçoit presque exclusivement la lecture opaque, quisquam possède aussi la lecture transpa­ rente, bien qu'elle soit très rare. Toutefois ce pronom est assorti d'une pré­ condition pragmatique négative: il ne s'agit pas seulement d'un doute sur la possibilité de l'existence de l'individu en question, mais surtout d'une implica­ tion du locuteur, selon laquelle il faudrait que cet individu n'existe pas. On peut donc attribuer à quisquam une lecture transparente, par exemple lorsqu'il se trouve dans des propositions au parfait, souvent introduites par des prédicats tels que: miror, admiror, queror, reprehendo, accuso, moleste fero: (24) Cic. Verr. 3,16 Illud reprehendo et accuso, cur in re tam vetere, tam usitata, quicquam novi feceris. La precondition pragmatique justifie qu'on puisse rencontrer quisquam dans des propositions positives, alors qu'en réalité il dénote quelqu'un qui pour le locuteur ne devrait pas exister: (25) Cic Catil. 1,6 Quam diu quisquam erit qui te defendere audeat, vives. Au contraire l'emploi de quisquam est très limité dans le cas où il n'y a pas de précondition négative du locuteur, par exemple dans les propositions suppositives (et par conséquent neutres au regard de la présupposition d'existence): (26) Cic. Epist. 6,14,1 si quisquam est timidus in magnis periculosisque rebus..., is ego sum, eqs. Mais le pronom quisquam est courant surtout dans un sens négatif pour la présupposition et c'est pourquoi qu'il a un emploi très étendu dans les pro­ positions négatives, dans les défenses, dans les interrogations rhétoriques négatives aussi bien qu'au second terme d'une comparaison: (27) Cic. Verr. 4,123 ac videte quanto taetrior hic tyrannus Syracusanis fuerit quam quisquam superiorum. Parmi les indéfinis singuliers qu'on peut formaliser par le quantificateur existentiel, il nous reste à examiner le pronom quisquis, qui joue le rôle de liai­ son entre les indéfinis que nous venons d'analyser et les indéfinis relativisés qu'on appelle "pluriels". A l'origine quisquis était un pronom indéfini, et on peut le déduire de la comparaison entre quisquis et le pronom correspondant

A. ORLANDINI

236

hittite kwiskwis, au début indéfini et ensuite relativisé (cf. Hahn 1933: 30 ss.). Quisquis non relativisé a la même signification que le pronom quivis,quilibet, qu'on peut traduire en français par n'importe qui, en anglais par anyone, et dénote un individu appartenant à une liste non exhaustive, nécessairement ouverte, caractérisée par la liberté de choix et l'inachèvement (cf. Vendler 1967: 80 ss.). Ce pronom n'a pas la présupposition d'existence et par consé­ quent il n'est presque jamais employé dans les propositions déclaratives ayant un prédicat d'aspect imperfectif: (28) Liv. 41,8 liberos suos quibusquibus Romanis mancipio dabant. Quisquis relativisé envahit le domaine sémantique des pronoms quicumque et quisque, en prenant la valeur du pronom anglais whoever: (29) Cic. Fin. 1,64 quicquid porro animo cernimus, id omne oritur a sensibus. La quantification se réalise dans une liste fermée, mais il n'est pas encore pos­ sible de nommer cette quantification "présuppositionnelle". Il y a entre quis­ quis relativisé et qui relatif non généralisant la même différence qui existe entre whoever et who, quiconque et qui. Le pronom indéfini relativisé mar­ que, du point de vue de la syntaxe, une subordination, au contraire des indéfi­ nis authentiques, mais il n'a pas la présupposition d'existence. En effet on pourra analyser la proposition: (30) Whoever says it, fails de cette façon: (31) If anyone says it, then he fails. C'est-à-dire comme une proposition conditionnelle, sans aucune présupposi­ tion d'existence, qui peut être formalisée par le quantificateur universel; en revanche, l'analyse de la proposition: (32) The one who is saying it, fails aurait la présupposition: (33) Someone says it. Le pronom le plus proche sémantiquement de quisquis est sûrement qui­ cumque. Il existe aussi un emploi non relativisé de ce pronom, avec une valeur qualitative (n'importe qui), même s'il est très rare. Cet emploi atteste un échange sémantique avec quisquis non relativisé (anyone):

PRONOMS INDEFINIS

237

(34) Caes. Gall. 6,33 qui quacumque de causa ad eos venerunt, ab iniuria prohibent. Mais quicumque est utilisé surtout dans le cas relativisé (whoever). C'est pour cet emploi du pronom que nous proposons la formalisation par le quantifica­ teur universel, en général sans aucune présupposition d'existence, car les énoncés où il est utilisé sont pour la plupart des propositions conditionnelles: (35) Cic. de Orat. 2,5 omnia, quaecumque in hominum disceptationem cadere possunt, bene sunt ei dicenda, qui hoc se posse profite­ tur, eqs. En l'absence d'une modalité d'attitude propositionnelle, dans les énoncés déclaratifs au parfait, on peut toutefois reconnaître à quicumque une présup­ position existentielle: (36) Cic. ad Q.fr. 1,2,2 quoscumque de te queri audivi, quacumque potui ratione placavi. On peut vraisemblablement rappeler à propos de quicumque relativisé ce que Z. Vendler (1967: 90 s.) remarque à propos de all c'est-à-dire que la proposi­ tion: (37) All the messages you sent were intercepted aussi bien que: (38) All messages you might have sent would have been intercepted sont toutes les deux correctes, bien que dans la deuxième proposition il n'y ait pas la présupposition d'existence. Au contraire, une proposition telle que: (39)* Each (every) message you might have sent would have been intercepted serait incorrecte, parce que every et each exigent cette présupposition. Ces exemples nous semblent expliquer la différence qu'il y a dans quelque cas entre le pronom quicumque et quisque. Celui-ci, dans des constructions relati­ visées particulières, semble rencontrer les mêmes conditions d'emploi que les indéfinis anglais each et every, et avoir donc une signification existentielle. On va traduire quisque par each, en soulignant sa valeur distributive, dans le cas où ce pronom est accompagné d'un pronom réfléchi: (40) Ov. Fast. 2,715 oscula quisque suae matri properata tulerunt. mais on va traduire quisque par every, avec une connotation collective, lors­ que il est accompagné d'un superlatif:

238

A. ORLANDINI

(41) Liv. 25,6,23 asperrima quaeque ad laborem periculumque deposcimus. C'est la valeur distributive de quisque qui lui donne la possibilité d'être ac­ compagné de numéraux ordinaux, lorsqu'il dénote une généralité (every): (42) Cic. Rab. Post. 34 tertio quoque verbo excitabatur. Dans cette signification every s'oppose à each, car on dit couramment: (43) He came every three days, alors qu'il ne serait pas correct de dire: (44)* He came each three days (cf.Vendier 1967: 76) En effet each a une valeur separative (commune au pronom indéfini unusquisque) et si on l'employait dans la phrase (44), on indiquerait les jours un à un séparément et non dans la totalité de leur suite temporelle. Voici enfin un tableau synoptique des pronoms indéfinis qu'on vient d'analyser par rapport à leur traduction anglaise: Pronoms indéfinis qui indiquent l'existence T spéc. quidam (a certain)

T/O non-spéc. aliquis (someone/ anyone)

O/T quispiam (anyone/ someone)

O/T quisquam (anyone)

O quis (anyone)

O quisquis (anyone)

T= lecture transparente 0 = lecture opaque T / 0 = lecture plus fréquemment transparente O/T= lecture plus fréquemment opaque Pronoms indéfinis relativisés En général sans aucune

toujours avec la

présupposition d'existence

présupposition d'existence

quisquis

quicumque

quisque

unusquisque

(whoever)

(whoever)

(each/

(each)

every)

PRONOMS INDEFINIS

239

NOTES 1) Les interrogatifs lient des variables qui se trouvent dans le scope d'un opérateur épistémique, alors que les relatifs impliquent une sorte de "quantificational cross-reference" (Hinttikka 1974:130), mais non une quantification dans des contextes épistémiques. 2) Il y a beaucoup de théories sémantico-syntaxiques qui soulignent la relation entre la relative et l'article défini. Ch.Touratier (1980: 368-386) nous en a donné un compte-rendu exhaustif et en particulier il nous propose l'interpretation de N.Chomsky (1971:141 n.26), d'après laquelle "alors que les relatives restrictives appartiennent au système du Determinant ( = l'article) il y a plu­ sieurs raisons de supposer que les relatives non restrictives sont plûtot des Complement du SN tout entier". En ce qui concerne la quantification Touratier nous rapporte l'interprétation de G.van Hout (1973-74,I:48-49, 95, 31-32) c'est-à-dire que l'article défini est un quantificateur aussi bien que la relative determinative, qui quantifie en délimitant "un sous-ensemble" de l'extension de son antécédent. 3) La signification d'indéfini spécifique disparaît lorsque l'adjectif quidam, employé avec un autre adjectif, exprime proprement une nuance intensive ou limitative: (45) Cic. Lael. 48 qui virtutem duram et quasi ferream esse quandam volunt; (46) Cic. de Orat. 2,10 timiditate quadam ingenua. 4) Cette distinction a déjà été signalée à propos des pronoms latins dans Lyons (1977,II:759) qui toutefois n'insiste pas sur leur nature pragmatique. 5) Il faut aussi rappeler que dans la même situation pragmatique on doit utiliser l'indéfini any, s'il y a un prédicat tel que warn, alors qu'on utilise some avec promise (R.Lakoff 1968:611).

BIBLIOGRAPHIE Chomsky, N., (1971), Aspects de la théorie syntaxique, trad. par J.-Cl. Mimer. Paris: Seuil. Cushing, S., (1977), The formal semantics of quantification. Bloomington: Indiana University Linguistics Club. Hahn, E.A.,(1933), 'Light from Hittite on Latin indefinites' . Transactions and proceedings of the American philological association 64, 28-40. Hinttikka, K.J.J., (1974), 'Questions about questions' . In: M.K.Munitz and P.K. Unger (eds.), Semantics and philosophy. New York: New York Uni­ versity Press. Hout, G. van, (1973-74), Franc math (Essai pédagogique sur les structures grammaticales du français moderne). Paris: Didier. Kasher, A.-Gabbay, D.M., (1976), 'On the semantics and pragmatics of spe­ cific and non-specific indefinite expressions' . Theorectical Linguistics 3, 145-190.

240

A. ORLANDINI

Lakoff, R.T., (1969), 'Some reasons why there can't be any some-any rule' . Language 45, 608-615. Lyons, J.,(1977), Semantics. Cambridge: Cambridge University Press. Quine, W., (1960), Word and Object. Cambridge: Harvard University Press. Touratier, Ch., (1980), La relative. Essai de théorie syntaxique. Paris: Klincksieck. Vendler, Z., (1967), Linguistics in Philosophy. Ithaca and London: Cornell University Press.

V. MORPHOLOGY/PHONOLOGY

LATIN DEGEMINATION: A N AUTOSEGMENTAL APPROACH

O.S. P I L L I N G E R Colchester SUMMARY This paper sets out to demonstrate the shortcomings of a purely segmen­ tal approach to phenomena such as degemination in Latin, and the need for some kind of formal representation of syllable structure. After a brief examination of the relevant Latin data and some of the prob­ lems connected with it, the first point that is made is that degemination can only be accounted for, in a segmental phonology, as a set of between two and four distinct processes. Reference to prosodic phenomena such as stress and metre, however, shows that stress assignment and metrical scansion are par­ tially determined by precisely those environments that give rise to degemina­ tion: namely, heavy syllables. This suggests that degemination is in fact a uni­ tary process, and should be formulated as such. Accordingly, after further ar­ guments to justify a unitary approach, an analysis incorporating the represen­ tation of syllable structure (and following the general conventions of Goldsmith's Autosegmental theory of phonology) is briefly presented. A.

Data

The kinds of data under consideration are as follows: 1. (i)

'Undisputed' cases of synchronic degemination:1

Word finally: (ii) Following a consonant: miles (←/miless/←/milit+s/) arsï (←/arssii/←/ard+sii/) comes (←/komess/←/komit + s/) concidï (←/koŋkkidii/←/kon+kekidii/) ager (←/agerr/←/ager+s/) pexus (←/pekssus/←/pekt+tus/) mānsus (←/manssus/←/mand+tus/)

244

O.S. PILLINGER

(iii) Following a long vowel or (iv) Preceding a non-liquid cons. : diphthong'. ascendō (←/asskendoo/←/ad+skendoo/) [fisus (←/fiissus/←/fiid + tus/)] agnoscō (←/aggnoskoo/←/ad + gnoskoo/) (See below.)

[?versus

(←?/werrsus/←?/werr+tus/)]

What I have described in (1) as the 'undisputed' cases of degemination are those generally accepted as such in traditional descriptions of Latin phonology, and which would be likely to be analysed as such in a 'standard' generative phonology of the language. This is not to say that they are unproblematic: there are difficulties connected with the derivations of falsus and ver­ sus, for instance — do they derive from forms with a geminate /1/ or /r/, as suggested here, or (as diachronic considerations would suggest) from forms with a geminate /s/? And the precise formulation of the process of assibilation that produces the -sus terminations is not altogether clear: the existence of al­ ternations like metō ~ messus argues in favour of the intermediate geminate /-ss-/ stage that I assume for forms like pexus & ascendō (rather than, for in­ stance, partial assibilation producing /-ts-/, followed by deletion of the re­ maining /t/). However, while one might argue over individual instances, I think there are enough clear cases to establish at least environments (i), (ii) and (iv) fairly uncontroversially as 'triggers' for degemination. (2)

Disputed cases of degemination :

(iii) Continuants f ollowing a long (iv) Preceding a liquid consonant: vowel or diphthong'. acclamö; effringō; attrahö; -ss- caussa; cāssus; QVAESSO... accrescō... -//- mille; villa; Stēlla; rāllum [SVPLICARE; SVBFRAGIA; SVPPLIcorölla; üllus ; paullus... CARE;...] -rr- nārrö. (v) Contracted perfects', amāsse; audïsse; ... The above cases of degemination that have been disputed in the litera­ ture raise a number of interesting problems; but in the space available I shall only be able to sketch in very briefly some of these problems and the position I have adopted with regard to them. As for environment (iii), the only clear cases of synchronic degemination following a long vowel or diphthong involve continuants—mainly /ss/ and /ll/. In the case of/ss/, we have the statement by Quintilian (1.7.20), supported by inscriptionai evidence (forms like QVAESSO, DIVISSIT, IMPERIOSSVS: cf. Sommer, 1902:295), to the effect that up to the beginning of the imperial

DEGEMINATION

245

period, /ss/ was not degeminated here; but that this began to occur thereafter. I shall accordingly restrict myself in this paper to the later period, when degemination was more general. As regards /ll/, all the traditional sources seem agreed that words like mille, villa, Stella, vallum, corolla and ūllus contained long vowels (and Som­ mer, p.296, seems to imply that this was the case in imperial as well as republi­ can times). If these truly contained a geminate /ll/, therefore, they would con­ stitute an exception to the rule of degemination following a long vowel. And for diphthongs, we have the spellings aulla, paullus, etc., to give further sup­ port to this apparent exception. However Niedermann makes the interesting suggestion here (1953:125ff.), that these double /'s represented, not a gemi­ nate /ll/, but a palatal/I/:and this palatal quality needed to be distinguished in the orthography from the velar quality of the /l/ in words like volō, etc. This analysis derives some support from Priscian 2.29, 8-12 K, where Pliny is quoted to the effect that L triplicem...sonum habet: exilem quando geminatur... plenum quando finit...syllabas; medium in aliis... This would require the assumption that at some stage Pliny's sonus exilis was distinctive enough to require a separate orthographic form (//) — perhaps phonemically distinct. However the sorts of evidence required to establish this — especially the later Romance developments — are somewhat intricate, and for the present I shall merely have to ask that it be taken on trust that Niedermann's proposal is worthy of serious consideration, and does receive some support from the Romance evidence. As for aulla, paullus, etc., I think the most straightforward explanation of the not-infrequent -//- in the spelling of these words is that they were often pronounced as olla, and the non-literary pollus. As for /rr/, there is only one problem case, and this is nārrō. This appears in inscriptions with both the geminate and the long ă, so that I shall either have to regard it as an idiosyncratic exception to the general rule, or as one of the not-infrequent cases where the inscriptionai evidence has redundantly re­ tained both long vowel and geminate consonant. In support of the latter view, Varro (quoted by Velius Longus) in fact suggests that it should be written nārō — without a geminate r. I therefore conclude that as a general rule degemination did occur, at least in post-republican times, following a long vowel or diphthong. The only remaining problem in this connection is that of the contracted perfects:

246

O.S. PILLINGER

amāsse, audïsse, and so on. There is again some disagreement, both among modern and ancient sources, as to whether these geminates were in fact re­ tained: but even if they were, there is a ready explanation in the very fact they were contracted forms, and therefore somewhat outside the normal con­ straints of the phonology. We move on now to the question of degemination preceding a liquid con­ sonant (/l,r/). The forms in question here are acclamö, effringō, attrahō, ac­ ­rescö , and so on. Niedermann claims that these are spelt thus by morphologi­ cal analogy: i.e., that degemination did in fact occur. However an examina­ tion of the inscriptionai evidence in C.I.L. reveals that the 'full' forms are more common, except in the republican period. Thus the earlier inscriptions contain forms such as AFLEICTA, SVFRAGIVM, SVPLICARE, etc., while we later tend to find SVPPEICARE, ATTRIBVO, SVPPRAEFECTVS,... The 'morphologically explicit' forms such as ADFLICTVS, SVBFRAGIA, SUBPRAEFICERE, etc., occur in all periods but are more common in later inscriptions. I therefore make the more difficult assumption (on the 'lectio difficilior' principle), that these forms constitute an exception to the otherwise rather general rule of degemination preceding a consonant. In summary, then, there appear to be four environments in Latin where degemination occurs, and two environments where it does not occur: (3)

(4)

Environments f or degemination: (i) (ii) (iii)

/ # / C / VV

(iv)

/

{miles...) (pexus...) (fisus...) [ - son, - cont] (ascendō ...)

Environments where degemination does not occur: (i) (ii)

/ V /

V [+son, + cont]

(missus, fissus...) (acclamō, effringō...)

B. Standard Theory Formulation The above four environments for degemination will, in the generative Standard Theory — and in any formally-explicit segment-based theory — re­ quire four separate rules of geminate consonant deletion. The use of abbreviatory conventions such as braces will reduce the four to two: Rule (a): C, → ø /

Ci

≠ [-son,-cont]

(= environments (i)&(iv))

DEGEMINATION

247

Rule (b):2 Ci → ø / C VV – Ci (= environments (ii) & (iii)) ( - and it is clear that the disparity between the environments will prevent any further device, such as mirror-image notation, from collapsing these into a single rule-schema). However such abbreviatory devices in generative phonology are quite clearly an expression of failure: what should have been expressed unitarily cannot be so expressed by the available formalisms. Rules (a) & (b), therefore, merely abbreviate without explaining the four cases of degemination. C. Degemination as a Unitary Phenomenon This state of affairs, however, may seem to be perfectly acceptable at first sight: Latin degemination is apparently quite adequately described in terms of four distinct processes. But this runs counter to the traditional analyses, and, I'm sure, to the intuitions of most Latinists. The point is that there is a factor unifying these four environments: namely, that in each case degemination oc­ curs in the context of a heavy syllable; or, more accurately, whenever the geminate is not required in order to maintain the 'heaviness' of a preceding syllable. This generalisation might be stated as follows: (5) Whenever a geminate cluster occurs at the rightward boundary of a heavy syllable, degemination occurs: provided that the heaviness of the syllable is not thereby destroyed. Or, informally, degemination occurs in the context of a preceding 'already heavy' syllable. Thus degemination does not occur with forms like /missus/, where its application would render the first syllable light; but it does occur with /pekssus/ and /asskendoo/, where the presence of a preceding or follow­ ing consonant ensures that with or without the geminate, the first syllable will be closed — i.e., heavy. And it also occurs with /kaassus/ and /kaissus/, where the presence of a preceding long vowel or diphthong ensures that the preced­ ing syllable will be heavy with or without the geminate. (I should make it clear at this point that I am assuming with Allen (1973:130-1), Zirin (1970:57), Kent (1948:303-8) and most other modern writers on the subject, that heavy sylla­ bles were either closed, or contained a long vowel/diphthong: i.e., that con­ trary to ancient statements, pre-pausal -CVC was regularly heavy, regardless of any following vowel. We therefore apear to be dealing with a process that has syllable weight

248

O.S.PILLINGER

— defined in terms of syllable structure — as a unifying factor across its vari­ ous manifestations. Such a conditioning factor clearly could not be captured in a segment-based theory that makes no provision for the representation of syl­ lable structure — and it is therefore important to establish that this is not acci­ dental, but that syllable weight is integral to the Latin phonological system. It does not however need emphasising here that there are two further areas of Latin phonology that cannot be accounted for in any coherent fashion without reference to syllable weight: namely, stress and metre. The Latin stress rule depends crucially on the weight of the penultimate syllable; and verse scan­ sion operates almost entirely in terms of syllable weight. It would therefore surely be counterintuitive to admit some representation of syllable weight in one's account of stress and metre, while barring it from the formulation of degemination. However, further confirming evidence is also available, in that all three phenomena apparently share an identical set of exceptions: namely, conso­ nant-plus-liquid (CL) clusters. It has already been concluded that degemination did not occur in this environment (acclamō, etc.); and we also find as re­ gards stress, that words with a CL cluster at the boundary between the penulti­ mate syllable and the ultima are generally not treated as if the penultimate were heavy: i.e., ténebrae rather than tenébrae; and similar syllables are fre­ quently treated as light for the purposes of verse scansion. Allen (1973:138-9) in fact concludes that the metrical treatment of such syllables as heavy was in conscious imitation of the Greek practice, and that their usual pronunciation, as reflected in the stress assignment, was 'light' — i.e., in his terms, as a com­ plex release rather than arrest + release. However Allen's adherence to Stet­ son's (1945) motor theory of syllabification also leads him to deny that se­ quences of ƒ plus a liquid behaved in the same way. This contradicts the tradi­ tional statements, and is asserted without supporting evidence (p. 137-8); in fact there does not appear to be any evidence one way or the other, and in what follows — based, as it is, on phonological rather than phonetic consider­ ations — I shall assume that all CL clusters regularly failed to render the pre­ ceding syllable heavy. The parallelism between the three phenomena of stress, metre, and degemination, may therefore be schematized as follows:

DEGEMINATION

-cc Stress: Metre: Degemination:

249

-CL renders penultimate heavy renders syllable heavy causes degem. if one C is part of geminate cluster

does not render penult. heavy does not render syllable heavy does not cause degem. if C is part of a geminate cluster.

This provides fairly convincing evidence that degemination should be treated as a unitary phenomenon governed by the presence of a preceding 'alreadyheavy' syllable. D. An Autosegmental Analysis Since it appears therefore that the generative Standard Theory, being segment-based, cannot provide such a unitary treatment, I shall consider an alternative analysis in terms of syllable structure. The model I shall be using is basically that of Goldsmith (1976) and Kahn (1976), i.e. 'autosegmental' in approach. This theory allows the possibility of more than one 'tier' of phonological structure. In order to capture the generalisations outlined above, I shall suggest that in addition to the basic segmental tier, two further levels of struc­ ture are required: a 'skeletal' tier on which the consonantal and vocalic struc­ ture is specified; and a 'syllabic' tier on which the syllables are delimited. The elements of each tier are associated with elements of the next tier both by specific association rules, and by a general Well-Formedness Condition (WFC); and the failure of any element on a given tier to be associated with an element on the next will result in the deletion (by specific phonological rule) or association (as a general result of the WFC) of the unassociated element. The label 'autosegmental', it should be noted, is used by Goldsmith to replace the term 'suprasegmental', which he regards as misleading: "A more accurate picture, we are suggesting, is [that of] parallel sequences of segments, none of which 'depend' or 'ride on' the others. Each is independent in its own right: hence the name autosegmental level." (1976:20-21.) In this system, therefore, a word like addō would have the following (somewhat simplified) structure: Segmental tier:

a d d o o

Skeletal tier:

V C C V ,V

Syllabic tier: Figure 1

l

2

250

O.S.PILLINGER

(-representing long vowels phonologically as a sequence of two short vowels, similar to the representation of geminate consonants. This is not meant to imply that either long vowels or geminate consonants are phonetically twice the duration of the corresponding single segments; only that in terms of sylla­ ble structure an extra mora is involved.) The need for a skeletal tier will become clearer in what follows. For the present I shall merely outline the mechanisms by which the various tiers are associated. First of all, on the segmental tier a rule applies associating segments with skeletal autosegments. As a preliminary approximation, this might read as follows: (6) Associate each [+syllabic] segment with a V autosegment, and each [-syllabic] segment with a C autosegment; co-index any ad­ jacent identical autosegments on the skeletal tier; and index any diphthongs on the skeletal tier. By this means vowels and consonants are associated on the two levels, and long vowels, diphthongs and geminates are identified on the skeletal tier, long vowels as V i V i , diphthongs as V i V j , and geminates as C i C i . The operation of rule (6) on addō would be represented as follows: a

d

d

o

o

Figure 2 Structures of this kind will then form the input to the syllabic association rules, which connect the skeletal and syllabic tiers. The first point to be deter­ mined here is where the syllable nucleus will lie: clearly, with long vowels and diphthongs represented as sequences of short vowels, this is not a straightfor­ ward matter of any V-segment constituting a syllable nucleus (as, for instance, in Kahn 1976). I therefore propose the following definition: (7) A syllable-nuclear vowel is one which is not immediately preceded by an indexed vowel on the skeletal tier. -i.e., a syllable-nuclear vowel will be any non-indexed vowel or the first of two indexed vowels.

DEGEMINATION

251

The first rule of syllable association therefore reads as follows: (8) Syllabic Association Rule [SAR] 7: Associate each syllable-nuc­ lear vowel with exactly one syllable. This will give us (omitting features on the segmental tier):

Figure 3 Having associated the syllable nuclei, we now need a rule to 'spread' the syllable association leftward and rightward from these nuclei. This can most simply be accomplished by careful specification of the priorities that deter­ mine the order in which segments are successively associated. (It will be clear from what follows that this must be a successive spreading, rather than, for in­ stance, simultaneous association of any skeletal autosegment with any sylla­ ble, governed only by the WFC.) The second rule of syllable association may be given in the following preliminary form: (9) SAR 2 [Preliminary - abbreviated] : (a) Priority of association is determined by proximity to syllablenucleus. (b) Priority is assigned to vowels over consonants. (c) Priority in sequences of consonants is assigned to rightward [following-syllable] association rather than leftward [preceding-syllable] association. SAR (2) is unrestricted as to the number of associations it performs ; and it will apply regardless (in the first instance) of any intervening morpheme or word boundaries [but on this, see below]. Returning to addō, it will proceed in the following stages (with clause (a) acting as the overriding determinant of the outward-spreading operation of the rule):

252

O.S. PILLINGER

Figure 4 With a longer consonant cluster, such as that found in a word like extrā, SAR 2(c) would in effect apply twice, on each occasion associating first to the right and then to the left:

Figure 5 The formulation of SAR 2, by giving priority to rightward consonant as­ sociation , incorporates the universal preference for CV syllabification. Thus a word like pestis could not be syllabified as /pest-is/ by this rule. Nor could it come out as /pe-stis/: despite certain traditional statements to the contrary, in­ scriptionai and metrical evidence shows the first syllable in such words regu­ larly to be heavy (cf. Allen 1973:137). S + Consonant (sC) clusters 'make po­ sition' even when a morpheme boundary precedes; the only exception is when a word boundary precedes the sC cluster (Allen, op.cit., p.139-40). Thus the syllabification produced by SAR 2 — i.e., /pes-tis/ — will be the correct one; and special conditions will in any case be operative across word boundaries (cf. below). However, as it stands SAR 2 would not prevent certain absurd output structures — such as

Figure 6

DEGEMINATION

253

It is precisely this sort of output, however, that the Well-Formedness Condi­ tion is intended to rule out. The WFC would ensure (at the least) that (10) [Well-Formedness Condition (Preliminary):] (i) Every segment on a given tier is associated with at least one segment on an adjacent tier; (ii) Association lines do not cross. Concerning the Well-Formedness Condition, Goldsmith emphasizes that it "is in the indicative, not the imperative. A derivation containing a representa­ tion that violates the Well-Formedness Condition is not thereby marked as illformed; rather the condition is interpreted so as to change the representation minimally ... so as to meet the Condition maximally" (1976: 27). Such 'mini­ mal changes' in the present context would consist in the deletion and addition of association lines in accordance with the principles governing syllable as­ sociation. Thus Figure 6 (a) & (b) would both end up with the correct output structure illustrated in Figure 5 (b). However, there is one further type of output structure that neither SAR 2 nor the WFC as presently stated would prevent, namely:

Figure 7 - with the /s/ represented as an ambisyllabic segment associated indetermi­ nately with either Σ1 or Σ2. Should this type of output be permitted? Kahn (1976) has made a case for representations of this sort, maintaining that it is not always possible, phonetically or phonologically, to determine the precise location of the syllable boundary in such cases — and that in fact it is more advantageous not to do so. This is an empirical issue, and much further study of Latin syllable structure will have to be undertaken before it can be de­ termined with certainty whether or not ambisyllabicity is required here as well; certainly, to exclude it will require an extra clause in the WFC, and in the interests of simplicity this should be avoided if possible. However such phonological evidence as is available in the present context seems on balance

254

O.S.PILLINGER

to argue against ambisyllabicity. We shall first consider CL and sC clusters in words like atra, astō: these would be syllabified as follows (with dotted lines indicating possible ambisyllabic associations):

Figure 8 In Figure 8 (a), the ambisyllabic analysis seems attractive, as it would allow the syllabification /a-tra/ — which, at least in the absence of intervening morpheme or word boundaries, appears to have been the regular situation in classical Latin. However in Fig. 8(b), the ambisyllabic analysis is distinctly un­ attractive, since a syllabification /a-stoo/ is predicted to be just as likely as /astoo/ — and, as noted above, the syllable preceding an sC cluster was regularly heavy. Thus ambisyllabicity does not appear to provide a unified account of these cases: and for atra, at least, an explanation for the strong tendency to­ wards a light initial syllable must be sought elsewhere. This is where the moti­ vation for the Skeletal tier becomes apparent. In order to distinguish atra & astö formally in terms of their differing syllabifications, I propose that the Skeletal Association Rule (6) above be modified so as to associate both mem­ bers of a CL cluster with a single skeletal autosegment (obligatorily in the ab­ sence of any intervening morpheme or word boundary). This would give the following revised structure to replace Fig. 8(a):

Figure 9 Clearly such a representation, if justified, will completely rule out the possi­ bility of ambisyllabic segments, since an ambisyllabic attachment of the skeletal C in Fig. 9 to Σ1 will predict */atr-a/ as an acceptable syllabification. This conclusion concerning ambisyllabicity is further reinforced if we

DEGEMINATION

255

consider words such as anxius /aŋksius/, arcturus /arkturus/, linctus /liŋktus/, etc. SAR 2 in these cases will produce structures of the form

Figure 10 (Note, in passing, that ambisyliabicity would wrongly allow /u/ to be as­ sociated with both Σ2 & Σ3 here — and similarly in any words containing ad­ jacent syllable-nuclear vowels. This consequence could be ruled out by other means, but would not arise without ambisyliabicity.) The main problem here is that Σ2 is made to begin with the cluster /ks-/ — which is impermissible word-initially in Latin words of native origin. (And similarly /kt-/ in arcturus & linctus.) This arises from the priority given in the statement of SAR 2 to the rightward association of consonants (which how­ ever does not violate word-initial constraints in most other cases). Ambisyl­ iabicity in this environment might appear attractive (i.e., having/k/ associated with both Σ1 & Σ2 — but again, it would predict both syllabifications as equally acceptable — which appears intuitively incorrect. So once more we are driven to a solution other than ambisyliabicity: in this case one that Kahn also employs, namely a restriction on clause (c) of SAR 2 to the effect that rightward association of consonants cannot be allowed to produce an imper­ missible wordinitial cluster in the onset of a syllable. Note that, unlike Kahn's analysis for English, it does not appear necessary to restrict SAR 2 in terms of word-final consonant clusters: since it seems to be the case in Latin that while syllable-initial clusters may be identified with word-initial clusters, the same is not true for syllable- and word-final clusters. Thus in the revised syllabifica­ tions of the words under considerations here, we have syllable-final clusters / ŋk/ aŋk-sius/ & /liŋk-tus/) and /rk/ (/ark-tu-rus/) — neither of which may ap­ pear word-finally. The exclusion of ambisyliabicity, therefore, will require some modifica­ tion of the WFC, perhaps an additional clause as follows: (11) Well-Formedmess Condition: (iii) A segment on the skeletal tier is associated with exactly one syllable. Given the above framework for syllabification, it remains to outline how

256

O.S. PILLINGER

degemination can be achieved in a unitary way and with reference to the light/ heavy-syllable distinction. It will be clear from what has been said above that heavy syllables in this system are easily definable as those that contain an association line to the right of the syllable nucleus. And thus in line with generalisation (5) above, degemi­ nation will occur whenever both members of a geminate cluster are not im­ mediately adjacent to the syllable nucleus. Thus in /missus/ —

Figure 11 - since both geminates are adjacent to a syllable nucleus they are retained and serve to render Σ1 heavy. In /peksus/, however,

Figure 12 only the second geminate consonant is adjacent to the syllable nucleus, so that the first remains unassociated (indicated by x). It will then be deleted by a phonological rule of degemination that applies after syllabification (for further discussion of this rule, see below). The restriction on geminate consonants that they must be adjacent to the syllable nucleus in order to receive syllable association, is best handled in the present system by means of a further priority statement in SAR 2. With this and the earlier modification concerning permissible sylllable-initial clusters, SAR 2 will read in full: (12) Syllable Association Rule 2: (a) Associate segments with syllables in an order determined by their degree of proximity to the syllable nucleus. (b) Associate indexed consonants before associating other segments.

DEGEMINATION

(c) (d)

257

Associate indexed vowels leftward (to a preceding syllable) before associating non-indexed consonants. In any continuous sequence of consonants, associate conso­ nants one at a time, first rightward (to a following syllable) and then leftward (to a preceding syllable); unless rightward association would produce an impermissible wordinitial cluster preceding a syllable nucleus: in which case associate leftwards.

SAR 2 will now correctly produce the structure in Figure 12, which will then form the input to a degemination rule deleting the unassociated segment. Similarly with cases (i) and (iii) of degemination: the final /-s/ of /miless/ and the first /-s-/ of/fiissus/ will be unassociated, and therefore deleted (in the latter case, because of the preceding non-nuclear vowel):

Figure 13 In case (iv) of degemination, the same process will apply to eliminate the second /g/ of /aggnoskoo/; but with CL clusters, as in /akklamoo/, the prior 'merger' of the second element of the geminate with the liquid to form a single C autosegment on the skeletal tier, would prevent degemination from occur­ ring. Thus after skeletal association we would have, not a geminate /kk/ fol­ lowed by /l/, but /k/ followed by a complex /kl/ consonant — and these clearly would not be co-indexed as geminates on the skeletal tier. The normal appli­ cation of SAR 2 will therefore produce what we are hypothesizing is the cor­ rect result, i.e.

Figure 14

258

O.S.PILLINGER

- to which degemination will be unable to apply. One final case of degemination needs to be considered, namely that of words like /eksspektoo/ - which might be regarded as a combination of cases (ii) and (iii), since the geminate is both preceded and followed by a consonant. The problem here is that simple application of SAR 2 will produce

Figure 15 - with both geminates unattached, since neither is adjacent to a syllable nuc­ leus. This requires a slight modification to the rule of degemination, so that it will not merely delete an unassociated geminate consonant, but will delete only one such geminate. With this revision, the rule might read as follows: (13) Degemination: If either or both members of a geminate cluster are unassociated with any syllable, delete one unassociated gemi­ nate consonant. This would then have the effect of deleting either of the s's in Fig. 15; and the WFC would then have its normal effect of causing the remaining /s/ to be as­ sociated with Σ2, as SAR 2(d) predicts:

Figure 16 It has been hinted at various points in this paper that morpheme and word boundaries must be allowed to affect syllabification, but nothing very specific has been said. However it is simple enough to envisage the necessary modifi­ cations, though I shall not have time to go into this area in any detail. Thus where a morpheme boundary falls between consonant & liquid within a word - as in /ab + ripii/, for instance — 'merger' of the CL cluster on the skeletal

DEGEMINATION

259

level will be prevented, and the syllabification will necessarily be /ab-ripii/. Conversely where a morpheme boundary precedes a CL cluster, as in /re+trahoo/, such merger will be obligatory. Similarly with sC clusters, a preced­ ing word boundary could be made optionally to coincide with the syllable di­ vision. And textual factors, such as the occurrence of the form in question in verse of certain types, could also permit the relaxation or strengthening of the normal conditions as appropriate. To conclude by returning to the central point of this paper: the autosegmental formulation of Latin syllable structure and degemination outlined here, if it can be maintained, has the advantage over segment-based theories of being able to capture with a rather simple formalism the generalisation that it is the presence of an 'already-heavy' preceding syllable (or of an extra 'mora') that causes degemination.

NOTES 1) It should be noted that I am not taking into account purely diachronic cases of degemination, i.e. cases where it seems reasonable to assume that no synchronic relationship would have been felt to exist between the supposedly degeminated and the (historically prior) non-degeminated forms. Examples of such excluded forms would be sarmentum (< *sarmmentum < *sarpmentum) or praec (< ?*praeccö < *prae+dicō). 2) The formulation of this rule is inadequate as it stands, since it would lead to the degemination of /kk/ in forms like reaccendō. This could be remedied by the use of appropriate indices & condi­ tion(s) on the rule to restrict the VV sequence to long vowels and diphthongs only. 3) Dotted association lines indicate the most 'recent' association made by the sequential applica­ tion of the rule.

BIBLIOGRAPHY Allen. W.S., (1973), Accent and Rhythm. Cambridge. Goldsmith, J.A., (1976), Autosegmental Phonology. Bloomington: Indiana University Linguistics Club. Kahn, D., (1976), Syllable-based Generalizations in English Phonology. Bloomington: Indiana University Linguistics Club. Kent, R.G., (1948), 'A problem in Latin prosody'. In Mélanges de philologie, de littérature, et ďhistoire anciennes... Paris :303-8. Niedermann, M., (1953), Phonétique historique du Latin. Paris: Klincksieck. Sommer, F., (1902), Handbuch der Lateinischen haut- und Formenlehre.

260

O.S. PILLINGER

Heidelberg: Carl Winter's Unhersitătsbuchhandlung. Stetson, R.H., (1951), Motor Phonetics: A Study of Speech Movements in Ac­ tion. (Second Edition.) Amsterdam: North-Holland. Zirin, R.A., (1970), The Phonological Basis of Latin Prosody. The Hague: Mouton. COMMENTS ON PILLINGER'S PAPER E. VESTER Free University, Amsterdam My first point is your use of the term 'synchronic'. In your paper this term is assumed to apply to cases of degemination in which 'a synchronic relation­ ship would have been felt to exist between the supposedly degeminated and the (historical prior) non-degeminated forms' as opposed to diachronic cases of degemination in which no such relation is felt (note 1). Your example of diachronic degemination is sarmentum. My question is: at which point of time in the long history of Latin do we have to speak of synchronic degemination? As you state yourself, a general rule of degemination occurred in post-repub­ lican times. You have to choose this period to be able to account for attested cases of caussa in the republican period. Now miles, one of your examples, is attested with a double s in the prosody of Plautus—which does not hold, as far as I know, for all the other examples you give —, but for a general rule of de­ gemination to occur in post-republican times, one should expect to find cases of miles with a double s in the period after Plautus as well. My second question is: in which way can degemination preceding a heavy non-final syllable, but following a short vowel be explained? I mean cases like: omitto (