Stability, Variation and Change of Word-order Patterns over Time (Current Issues in Linguistic Theory)

STABILITY, VARIATION AND CHANGE OF WORD-ORDER PATTERNS OVER TIME

AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E. F. KONRAD KOERNER (University of Ottawa) Series IV - CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board Raimo Anttila (Los Angeles); Lyle Campbell (Christchurch, N.Z.] Sheila Embleton (Toronto); John E. Joseph (Edinburgh) Manfred Krifka (Berlin); Hans-Heinrich Lieb (Berlin) E. Wyn Roberts (Vancouver, B.C.); Hans-Jürgen Sasse (Köln)

Volume 213

Rosanna Sornicola, Erich Poppe and Ariel Shisha-Halevy (eds) Stability, Variation and Change of Word-Order Patterns over Time

STABILITY, VARIATION AND CHANGE OF WORD-ORDER PATTERNS OVER TIME Edited by ROSANNA SORNICOLA Università di Napoli Frederico II ERICH POPPE Philipps- Universität Marburg ARIEL S H I S H A - H A L E V Y The Hebrew University of Jerusalem with the assistance of Paola Como

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA

/^~~\ TM (OOÍ V_^'

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences — Permanence of Paper for Printed Library Materials, ANSI Z39.48-1984.

Library of Congress Cataloging-in-Publication Data Stability, variation, and change of word-order patterns over time / edited by Rosanna Sornicola, Erich Poppe, Ariel Shisha-Halevy, with the assistance of Paola Como. p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 213) Chiefly papers presented at a workshop held within the 13th International Conference on Historical Linguisitics, which was held Aug. 14,1997, Düsseldorf, Ger. Includes bibliographical references and index. 1. Grammar, Comparative and general-Word order-Congresses. 2. Linguistic changeCongresses. 3. Language and languages-Variation-Congresses. I. Sornicola, Rosanna, 1953-. II. Poppe, Erich. III. Shisha-Halevy, Ariel. IV. International Conference on Historical Linguistics (13th : 1997 : Düsseldorf, Germany) V. Series. P295.S69 2000 415--dc21 00-050830 ISBN 90 272 3720 4 (Eur.) / 1 58811 037 o (US) © 2000 - John Benjamins B.V. 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. John Benjamins Publishing Co. • P.O.Box 75577 • 1070 AN Amsterdam • The Netherlands John Benjamins North America • P.O.Box 27519 • Philadelphia PA 19118-0519 • USA

CONTENTS

Acknowledgements Introduction Part I. Stylistic and Pragmatic Principles in Stability, Variation and Change Jean-Pierre Benoist La formation des principes de l'ordre des mots du russe moderne en tant que problème de stylistique littéraire et de linguistique Antonio Loprieno From VSO to SVO? Word Order and Rear Extraposition in Coptic Erich Poppe Constituent Order in Middle Welsh: the Stability of the Pragmatic Principle Part II. Stability and Diachrony Guy Deutscher Stability Against the Odds? The Survival of Verb Final Order in Akkadian Ariel Shisha-Halevy Stability in Clausal/Phrasal Pattern Constituent Sequencing: 4000 Years of Egyptian (with Some Theoretical Reflections, also on Celtic) Rosanna Sornicola Stability, Variation and Change in Word Order: Some Evidence from the Romance Languages

vii ix

3

23

41

55

71

101

vi

Part III. Reanalysis, Grammaticalization and Change Jan Terje Faarlund Reanalysis in Word Order Stability and Change Alice C. Harris Word Order Harmonies and Word Order Change in Georgian Carol F. Justus Word Order and the First Person Imperative Part IV. Variation and Change Marcello Cherchi Variant Order of Surface Segmentables on the Border between Morphology and Syntax: the Case of Preradicai Verbal Morphology in Kartvelian Oliver Currie Word Order Stability and Change from a Sociolinguistic Perspective: the Case of Early Modern Welsh Mark Janse Stability and Change of Word Order Patterns Over Time Willem Koopman & Wim Van Der Wurff Two Word Order Patterns in the History of English: Stability, Variation, and Change Anette Rosenbach & Letizia Vezzosi Genitive Constructions in Early Modern English. New Evidence from a Corpus Analysis Index of names Index of subjects Index of languages

119 133 165

187

203

231 259

285

309 315 321

ACKNOWLEDGEMENTS

The editors would like to thank Clare Orchard and Giovanna Pianese for editorial assistance, and to acknowledge financial support from the Department of Modern Philology of the University of Naples "Federico II".

INTRODUCTION

The subject discussed in this volume was originally proposed for a workshop on "Stability, variation and change of word-order patterns over time" held at the XIIIth Conference of the International Society for Historical Linguistics (Düsseldorf, 14th August 1997). Guidelines were prepared centred around two main problems: (a) the relationship between stability and variation of word-order (WO) patterns and (b) the relationship between synchronic variability and change of WO patterns. This partition, which as a mere device for organizing the debate certainly had an element of artificiality, nonetheless reflects the implicit assumption that variation and variability are the intermediate link between stability and change in a manner that is far from being clear and uncontroversial. Each problem prompted further issues. Thus participants were asked to present criteria for assessing stability and variation and to discuss the nature of stability and variation; in particular, the impact of internal and external factors and the role of grammaticalization of WO patterns in determining stability were suggested as possible factors affecting the nature of stability and variation. As to the relationship between synchronic variability and change, two factors were proposed as further themes for debate, i.e. the relationship between internal flexibility of WO patterns and change and the influence of pragmatic factors on trends of change. The present collection of papers includes some of the contributions presented at the workshop in Düsseldorf, as well as additional invited articles. One of the criteria inspiring the preparation of the workshop and of the volume was gathering an array of studies covering a significant range of languages, in order to have problems and conceptions specific to the various linguistic and philological traditions reflected in the discussion of synchronic and diachronic 'movements' of WO. It is, in fact, a basic belief shared by the editors that themes like WO flexibility and change, which are currently being debated in contemporary research on diachronic syntax, can benefit from detailed studies and fresh theoretical and historical visions stemming from diverse scholarly traditions. The range of languages investigated here includes Hamito-Semitic languages like Egyptian with its later stage, Coptic, an ancient Semitic language like Akkadian, Caucasian languages like Georgian, Mingrelian, Svan,

x

INTRODUCTION

Laz and Indo-European languages. Among the latter there are ancient languages like Hittite, Greek, Latin and modern ones like Russian, Welsh and various Romance languages. Germanic languages are represented by Old Norse, Modern Scandinavian languages and by English. The detailed study of WO patterns from this wide range of languages can balance, test and perhaps allow a critical reconsideration of large-scale typological pictures that - from Greenberg's, Vennemann's and Givón's works onwards - have characterized a non-negligible part of contemporary research in diachronic syntax. Moreover it can also prompt a critical consideration of formal models of syntactic change like the ones put forward by Lightfoot. Almost all contributions concern languages which are no longer spoken. The special theoretical and methodological foundation and properties of what one may therefore term 'dead-language linguistics' are as yet relatively underplayed in linguistic analysis. Three aspects seem particularly relevant to the issues under discussion in the present volume, and call for caution in appreciating linguistic features: (a) The properties of written texts as 'testi di lingua': considerations of genre and style, possible hyper-dialectal or interdialectal status, relationship with spoken language, even ecdotics (variatio here acquires a very different meaning). (b) The problem of the written text, and of its transmission, in linguistic diachrony: a chronological sequence of written texts does not necessarily constitute a diachronic continuity that can be taken, without special unscrambling, as a database for tracing grammatical change. (c) The methodological and 'technical' problems of analytic and heuristic work on ancient languages which are attested only in a limited corpus of written texts, and in particular the methodological position of non-attestation of linguistic features. The book exhibits a lively concordia discors. Although the overall perspective is functionalist, formal paradigms are assumed in some papers as a stimulating source for more general reflections. Some articles show a pronounced orientation towards a truly 'historical' analysis of data, while others have a stronger theoretical inclination. On the whole, however, a nondogmatic spirit pervades all papers: rather than proposing all-encompassing models or sweeping generalizations, the authors tend to raise questions, point out problems deriving from current models and give outlines for further possible directions of research. Non-dogmatism and a certain amount of skepticism may not be unwelcome. If the reflection on stability, variation and change of every single aspect of syntax inevitably leads to general problems of both synchronic and diachronic linguistics, with a still open bicentennial debate behind them, the

INTRODUCTION

xi

state of the art in WO studies does not unravel less controversial scenarios. A centuries-old subject of research, WO (and perhaps it would be better to adopt the more comprehensive term 'constituent order', which would also better suit the range of phenomena covered in this volume) embraces a series of phenomena of primary importance for the understanding of syntactic structure. The models elaborated in the last decades show conceptual similarities but also differences in their assumptions and results that make a comparison difficult. In recent years, rather than producing a cumulative growth of knowledge, research on WO has instead leapt forward partially neglecting former results. This is hardly a unique situation in the history of sciences, and linguistics is no exception. Sometimes recent models opened new views, like typological studies on correlations or 'harmonies' of WO patterns, which have revealed regularities linking individual structures and produced an attempt to globally represent linearization processes. Yet this result had its costs, as linearity was perhaps inevitably - singled out as a parameter independent of structure and this was at times represented simplistically with consequent misrepresentation of data (cf. the observations in Deutscher's, Loprieno's, Poppe's, ShishaHalevy's, Sornicola's papers). Although a conspicuous mass of reasearch, with data from a vast number of languages, has been carried out over the last twenty years (cf. Hawkins 1983, Hawkins 1994, Harris & Campbell 1995), it is perhaps possible to question the soundness of the overall representation of diachronic 'cycles' and single 'stages' of a given cycle. This may be due, not only to the difficulty in determining a specific linear property, for any given period of a language, but also to the more general theoretical problem of how to represent individual structures and correlations of structures over time (see further on). Thus it would not be too pessimistic to say that we still do not know much about continuities and discontinuities of WO patterns over time and their range of synchronic flexibility. Even after twenty-five years Charles Li's evaluation of Greenbergian typology does not seem out-of-date. In his view, although "invaluable as a pioneering source of stimulation in modern studies of language typologies", it "does not constitute an unshakable foundation for diachronic syntax". Moreover "the SVO, SOV, VSO typology does not provide a trichotomy of the vast majority of languages in the world. At best, it represents three points of idealization in the continuum of word orders and a wide array of ancillary grammatical properties. The precise nature of this continuum remains somewhat of a mystery to date" (Li 1977:xiv).

xii

INTRODUCTION

Further problems with a more specific explanatory dimension can be considered. For example, the explanations of typological 'cycles' and correlations of WO patterns are often based on functional factors of a semantic, pragmatic or cognitive nature (like Hawkins's principle of ease of perception), which are singled out as having an independent and unconditional role. Functional factors - and in particular semantic and pragmatic ones - certainly play a central role in the functioning of syntax. Yet, it is possible to entertain doubts as to whether they can per se explain the complexities of macroscopic diachronic processes like the setting up of a new type or the alignment of a pattern with another within the same language. These phenomena can hardly be observed by linguists, as they go through large-scale historical processes like those that bring about new cultural norms within a society. The locus of the cognitive factors considered in functional-typological models is instead the speaker conceived as an abstract individual who is always and everywhere the same and as such is placed 'beyond history'. It seems therefore profitable to reconsider the problem of 'movements' of WO patterns over time from a perspective which combines the historicalphilological approach with the functional-typological one. In this view, the search for 'cycles' and harmonies can be usefully integrated with studies on individual structures/patterns. Yet an attempt to bring together the two traditions of research meets the preliminary difficulty of having to relate assumptions, methods and results that may be very different and in some cases even at variance. On the other hand, a reflection on stability, variation and change has to face the additional difficulty of having to handle concepts of broad epistemological interest that are not unique to linguistics and that - although frequently employed in human sciences - seem to escape unitary definitions. In fact, they are found more or less implicitly in various historical, sociological and economic models, with different meanings. Even their use within the same discipline may considerably differ. In recent historiographical models invigorated by a critical reception of structural ideas, the notion of stability is connected with 'permanence', 'duration' of a historical structure. Some French scholars identify periods in European history described as "grandes plages d'équilibre, qui, dans le long terme, anulent le mouvement" (Burguière 1971:v), which nonetheless may not necessarily include a true 'stationary state' but also minor periodical cycles of change. What seems particularly interesting, however, is that within contemporary historiographical models 'stability' - be it conceived as permanence /

INTRODUCTION

xiii

duration or recurring cycles of change - is part of a more general interest in change and the times of change. It is in mathematical and physical sciences that stability is considered per se as a first rank theoretical problem. The properties of being in 'stable' vs 'unstable' equilibrium are defined for a simple pendulum in lower upright position: "pour tout voisinage arbitrairement petit de l'équilibre, le pendule demeure dans ce voisinage pour toutes les perturbations suffisamment petites (les perturbations affectent la position et la vitesse, toutes deux nulles à l'équilibre). Le pendule est aussi en équilibre en position verticale supérieure, mais à l'opposé du cas précédent, cet équilibre est instable" (EPHU, 2: 2449a). This meaning, which has an obvious intuitive content, seems very close to the one that is found directly or indirectly associated with the terms 'stability, 'stable' in various human sciences. Likewise in economics the notion of 'equilibrium state' implies the possibility of minimal fluctuations not substantially affecting it. Here 'stability conditions' are defined as "the conditions for a system to tend to revert to its original conditions after a disturbance" (Black 1997:440). The equilibrium state to which a system reverts is not necessarily a stationary state, but may also be "a steady-state growth path or some form of limit cycle" (Black 1997: 440). Incidentally, the notion of 'cycle' is one of the economic models that have admittedly influenced conceptual tools of contemporary historical reasearch. More than other sciences, mathematics has assigned a central place to the theoretical interest in stability developing many formal models. Yet even mathematicians admit that "there is no unitary rigorous definition of stability encompassing all the meanings of this word in contemporary scientific research" (Milani 1981:418 [translation ours]); however, various models have been worked out, each amenable to a particular theory. This is why some scholars - following Thorn - suggest a definition of meta-mathematical nature: "a property defined within a mathematical structure is a property of stability or instability if it either positively or negatively determines the susceptibility of a mathematical object of the structure to be employed as a model of reality" (Milani 1981:418 [translation ours]). In this sense the notion of stability is connected to the traditional scientific problem of the possibility to predict events by mathematical models. In Milani's words, "the problem of stability is to predict the maximum limit to errors induced by measurements, simplifications of calculations, and schematizations of mathematical models" (Milani 1981:419 [translation ours]).

xiv

INTRODUCTION

In post-second-world-war diachronic linguistics, different models of stability seem to coexist, some implying the simple intuitive notion of 'stationary state', others the idea of minimal oscillations that do not alter the overall equilibrium of a system or - in a different version - of a grammar (cf. also Currie's considerations on stability and its relation to variability, in this volume). If elementary physical intuitions may ultimately be latent in such representations, the decisive role played by traditional linguistic ideas (like the notion of a 'dynamic statics of language' of the Kazan School and of the Prague School, or the more recent sociolinguistic conceptions of variability) is certainly indisputable. The meta-mathematical conception of stability also seems to have found its way into diachronic linguistics. It can be felt in Lightfoot's view that "languages are consistently changing gradually and in a piecemeal fashion, but meanwhile grammars remain in equilibrium, unchanged in their structural properties" and that nonetheless from time to time grammars "undergo more radical, catastrophic restructuring, corresponding to new parameter settings" (Lightfoot 1991:173). In fact, what Lightfoot has presented in his works is a conception of grammar as a meta-mathematical model of stability, in that the setting of parameters can be considered to determine the limits of possible fluctuations or variations with respect to predictions made by grammar. Conceptions of change are no less diverse than those of stability. General critical perspectives come from the productive debate on the role of structure in the study of change, which has affected historical research of the second half of XXth century. For example, it has been observed that there is an inevitable ideological component in every representation of change. As the French historian Burguière said (1971:iv), the dynamic models of change have developed out of the philosophical context of evolutionism and meet the expectations "d'une pensée militante qui attend de l'histoire une vérification expérimentale de l'idée de progrès"; besides, they project the system of values of the industrial world, "son culte du changement et de l'innovation". This ideological aspect, which could have the salutary effect of relativizing representations considered absolute and objective, has been almost completely overlooked in contemporary diachronic linguistics. Whatever the definitions of stability and change may be, the notion of variation is associated with both, in a way that raises the question of criteria for studying the amount and/or the nature of variation itself. Insofar as it can be 'kept under control' or even disregarded (either because it is of modest dimension or because it is unordered and chaotic) no real change is believed to be able to set in. An example of this view can be found in Lightfoot's models,

INTRODUCTION

xv

where variation seems to be understood as a minor chaotic change. But such representations are not exempt from problems. More in general, it is possible to wonder whether 'variation' is a notion with a clear and strict definition in human sciences and in particular in linguistics and whether we explored all its possible implications for stability and change. The issues discussed so far call into question the position of both synchronic and diachronic linguistics with respect to themes of general epistemological nature. In the last decades linguistics seems to have been untouched by the vast epistemological debate that has shaken and perhaps rejuvenated other human sciences, like history and geography, and even "hard sciences" like mathematics. Yet linguistics could play a significant role in the general reflection on themes like stability, variation and change, in spite of biases like the prominence given by structural research to the aspect of structural invariance and the delay in facing as central the study of variation within structure. However, considering that the notions of stability, variation and change belong to a general inventory of conceptual tools that are not specific to linguistics, they must be interpreted, before being used, within this field of research. If they are considered properties or attributes, it is necessary to define the exact domain in which they hold. Here lie problems that occasionally come to the surface in various linguistic disciplines. A first dilemma is whether stability, variation and change should be considered internal or external properties of structures. In the first interpretation stability, variation and change are related to the shape of the structure, its inner balance or unbalance, its potentialities to persist or be altered, apart from the time factor (this meaning is particularly evident in Cherchi's, Deutscher's, Janse's, Rosenbach & Vezzosi's and Sornicola's papers, where it is also found in various combinations with the other). In particular, an example of an internal property is the notion of typological stability referred to, from different perspectives, in Deutscher's, Harris's and Justus's articles. On the other hand, the consideration of stability, variation and change as external properties leads to focus 'structure' as a function of time, i.e. in this case time is the main factor determining stability, variation and change of structure (this meaning seems to be predominant in Currie's, Loprieno's, Poppe's, Shisha-Halevy's and Sornicola's papers). Although the two alternatives can be found intertwined in various studies, the distinction has a logical interest that can help to clarify the notions in question. The two perspectives perhaps also imply different views of classic problems in diachronic syntax, which have been variously addressed in this volume, like the abrupt vs gradual nature of syntactic change and the

xvi

INTRODUCTION

possibility of syntactic change being caused by external factors (such as borrowings, literary influences, etc.). A second dilemma is whether stability, variation and change are properties of structures conceived as 'traditional techniques' (in the sense of Coseriu 1982), i.e. 'external' products of the discourse activity that can be observed in historical texts, or of structures conceived as mental objects belonging to grammars as theories of speakers' competences. An examination of the bulk of literature on linguistic change shows that the two perspectives, pursued in different scientific environments, involve different systems of assumptions, methods and results. Again, it may be useful to be aware of this bifurcation in order to better understand and compare the various approaches. The papers in this volume reflect either option. Benoist, Janse, Poppe, Rosenbach & Vezzosi, Shisha-Halevy, Sornicola refer to stability, variation and change as properties of structures conceived as traditional techniques, whereas Faarlund and Harris view these notions as properties of structures conceived as mental objects of grammars. Koopman & vand der Wurff, on the other hand, provide a comparison of the different results obtained from applying the two approaches. It may be useful in this context to review a number of theoretical implications of the two options that seem to be strictly interrelated, as they are relevant to the overall discussion. In the first approach, which usually adheres to historical models, (a) structure is considered as the product of the impact of a given socio-cultural context on discourse and indirectly on language; (b) a gap is envisaged between text and the underlying grammar/competence that is not believed easy to bridge; (c) the syntactic, semantic and pragmatic analysis of the phenomena of order does not necessarily assign any model of mental representation to structure, nor a fortiori is any attempt made at reconstructing diachronic sequences of mental representations of structures. Besides, following the route of Geneva structuralism (d) in this approach the relationship between synchrony and diachrony is often viewed as a problem per se. An interesting symptom of the latter opinion can be detected in the various indications of theoretical and methodological difficulties in reconstructing diachronic paths of order (cf. Benoist's, Poppe's, Shisha-Halevy's papers in particular). On the methodological level the articles of this group are also characterized by a more conspicuous interest in the analysis of registers and styles, as well as of possible cultural influences on continuities and discontinuities of WO. Another consistent set of assumptions seems to underlie the approach envisaging structures as mental objects related to grammars. Note that no

INTRODUCTION

xvii

necessary endorsement of the formal models in the generative framework is implied here. In fact, syntactic change can also be conceived in various functional models in terms of change in cognitive representations (see for example here Faarlund's and Harris's papers). Interestingly, however, scholars supporting this perspective from a functionalist stand often criticize the innatism of generative models and claim, with Harris' words, that "change be understood... rather as the result of language fulfilling its discourse and communicative functions" (Harris & Campbell 1995:45). To the extent that this approach is centred on structures as mental objects (or - in a different version - as cognitive representations) grammar is considered to be the locus of change. As a consequence, some important assumptions are made about two problems of unending theoretical interest and vital importance for the study of stability, variation and change, i.e. the conception of diachrony and the relationship between synchrony and diachrony. Diachrony is conceived as a sequence of grammars/systems of cognitive representations. Even more far-reaching is the idea that (apart from resorting to a model like reanalysis, which has the specific aim to explain syntactic change) diachrony can be studied in principle with the same theoretical tools employed for contemporary synchrony, where mental images of linguistic phenomena can be inferred by testing native speakers. Here lies a major dividing line in the approaches to understanding stability, variations and change. The adaptation of diachronic to synchronic models is another aspect where some functional approaches may come close to formal ones like generative grammar. In fact, in both directions of research, whatever their real or assumed discrepancies, an influence of the Prague School on the relationship between synchronic and diachronic methods can be detected, either directly (possibly, through Roman Jakobson) or indirectly (perhaps through the influential paper of Lehman 1968). A well-known passage from the Prague School Thèses seems relevant to this supposition. Drawing attention to the idea that the conception of language as a functional system must be applied not only to synchronic analysis of present-day phenomena, but also to the study of past stages of language, the Prague School scholars stated:

xviii

INTRODUCTION

On ne saurait poser de barrières infranchissables entre les méthodes synchronique et diachronique comme le fait l'école de Genève. Si l'on envisage en linguistique synchronique les éléments du système de la langue du point de vue de leurs fonctions, on ne saurait juger non plus les changements subis par la langue sans tenir compte du système qui se trouve affecté par lesdits changements. Il ne serait pas logique de supposer que les changements linguistiques ne sont que des atteintes destructives s'opérant au hasard et hétérogènes du point de vue du système. Les changements linguistiques visent souvent le système, sa stabilisation, sa reconstruction, etc. Ainsi l'étude diachronique, non seulement n'exclut pas les notions de système et de fonction, mais tout au contraire, à ne pas tenir compte de ces notions, elle est incomplète {Thèses, 8-9, [italics ours])

The emphasis on function rather than structure in the study of linguistic change is certainly a common heritage of all traditions of functionalism, but the idea that structural rearrangements determining a more stable equilibrium of the system underlie most changes - an idea that contains an element of teleology - seems to have more deeply affected the functional-typological approaches to diachrony. Reanalysis is a pivotal issue in contemporary views of change as aiming at a more stable equilibrium of grammar. Although the model had already been described in Delbrück's and Paul's works, it has gained a fundamental place in the study of diachronic syntax only in the last decades. Faarlund and Harris independently assert the double dimension of stability and change which characterizes reanalysis. It seems interesting, however, that - as in other contemporary studies - the main implicit perspective of investigation seems centred on change rather than stability. Of special interest in this respect is the paper by Koopman & van der Wurff. It is an attempt by scholars inclined to see structures as textual objects to discuss the alternative point of view which emerges in recent research on change in grammar. The authors address important theoretical and methodological problems, such as the abrupt vs gradual nature of change, and in particular the possibility to identify cut-off points in change of grammar. Consistent with the textual orientation of their approach, they detect - for the patterns of order investigated - a much slower path of change than is usually admitted and, more generally, argue in favour of the importance of gradualness and long-term coexistence of competing structures in processes of change. But the divergencies in modelling change might be - at least in part - an effect of the clash of theoretical paradigms. An interesting range of issues concerning the notions of stability, variation and change and their interrelations emerges from the articles in the volume. In some stability is considered a key concept, though to different degrees, with a focus either on its synchronic (as in Currie's and Poppe's papers) or diachronic (as in Deutscher's, Shisha-Halevy's, Sornicola's papers) dimensions. Despite

INTRODUCTION

xix

these minor divergences, however, there is a common interest in pragmatic principles that contribute to stability and in factors which limit or resist change. Baudouin de Courtenay's conception of a 'static' vs. 'dynamic stability', which was defined with respect to phonetics, could find a continuation and a reformulation here for WO: "Static stability, i.e. resistance to changes at a given moment in the existence of a language. Dynamic stability, resistance to historical changes. A limit is placed on the changeability of WO" (for the original passage dealing with 'sounds' cf. Baudouin de Courtenay 1876-77:94). The notion of 'WO change' is addressed from different functional perspectives in various contributions, which raise a multitude of problems (cf. sections III and IV). It seems worth noting that, whatever the theoretical approach adopted, change is unanimously considered to be related to stability. In particular it is independently argued by many authors that in any process of change continuities and discontinuities are strictly interlaced, an idea that results from detailed observations of historical data. In some papers, however, the complicate interplay of permanence and alteration (or substitution) of structures is interpreted in the light of two models that have played a fundamental role in contemporary syntactic research, i.e. reanalysis and/or grammaticalization (for a discussion of their relations cf. Harris & Campbell 1995:92). A further leit-motiv of theoretical interest is the impact of the form function relation on WO change (cf. Faarlund's, Harris's, Justus's and Rosenbach's & Vezzosi's papers). Currie, Janse and Koopman & van der Wurff draw attention to the problematic definition of WO change from various perspectives. They indicate the crucial importance of analyzing textual and cross-textual WO variations in synchrony and diachrony for understanding change. A number of aspects of variation are discussed, like synchronic oscillations, multiple diachronic and diatopic variation, the permanence of patterns and their competition with others. The problematic relation between stability and variation is especially dealt with by Currie, whose contribution brings to mind Coseriu's thoughtprovoking aphorism "linguistic change doesn't exist". An issue addressed in all contributions is the impact of semantic and pragmatic functions (which are variously interpreted as 'theme', 'topic' and 'focus', or as perceptual and cognitive factors) on stability, variation and change of WO patterns. While there is a consensus on the general importance of these functions in determining stability, variation and change, the role ascribed to them is a matter of debate. Two main options emerge. According to some, pragmatic factors are initiators of WO change. This idea, which finds broader support in the volume, seems to reflect a widely accepted view in

xx

INTRODUCTION

functional-typological literature, where it can be traced back to the influential paper of Vennemann 1974. Other contributors, however, maintain that semantic and especially pragmatic functions do not trigger WO change, but rather principles of linear structuring always present across time, and delimiting - in the Prague School terms - the 'field of tension' of a structure, i.e. its range of variability or flexibility; therefore as such they are factors of synchronic and/or diachronic stability. This dichotomy is interesting and poses a problem that perhaps deserves further scrutiny. Taking into account the issues discussed so far, the contributions have been arranged in four sections, respectively: I. Stylistic and pragmatic principles in stability, variation and change, II. Stability and diachrony, III. Reanalysis, grammaticalization and change, IV. Variation and change. The three articles in Part I. share the leit-motiv of the fundamental importance of stylistic and pragmatic principles for the understanding of synchronic variability and possibly trends of change. They reveal various interesting theoretical and methodological approaches and different nuances in their conclusions. Benoist's contribution examines - from a philological and functional standpoint - the fundamental stages of 'internal' and 'external history' in the formation of Modern Russian WO. According to the criteria of Prague School structuralism, diachronic representation is conceived as the analysis of synchronic states, each of which constitutes a stratum. The Praguian and more generally Pan-Slavic functionalist approach is evident in the attention paid to differences in register and style, as well as in the insistence on the interplay of syntactic and pragmatic functions and particularly in the central role given to intonation. The historical dimension emerges from the detailed presentation of the slow and complex development of the principles governing the WO in Modern Russian. With particular regard to the literary language, this process is studied considering the influence of external factors such as rhetorical normative models as well as foreign models - above all French - still in use up to the first decades of the XIXth century. Benoist underlines the limitations in the descriptions of the historical process under study, which have not taken into account these factors. The internal dimension of the process is also considered by analyzing the arrangement of the constituents of the main sentence, of the subordinate sentence and of phrases. Benoist questions the notion of 'freedom of WO' in Modern Russian as a general typological feature. He claims, on the contrary, that its validity is limited to the Lomonosov period.

INTRODUCTION

xxi

Loprieno's article considers a possible shift from VSO to SVO order in the long-term diachrony of Egyptian. This gradual shift reaches its peak in Coptic, the final stage of Late Egyptian. Loprieno makes use of current synchronic and diachronic typological models. However, he mentions some difficulties in describing Coptic as a SVO language. Furthermore, he addresses the theme of 'predictability' of change, and compares the trends of movement of Egyptian WO to analogous trends of other VSO languages that have likewise developed SVO patterns. Important historical problems are also discussed, such as the possibility of Greek influence on the WO patterns of Coptic, which makes Loprieno favour a productive and non-mechanical influence. The emphasis in this work is on the synchronic flexibility of the SVO structure in the life-span of Coptic. The author maintains that the SVO pattern has in synchrony a paradigmatic nature, which is not weakened by the existence of alternative structures with the VS pattern. In fact, these constructions have a relatively limited synchronic functional load. Special attention is given to a particular structure with the VSO pattern, the rear extraposition of S by means of the nci-marker. This occurs frequently and has the function to rhematize an informationally heavier S, with a symmetrical increase in the thematicity of the SV-predicate phrase. According to Loprieno, the nci-construction is no exception to the unmarked SVO order. In fact, it reorganizes patterns of information distribution in the sentence "according to the same pragmatic pressures that motivated the change from VSO to the SVO order in the history of Egyptian". This suggestion seems to imply the existence of mechanisms which conspire to recompose the theme rheme sequence of the sentence in synchrony and which also work in diachrony. This amounts to saying that pragmatic factors have a predominant explicative power. The symmetry suggested here between synchronic and diachronic forces is also interesting. Poppe's article highlights the importance of an 'overall pragmatic principle' for stability of WO within a given synchronic stage of a particular language. This point is demonstrated through an analysis of syntactic structures and their relative patterns of order in Middle Welsh, which are shown to be consistently governed by pragmatic principles (except in certain types of clauses). A special emphasis is given to performance-related factors; in particular, Poppe underlines that topichood is not an automatic process independent of the speaker's communicative intention, but that - on the contrary - the speaker's choice is crucial in selecting a constituent in order to assign topichood. Thus the overall pragmatic principle is to be understood as distinct from stylistic and random variatio, as well as from automatic selection

xxii

INTRODUCTION

for the fronting of constituents, or types of constituents, in specific syntactic contexts. Based on a corpus of stylistically differentiated texts, Poppe's analysis shows the considerable structural complexity of WO variation within a specific synchronic stage and in so doing opens problematic scenarios for the reconstruction of diachronic trends. In fact, the statistical data presented point to a conspicuous amount of variability in neutral main clauses, in which neither a basic nor a dominant order can be identified. 'Preferred' orders only emerge for negative main clauses and subordinate clauses with the format Conjunction + finite V. The only regularity identifiable for neutral affirmative main clauses is that V is preceded by (at least) one other constituent and a pre-verbal particle (or 'Verb-second' order according to another descriptive approach, Willis 1998). This structure, which has been a matter of hot debate in Welsh grammaticography (where it has been called 'abnormal sentence'), is of interest to diachronic typology, in that it marks an apparent discontinuity in the development of Welsh WO patterns. Poppe maintains that this structure "is abnormal only in comparison with dominant neutral VSO order in Modern Welsh, and presumably in Old Welsh as well", an opinion of methodological and theoretical consequence for the study of this supposed discontinuity, and of discontinuities in general. Being aware of the possibly deceptive character of some discontinuities - which in fact may fade away when they are investigated with attention to stylistic variation across text (cf. Currie's contribution), Poppe cautions against premature reconstructions of diachronic paths of WO and reinforces the point of view that - before sketching trends of change - syntactic phenomena should be studied synchronically and with reference to the complexities of their textual distribution (for some important refinements of Poppe's positions, mainly with regard to 'multiple frontings' and Verb-initial clauses, compare now Willis 1998). The articles in Part II. argue in favour of stability being considered central in the diachronic study of WO. On the basis of data from the Restsprache Akkadian, Deutscher addresses a problem which reverses a traditional approach of historical linguistics, predominantly oriented towards change, by drawing attention to why in the case he investigates no change at all occurs over two millennia. According to current typological models, Akkadian should be an ideal candidate for change, as it shows very inconsistent patterns of order. The fact that Akkadian goes against expectations is all the more puzzling since the language underwent a loss of case-endings, a factor traditionally (though not unanimously) believed a possible determinant of change from SOV to SVO order. Deutscher claims

INTRODUCTION

xxiii

that, despite the 'unfavourable' conditions mentioned above, stability of WO was protected by other factors, such as semantic and pragmatic ones, which prevented ambiguities in the identification of S and O, as well as a syntactic factor like verb-agreement. Of special interest here is the idea that typological 'inconsistency' does not imply instability. The author refutes three lines of argument which attempt to predict and justify the development of harmonic patterns from conditions of typological inconsistency. In particular, he rejects the explanation based on processing mechanisms, arguing that the inconsistent and supposedly difficult-to-process Akkadian structures investigated can function perfectly (i.e. they are common and productive) - over the whole documented history of the language because of some notable syntactic factors, which are then carefully examined. Observing that "isolating word order from other structural properties of the language can create a misleadingly simplified picture", Deutscher concludes that the Akkadian case shows that stability can be accounted for by the complex functioning of structural devices. The fact that this complex functioning has often been ignored helps to understand "why so many of the crosslinguistic word order universals are observed more in the breaking than in the keeping". Shisha-Halevy's article is characterized by an approach that is decidedly critical of functional-typological models and their implications for the study of stability, variation and change of WO. The author's arguments are based on his philological experience of Egyptian and in particular Coptic. From the perspective of a linguist engaged in the study of 'linguistic diversity' rather than of 'universals', he brings up once more theoretical and methodological problems of syntactic analysis considered classic in various scientific traditions. He warns against the use of supposedly universal categories, as this may create the deceptive idea that languages can easily be compared; additionally this assumption leads to the levelling of specific features of languages. A general problem is addressed here that calls into question the possibilities and limits of linguistic description and comparison. This point is exemplified by the analysis of a number of structures from the long-term diachrony of Egyptian. The author suggests that a language like Egypyian can only be analyzed through a descending distributional procedure "from full-text level down". Another theoretical issue dealt with - which is more specifically related to WO - is the distinction between 'syntagmatic sequencing' and 'structural sequencing'. Order of constituents is denied the status of a "prime, independent, overruling, governing or motivating feature of syntax" that is endowed in itself with typological value. This is a radical position, with some

xxiv

INTRODUCTION

important consequences of both theoretical and methodological import. One of particular interest is the impossibility to extrapolate WO patterns from structural patterning, a thesis that is in agreement with Deutscher's and Sornicola's appeal for the need to consider order and structure together. After reviewing constructions that are characteristic of various historical stages of Egyptian, Shisha-Halevy detects a "remarkable stability of sequencing within the enveloping diachronic system", though he underlines that the notion of 'stability' is not to be perceived as absolute. However, his main conclusions have implications for the possible time span of linguistic stability, insofar as the property described here can be observed over a period of four millennia. Stability is the key theme of Sornicola's article, which is devoted to an examination of WO patterns in main clauses in the history of the Romance languages, in particular French, Spanish, and Italian. The discussion sums up the results of syntactic-pragmatic and statistical analysis of corpora of texts which cover different chronological stages of the three languages, from the Middle Ages onwards. The results are discussed against the background of standard accounts from Romance linguistics. In addition, differences in register and genre are taken into account (for the contemporary stages, the variation of WO between spoken and written text is also considered). The author addresses various theoretical and methodological issues relevant to the study of stability, variation, and change in WO patterns. She argues, moreover, that such studies demand the use of dynamic models, which have to satisfy various prerequisites, mainly the necessity to consider order in close conjunction with structure and to adopt a perspective defined both 'microscopically' and 'longitudinally', in order to conduct detailed studies of WO in single languages or across a language family and over different chronological periods. Such a perspective, which stands in opposition to current typological models, will allow the description and evaluation of the nature and extent of oscillations in WO patterns over time. The microscopic and longitudinal analysis of the three Romance languages demonstrates a remarkable stability of the SVO order and of the pragmatic functions with which it is associated. This convergence of WO patterns between French, Spanish, and Italian over a long period a time would seem to contradict their conspicuous differences, and leads to a reconsideration of some problems of diachrony with their own long history, such as the path of change from Classical Latin to Romance. The articles in Part III. have in common the centrality assigned to reanalysis processes and/or paths of grammaticalization in WO change. Furthermore they all follow models that could be called 'pragmatic-typological', like those

INTRODUCTION

xxv

developed by Vennemann and Givón. They are characterized by the fact that pragmatic factors are considered main determinants for the initiation of mechanisms of change. In fact, the articles in this section, compared to others, show greater confidence in the view that pragmatic factors can decisively influence historical change. Faarlund analyzes two changes in North Germanic which he believes are both due to reanalysis. He assigns considerable importance to the discourse function of, and to the more general functional conditions for, a structure of order. In particular, he thinks that in many languages the weight of constituents and/or their discourse functions are responsible for the deviations from the basic and unmarked WO. The first phenomenon analyzed is the shift from OV to VO in North Germanic. According to Faarlund, the factor which has determined the process of change is the extraposition of O to the right of V when O is focalized and the O constituent is heavy. The author maintains that it is the relaxation of the conditions in the use of VO order which causes an increase in the frequency of this pattern, and the consequent reanalysis of VO as the basic form. After this process, the movement of O in the opposite direction is no longer possible and therefore the OV order disappears. A second point which deserves to be considered is the fact that in this case the change proceeds through an alteration of the order and not of the structure. In particular, the change leads to a better match of the new linear order with the principles of information structure. In contrast, what occurs in the second phenomenon investigated is a change of the structure, with the order remaining unaltered. The shift, observable in the history of Norwegian from Old Norse, is represented by Faarlund as a reanalysis, whereby an extraposed S was reinterpreted as complement of V, a process which was facilitated by the loss of case-endings. Faarlund also tackles an interesting theoretical issue which concerns the permanence of residual properties of Subjecthood in the post-verbal NP of the VS structure of Modern Scandinavian languages. His suggested solution uses a model of modularity from Autolexical Syntax. The harmonies of WO and their relevance for processes of change are at the core of Harris's work. This is a strictly typological theme with a broader cut and is tackled by combining typological methods with textual analysis. Two objectives are proposed: (a) to describe WO change from Old to Modern Georgian, (b) to show the relevance of 'construction reanalysis' for WO harmonies. In particular, this process is claimed to be the origin of the change in WO in two dyadic configurations in Kartvelian languages, i.e. the comparative adjective construction in Georgian and the V - Aux structure in Laz.

xxvi

INTRODUCTION

Notwithstanding the difficulty in ascertaining the order of the major constituents in Old Georgian, Harris demonstrates that "most of the unmarked orders are those harmonious with VO", with Po and Vaux the only exceptions. In Modern Georgian, on the other hand, the main constituents appear in SOV order. The sequencing of the other constituents is far more complex, with numerous cases of ambivalence and multiple options. In the case of the linear relationship between the Standard of comparison and Adjective, where Old Georgian showed an Adjective - Standard order, in harmony with VO, Modern Georgian has an additional Standard - Adjective order. Harris claims that such a pattern has originated through a process of 'construction reanalysis', insofar as the comparative construction conforms to the pattern of the Adjective + Complement order. The same general process is said to be responsible for the V - Aux order in Laz. In this case the pattern develops in correlation with the order of the embedded clause and of the verb of the main clause. In effect, the reanalysis of the verb of the matrix clause as an auxiliary is a diachronic process found in various languages. The author gives examples from Japanese and compares them to the Laz ones. Of significant theoretical interest is the suggestion that processes of reanalysis like the ones described not only create, but also preserve and even extend WO harmonies. Justus's work raises a stimulating problem, which has not been explored in depth previously, i.e. grammaticalization of the first person imperatives functionally defined as 'directives' - in Indo-European languages. It concerns a complex process of change, studied along two paths: (a) the emergence of new mechanisms for the coding of first person imperatives from older verbal paradigms with different imperative suffixes for the grammatical person; this development was achieved through the creation of new auxiliaries gravitating around semantic nuclei such as 'let', 'get', 'give'; (b) the change of WO correlated to such a slow shift. This process of grammaticalization is accomplished over an extremely long period of time, as it spans from Hittite texts of the second millennium B.C. right up to contemporary English. Justus observes that it is part of a broader trend of innovation, which also affects the strategies of codification of the second and third persons of the imperative. The overall trend shows a long-term persistence of multiple coexisting strategies, which are often realized by variant morphosyntactic and semantic structures. Comparing the path of grammaticalization followed by directive auxiliaries in Indo-European languages to that of directive imperative structures of Japanese, Justus notes that in the former the new auxiliaries are placed at the beginning of the sentence, while in the latter directive auxiliaries are placed at the end. This conforms to typological expectations, insofar as the Japanese order is

INTRODUCTION

xxvii

consistent with that of a V-final language, while the order of Indo-European languages reflects "the movement away from verb-final to verb-initial patterns". However, the author cautions against all-too-easy comparison of languages as well as against premature sketches of long-term trends, stressing the importance of taking into account the form - function relationship in diachronic analysis. Part IV deals with the relationship between variation and change. Synchronic variation and parameters underlying variability in linearization are the main themes addressed by Cherchi, in his discussion of one aspect of Kartvelian verbal morphology, i.e. the order of elements preceding the verbal root. The article takes up a theoretical issue of great consequence for modelling variation and change, i.e. the consideration of relationships between grammatical levels, and more particularly of phenomena in transition from one level to another. The author investigates alternative arrangements of pre-radical morphemes of the verb complex in Old and Early Middle Georgian, Mingrelian and Svan. These reflect aspects of the broader phenomenon of tmesis. However, the chief aim of the analysis is an investigation of orders deviating from those standard in tmesis, insofar as their study can help to understand "the limits and types of flexibility in morpheme ordering". This in turn constitutes a prerequisite to determining the parameters underlying variability. The range of variation in pre-radical constituent ordering is considerable. Similarities and differences emerge for the patterns of order in the three languages, which are partly due to differences in the inventory and function of pre-verbs. In particular, the author suggests that the flexibility of ordering is directly related to the size of the inventory of pre-verbs and their combinatorial possibilities. Variation of order may also affect the structures of individual languages - such as the asymmetry in Svan between the inner and the outer layer of pre-verbs, which the author believes to be related to a difference in the respective paths of grammaticalization. From a theoretical perspective, his conclusion is also relevant that the languages investigated cannot be placed along a single scale of grammaticalization insofar as the phenomena in question are "governed by multiple (and different) grammatical categories". Currie explores the implications of various sociolinguistic models for the understanding of WO stability and WO change. On the basis of a detailed syntactic, pragmatic and stylistic analysis of a corpus of Early Modern Welsh texts, Currie also calls attention to the importance of studies of the internal stylistic coherence not only within textual genres, but also within single texts. In fact, the frequency of absolute V-initial structures in his Early Modern Welsh corpus demonstrates differences not only across textual genres (like in

xxviii

INTRODUCTION

Modern Welsh), but also within the same textual genre, and even in the works of authors sharing the same cultural background. The author also considers the impact of discourse-pragmatic factors and the possible functional equivalence of absolute V-initial order with two other constructions, the dummy subject construction and the S(pers.pron.)V order. He believes that this interchangeability may have constituted a syntactic and stylistic resource that authors "could use to extend radically the range of absolute-initial verb order, and so develop a new prose style". Currie raises various interesting theoretical issues. A large portion of his article is assigned to discussions on the concept of stability. In particular, he emphasizes that this is difficult to define and not necessarily related to the concept of invariance, since - as has been proven in sociolinguistics - "stable patterns of variation" can exist. For Currie the interest of such a concept consists in providing "a starting point for investigating WO variation and change from a range of different perspectives". In his contribution Janse gives a multi-faceted description of convergences and divergences in the ordering of pronominal clitics from Ancient Greek and Classical Latin to Modern Greek and Romance languages, within a framework which combines the philological approach with grammaticalization models. His emphasis is on the multiple pathways of grammaticalization of the position of clitics. Orders are compared not only with regard to different structural environments in different chronological phases of a language, but also with regard to synchronic dialectal variations within a single language. Janse's analysis shows that the same processes may recur at different times in different languages, but also that a cascade of innovative patterns of order - with respect to a structure assumed as the starting point for the investigation - may sometimes emerge even within a single language. Variation therefore assumes a fundamental role in Janse's contribution. It implies not only synchronic oscillation, but also oscillations that will become grammaticalized diachronically. Instead of 'change', the author uses terms such as 'gradual evolution', 'development', 'survival', as he declares himself to be sceptical about models of abrupt change. The overall process described illustrates the divergences from Greek and Latin patterns of order - conceptually represented by the well-known 'Wackernagel Law', which Janse reinterprets in the light of pragmatic principles. The development of the multiple structural options of Modern Greek and of Romance languages is a process conforming to two main principles: (a) the evolution towards a system with a verb at its centre and

INTRODUCTION

xxix

pronouns floating around the verb, immediately before or after it; (b) the tendency to grammaticalize the position of clitics with respect to the verb. Koopman & van der Wurff address the theme of variability and optional--y in grammar, on the basis of two case-studies from the history of English syntax, i.e. the relative order of the Direct and Indirect Object and the relative order of Direct Object and verb. The authors examine a characteristically historical problem, that is to say the need to study the survival of older patterns in a phase where they are no longer dominant. This necessitates not only a reconsideration of the problem of gradualness, but also a discussion of models of variability and grammatical optionality different from those suggested within the framework of generative grammar. Koopman & van der Wurff present a detailed description of the two changes from Old to Modern English. This results in an interesting theoretical discussion on the adequacy of the models of WO change which use the notion of 'grammar competition'. According to the authors, this notion is far from being satisfactory, for more than one reason: it does not allow an in-depth study of the intricacy of historical data and does not stimulate research into the reasons for discontinuities of change. Instead, they identify two areas deserving further scrutiny: "the nature and locus of the competition between the two grammatical systems and the length of timespans that can be involved". The latter problem, involving the relationship between abruptness and gradualness, has important implications for Koopman's & van der Wurff s argument. In fact, the two changes examined are much slower than adherents of the theory of grammar competition would admit. The permanence of patterns and non-linear pathways in linguistic change is also focussed in the paper by Rosenbach & Vezzosi which concerns the genitival constructions in Early Modern English. On the basis of rich textual documentation, the authors detect a trend in the history of the s-genitive which contradicts the widely accepted account, and argue for a very gradual substitution of the inflected genitive by the of-prepositional phrase. Their results lead not only to a new proposal for the overall configuration of the change - with a stagnation and a successive rise in the frequency of the structure - but also for a new chronology of the entire process. This confirms the need for great care when determining chronological boundaries for linguistic developments, and for further reflection on the relationship between stability and change. The authors claim that stability and change are strictly interconnected and substantiate their conclusion with an analysis of formal and functional aspects of the genitival construction. They identify stability as a

xxx

INTRODUCTION

property related to form and change as a property related to semantic and pragmatic functions. The issues debated certainly do not exhaust the range of problems which may emerge from the study of stability, variation and change of WO patterns. The editors, however, hope that this volume will stimulate further research on these themes.

Rosanna Sornicola, Erich Poppe, Ariel Shisha-Halevy Naples, Marburg and Jerusalem, September 2000

REFERENCES Black, John. 1997. A Dictionary of Economics. Oxford & New York: Oxford University Press. Burguière, André. 1971. "Histoire et structure". Annales 3-4.i-vii. Baudouin de Courtenay, Jan. 1876-77 [1972]. "Podrobnaja programma lekcij... v 1876-1877 uč godu, IKU, 1877" [A Detailed Program of Lectures for the Academic Year 1876-1877], quoted from the English translation in A Baudouin de Courtenay Anthology ed. by Edward Stankiewicz, 88-107. Bloomington & London: Indiana University Press. Coseriu, Eugenio. 1982 [1988]. "Linguistic Change does not Exist". Energeia und Ergon. Schriften von Eugenio Coseriu (1965-1987) ed. by F. Albrecht, J. Lüdtke & H. Thun 1,147-157. Tübingen: Narr. EPHU = Encyclopédie Philosophique Universelle, publié sous la direction d'André Jacob, Paris: PUF, II: Les notions philosophiques, volume dirigé par Sylvain Auroux, Tome 2. [1990]. Hawkins, John A. 1983. Word Order Universals. New York: Academic Press. . 1994. A Performance Theory of Order and Constituency. Cambridge: Cambridge University Press. Harris, Alice C. & Campbell, Lyle. 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: Cambridge University Press. Lehmann, Winifred P. 1968. "Saussure's Dichotomy Between Descriptive and Historical Linguistics". Directions for Historical Linguistics ed. by W. P. Lehmann & Y. Malkiel, 3-20. Austin: University of Texas Press. Li, Charles N. 1977. "Introduction". Mechanisms of Syntactic Change ed. by Ch. N. Li, xi-xix. Austin: University of Texas Press. Lightfoot, David. 1991. How to Set Parameters: Arguments from Language Change. Cambridge (Mass.): The MIT Press. Milani, Andrea 1981. "Stabilità/instabilità". Enciclopedia Einaudi 13.418-452. Torino: Giulio Einaudi.

INTRODUCTION

xxxi

Thèses = Mélanges Linguistiques dédiés au Premier Congrès des philologues slaves. Travaux du Cercle Linguistique de Prague 1 [1929]:7-29. Vennemann, Theo. 1974. "Topics, subjects and word order: from SXV to SVX via TVX". Historical Linguistics I: Proceedings of the First Lnternational Conference on Historical Linguistics (Edinburgh 2nd-7th September 1973) ed. by J. M. Anderson & C. Jones, 1. 339-376. Amsterdam: North Holland. Willis, David W. E. 1998. Syntactic Change in Welsh. A Study of the Loss of VerbSecond. Oxford: Clarendon Press.

PART I STYLISTIC AND PRAGMATIC PRINCIPLES IN STABILITY, VARIATION AND CHANGE

LA FORMATION DES PRINCIPES DE L'ORDRE DES MOTS DU RUSSE MODERNE EN TANT QUE PROBLEME DE STYLISTIQUE LITTERAIRE ET DE LINGUISTIQUE

JEAN-PIERRE BENOIST Institut National des Langues et Civilisations Orientales (Paris)

0. Introduction Nous proposant de décrire la façon dont se sont constitués les principes qui régissent l'ordre des mots en russe moderne, plus que ces principes en euxmêmes, leur formation et leur évolution, plus que leur état actuel, nous avons adopté une démarche diachronique et analytique, progressive, comme le furent les recherches linguistiques russes dans ce domaine, ce qui provoquera dans notre exposé quelques allers et retours. Mais cette approche permettra de saisir dans sa complexité la constitution du système linéaire du russe moderne dans son ensemble, où sont liées syntaxe, disposition relative et absolue des mots de la phrase (linéarité), intonation (moyens prosodiques) et catégories pragmatiques. 1. Naissance du russe moderne Ce qu'il est convenu d'appeler Sovremennyj Russkij Literaturnyj Jazyk ("Langue russe littérale moderne" - littérale plutôt que littéraire -, comme pour l'arabe, car à l'époque où le terme a été formé il s'agissait en fait de distinguer cette langue de culture des différents dialectes russes qui coexistaient avec elle) apparaît comme une synthèse du slavon (vieux slave d'Eglise) et du vieux russe, synthèse qui subit, selon les époques, l'influence des langues anciennes ou occidentales (en particulier le français au XVIII). Sa formation dans l'ensemble de ses systèmes remonte à la deuxième moitié du XVIII et est liée à la naissance dans l'utilisation de la langue écrite de besoins langagiers nouveaux, littéraires ou thématiques, qui apparaissent avec le règne des souverains éclairés Catherine II et Alexandre I. Cette synthèse entraîne le rejet des traits extrêmes du slavon, trop étrangers à la spécificité du russe, qui resteront propres à la langue de l'Eglise, et de certaines formes du vieux russe,

4

JEAN-PIERRE BENOIST

liées uniquement à la langue parlée, qui eux deviendront l'apanage du russe populaire. Cette formation de la langue moderne ne sera pas l'affaire de la seule seconde moitié du XVIII mais concernera également toute la première tiers du XIX, marquant la naissance de la littérature russe classique. 2. La théorie des "Trois styles" Le début du XVIII connut une tentative de mise en ordre de la langue sur une base stylistique, celle des "Trois styles" de Lomonosov, qui intégrait dans l'ensemble de la langue toutes les formes et structures attestées à l'époque, quelle que fût leur origine, en les catégorisant rigoureusement selon des critères 'stylistiques', en fait 'fonctionnels' puisque définis par les types de textes où elles devaient être utilisées. Outre les oeuvres de Lomonosov, exemples du style élevé, ce système global et diversifiant eut des applications dans les autres styles, même si le style élevé fut, quantitativement et qualitativement, le plus représenté, car c'était surtout lui qui était visé par le système, en tant que langue de l'abstraction et de la réflexion. Il faut dire que, parallèlement à cette tentative très contraignante de réglementation de la langue, continuait de régner, dans les styles inférieurs en particulier, l'extrême confusion des deux siècles précédents, caractéristique du moyen russe. Néanmoins le système de Lomonosov inspirera encore pendant longtemps bon nombre de distinctions à l'intérieur des différents registres de style du russe moderne. 3. Le problème de la phrase russe La nouvelle codification et l'uniformisation par élimination des variantes par trop slavonnes ou populaires portent sur les plans les plus discrets de la langue, ceux qui sont liés aux formes (lexicales, morphologiques) et à leur emploi (syntaxe de la phrase simple). Tout un domaine reste en dehors de cette réforme linguistique, c'est celui de la phrase (phonologie de la phrase, ordre des mots, syntaxe de la phrase complexe), où la conception de Lomonosov continuera de prévaloir pendant plusieurs décennies encore: se fondant sur le principe que l'ordre des mots ne remplit pas de fonction grammaticale (expression des relations syntaxiques) dans les diverses hypostases historiques du russe, langue à flexion nominale et à ordre des mots "libre", Lomonosov avait abordé cette question dans sa Rhétorique et non dans sa Grammaire; ainsi, selon lui, elle relevait de la "période" et non de la "phrase". En fait, il ne l'étudiait que dans les limites du style élevé, le seul style où la langue, outre sa fonction de communication, remplissait une fonction ornamentale, au service de l'art de l'écrivain. En tant que procédé oratoire du style élevé, la disposition des mots devait avant tout créer des figures, telles que la dislocation,

ORDRE DES MOTS DU RUSSE MODERNE

5

l'imbrication, le chevauchement, le chiasme, ce qui excluait tout fonctionne­ ment linguistique. De ce point de vue, la confusion la plus grande règnait dans les autres styles aussi, puisqu'on y trouvait les séquences les plus diverses; par exemple, pour le syntagme nominal, l'ordre Substantif Adjectif caractéristique du vieux slave coexistait avec la séquence du vieux russe Adjectif Substantif, sans oublier les diverses dislocations des deux termes du syntagme (nous verrons plus bas que le russe moderne admet de pareilles séquences, mais pour une fonction linguistique pertinente, en cas de différence du rôle pragmatique de chacun d'entre eux). Sans doute est-ce la raison pour laquelle la conception lomonosovienne de la "liberté" de l'ordre des mots restera si vivace, même lorsque la distinction des Trois Styles et ses conséquences auront cessé de faire autorité. Elle était la seule référence dans ce domaine, et comme l'a dit le linguiste Vinogradov (1950), dans la seconde moitié du XVIII l'ordre des mots tel qu'il se manifestait dans les différents corpus de l'époque était la "question malade" de la langue russe nouvelle, alors que ses autres systèmes, eux, s'étaient déjà constitués dans leurs principes essentiels, même s'ils pouvaient encore susciter des polémiques avec les tenants de la tradition vieux slave. Les normes grammaticales du russe moderne (qui définissent la place du mot par sa fonction syntaxique) portent essentiellement sur les membres secondaires de la proposition, qui au cours du XIX perdent la liberté dont ils jouissaient dans la langue de Lomonosov et de ses contemporains, liberté qui ne se retrouve actuellement que dans la poésie et la langue orale. En parlant de normes grammaticales nous reprenons la distinction de l'Ecole de Prague entre position "grammaticale" et position "grammaticalisée" (on a affaire à l'ordre grammatical quand la fonction syntaxique du mot est définie par sa place; c'est le cas du français mais pas du russe). La première série de normes est restrictive: le syntagme contenant un membre secondaire ne peut être disloqué. Ainsi la phrase (1) (1)

On dal bilet Pavlu Il donna billet (ac.) Paul (dat.) "Il donna billet à Paul de théâtre"

v teatr dans théâtre (ac.)

est inacceptable pour le russe moderne, car le membre secondaire de la proposition v teatr doit être enchaîné à bilet, membre principal lié directement au sommet syntaxique, le verbe, dont il dépend. Autrement dit, du point de vue de la correction grammaticale, seront acceptables les séquences (2), (3) et (4):

6

JEAN-PIERRE BENOIST

(2)

Bilet "Billet

v teatr dans théâtre

(3)

On "Il

bilet billet

(4)

On "Il

dal donna

v dans

on il

dal Pavlu donna à Paul"

teatr dal théâtre donna

Pavlu à Paul

bilet billet

Pavlu à Paul"

v dans

teatr théâtre"

et d'autres séquences plaçant le sujet On en fin de phrase par exemple, à condition que le membre secondaire v teatr ne soit pas séparé du membre dont il dépend par un autre membre de la proposition, quel qu'il soit. Cela signifie que les séquences chevauchantes sont interdites quand elles impliquent des membres secondaires. Néanmoins, si le membre secondaire est lui-même une tête de syntagme pour un autre membre secondaire, d'une profon­ deur syntaxique encore plus grande sur l'arbre de dépendance, il pourra être séparé du membre principal dont il dépend, comme dans la phrase suivante (5): (5)

Na nëm "Sur lui

byla était

seraja grise

ot à-cause-de

pyli kurtka poussière veste"

Une séquence comme (6): (6)

Na nëm "Sur lui

byla était

seraja kurtka ot pyli grise veste à-cause-de poussière"

serait inacceptable. On voit donc que les séquences d'encadrement sont possibles, à condition que le membre dominant absolu (ici kurtka) ne soit pas placé à l'intérieur du cadre. Les normes restrictives sont complétées par des normes de position, toujours pour les membres secondaires. Les membres secondaires non accordés (compléments) sont postposés, c'est le cas de v teatr et de ot pyli des exemples précédents; l'antéposition est caractéristique du membre secondaire détermi­ nant (accordé, quand il s'agit d'un adjectif ou d'un participe) comme pour seraja plus haut. Une séquence régressive de grande profondeur est fréquemment réaménagée en séquence progressive, mais une ponctuation spécifique s'impose, qui disjoint le substantif de son déterminant (7): (7)

Na "Sur kotoraja qui

nëm byla lui était pokryvala couvrait

kurtka, veste, vsë tout

seraja ot pyli, grise à-cause-de poussière, vokrug autour"

7

ORDRE DES MOTS DU RUSSE MODERNE

Il faut dire que ce réaménagement est devenue pratiquement une norme d'usage au cours du XIX siècle lorsque le déterminant est constitué par un participe tête de syntagme complétif, comme dans (7b). b.

Ljudi, "Gens

postradavshie aiant souffert

ot à-cause-de

xoloda froid"

La variante (7c), Postradavshie ot xoloda ljudi... semblerait aujourd'hui vieillie. Peut-être est-ce dû aussi au fait que le participe équivaut fonctionnellement à une proposition relative, qui elle est toujours postposée à son antécédent (au moins dans le discours écrit)? La rigueur des normes grammaticales connaît des dérogations dues à l'intervention du principe pragmatique, dont nous parlerons plus loin en détail, lorsque le membre secondaire et le membre principal dont il dépend ont des rôles pragmatiques différents: ainsi le syntagme bilet v teatr pourra être disloqué, avec position du membre secondaire au début absolu de la phrase (ce qui s'accompagnera d'une mélodie montante spécifique et même d'une pause). On aura ainsi la variante pragmatique (8): (8)

V "Dans/de

teatr on théâtre il

dal donna

bilet billet

Pavlu à Paul"

dans laquelle le substantif teatr constitue l'objet de l'énonciation, "ce dont on parle" (un Thème'). Il peut prendre une valeur contrastive, en admettant que l'auteur du don, "il", ait un autre billet, pour le cirque par exemple, où, lui, il ira. La phrase (8) se traduirait à peu près par "Pour ce qui du théâtre il donna son billet à Paul", ou même (voir plus loin) "Pour ce qui est du théâtre c'est à Paul qu'il donna son billet". Le membre secondaire accordé peut lui aussi être séparé du membre principal qu'il détermine; d'ordinaire ce dernier est placé au début absolu pour être thématisé, alors que l'adjectif reste en finale, comme dans la phrase (9): (9)

Sigary "Cigares

on il

vsegda toujours

pokupaet achète

samye les-plus

dorogie chers"

Cette phrase pourrait opposer des cigares et des cigarettes par exemple, ces dernières seraient bon marché (nous verrons plus loin la façon canonique dont le français exprime la thématisation). Cette différence des rôles pragmatiques de l'adjectif et du substantif est nette, elle peut être exprimée par une simple postposition de l'adjectif déterminant, sans la ponctuation que nous avons décrite plus haut: il n'a plus de thématisation du substantif, mais celui-ci n'appartient pas néanmoins au Focus de la phrase, puisque l'information que le

8

JEAN-PIERRE BENOIST

locuteur veut donner ne porte que sur la valeur des cigares que le sujet "il" a l'habitude d'acheter (la valeur contrastive serait tout à fait possible "chers et pas bon marché", comme le destinataire du message semble le croire). On aurait la variante (10): (10)

On vsegda pokupaet "Il toujours achète

sigary samye dorogie cigares les-plus chers"

4. La structure linéaire et prosodique de la phrase nouvelle Même si la conception nouvelle qui va fonder par la suite le fonctionnement de l'ordre des mots en russe, celle de Karamzin, est proprement linguistique dans ses définitions puisqu'elle s'appuie sur l'utilisation de la phonologie et de la linéarité de la phrase dans l'expression des rôles pragmatiques de Focus et de Topic, de Base et de Théme, elle aura une visée stylistique en soi, essentiellement littéraire: il s'agit pour l'écrivain et historien Karamzin de faciliter la compréhension du message langagier, de la phrase dans sa totalité linguistique, à son destinataire, le lecteur des oeuvres romanesques de la littérature russe nouvelle, dont il est lui-même un des représentants, le plus illustre en cette fin du XVIII siècle. Cette conception évidemment peu explicitée par Karamzin lui-même sera confortée au cours du siècle suivant par différents apports de la linguistique russe ou étrangère (allemande surtout de Paul, Steinthal, von Gabelentz et Amman avec les distinctions entre Sujet logique, puis psychologique, et Sujet grammatical, entre Prédicat logique, psychologique et Prédicat grammatical, à partir des positions aristotéliciennes; française de Weil sur la "liberté" de l'ordre des mots des langues anciennes, langues à flexion nominale, comme le russe). Néanmoins, pendant longtemps encore, la plupart des descriptions russes de la question, que ce soit dans des grammaires ou des études spéciales (Davydov 1816 ou Grech 1840), continuera à ignorer la nature linguistique de l'ordre des mots et en restera à une conception figée proche de celle de Lomonosov. En fait les catégories de la pragmatique moderne appliquées à l'analyse de l'ordre des mots (essentiellement la position réciproque des membres primaires de phrase) ne peuvent en rester à la dichotomie du Sujet et Prédicat logiques ou psychologiques comme chez Paul ou von Gabelentz: le linguiste doit montrer leur pertinence pour la langue qu'il décrit, en particulier dans une perspective typologique et contrastive. Mais pour que l'on puisse affirmer que les catégories pragmatiques sont pertinentes pour telle ou telle langue, il faut que les divers moyens qui les expriment dans cette langue (l'ordre des termes de l'énoncé et l'intonation, séparément ou en combinaison, pour le russe et le

ORDRE DES MOTS DU RUSSE MODERNE

9

tchèque; des particules spécifiques, pour le japonais; certains emplois des articles définis ou indéfinis, pour le français; la diathèse passive, en général) constituent des oppositions discrètes. C'est en ceci qu'une description réellement linguistique se différenciera de l'approche purement logicopsychologique héritée de l'Antiquité. 4.1 Les hésitations sur les principes nouveaux Enfin, pour clore ce tableau rapide de la naissance du russe moderne et en particulier de la façon dont se résoud la problématique confuse de l'ordre des mots tant dans sa pratique que dans sa théorie, il faut signaler le rôle joué par telle ou telle principe d'organisation linéaire propre aux langues étrangères, dont l'influence va se succéder dans lévolution du russe, latin, allemand et français. Nous nous arrêterons sur cette dernière langue car le modèle de la structure et de la linéairité de la phrase française s'imposera à bien des auteurs russes du début du XIX siècle en contradiction avec les principes mêmes de l'expression linéaire des rôles pragmatiques, créant un système parallèle difficile à analyser à l'heure actuelle faute de savoir comment les contemporains lisaient les énoncés en question. C'est le cas des syntagmes subjectal et objectai qui organisent toute la phrase quant à la place du sujet, en russe comme en français: ainsi jusqu'au premier tiers du XIX la séquence du français à fonction d'expression des relations syntaxiques Sujet Verbe Complément transitif, et sa conséquence, la séquence simple Sujet Verbe (Benoist 1993), pourront être employées en russe, alors que le sujet ne constitue pas le Topic, ce qui est en contradiction avec le principe d'organisation pragmatique de Karamzin, du Topic au Focus. Comme le fait remarquer Kovtunova, auteur de la première thèse de doctorat d'Etat sur l'ordre des mots du russe contemporain (1973), ces 'dèvian­ ces' se rencontrent fréquemment dans la prose de Pushkin, nourri, comme on le sait, de la tradition littéraire française, de Voltaire en particulier. Par exemple, dans le passage suivant extrait de la nouvelle Vystrel (Le coup de fusil): (11)

Silvio stal v "Silvio commença dans otvorilis', Masha vbegaet i s'ouvrirent, Macha entre et na sheju. sur cou".

menja pricelivat'sja. Vdrug dveri moi viser. Soudain portes s vizgom brosaetsja mne avec cri se-jette à-moi

Dans la deuxième phrase de (11) l'ordre neutre (d'expressivité nulle) serait, en russe contemporain, la séquence Verbe Sujet, c'est-à-dire otvorilis' dveri et vbegaet Masha: c'est l'organisation linéaire des énoncés qui décrivent

10

JEAN-PIERRE BENOIST

la manifestation nouvelle d'un individu ou d'un objet du texte, même s'il a déjà été mentionné auparavant; il s'agit ici du changement de la position des portes et de l'arrivée de la jeune fille, bien connue du lecteur mais absente de l'épisode précédent. Les deux items sont contextuellement 'nouveaux', la nouveauté de leurs référents est définie par la nouveauté de la situation dans laquelle ils apparaissent. Ils appartiennent donc au Focus, avec le verbe, et sont affectés par la chute de la mélodie en position finale de la phrase. On doit se demander comment Pushkin, Karamzin et leurs contemporains intonnaient le gallicisme que constitue dans ces phrases la séquence Sujet Verbe (Complément): il ne s'agit pas d'une inversion à valeur expressive comme ce serait d'ordinaire le cas en russe contemporain pour les phrase où le Focus précède le Topic (l'expressivité n'est pas difficile à identifier parce qu'en général elle n'est pas limitée au seul ordre des mots mais fait appel à d'autres plans de la langue, le lexique en particulier). Les deux intonations sont théoriquement possibles: soit une mélodie analogue à la mélodie expressive, sur le Sujet, sans en avoir la fonction, simplement parce que le schème prosodique de la phonologie de la phrase nouvelle l'exigeait, soit une mélodie "automatisée" (Danes 1967) faisant porter la chute sur le Verbe final (à l'instar du français). Le plus souvent les séquences inspirées du français sont caractérisées par la présence en fin de phrase d'un syntagme complétif, dont le rôle pragmatique ne relève pas le moins du monde de la focalisation (la connaissance de son référent est donnée par le contexte précédent immédiat). Ainsi, dans Puteshestvie v Arzrum ("Voyage à Arzrum"), du même Pushkin, on trouve le passage suivant (12): (12)

Priroda "Nature xoloden, froid, lezhal s'étendait

okolo autour-de gory montagnes v dans

nas byla ugrjuma. Vozduh byl nous était maussade. Air était byli pokryty pechal'nymi sosnami i sneg était couvertes de tristes sapins et neige ovragah ravins".

La dernière phrase du passage (12) aurait sans doute à l'heure actuelle l'ordre des mots V ovragah lezhal sneg, qui correspond à la nouveauté de la "neige" dans la situation, pour les voyageurs. En revanche, les "ravins" qui appartiennent au paysage sont 'connus' par relation métonymique avec les "montagnes" environnantes, ils constituent normalement le Topic de la phrase. La 'vacuité' de l'information apportée par le prédicat de ces phrases les rapproche de la phrase nominale, phrase sans verbe, fréquente en russe dans les descriptions. On peut donc supposer qu'à l'époque de Pushkin elles étaient

11

ORDRE DES MOTS DU RUSSE MODERNE

intonnées comme les phrases contemporaines le seraient, avec la mélodie conclusive sur les Sujets, sans qu'il s'agisse d'inversion expressive (ce qui serait le cas actuellement). Il peut arriver d'ailleurs que le Complément, même si son 'ancienneté' dans le texte lui confère une 'vocation' topicale, constitue néanmoins le Focus de la phrase, lorsqu'il est exprimé par un groupe de mots de structure syntaxique complexe (syntagme déterminatif) et de poids lexical riche (adjectif descriptif, expressif). Dans ce cas, il est clair que la phrase était lue avec une chute de la mélodie sur ce groupe, le Complément ou ses Détermimants. Ainsi lorsque dans Pikovaja dama ("La dame de pique") Pushkin décrit la réception à laquelle assiste le héros, German, après une succession de phrases à séquence Complément Verbe Sujet, qui introduisent dans le récit les personnages nouveaux, il passe à une séquence Sujet Verbe Complément pour un énoncé analogue syntaxiquement et sémantiquement, sans doute parce que le Com­ plément est accompagné d'un adjectif important du point de vue narratif (13): (13)

Shuby i plashchi "Pelisses et capes

mel'kali passaient

mimo près de

velichavogo majestueux

shvejcara. portier".

La séquence contraire, possible, qui ferait du "portier" même 'nouveau' ('nouveauté' exprimée par l'article indéfini en français, "un majestueux portier") un Topic explicite (justifié Par le fait que ce personnage est lié métonymiquement à la demeure où il officie, et dans ce cas un article défini conviendrait aussi, à moins qu'il n'y ait plusieurs portiers!), 'gommerait' l'allusion à l'importance de la fonction de cet individu qui constitue une barrière devant le jeune homme pauvre mais ambitieux. S'y ajoute la description du luxe qui s'étale devant lui et qui ne lui est pas accessible. Malgré son origine étrangère, l'influence de l'ordre des mots français sur la disposition réciproque des membres primaires de la phrase ne sera pas limitée à une pratique ancienne. Elle faussera bon nombre de descriptions tout au cours du XIX siècle (et même dans de nombreuses études de la première moitié du XX siècle), jusqu'à ce que certains linguistes russes comme Krushel'nickaja, Sirotinina, Kovtunova, Raspopov et d'autres, à la lumière des travaux de différents représentants de l' Ecole de Prague d'après-guerre, tels Mathesius, Firbas, Danes, Sgall, ne prennent conscience de la spécificité de l'ordre des mots du russe, si proche de celui du tchèque, langue à flexion et de la même famille. Les recherches menées dans ce direction par des linguistes russes de la deuxième moitié du XIX comme Fortunatov (1856) ou Slanskij (1887) semblent n'avoir laissé aucune trace malgré leur lien avec les linguistes 'psychologistes' allemands auxquels elles font référence; il faut préciser

12

JEAN-PIERRE BENOIST

qu'elles étaient destinées plus à des pédagogues qu'à des philologues et avaient une visée essentiellement scolaire. Encore imprécises quant à leur approche théorique, malgré l'appel aux procédés linguistiques qui seront plus tard spécifiques de l'étude systématique de la "fonction communicative de l'ordre des mots" (panel de questions, permutations, variation pertinente du contexte, traduction interne, paraphrasage), ces études ne pouvaient retenir l'attention des représentants du courant dominant de la linguistique russe de l'époque tourné davantage vers la diachronie et la description des formes. Ainsi seule l'analyse fonctionnaliste héritée de l'Ecole de Prague, de linguistes comme Jakobson et Karcevskij, a permis de définir la fonction de l'ordre des mots dans les langues slaves en général, en rompant avec toute conception extérieure ou a priori. Avec la théorie de la "Division actuelle" (traduction proposée par les linguistes praguois eux-mêmes), Mathesius (1942, dans son article novateur) est à l'origine de l'interprétation linguistique de l'organisation linéaire de la phrase des langues slaves. Il oppose ainsi l'articulation pragmatique de l'énoncé à son articulation syntaxique. Dans cet article il montre que l'ordre des mots de la phrase simple des contes fantastiques populaires slaves est régi par le principe de la succession Topic Focus (appelés différemment par lui), indépendamment de la fonction syntaxique de ces éléments, et, en particulier, qu'il est caractérisé par la succession Circonstance ou Complément Verbe Sujet (et non Sujet Verbe Circonstance ou Complément, comme ce serait le cas en français ou en anglais) quand il s'agit de présenter pour la première fois dans le récit tel ou tel personnage qui est exprimé par l'item du Sujet. Il faut souligner l'abondance des études contrastives avec les langues sans flexion et à article telles que le français et l'anglais. Ainsi Firbas (1966) insiste sur la différence de l'ordre des mots de la phrase simple tchèque et l'anglaise particulièrement dans le cas où le Sujet a le même rôle pragmatique que le Verbe (Focus complexe), compte tenu de la signification de ce dernier (existence, présence, apparition, arrivée, surgissement, c'est-à-dire les phases ingressives de l'existence): la 'nouveauté' du référent du Sujet est marquée en tchèque par sa postposition au verbe, en anglais par l'article indéfini. A partir d'une comparaison entre le russe, le français et l'italien (langue sans flexion nominale et avec article comme le français et l'anglais) nous-même avons montré (1993) qu'il fallait sans doute faire intervenir la 'pauvreté' des désinences verbales en anglais et en français (pour ce dernier au moins à l'oral), face à la 'richesse sonore' de la morphologie verbale de l'italien (qui permet également l'ellipse du pronom sujet).

ORDRE DES MOTS DU RUSSE MODERNE

13

5. Ordre des mots et prosodie dans la phrase russe contemporaine Les analyses des linguistes fonctionnalistes russes, inspirées par les recherches praguoises, insistent volontiers sur le rôle actif joué par Karamzin dans l'organisation linéaire de la phrase russe moderne. Celui-ci, à la foi praticien et théoricien (évidemment dans les termes généraux et philosophiques de son époque mais suffisamment clairs), assied le fonctionnement pragmati­ que de l'ordre des mots en russe moderne sur une phonologie de la phrase liée à la fonction stylistique nouvelle qu'il entend lui donner (après le système de Lomonosov, pesant et inadapté aux besoins nouveaux de la société russe): pour que le lecteur des genres nouveaux éprouve du plaisir, la phrase n'est pas déclamée mais est prononcée de façon naturelle, ainsi dirons-nous en termes linguistiques modernes que sa prosodie suit la courbe mélodique naturelle, fondée sur l'effort physiologique de la phonation, qui débute par la montée de la mélodie et s'achève par sa chute. Cette montée (identifiée comme intonème non conclusif par les théories contemporaines de l'intonation) coïncide avec le Topic (ou la zone topicale), alors que la chute (intonème conclusif, "accent de phrase" dans la terminologie linguistique russe et tchèque) marque le Focus (ou l'élément nominal du Focus complexe), c'est ce que Danes (1967) appelait la "place automatisée de l'accent de phrase". Ainsi se constituent peu à peu les principes de la phonolo­ gie de la phrase écrite du russe moderne. Lorsque le Focus repose sur un contraste, le contour conclusif s'associe à une forte intensité articulatoire de la syllabe et parfois à une durée notable de sa réalisation, c'est ce que traditionnellement les linguistes russes nomment "l'accent logique" ou "l'accent d'insistance". Certes, dans une approche littéraire sentimentaliste, Karamzin rompait parfois cette prosodie prosaïque par certaines mélodies particulières visant à créer à l'intérieur de la phrase un certain rythme, celui de la'blleprose", mais ceci est resté un trait spécifique de son écriture, qui avec le déclin du sentimentalisme disparaîtra de la prosodie de la phrase russe standard écrite. C'est donc à partir de la phonologie phrastique de Karamzin qu'il convient d'analyser les différents aspects de l'organisation linéaire du russe moderne et non à partir de schéma préconçu. En particulier, comme nous l'avons déjà vu, il faut cesser de considérer qu'en russe l'ordre des mots est libre, alors qu'il a une fonction contraignante, et si on peut parler d''inversion' ou d''ordre des mots expressif, émotionnel' ce n'est que lorsque les séquences en question

14

JEAN-PIERRE BENOIST

constituent une déviance du schème intonatif ' automatisé', neutre de la langue écrite. On comprend que la phrase encore traditionnelle dans la deuxième moitié du XVIII, la phrase lomosovienne, destinée à une perception plus visuelle que sonore, plus mentale et plus abstraite que lisible et informative était indifférente à la composante intonative, donc à la linéarité fonctionnelle. 6. Les rôles pragmatiques en russe Les premières études fonctionnalistes russes de la deuxième moitié du XX siècle ne se préoccupaient que des différences particulièrement discrètes entre la 'connaissance', 'l'ancienneté' (contextuelle ou situationnelle, répétitive ou métonymique) caractéristique du Topic, et la 'nouveauté' (absolue ou relative dans un entourage nouveau) du Focus, elles retrouvaient ici mécaniquement les catégories grammaticales des langues à article comme les phrases suivantes (14) et (15) l'illustrent bien: (14)

Molodoj Jeune

chelovek ugnai homme(nom.) vola

mashinu voiture(accus.)

avec les distinctions référentielles du français "Le jeune homme vola une voiture", et (15)

Mashinu ugnai molodoj Voiture(accus.) vola jeune "La voiture vola un jeune homme",

chelovek homme(nom.)

énoncé qui est un non-sens en français. Il faut faire appel au passif: "La voiture fut volé par un jeune homme" ou bien "C'est un jeune homme qui vola la voiture" (s'il y a insistance, contraste possible sur l'identité du voleur). Le Focus contrastif est d'ailleurs fréquemment constitué par un item 'connu', à référence définie, qui renvoie à un individu mentionné auparavant et qu'il s'agit d'opposer à un autre individu également 'connu': "C'est le jeune homme qui vola la voiture (et non mon voisin)". La 'connaissance' n'est pas un obstacle en soi à la focalisation d'un élément. Si elle n'est pas immédiate (le plus souvent signalée dans ce cas par l'emploi d'un pronom anaphorique) et qu'elle participe à la situation nouvelle décrite par le verbe de la phrase, son référent constituera facilement avec celuici un Focus complexe et occupera de plein droit la zone finale, sous l'accent de phrase, qu'il soit sujet ou complément. On pourrait dire que son référent est 'réintroduit' dans l'univers du discours. Il est évidemment possible que les deux membres primaires Sujet et Complément de la proposition soient topicaux, le Focus est alors constitué par

ORDRE DES MOTS DU RUSSE MODERNE

15

le seul Verbe: les deux Topics occupent la zone initiale dans une disposition réciproque sur laquelle nous ne nous étendrons pas ici, le Verbe est mis en position finale sous l'accent de phrase, soit une phrase comme (16): (16)

Unizitel'naja procedura "Humiliante procédure(nom.)

Grigorja volnovala Grégoire(ac.) émouvait"

où les deux substantifs sont des Topics (le premier de connaissance 'immédiate' qui renvoie à l'énoncé précédent, la fouille à laquelle le héros est soumis; le second de 'connaissance' constante), c'est la disposition réciproque la plus fréquente. Mais cette phrase apparaît chez son auteur, l'écrivain bien connu Sholoxov, sous une forme différente, où seul le substantif de connaissance 'immédiate', 'connaissance' la plus liée au contexte précédent, qui assure la cohésion du texte, UniziteVnaja procedura est placé en position initiale. La phrase a l'aspect suivant (16a): (16) a. UniziteVnaja procedura "Humiliante procédure

volnovala émouvait

Grigorja Grégoire"

Ce qui ne signifie pas que le nom du héros est focalisé, l'accent de phrase affecte toujours le verbe, et il ne s'agit pas d'une séquence 'inversive', 'expressive': dans le cas de plusieurs Topics on évite fréquemment de charger la zone initiale afin de ne pas augmenter le degré d'enchaînement des phrases du texte et laisser plus de liberté au lecteur. La position médiane du verbe accentué est la solution la plus simple. Le russe du début du XVIII siècle, marqué par l'influence de l'allemand, dans des corpus de genre administratif, officiel, affectionnait la position finale du verbe (il s'agissait évidemment d'un système autre que celui de notre époque, en fait c'était une imitation de la position finale du verbe allemand). Il s'avéra rapidement qu'il fallait dépasser le cadre de la 'connaissance' et de la 'nouveauté' qui ne suffisaient pas à expliquer la disposition des membres primaires de l'énoncé dans la même phrase (14) par exemple. En effet Molodoj chelovek ugnai mashinu peut constituer une phrase (14a) dont les éléments nominaux sont tous 'nouveaux', comme le montrera sa traduction en français avec des articles indéfinis: "Un jeune homme vola une voiture". Le verbe lui aussi est 'nouveau' puisqu'il dénote une action dont l'auteur, un jeune homme, est 'nouveau', et dont l'objet, une voiture, l'est également. Il s'agit de ce que nous appellerons la séquence canonique de ce type d'énoncé, définie par son contenu sémantique et syntaxique, ainsi que le prouve l'impossibilité de permuter tel quel les deux substantifs et de faire porter la mélodie conclusive

16

JEAN-PIERRE BENOIST

("l'accent de phrase" dans la terminologie russe) sur le sujet molodoj chelovek, comme dans (15). La permutation, avec sa conséquence intonative, n'est acceptable pour cet énoncé qu'à condition que le complément mashinu soit affecté d'une référence définie ('connaissance' contextuelle ou situationnelle), c'est nécessairement la phrase (15). Nous voyons bien que l'ordre des mots de (14a), à la différence de celui, pourtant identique de (14), ne relève pas des distinctions pragmatiques que nous avons définies plus haut à la suite des linguistes tchèques. Le sujet molodoj chelovek de (14a) ne peut qualifié de Topic, il est d'ailleurs affecté d'une légère montée mélodique, moins forte que celle du Thème mais qui signale sa 'nouveauté' et sa place particulière dans les catégories de la "division actuelle" de la phrase. Ces caractéristiques sémantique et intonative ont été mises en évidence dans les travaux russes, qui ont conclu que dans un énoncé comme (14a) c'est son Rôle sémantique d'Agent (Cas profond des Grammaires de Cas) qui lui confère une fonction pragmatique spécifique, distincte du Topic. Dans nos propres recherches (1979, 1992) nous avons proposé le terme de Base: dans les phrases dont tous les éléments sont 'nouveaux' l'Agent (s'il y en a un dans l'énoncé), indépendamment de sa signification lexicale, est le point de départ de la communication et occupe donc la position initiale de cette phrase. Pour éviter toute confusion avec la théorie syntacticienne étrangère à l'esprit du russe, que nous avons d'ailleurs rejetée plus haut, rappelons que cette position n'est pas due à la fonction syntaxique centrale du Sujet dans l'énoncé en général et à l'emploi du nominatif pour l'exprimer, mais à la place la plus élevée que l'Agent occupe dans la hiérarchie sémantique de l'énoncé donné: selon une projection linéaire de cette hiérarchie, il a donc la priorité sur tous les autres Rôles. Les opérations linguistiques qui créent le Topic et la Base sont absolument différentes: le Topic est choisi à l'intérieur du contexte précédant la phrase où il occurre (ou bien dans la situation extralinguistique à laquelle il participe): il relève de l'avant-texte; la Base au contraire est imposée par le contexte interne de l'énoncé, à partir de son contenu sémantique (le sens du verbe détermine les Rôles sémantiques assurés par ses différents arguments, compte tenu d'un certain nombre de souscatégorisation lexicale de leurs référents '+ ou - animé', en particulier dans le cas de l'Agent). Dans les énoncés que nous avons qualifiés plus haut de 'phrases d'existence', dont le verbe exprime l'existence, la présence, l'arrivée dans un lieu, ce sont les divers Localisateurs, même 'nouveaux', qui ont la priorité (le Rôle des Agents s'estompe d'ailleurs puisqu'il ne s'agit que d'asserter la pré­ sence de leurs référents dans la situation, à l'instar d'une proposition

ORDRE DES MOTS DU RUSSE MODERNE

17

nominale): ces Localisateurs occupent par conséquent la zone initiale de la phrase. Pour être complet dans notre description des éléments non topicaux qui peuvent apparaître au début de la phrase, il faut citer le Thème, dont nous avons donné des exemples auparavant en parlant des cas de dislocation autorisée du syntagme nominal. La thématisation a un statut à part dans l'ensemble des catégories pragmatiques non focales du russe puisqu'elle est capable de briser des liens syntaxiques plus profonds dans la structure dépendancielle que les liens des constituants primaires de l'énoncé. Le Thème par son rôle pragmatique à part est d'ailleurs parfois confondu avec un Focus inversé tant il se distingue du Topic, qui relève lui du simple enchaînement du texte, et de la Base, automatisme sémantique. Enfin il faut dire que le russe oral, dont il sera question avec le paragraphe suivant, admet, dans une variante assez 'relâchée', une véritable anacoluthe quant à l'expression du Thème, puisque la thématisation peut entraîner une rupture complète des liens syntaxiques de la phrase: le substantif thématisé est au nominatif (cas asyntaxique ici). Si la langue écrite ne peut dire que (17): (17)

Etu devochku ja "Cette fillette(ac.) j '

ljublju aime"

en français courant "Cette fillette je l'aime", la langue orale, elle, admettra des variantes (18) et (19): (18)

Eta devochka "Cette fillette(nom.)

ejë ja elle(ac.) j '

ljublju aime"

où il y a une reprise par le pronom personnel au cas voulu par la construction du verbe, et (19)

Eta devochka ja "Cette fillette(nom.) j '

ljublju aime"

sans reprise. 7. Ordre des mots et oralité Se fondant sur la fréquence dans les réalisations parlées de phrase à accent de phrase non final (sauf s'il affecte le verbe), médian dans Sujet Objet accentué Verbe ou Objet Sujet accentué Verbe, et initial dans Objet accentué Sujet Verbe ou Sujet accentué Objet Verbe, certains linguistes russes contemporains, comme Kovtunova (1973, 1976), ont conclu que Vinversion,

18

JEAN-PIERRE BENOIST

définie comme séquence de la place non finale de cet accent, a une double fonction, expressive pour la langue écrite, et distinctive de la langue parlée. Une constation s'impose; lorsqu'un écrivain veut faire 'parler' ses personnages, il imite la langue parlée et un des premiers traits langagiers auxquels il fait appel est précisément cette position non finale du sujet et du complément accentués. Néanmoins il est clair qu'il y a ici une confusion fréquente entre le registre de style (style 'parlé') et le code spécifique de l'oralité, orientée le plus souvent sur un échange de messages, c'est-à-dire le dialogue. Dans le discours littéraire qui vise à imiter le discours d'autrui (''discours direct") les marques du code oral fonctionnent comme marques du style parlé. La phonologie du code oral est différente de la phonologie de l'écrit, essentiellement monologique. Des études récentes, russes (Nikolaeva 1982) et françaises (Bonnot et Fougeron 1982) ont montré que la séquence non "automatisée", définie comme "inversive" par Kovtunova, relève d'un fonctionnement pragmatique spécifique et non expressif, que nous avons défini nous-même au Congrès slavo-roman de Florence comme une insistance phrastique (puisqu'elle porte sur toute la phrase et pas sur un mot seul comme dans le cas d'un Focus contrastif, Benoist 1985). Nous avons également souligné que l'emploi de cette mélodie non finale est liée d'ordinaire à celui de structures phrastiques particulières comme la parataxe, et que l'insistance en soi est une intention pragmatique propre au dialogue, dont l'oralité est le canal naturel. La séquence Sujet accentué Verbe de la phrase (20): (20)

Papa "Papa

prishël est-arrivé",

dans le petit corpus oral (donc sans ponctuation) proposé par Nikolaeva (20a): (20) a. Otkroj "Ouvre

papa papa

prishël est-arrivé"

relève de la valeur de justification qu'a la deuxième partie de cette énonciation complexe, sans qu'elle soit exprimée par une subordination: il s'agit d'attirer l'attention de l'interlocuteur pour qu'il ouvre à son père. Dans un corpus écrit dépourvu de cette insistance, qui se contenterait de constater l'arrivée du Papa, on retrouverait l'ordre Verbe Sujet accentué, qui est la séquence canonique de la phrase d'existence dans ce code (21): (21)

Prishël Papa "Est-arrivé Papa"

ORDRE DES MOTS DU RUSSE MODERNE

19

en français "Papa est arrivé". Dans le code écrit, la causalité exprimée par la conjonction de subordination potomu chto ("parce que") se combine régulière­ ment à la séquence canonique (22): (22)

Otkroj, "Ouvre

potomu chto parce que

prishël est-arrivé

papa papa"

La phrase du discours oral est souvent insistante en soi, par exemple, quand la situation est inattendue, voire incompréhensible comme dans (23): (23)

De Goll' umer "de Gaulle est-mort" (exemple emprunté à Nikolaeva)

(dans le discours écrit on mettrait sans doute un point d'exclamation, mais avec la séquence canonique). C'est que le locuteur fait part d'une nouvelle à la fois inattendue et éprouvante à son interlocuteur (on connaît l'admiration et la sympathie que les Soviétiques ressentaient pour le Chef de la France libre). Au Colloque de Florence nous avons proposé comme équivalents français les particules d'insistance phrastique "C'est que", "Voilà que, Ne voilà-t-il pas que", qui expriment l'étonnement du locuteur face à la situation qu'il décrit. Pour un contenu analogue mais moins chargé d'émotivité (l'événement "tout seul" en quelque sorte), on aurait la séquence canonique des phrases d'existence (24): (24)

Umer Petrov Est-mort Petrov "Petrov est mort"

C'est la fréquence dans le discours oral, dialogique, de ces messages d'insistance phrastique qui explique la fréquence des séquences avec accent non final, à côté des séquences analogues à celles de la langue écrite, dont l'emploi est régulier dans les phrases simplement assertives.

20

JEAN-PIERRE BENOIST

ABSTRACT 1) In the middle of the XVIIIth century the stylistic distinction, still traditional for many essential linguistic levels (morphological, syntactical, lexical) at the beginning of this century, between the prevailing Church-Slavonic (Old-Slavonic) elements in poetic and abstract works and the properly Russian (Old-Russian) forms in prosaic novels begins to give way to a uniform ('standardized') language, the Modern Russian Language, with the exception of sentence organization (speech linearity), particularly word order. For example, for Lomonosov, the famous Russian philosopher and scientist, and author of Russian Grammar and Rhetorica, WO is a rhetorical device of "high style", characterized not only by Church-Slavonic sequences but also by ornamental figures such as chiasmus and dislocation. This rhetorical approach, just like the analysis of XIXth century Russian linguists, rests on the nongrammaticalization of Russian WO (syntactical functions are expressed by inflexional ends, WO is "free"). 2) A real revolution in Russian WO is accomplished at the end of the XVIIIth century by the Russian writer and historian, Karamzin: WO is no longer an ornament, but it must make for easier reading and understanding: it is the principle of WO communicative function, from Topic to Focus, which is founded on sentence phonology with conclusive intoneme denoting Focus in the sentence end, independently of syntactical functions, Subject or Object. If this principle is a linguistic device, its original aim is still stylistic. In fact it is the same narrative principle of WO of oral Pan-Slavonic tales. 3) Nevertheless the power of word order grammatical principle, particularly under the influence of the French sequence Subject Verb, is so great that the communicatively relevant sequence Verb Subject is never used by writers in the early XlXth century. That is to say WO was perceived later by XlXth and XXth century Russian linguists and grammarians as a relevant linguistic level: ordinarily they distinguish only 'natural', 'neutral' word order and 'subjective', 'inversive' only for syntactical and morphological reasons. 4) It is important to say that the stylistic origin of communicative WO in Modern Russian appears even now when we differentially study the totality of its language manifestations in the enunciative opposition written - monologue - historical and oral - dialogue - discursive; the latter term has another sentence phonology that does not allow communicative WO as in the written variant.

ORDRE DES MOTS DU RUSSE MODERNE

21

BIBLIOGRAPHIE Benoist, J.-P. 1979. Les fonctions de Vordre des mots en russe moderne. Thèse de doctorat, Institut d'Etudes slaves, Paris. . 1985. "La division actuelle dans la communication orale en russe contemporain", II Colloque slavoroman de Florence, Atti del Colloquio Lingue slave e lingue romanze. Firenze: Università degli Studi. . 1992. "La position initiale dans la phrase russe", Etudes de linguistique romane et slave dédiées à S. Karolak, Krakow. . 1993. "La séquence Verbe Sujet en russe, français et italien, ou les Icônes de la linéarité", Slovo 15, Paris. Bergel'son M. ., Kibrik A. E. 1981. "Pragmaticheskij princip prioriteta i ego otrazhenie v jazyke", IAN SSSR SUA 40, Moskva. Bonnot C , Fougeron I. 1982. "L'accent de phrase initial en russe est-il toujours un signe d'expressivité ou de familiarité?", BSLP, LXXVII, 1982-1, Paris. Dahl O. 1969. "Topic and Comment. A Study in Russian and General Transformational Grammar", Slavica Gothoburgensia, 4. Göteborg. Danes F. 1967. "Order of Elements and Sentence Intonation". To Honor R. Jakobson, The Hague-Paris. Davydov I. I. "Opyt o porjadke slov", Trudy Obshchestva ljubitelej rossijskoj slovestnosti, Ch. IV (1816), Ch. VII (1817), Ch. XIX (1819), Sankt-Peterburg. Ebeling C. L. 1958. Subject and Predicate, especially in Russian. s'Gravenhage. Firbas J. 1966. "On defining the Theme in Fuctional Sentence Perspective". TLP, 1, Praha. Fortunatov F. F. 1856. "O prepodavanii grammatiki russkogo jazyka v srednej shkole", Izbrannye trudy, 1957, Moskva. Fougeron I. 1989. Prosodie et organisation du message. Analyse de la phrase assertive en russe contemporain. Paris: . Klincksieck. Grech N. I. 1840. Chtenija o russkom jazyke. Sankt-Peterburg. Isachenko A. 1967. "Frazovoe udarenie i porjadok slov". To Honor R. Jakobson II. The Hague-Paris. Karamzin N. M. 1792. "O russkoj grammatike francuza Modrju". Moskovskij zhurnal. Moskva. Kovtunova I. I. 1973. Porjadok slov v sovremennom russkom literaturnom jazyke i formirovanie norm slovoraspolozhenija v XVIII- pervoj treti XIX v. IRJA AN SSSR, Moskva. . 1976. Sovremennyj russkij jazyk, porjadok slov i aktual'noe chlenenie. Moskva. Krushel'nickaja K. G. 1969. "O sintaksicheskoj prirode aktual'nogo chlenenija". Invariantnye sintaksicheskie znachenija i struktur a predlozhenija. Moskva. Lomonosov M. V. 1748. Ritorika. Sankt-Peterburg. Mathesius V. 1942. "Zakladni funkce poradku slov v cestine". SAS 7. Praha. Nikolaeva T. M. 1982. Semantika akcentnogo vydelenija. Moskva: Nauka. Sgall P. 1967. "Functional Sentence Perspective in a Generative Description". Prague Studies in Mathematical Linguistics 2.203-225.

FROM VSO TO SVO? WORD ORDER AND REAR EXTRAPOSITION IN COPTIC

ANTONIO LOPRIENO University of California, Los Angeles

0.

Historical introduction The history of Egyptian is usually divided into two main phases, 'Earlier Egyptian' and 'Later Egyptian', which are characterized by a major change from synthetic to analytic patterns (Junge 1985). Earlier Egyptian is the language of all written texts from 3000 to 1300  and survives in formal religious texts until the IIIrd century CE. It displays morphological affixes indicating gender and number in the noun as well as flectional deictic features in the verb. In verbal sentences, Earlier Egyptian exhibits the VSO-order: (1)

dd.jn shtj pn n hjm.t=f tn say.NARR peasant this to woman.FEM=his this[FEM] "Then the peasant said to his wife" (Eloquent Peasant R1.5, cf. Parkinson 1991:1)

Later Egyptian, on the other hand, is documented from the nineteenth dynasty to the Middle Ages (1300 BCE-1300 CE). Its latest stage is Coptic (Till 1970), the language spoken in Christian Egypt from the fourth through presumably the XIVth century CE, written in a variety of the Greek alphabet with the addition of a few Demotic signs to represent Egyptian phonemes absent from Greek. Coptic also differs from the earlier stages of the language because of the high percentage of lexical borrowing from Greek (Kasser 1991), whereas the impact of Greek on the shaping of Coptic syntax is still a matter of considerable scholarly debate.1 As a spoken, and gradually also as a written language, Coptic was superseded by Arabic from the IXth century onward, but it survives to the present time as the liturgical language of the Christian church of Egypt (Till 1970:29-39) and in the few linguistic traces it left in spoken Egyptian Arabic (Vittmann 1991). In general, the structuralist tradition of which Hans Jacob Polotsky is the most representative exponent tends to consider this impact relatively limited, cf. Polotsky 1950; Shisha-Halevy 1986:52. Within the generative or functionalist frame, on the other hand, a certain number of Coptic constructions are seen as the result of code-switching with Greek: Reintges, in press. For general observations cf. Weiss 1968; Nagel 1971.

24

ANTONIO LOPRIENO

Besides displaying a number of phonological evolutions, Later Egyptian, particularly in its latest phase, tends to develop analytic features: periphrastic SVO-patterns supersede the older VSO-formations, suffixal markers of morphological oppositions are dropped and functionally replaced by prefixai classifiers such as articles or relative constructions (Hintze 1950):2 (2)

a=k-tôrp m-p-eiôhe m-p-ent-a=k-jit=f n-cons PRET=you-steal ACC-the-field from-he-who-PRET=you-take=him in-violence "You stole the field from the one against whom you have committed violence" (Behlmer 1996:62)

(3)

Earlier Egyptian participle > Later Egyptian relative construction sdm. t > t3-ntj hr-sdm >t-et-sôtm hearing.FEM > the[FEM]-who on-hearing>the[FEM]-who-hear "the hearer"

Finite VSO-forms, therefore, appear replaced by a paradigm of SVOpatterns called "sentence conjugations" or "clause conjugations" (Polotsky 1960), depending on whether they function as main or as dependent clauses, which obtain from the grammaticalization as tense markers (Hopper & Traugott 1993:8-10) of what was originally a form of the auxiliary verb to do followed by the infinitive: (4)

Earlier Eg. VSO sdm.hr-f hear.AOR=he "he usually hears"

> > >

Late Eg. AUX-SVO hr-jr-f-sdm AOR-do=he-hearing

> Coptic TENSE-SVO > ša=f-sôtm > AOR=he-hear

In the unmarked order of verbal sentences, whether in the Earlier Egyptian VSO- or in the Later Egyptian SVO-order, the hierarchy of arguments is always SUBJ < OBJ < OBL (cf. Croft 1990:101-111). This order is modified when the NP-argument is conveyed by a clitic pronoun, which tends to move to the position immediately following V, regardless of its semantic role: (5) a.

b.

rdj.n wj h3s.t n h3s.t give.PRET me foreign country .FEM to foreign country .FEM "Foreign country gave me to foreign country" (Sinuhe  28-29, cf. Koch 1990:23) ntk dd n=f you giving[IMPERF] to=him "You are the one who brings it to him" cf. Parkinson 1991:21)

st it (Eloquent Peasant R 18.6,

In order to provide a more corpus-based, and therefore uniform description, I shall try in the following paragraphs, whenever possible, to choose my examples from a lengthy classical Coptic text, namely Shenute's De iudicio (Behlmer 1996).

25

FROM VSO TO SVO?

c.

mê e-ša=k-kô nsô=k an n-ne-khrêma QUEST that-AOR=you-leave after=you not ACC-the[PLUR]-property hm-pe-hoou m-p-mou in-the-day of-the-death "Do you not leave your wealth behind you on the day of death?" (Behlmer 1996:54)

Throughout the history of Egyptian, the predicate of verbal clauses, in its so-called 'second' or 'that-form', may function as the 'theme' of the utterance: (6) a.

b.

wnn jrf ti pf mj-m m-hmt=f ntr pf mnh that-be indeed land that how without=him god that excellent "And how is that land faring without him, namely this excellent god?" (Sinuhe  43-44, cf. Koch 1990:30) e=k-na-cine hôô=f tôn m-p-ete-mp=k-fit=f that=you-FUT-find self=it where ACC-the-which-NEG.PRET=you-bring=it nmma=k with=you "Where will you find what you have not brought with you?" (Behlmer 1996:54)

In general, Egyptian verbal syntax displays a comparatively high degree of topicalization and focalization phenomena (Loprieno 1988:41-52). Unmarked verbal phrases, i.e. VP that are neither topicalized as the marked theme of the utterance nor rhematized by the presence of a topic or of a discourse marker are, therefore, less frequent than in the majority of natural languages and mostly function as embedded or as modal clauses: (7) a.

b.

mw m jtrw zwr.t(w)=f mrj=k water in river drink.PASS=it wish=you "The water of the Nile is drunk (only) if you wish it" (Sinuhe  233-234, cf. Koch 1990:68) mh hr h3.t jw.t=k think [IMP] on rpse.FEM come.PROSP=you "Think of your corpse, and come back!" (Sinuhe  199, cf. Koch 1990:63)

In Coptic verbal sentences, the tendency to have verbal phrases function as marked theme or rheme of the utterance reaches its full development: in the former case, the VP is preceded by a relative marker e- or nt- and is described in Coptological literature as 'second tense' (Polotsky 1987:129-140); in the latter, the form is preceded by the subordination marker e- and is described as 'circumstantial' (Polotsky 1990:225-260): (8)

nt-a=n-jpo-f e=f-o that-PRET=we-beget=him while=he-be "He was born to us blind" (John 9,19)

n-blle as-blind

26

ANTONIO LOPRIENO

1. The SVO-order in Coptic The gradual historical affirmation of the SVO-order can indeed be seen as a predictable typological trend which Egyptian shares with a certain number of VSO- (and SOV-) languages:3 to choose an example within the same linguistic family, in Modern Standard Arabic the VSO-order of the classical language is being challenged by the SVO-order of both the Arabic dialects and of Standard European (Holes 1995:203-213); in a genetically different family, the SOV-order generally assumed for Proto-Indo-European and Proto-Germanic has left room to the flexible SVO-order of many Germanic languages (Hock 1991:618-621).4 The dominance of the SVO-order in Coptic, however, is not as generalized as a cursory look at the basic clause structures would lead one to believe. On the one hand, a pure SVO-order only underlies the present conjugation, which in fact displays the etymological and syntactic structure of adverbial clauses. Compare the verbal clause in (9) and the adverbial clause in (10): (9)

f-sôtp hôô=f na=f m-p-šône m-p-nobe m-p-diabolos he-choose self=he to=him ACC-the-disease of-the-sin of-the-devil "He indeed chooses for himself the disease of the devil's sin" (Behlmer 1996:28)

(10)

epeidê se-hijn-rôme nim et-r-nobe because they-on-man every who-do-sin "because they are on every man who commits sins" (Behlmer 1996:14)

On the other hand, the SVO-order can be said to synchronically character­ ize Coptic verbal clauses only if we view the subject of the present conjugation as displaying "a synthesis of lexical and grammatical components" (ShishaHalevy 1986:159) and the conjugational basis as being located outside of the word order sequence of basic sentence components-in other words, if we treat the conjugational basis as a grammaticalized tense marker (Hopper & Traugott 1993:50-56) which does not modify the SVO-order: (11)

a=k-ouôm n-n-sarks m-p-laos m-p-noute TENSE[PRET]=you-eat ACC-the[PLUR]-flesh of-the-people of-the-god hn-ou-hre-n-oeik in-a-food-of-bread "You have eaten the flesh of God's people as bread" (Behlmer 1996:66)

For similar reasons, SVO is the most predictable alternative order in both non-rigid VSOlanguages (as is the case in classical Northwest Semitic languages), in which S is unstable because it occupies the slot between the two components of the VO-predicate, and in non-rigid SOV-languages (as in reconstructed Proto-Indo-European), in which this alternative allows the dominant SV-order to be restaured, cf. Croft 1990:53-63; Hock 1991:313-329. 4 The SOV-order of subordinate clauses in German or Dutch may be seen either as a "frozen" Proto-Germanic remnant or as an innovation motivated by nominalization, cf. Givón 1990: 212-214; 510-511.

FROM VSO TO SVO?

27

2. VSO-constructions There are nonetheless many Coptic constructions which indeed exhibit the VSO-order. These patterns are (a) the predication of existence of non-specific subjects: (12)

oun~hoeine e=u-raše e=u-mpša m-p-rime EXIST-some while=they-rejoice while=they-in-the-value of-the-cry "There are some who rejoice, although they would deserve to cry" (Behlmer 1996:87)

(b) the predication of possession, which is etymologically derived from the latter: (13)

mnta=f ce- mnta=f ke-cnmise NOT HAVE=he other-might NOT HAVE=he other-fighting power "He has no other might, he has no other fighting power" (Behlmer 1996:17)

and (c) the remnants of the Egyptian suffix conjugation: (14)

nanou-t-mntnoc eršan-ne-hbêue hôtr nmma-s be good-the-greatness if-the[PLUR]-thing[PLUR] join with=it "Greatness is good if deeds are combined with it" (Behlmer 1996:108)

Of these constructions, the easiest to explain from a typological point of view is the predication of existence of non-specific subjects, which, because of their relatively low status on the hierarchy of animacy (Croft 1990:111-117), tend in many languages to avoid the head position within the sentence: cf. in the English translation the presence of the dummy deictic adverb "there", which reinterprets the sentence in a localistic perspective and prevents the subject's access to the unmarked topic position (Givón 1984:139-145); cf. also "there is a man at the door" as the more usual form of "a man is at the door", as opposed to the unmarked character of the corresponding sentence with the specific subject in topic position: "my brother is at the door" vs. the marked status of the localistic alternative "there is my brother at the door". To a certain extent, the same explanation applies to the patterns of pos­ session, etymologically derived from the verb of existence (oun-, mn-) followed by a prepositional phrase indicating the possessor (ounta=f "he has" < "there-is in-his-hand", mnta=f "he has not" < "there-is-not in-his hand"). While these constructions retain a localistic feature in the form of a dummy adverb mmau "there", they appear to have completed the process toward a full lexicalization as verbal patterns, since the original subject of the existential

28

ANTONIO LOPRIENO

pattern is now reinterpreted as direct object and preceded by the accusative marker:5 (15)

p-ete-ounta=f mmau n-ou-noub mn-ou-hat he-who-HAVE=he there ACC-a-gold with-a-silver "he who has gold and silver" (Behlmer 1996:128)

As to the Coptic 'vestiges' of the Egyptian suffix conjugation, similar typological factors contributed to their preservation of the VS(0)-structure. In the case of adjective verbs, such as nanou- in (14) above, the reason is once again historical, since they too represent the grammaticalization of the existential predicate followed by a nominalized verb form (nanou=f "he is good" < "there-is that-he-is good"). Other instances of VS-verb forms also derive from underlying nominal patterns: the verb peja=f is a lexicalized relative form: "he said" < "that which he said (is)" (Reich 1924:285-288), the verb hna=f "he desires" finds its origin in the adjectival sentence "it is profitable for him" > "it is his desire (that)" (Sethe 1910:136-141). Finally, the verb meša=fis a relict of a negative pattern on its way to full lexicalization in classical Coptic: mešak ("you don't know") > "perhaps" (cf. Shisha-Halevy 1986:211): (16)

mešak e-u-meeue e-pai nci-net-ti je perhaps that=they-think to-that EXTR-those-who-give that "Perhaps the buyers think that... " (Behlmer 1996:52)

3. Rear extraposition : a multifunctional feature ? If all these forms can ultimately be considered to have a relatively limited functional yield at the synchronic level and thus not to infringe on the paradig­ matic nature of the SVO-order in Coptic, (16) introduces us to the synchronically most productive VSO-pattern which, because of both syntactic structure and pragmatic frequency, deserves the linguist's attention: the rear extraposition of the subject by means of the marker nci- in Sahidic and nje- in Bohairic (Westendorf 1965:137-138). In this type of verbal clause, the subject pronoun is left in its original syntactic slot but is cataphorically expanded by an extraposed nominal subject introduced by nei- (Shisha-Halevy 1986:157; Polotsky 1987:148-150; Polotsky 1990:179):

The same evolution is observable in contemporary Hebrew, where the constructions with yes l-"there is to X" > "X has," enl-"there is not to X" > "X has not" are followed by a direct object introduced, when specific, by the accusativ marker et: yihye lanu et ha-nesek bkarov "We will have the weaponry shortly" (Glinert 1989:179-181).

FROM VSO TO SVO?

(17)

29

se-ôš ebol nci-n-et-nehpe e-p-jinjê they-cry out EXTR-those-who-mourn to-the-vanity "The mourners cry out in vain" (Behlmer 1996:9)

In view of the extraposing function of nci-, which moves the subject to the right of the verbal predicate, it may seem surprising that this construction is wholly compatible with verb-first clauses of the type we touched on in the preceding paragraphs: (18)

ount=f-eksousia gar nci-p-logos m-p-noute e-joo=s je HAVE=he-power then EXTR-the-word of-the-god to-say=it that "for the Word of God has the power to say that... " (Behlmer 1996:49)

(19)

našô=ou be many=they našô=ou be many=they e-tre=u-bôk to-let=they-go et-hm-m-pêue

nci-n-et=f-na-r-joeis ero=ou hitm-p-kôht EXTR-those-who=he-FUT-do-lord to=them by-the-fire on nci-n-et-na-r-bol ero=f also EXTR-those-who-FUT-do-outside to=it ehoun e-m-ma n-mton inside to-the[PLUR]-place of-rest

which-in-the[PLUR]-heaven [PLUR]

"Numerous are those over whom it (= death) will have control by means of the fire, and numerous are also those who will escape it in order to go to the places of rest that are in heaven" (Behlmer 1996:20)

The nci-construction has often been considered to be particularly suited to three functional or especially stylistic environments (Morenz 1952:1-15). First, the presence of a prosodically rather long subject: (20)

f=shouort nci-p-meeue et-hoou nte-p=satanas et-hôš it-be cursed EXTR-the-thought which-evil of-the-Satan who-bring e-p-hêt m-p-rôme mmo=f ehoun ACC=it inside to-the-heart of-the-man "Cursed is the evil thought of Satan, who brings it surreptitiously into the man's heart" (Behlmer 1996:29)

(21)

f=smamaat nci-p-meeue et-nanou=f nte-p-joeis it-be blessed EXTR-the-thought which-be good=it of-the-lord e-p-hêt m-p-rôme et-neu ehoun which-come inside to-the heart of-the-man "Blessed is the good thought of God that enters the man's heart" (Behlmer 1996:31)

Second, a stress on the subject of the verbal clause: (22)

našô=ou e-pe-houo nci-n-šaje têr=ou be many-they to-the-excess EXTR-the[PLUR]-word all=them m-p-noute nt-a=f-taa=u na=n e=u-boethei of-the-god which-PRET=he-give=them to-us while=they-help

30

ANTONIO LOPRIENO

e=u-jimoeit hêt=n e-p-agathon while=they-lead heart=us to-the-good "Even more numerous are all of God's words, which he gave us for they are helpful and lead our heart to the good" (Behlmer 1996:163)

And finally, the desire to imitate a Greek syntactic model; cf. the adherance to the Greek model in the Boharic version of a biblical passage in (23a) as opposed to the unmarked word order of the same verse in Sahidic (23b): (23)

agiasthētō tò ónomá sou a. Bohairic (< Greek) b. Sahidic

(unmarked)

mare=f-toubo nje-pek-ran OPT=it-become holy EXTR-your-name mare-pek-ran ouop OPT-your-name become holy "Hallowed be thy name" (Matthew 6,9)

4. The nci-construction as rheme In the following paragraphs, we shall observe on the one hand that there is only a single functional reality underlying these apparently diverse environ­ ments of the nci=construction (cf. Zakrzewska 1993); on the other, that the nciextraposition does not represent an exception to the unmarked typological SVO-sequence; rather, it reorganizes the distribution of information patterns within the sentence according to the same pragmatic pressures that motivated the change from the VSO- to the SVO-order in the history of Egyptian. Let us start with the alleged desire to imitate the Greek word order. While a certain impact of Greek syntactic models on the development of Coptic stylistic patterns can indeed be detected (Nagel 1971; Reintges, in press), it remains nonetheless clear that linguistic interference generally operates productively within the grammatical and typological boundaries of the target language in a particular stage of its development,6 permeating the system through the more unstable components of grammar and lexicon (Kammerzell 1998:11-23). Consider (24), another example of difference between Bohairic (24a) and Sahidic (24b) in the rendition of a Greek verse: (24)

egéneto rhēma theoû a. Bohairic (< Greek)

a=f-šôpi nje-ou-saji nte-ph-nouti PRET=it-happen EXTR-a-word of-the-god "There occurred a word of God"

There are of course exceptions, mostly due to the desire to reproduce exactly the patterns of the linguistic model of the source language, especially when the latter is considered sacred: cf. e.g. the IInd century CE calque translations of the Bible into Greek by Aquila, Symmachus and Theodotion, as "orthodox corrections" of the IIIrd century  Septuagint version, which was written in productive Hellenistic Greek; cf. Sáenz-Badillos 1975:107-130.

FROM VSO TO SVO?

b. Sahidic

31

(unmarked) a-p-saje m-p-noute šôpe PRET-the-word of-the-god happen "The word of God occurred" (Luke 3,2)

While the argument can indeed be made that the Bohairic choice of the nje-construction was inspired by the Greek sentence structure (Nagel 1971:350-353; Kammerzeil 1998:19), we should observe that the information structure of the two Coptic versions is not identical: in the former sentence, the subject is non-specific ("a word of God"); in the latter, it is specific ('the word of God'*). We saw above that non-specific subjects tend to occupy a lower level on the hierarchy of topicality and to be, therefore, cotextually less expected or less predictable than specific subjects (Chafe 1976:38-43). Thus, whereas the Bohairic translation imitates the Greek version (rhēma theoû) in this feature as well, the definiteness of the subject in the Sahidic text is certainly connected with the preference for the SV-order displayed by the sentence (TENSE MARKER - pšaje - šôpe). From the informational point of view, it looks as if the extraposed subject of the nje-construction, at least in (24a), is less topical than its SV-counterpart: it has surrendered a portion of the expectedness of the subject in unmarked environments to the verbal phrase which occupies the head position in the sentence and which represents a combination of cataphoric subject and predicate (a=f-šôpi "it occurred"). A passage already quoted by Polotsky (1990:179) corroborates this finding: (25)

mar=ou-ouônh na=k ebol nci-n-laos p-noute OPT=they-reveal to=you out EXTR-the[PLUR]-people the-god mare-n-laos têr=ou ouônh na=k ebol OPT-the[PLUR]-people all=them reveal to=you out "Let the nations praise you, o God, let all the nations praise you" (Psalms 66,6)

In this case, it cannot be claimed that the Greek distribution of subject and predicate had any impact on the stylistic choice of the nci-construction in the first hemistich as opposed to the second. In fact, the Greek original exhibits a very similar pattern in the two portions of the verse, with the quantifier pántes being informationally salient in the second hemistich: (26)

eksomologēsásthōsán soi laoí ho theós eksomologēsásthōsán soi laoì pántes

Rather, the reason for the different treatment of the two forms in Coptic is to be sought in the cotextual distribution of attentional foci within the Coptic sentence. In the first hemistich, the theme is the "praise to God", and the extra-

32

ANTONIO LOPRIENO

position of the subject "the nations" allows it to be ascribed a certain degree of rhematic unexpectedness. In the second hemistich, the subject is now known and occupies, therefore, the unmarked thematic position at the head of the sentence ("all the nations"); chiastically, the rheme of the sentence is now the predicate ("give praise to you"), which resumes the theme of the first hemistich: (27) a. [give praise to you]THEME b. [all the nations]THEME

-

[the nations]RHEME [give praise to you]RHEME

Since this distribution of attentional foci is not present in the Greek text, we must posit that in a variety of cases functional reasons were at work in shaping the structure of a Coptic clause with nci-extraposition. This allows us to move to the second function scholars have attributed to its use, namely to stress the presence of a series of antithetical or parallel subjects (Morenz 1952:5-6). Now, the term 'stress' (Betonung) has of course a rather vague, or better diverse semantic scope. One can identify at least two types of linguistic 'stress' or 'emphasis', which we might label 'distinctive' vs. 'isolating' (Depuydt 1991), or perhaps 'assertive' vs. 'contrastive' (cf. Givón 1990:699-778), the supplementary prominence investing in the former case an informationally more expected, in the latter an informationally less expected portion of discourse. The higher information yield in the case of the subject on an nciconstruction belongs to the first type: 'emphasis' in the sense of communica­ tive salience, i.e. of rhematicity, not of communicative contrast, i.e. of focality. All foci are rhematic, but not all rhemes are focal (cf. Givón 1990:699-700). Rather than of emphasis on the subject, we should therefore speak in the case of nci- of its rhematization, with a parallel reduction of the stress on the predicate and its confinement to a more thematic function: [TENSE MARKERSV]THEME - nci-[S]RHEME. This redistribution is the reason why the nciextraposition appears particularly suitable when a series of identical or similar verb forms is thematically predicated of a sequence of different subjects. Compare (19) above, where the thematic continuity of the predicate is balanced by the informational novelty of the subject, which in the second sentence is pointed by the use of the adverb on "also", with (28), in which the SV-phrases represent the carriers of cotextual continuity as opposed to the assertive character of the extraposed subjects:

FROM VSO TO SVO?

(28)

33

se-ôš ebol nci-n-et-nehpe e-p-jinjê they-cry out EXTR-those-who-mourn to-the-vanity se-kôrš ero=f e-tre=f-ouošb-ou nci-net=f-oi they-ask to=him to-let=he-answer=them EXTR-those who=he-be e-kaa=ou n-toot=ou e=f-na-bôk in-hand=them while=he-FUT-go to-leave=them "The mourners cry out in vain, those who value him ask him to answer them, although he is going to leave them" (Behlmer 1996:9-10)

and (29), in which the nci-extraposition expands the subject of two semantically linked predicates: (29)

f-hejhôj ê f-hose nci-pet-eire m-p-agathon he-ache or he-suffer EXTR-he who-do ACC-the-good "He who does good is in pain and suffers" (Behlmer 1996:22)

5.

Topic and focus The higher thematicity of 'serial' portions of discourse may also underlie the preference for the topicalization of recurring subjects, with a structure [S]TOPIC - [TENSE MARKER -SV]COMMENT: (30)

p-mou gar ša=f-r-n-dikaios m-makarios (...) the-death in fact AOR=he-do-the[PLUR]-righteous as-blessed p-mou de on ša=f-r-n-refrnobe n-ebiên (...) the-death indeed also AOR=he-do-the[PLUR]-sinner as-poor "For death makes the righteous blessed, (...) and death also makes the sinners miserable" (Behlmer 1996:26-27)

and even for the fronting of focal morphemes such as etbe-ou or ahro= 'why?' in similar contexts. In these cases the verb, which no longer occupies the head position, tends to be conjugated in the 'non-topical' first tense rather than in the 'topical' second tense:7

7

This linguistic behavior also enables us to establish a syntactic hierarchy in the process which leads to the use of the nominal (or topical) transposition of the verb (Polotsky 1987:129-140): the head position of the verbal phrase seems to be the primary feature vis-à-vis the focality of the adverbial phrase, because a verb form which has the same pragmatic function but does not occupy the first position in the sentence tends to maintain its unconverted tense. Cf. ShishaHalevy 1986:88-93.

34

ANTONIO LOPRIENO

(31)

etbe-ou mp=i-r-hbêue ejn-na-nobe hn-henbecause-what NEG.PRET=I-do-thing[PLUR] on-my[PLUR]-sin in-some nêstia mn-hn-hen-šlêl etbe-ou a=i-ti fast with-in-some-prayer because-what PRET=I-give n-ta-mntrmmao n-ne-hiome auô ACC-my [FEM]-wealth to-the[PLUR]-woman[PLUR] and na-khrêma e-hen-titôn etbe-ou mp=i-ti my [PLUR]-property to-some-suit because-what NEG.PRET=I-give m-pa-oeik m-pet-hkaeit ê a-i-kte-pa-ho ebol ACC-my-bread to-he who-be hungry or PRET=I-turn-my-face out m-pet-šaat ê ahro=i a=i-obš=t e-hobnim of-he who-be needy or why=I PRET=I-forget=me to-thing every e-nanou=ou while-be good=they "Why have I not repented over my sins with fasts and prayers? Why have I given out my wealth for women and my properties for legal suits? Why nave I not given my bread to the hungry? And why have I turned my face away from the needy? And why have I forgotten all good deeds?" (Behlmer 1996:2).

Incidentally, it fits this analysis quite nicely that in the case of ton "where" the fronting of the focus, which is always the marked rheme of the sentence, is, if at all, only extremely rarely combined with the nci-construction (Till 1970:217; Shisha-Halevy 1986:158), which itself rhematizes the subject by transferring a portion of its thematic potential to the entire verbal predication: (32)

e=f-tôn tenou p-hêu m-p-nobe je that=he-wherenow the-advantage of-the-sinthat e=n-na-aa=f (Behlmer 1996:46) PRES=we-FUT-do-it "Now, wherein lies the value of the sin, that we should commit it?"

Similarly, the higher status of the subject of causative verbs on the socalled binding hierarchy-which is closely linked to the topicality hierarchy (Croft 1990:174-183)-, i.e. its more intimate connection with the verbal action8 also prevents them from licensing the nci-extraposition (Polotsky 1987:149). The latter, as we saw, triggers a reorganization of the information within the sentence by breaking the unmarked solidarity between the subject and the predicate to the advantage of the former. Thus, in spite of the semantic complexity and cotextual novelty of the subjects of the causative verb in (33), they do not undergo extraposition: As another example of the higher status of causative subjects on the binding hierarchy cf. the widespread grammaticalization of the third person masculine pronoun (*su-) into the verbal sprefix of causative stems in Afroasiatic (Eg. s-zwr "to let drink"), and especially of the causative conjugation in the Semitic domain (Akkadian saprus, Hebrew hif'il, Arabic 'af'ala): Loprieno 1986:142-144.

FROM VSO TO SVO?

(33)

35

p-kôht hôô-f ete me=f-ôšm mn-p-fnt ete the-fire self=it which NEG.AOR=it-extinguish with-the-worm which me=f-mou n-t-he et-sêh na-tako NEG.AOR=it-die in-the[FEM]-manner which-be written FUT-destroy n-tek-psuchê mn-pek-sôma hrai hn-amnte ACC-your[FEM]-soul with-your-body under in-hell "The unextinguishable fire and the immortal worm will indeed destroy, as it is written, your soul and your body in hell" (Behlmer 1996:63)

6. Prosodic length and informational weight The nature of the subject in (33) leads us to a discussion of the third function usually attributed to the nci-extraposition, namely its preferred use in presence of a 'long' or 'complex' subject (Morenz 1952:4-5). Is there a connection between functional rhematicity and prosodic weight? There is indeed a strong typological link between these two apparently unrelated features: "Moreover, in many languages, extraposition serves merely 'rhythmic' functions, putting 'heavy' or long NPs, which otherwise might excessively interrupt the flow of speech, at the end of the clause" (Hock 1991:316). What Hock calls "the flow of speech" tantamounts to the cotextually acceptable distribution of attentional foci within the spoken or written utterance. This distribution must provide a balance between thematic (or presuppositional, or topical) and rhematic (or unexpected, or focal) elements which enables the speaker or hearer to retrieve information in an economic way. When the subject is prosodically challenging-the quantification of this challenge being a combination of language-internal conventions, universal pragmatic tendencies, and cotextual constraints-its likelihood to function as a theme decreases, and Coptic tends in this case to reorganize the information in such a way that the predicate absorbs a portion of the subject's topicality. The subject, by virtue of its prosodie complexity, necessarily requires an amount of rhematic attention otherwise ascribed to the more unexpected or informative portion of the sentence, i.e. the predicate or its complements. Prosodie length and informational weight, therefore, engage in a cotextual dialectic relationship. In (25), for example, the subject of the sentence in the second hemistich, although prosodically longer (n-laos têr=ou "all the nations") than the subject of the first hemistich, nonetheless functions as theme, whereas the latter is extraposed as rheme (nci-n-laos "the nations"). This is because in spite of the additional information provided by the quantifier, n-laos têr=ou is treated in the Coptic version as cotextually more expected information than the bare n-laos in the first hemistich. In the Greek original, on the other hand, the quantifier functioned as the rheme of the second

36

ANTONIO LOPRIENO

hemistich. We should bear in mind that the promotion of the otherwise most likely topic of the sentence to rhematic prominence is a lexicalizing device (Shisha-Halevy 1986:157) which isolates the semantic subject without affecting the syntactic position of the basic components of the clause. Since it operates above the basic sentence structure, it is indeed compatible with second tense constructions, in which the predicative phrase is topicalized: (34)

nt-a=f-sibe gar an nci-p-noute hn-nef-hap that-PRET=he-change in fact not EXTR-the-god in-his[PLUR]-law "God did not change in his judgments" (Behlmer 1996:31)

The use of the nci-extraposition in second tense constructions is, however, predictably rare, since this pattern serves precisely the function of rhematizing an adverbial element. The extraposition, therefore, creates an informationally infelicitous tension between two rhematic elements, namely the extraposed subject of the second tense phrase and the adverbial rheme of the entire clause: "He, namely God (rheme 1), did not change in his judgments (rheme 2)".9 The nci-extraposition is generally not compatible, however, with focus constructions such as in (32) above, since the focus, as a promoted, or marked rheme, does not tolerate a competing overt rhematization. The lexicalizing expansion, which invests in this case a presupposed element, remains therefore unmarked. 7. Irregular uses of the nci-extraposition The so-called 'irregular uses' of the nci-construction (Crum 1939:252b) remind us that functional movements are always organized according to hierar­ chies (of animacy, topicality, etc.), rather than rigid rules. While the subject is the most likely argument to be rhematized, one can occasionally witness the nci-extraposition of a direct or prepositional object (the examples are from Crum): (35)

ti-sôtm ero=f I-hear to=him "To John I listen"

nci-iohannês EXTR-John

(36)

B a=i-têit=ou nje-ne-šêri e-ph-ouei PRET=I-give=them EXTR-your[FEM.PLUR]-child to-the-far "Your children I placed far away"

(37)

B thê et-a=u-ouôrp šaro=s nje-gabriêl this[FEM] who-PRET=they-send until=her EXTR-Gabriel "She to whom Gabriel was sent"

For examples of nci-extraposition with (both interrogative and non-interrogative) focus cf. Shisha-Halevy 1986:99.

FROM VSO TO SVO?

37

Example (37) is particularly interesting in that the lexicalizing expansion by means of Bnje- affects the object of a general (i.e. non-referential) third person plural verb form, which appears omitted in the surface structure of the sentence (*a=u-ouôrp mmo=f šaro=s "they sent him to her"). This is the Coptic surrogate for the passive construction (Polotsky 1990:181-184): 'they (non-referential) sent him" = "he was sent". The extraposition of the grammatical object and its deletion under relevance in the predicate phrase shows that its role as semantic subject was gradually invading the grammatical sphere. It also provides an insight into the next typological phase in the creation of an autonomous passive conjugation after the standard Coptic pattern in which the semantic subject of a passive surrogate is extraposed by means of the preposition hitn- ("through the hand of" >) "b" and occupies, therefore, a lower level than the subject in the hierarchy of salience: (38)

rôme nim -mou kata p-ent-a=u-toš=f man every FUT-die according to that-which-PRET=they-command=it na=n ebol hitm-p-noute (Behlmer 1996:19) to=us out through the hand of-the-god "Every man shall die according to what was commanded to us by God"

8. Conclusion Coptic displays a relatively flexible SVO-order, in which the basic syntactic sequence can undergo a variety of syntactic movements aimed at reorganizing the information foci within the sentence. In this article, we observed that one of these movements, the extraposition by means of nci-/nje-, rhematizes an informationally heavier subject and symmetrically increases the thematicity of the SV-predicate phrase. Unlike focal constructions (with a sequence FOCUS-PRESUPPOSITION or PRESUPPOSITION-FOCUS) and other rhematizing devices such as the use of the second tense of the verb, nci-extrapositions represent a semantic expansion of a pronominal element which maintains its original syntactic location in the basic sentence. By virtue of this expansion, however, the subject loses a portion of its usual cotextual expectedness by surrendering it to the verbal phrase. In this respect, extrapositions share many discourse features with other Coptic VS-patterns, such as the predication of existence of non-specific subjects (cf. English "there is a man at the door" vs. "my brother is at the door"), the predication of possession, which is etymologically derived from it ("he has" < "there is in his hand"), and a few remnants of the ancient Egyptian suffix conjugation in which various forms of nominalization seem to be at work ("he desires" < "it is profitable for him"). These constructions also show a higher degree of topicality than paradigmatic

38

ANTONIO LOPRIENO

verbal phrases and are frequently accompanied by the nci-extraposition. In general, Coptic verbal patterns that do not display the SV(O)-order tend to counterbalance the pragmatic impact of the word order change to VS(O) by redistributing the information patterns in the predictable sense of a THEMERHEME sequence, i.e. by increasing on the one hand the informational weight of the subject and on the other hand, symmetrically, the thematic expectedness of the head predicate.

REFERENCES Behlmer, Heike. 1996. Shenute von Atripe: De iudicio (Museo Egizio di Torino, Cat. 63000, Cod. TV). (= Catalogo del Museo Egizio di Torino, 8). Torino: Soprintendenza al Museo delle Antichità Egizie. Chafe, Wallace L. 1976. "Givenness, Contrastiveness, Definiteness, Subjects, Topics, and Point of View". Subject and Topic ed. by Charles N. Li, 25-55. New York & San Francisco & London: Academic Press. Croft, William. 1990. Typology and Universals. Cambridge: Cambridge University Press. Crum, Walter E. 1939. A Coptic Dictionary. Oxford: At the Clarendon Press. Depuydt, Leo. 1991. "On Distinctive and Isolating Emphasis in Egyptian and in General". Lingua Aegyptia 1.33-56. Givón, Talmy. 1984-1990. Syntax. A Functional Typological Introduction. 2 volumes. Amsterdam: John Benjamins. Glinert, Lewis. 1989. The Grammar of Modern Hebrew. Cambridge: Cambridge University Press. Hintze, Fritz. 1950. "Konversion und analytische Tendenz in der ägyptischen Sprach­ entwicklung". Zeitschrift für Phonetik und Allgemeine Sprachwissenschaft 4.41-56. Hock, Hans Heinrich. 1991. Principles of Historical Linguistics. Berlin & New York: Mouton de Gruyter. Holes, Clive. 1995. Modern Arabic. Structures, Functions and Varieties. London & New York: Longman. Hopper, Paul J. & Elizabeth Closs Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press. Junge, Friedrich. 1985. "Sprachstufen und Sprachgeschichte". Zeitschrift der Deut­ schen Morgenländischen Gesellschaft. Supplement VI, 17-34. Stuttgart: Franz Steiner Verlag. Kammerzeil, Frank. 1998. Sprachkontakte und Sprachwandel im Alten Ägypten. Habilitationsschrift Universität Göttingen. Kasser, Rodolphe. 1991. "Vocabulary, Copto-Greek". The Coptic Encyclopedia, vol. 8, 215-222. New York: Macmillan Publishing Company. Koch, Roland 1990. Die Erzählung des Sinuhe. (= Bibliotheca Aegyptiaca, 17). Bruxelles: Éditions de la Fondation Égyptologique Reine Élisabeth.

FROM VSO TO SVO?

39

Loprieno, Antonio. 1986. Das Verbalsystem im Ägyptischen und im Semitischen. Zur Grundlegung einer Aspekttheorie. (= Göttinger Orientforschungen, IV.17). Wiesbaden: Harrassowitz. . 1988. "On the Typological Order of Constituents in Egyptian". Journal of Afroasiatic Languages 1.26-57. . 1995. Ancient Egyptian. A Linguistic Introduction. Cambridge & New York: Cambridge University Press. Morenz, Siegfried. 1952. "Die nci-Konstruktion als sprachliche und stilistische Er­ scheinung des Koptischen". Annales du Service des Antiquités de l' Égypte 52.1-15. Nagel, Peter. 1971. "Die Einwirkung des Griechischen auf die Entstehung der koptischen Literatursprache". Christentum am Roten Meer ed. by Franz Altheim & Ruth Stiehl, vol. I, 327-355. Berlin & New York: Walter de Gruyter. Parkinson, Richard B. 1991. The Tale of the Eloquent Peasant. Oxford: Griffith Institute & Ashmolean Museum. Polotsky, Hans Jacob. 1950. "Modes grecs en copte?". Coptic Studies in Honor of Walter Ewing Crum ed. by Michel Malinine, 73-90. Boston: The Byzantine Institute. . 1960. "The Coptic Conjugation System". Orientalia 29.392-422 . 1987-1990. Grundlagen des koptischen Satzbaus. (= American Studies in Papyrology, 27-29). Atlanta: Scholars' Press. Reich, Nathaniel 1924. "A Grammatical Exercise of an Egyptian Schoolboy". Journal of Egyptian Archaeology 10.285-288. Reintges, Chris. in press. "Code-Mixing Strategies in Coptic Egyptian". Lingua Aegyptia 6. Shisha-Halevy, Ariel. 1986. Coptic Grammatical Categories. (=Analecta Orientalia, 53). Rome: Pontifical Biblical Institute. Sáenz-Badillos, Angel. 1975. "El hebreo del s. II d.C. a la luz de las transcriptiones griegas de Aquila, Simmaco y Teodoción". Sefarad 35.107-130. Sethe, Kurt. 1910. "Über einige sekundäre Verben im Koptischen". Zeitschrift für Ägyptische Sprache und Altertumskunde 47.136-146. Till, Walter C. 1970. Koptische Grammatik (Saidischer Dialekt). (= Lehrbücher für das Studium der orientalischen und afrikanischen Sprachen, 1). Leipzig: Verlag Enzyklopädie. Vittmann, Günther. 1991. "Zum koptischen Sprachgut im Ägyptisch-Arabisch". Wiener Zeitschrift für die Kunde des Morgenlandes 81.197-227. Weiss, Hans-Friedrich. 1968. "Beobachtungen zur Frage der griechischen Komponente in der Sprache des Schenute". Probleme der koptischen Literatur ed. by Peter Nagel, 173-185. (= Wissenschaftliche Beiträge der Martin-LutherUniversität Halle-Wittenberg, 1968.1). Westendorf, Wolfhart. 1965. Koptisches Handwörterbuch. Heidelberg: Carl Winter Verlag. Zakszewska, Ewa D. 1993. "The Inner-Verbal Subject in Bohairic Coptic". Discussions in Egyptology 26.71-90.

CONSTITUENT ORDER IN MIDDLE WELSH THE STABILITY OF THE PRAGMATIC PRINCIPLE

ERICH POPPE Philipps- Universität Marburg

1.

The data, and some preliminaries Statistical data about the attestation of different word-order patterns in a number of Middle Welsh prose texts - Middle Welsh is roughly the period from 1200 to 1500 - indicates that with regard to the ordering of the constituents subject (S), verb (V), object (O), and adverbial phrase (A) in positive and pragmatically neutral main statements, no consistent, basic, or dominant, pattern emerges (for the terms, see Sornicola 1994:25). In negative main clauses the preferred order is NEG + V + (S) + ( nom ) (Watkins 1990, see (la)); in subordinate clauses with a conjunction (CON) and a finite verb it is CON + V + (S) + (Onom) (see (lb)). 1 (1) a.

b.

'ny chyuarchaf i well it ' NEG V "I shall not greet you". (PKM 2.8) Ac ual  byd yn llithiau y cwn, ef a welei uarchauc CON V "and while he was feeding the dogs, he could see a rider".

(PKM 2.3)

For positive and pragmatically neutral main statements one regularity emerges on the structural level beyond the obvious variability of the basic constituents, namely that the verb is preceded by (at least) one other constituent (C) and a pre-verbal particle (P) - the realization of which as a or  is conditioned by the syntactic status of the immediately preceding constituent as either S/O or A (see (2)).

Conjunctions which are in origin prepositions are combined with a subordinating particle  (as in (lb)); this is ignored for the purpose of my schematic presentation here.

42

ERICH POPPE (2) a.

Ac ar hynny y deu urenhin a nessayssant y gyt A S PV "and after this the two kings approached each other". (PKM 5.19) Ac ef a welei lannerch yn y coet S PV "and he could see a clearing in the forest".

(PKM 1.13-14)

b.

A chyghor a gymerth Madawc  PV "and Madog took counsel". (BR 1.18)

c.

A threigylgweith yd oed yn Arberth A P V "and one time he was in Arberth". (PKM 1.2-3)

This format is schematically represented in (3). (3)

(C4/3/2) C1 P V (S) (Onom)(A)2

A second format, the so-called cleft sentence, about which more will be said later (see § 2.3), is similar on the surface, but pragmatically different. My formulation here stresses the syntactic and pragmatic status of the pre­ verbal constituent(s), which is also reflected in the functional analysis suggested below; it has recently been re-formulated by David Willis (1997:151) with a different theoretical bias: "it is essentially a verb-second (V2) rule in that it requires the finite verb to appear in second position following an arbitrary phrasal constituent". As I hope to vindicate in § 2, a pragmatic principle governs the selection of the pre-verbal constituent(s). The pattern in (3) is traditionally called the 'abnormal order' in Welsh grammaticography, but it is abnormal only in comparison with dominant neutral VSO in Modern Welsh, and presumably in Old Welsh (see Watkins 1987) as well.3 The statistical analyses referred to above have shown that the pattern in (3) appears to be the dominant pattern in (at least) one specific textual register of Middle Welsh, the language of literary prose. Here it has been demonstrated to occur in texts of the so-called A personal pronoun as the direct object is regularly realized as an enclitic on P, as in the following example: minheu a'e [= P+Opron] kymmeraf yn Nawen (PKM 17.25-26) "I shall accept it gladly". For a further 'irregularity' of the abnormal sentence - from the perspective of Modern Welsh - in comparison with the so-called 'mixed sentence' with emphasis on a fronted constituent, i.e. the cleft-sentence (see § 2.3), cf. Richards (1938:105).

MIDDLE WELSH CONSTITUENT ORDER

43

Mabinogion-corpus, a body of eleven tales which deal with incidents in the legendary history of Wales, in historiographical texts (as indicated by a preliminary survey), and in hagiographical texts adapted from foreign sources.4 For a different textual register, the early poetry in the Book of Aneirin, Graham Isaac (1996:38) has recently suggested that [o]n the basis of both absolute statistical frequency, and grammatical distribution with respect to principle and subordinate clauses, the basic order of the poetry can be established as verb-initial, and specifically of the form VSO.

Since his data-base includes sentence-types other than positive main clauses, his results are not strictly comparable to those of the studies of the prose texts. Oliver Currie (this volume) has adduced further evidence for a split between the word-order systems of Middle Welsh prose and poetry respectively; he has, for example, identified thirty-seven examples of absoluteinitial word order in one-hundred positive declarative main clauses in the cywyddau of Dafydd ap Gwilym. 2. Functional analysis - the pragmatic principle in performance 2.1 Topicalization In the following I shall present the outlines of my hypothesis to explain the mechanisms underlying the selection of the pre-verbal constituent(s) in (3). I argue within a functional framework which posits a division of the clause on the pragmatic level into topic (or theme) and comment (or rheme).5 My contention is that the constituent(s) selected for the pre-verbal position(s) is (are) assigned topic function by the 'speaker'. In my understanding of the notion 'topic' I follow Tanya Reinhart (1983:6) who defines sentence-topics in terms of "pragmatic aboutness": [A]n expression will be understood as representing the topic if the assertion is understood as intending to expand our knowledge of this topic, in accordance with the second principle [viz. the PRINCIPLE OF RELEVANCE [which] states that discourse does not proceed arbitrarily but rather relates itself to and makes use of what is presumed to be known and 'intends, in general, to give or add information about what is a matter of standing or current interest or concern']. So the crucial thing here is not what can be assumed to be already known, but what can be assumed about the purpose of the utterance. 4

Cf. Watkins (1977/78), (1983/84), (1988), (1993), (1997) and Poppe (1989), (1990), (1991a) for texts from the Mabinogion-corpus; Poppe (1991a) for a preliminary survey of a text from a historiographical context; and Poppe (1993) and Fife (1996) for religious texts adapted from foreign sources. The pre-verbal particle would appear to function typically as the demarcation between the topic and the comment, but there may be problems here with the status of postverbal definite subjects, which require further investigation (see also fn 15).

44

ERICH POPPE

In her - and my - view a speaker has (relative) freedom to select a constituent in order to assign topichood to it - this is not an automatic process independent of the speaker's communicative intention, or in the words of Morgan (1975, quoted Reinhart 1983:7): "It is not sentences that have topics, but speakers". Admittedly there are some features that are prototypically associated with topics - thus they tend to represent old/given information,6 and certain roles and categories seem to be ordered hierarchically with regard to their potential for topichood, viz. subject > direct object > indirect object > N in PP (Reinhart 1983:34) or 1st person > 2 nd person > 3 rd person (Mallinson & Blake 1981:158). I would suggest, however, that such hierarchies reflect the statistical probability of these constituents' selection as topics, rather than an automatism or determinism according to which topics are selected.7 In (3) the pre-verbal C-position(s) can be filled with nominal or adverbial constituents; and nominal and adverbial topics have broadly similar, but not identical textual functions: Both establish points of reference and orientation.8 Adverbial topics typically provide the scene for the following assertion including, if present, a nominal topic - by defining its temporal and local, but also conceptual, background (scene-setting). The typical order of topical constituents (T) is therefore as in (4).

(4)

TA TN P V (S)(O)

The following table (5) uses two parameters, information status represented by the poles 'old/given' versus 'new' - and syntactic status - TN versus TA - to characterize in a somewhat simplified form the core or prototypical textual functions of topical constituents. (5) old/given new

TN topic continuity topic change

TA connective paragraph marker

But as Reinhart (1983:18) has reminded us "although in most cases the topics tend indeed to represent old information, this is neither a sufficient nor a necessary condition for topichood". See also Isaac (1996:39-106) for a discussion of a topic hierarchy and its applicability to the analysis of Middle Welsh word order. Compare Reinhart (1983:24): "Sentence-topics [...] are one of the means available in the language to organize, or classify the information exchanged in linguistic communication - they are signals for how to construct the context set [viz. assess by what you already know about the topic; store under an entry corresponding to this topic], or under which entries to classify the new proposition".

MIDDLE WELSH CONSTITUENT ORDER

45

In actual performance, some topics will be, so to speak, better representa­ tives of this category than others (compare, e.g., Duszak 1987; Sornicola 1988: 370-371, 376-378); and if one accepts the premises of prototype theory as a valid hypothesis to describe the structure of linguistic categories the less typical, marginal examples will not constitute evidence against the overall viability of the category (compare Kleiber 1993:76). On the whole, the interpretation of pre-verbal nominals and adverbials as topics with a range of specific intra- and extra-sentential functions is, I think, relatively straightfor­ ward, on the level both of the abstract pragmatic principle and of the actual textual performance.9 There is one special format which may require a brief comment here, the so-called gwneuthur-periphrasis in which the verbal phrase is broken up into a non-finite element, a verbal noun, in topic position - syntactically the object and a finite form of gwneuthur "to do" as auxiliary;10 the semantic core of the verbal phrase thus appears in sentence-initial position (see (6)). (6)

A guahanu a wnaethant OVN P Vaux "and they separated". (PKM 13.12)

I have suggested elsewhere (Poppe 1993:110) that this pattern, which typically occurs in narrative contexts and with a past tense,11 "presents a special format for continuous neutral narration in which no nominal participant of the verbal action has to be topicalized and where the subject typically will be recoverable from the immediately preceding sentence or a fronted adverbial expression". 2.2 Sentence-initial verbs Although the pattern given in (3) is statistically clearly dominant in the corpus under consideration here, there occurs in some, but not all, texts a small percentage of sentences with the verb in sentence-initial position, from about 2% to 10% (see (7)).

For examples of such analyses cf. Poppe (1989/90), (1989:52-58), (1991: 36-122, 269-303), (1991a:184-200), (1993:105-114); see also Isaac (1966:45-56) and, for a more general discussion, Fife (1988) and Fife & King (1991). For a dissenting view, see Shisha-Halevy (1995:149-156). For the semantics of this construction in Modern Welsh, with some remarks on Middle Welsh, see Fife (1986). See also Shisha-Halevy (1998:22-54) for further pertinent remarks on the use of this construction in Middle and Modern Welsh. The fronted verbal noun can also be the focus in a cleft-construction (see § 2.3); cf. Poppe (1993:111-112) for an example from direct speech.

46

ERICH POPPE

(7) a. 'y huant ulwydyn gyt a mi' PV "they were with me for a year". b.

(PKM 35.24)

A gwr a oed ar  keis hwnnw, sef oed y enw, Ronabwy.   doeth Ronabwy a PV S Chynnwric Vrychgoch, gwr  Vawdwy, a Chadwgawn Vras, gwr o Voelure yg Kynlleith,  ty Heilyn Goch "and a man was on this quest, Rhonabwy was his name, and Rhonabwy and Dynwrig Frychgoch, a man from Mawddwy, and Cadwgawn Fras, a man from Moelfre in Cynlleith, came to the house of Heilyn Goch". (BR 2.4-8)

This verb-initial order is found typically with forms of the verb bot "to be" (see (7a)) or, as David Willis (1997) has argued, "in the second pair of conjoined sentences" - as in (7b). The latter type he has convincingly explained as a special device for clausal coordination "in which the topic position is occupied by a null topic operator" (Willis 1998:13). Such apparent violations of V2 occur in other V2-languages as well, for example in Old Icelandic (see (8a)), with the same underlying structure as in Middle Welsh, and in Modern High German (see (8b)), with a different underlying structure (compare Willis 1997:156-169).12 (8) a.

b.

Olaf konungr hafpi stefny við lið sitt oc við bøndr oc reðv landraðym. SSg V3rd V 3 r d pl sg "King Olaf had a meeting with both his men and the farmers, and they all considered government policy". Das Gepäck ließ er fallen und rannte zum Ausgang. "He dropped the luggage and ran to the exit".

2.3 Cleft-constructions A construction which on the surface is in most cases similar to (3), is the cleft- or focus-construction as in (9).13 (9)

(form of bot) CFOCUS P rel V (S) (O)

Here one constituent is moved to the front and optionally marked by a form of bot "to be", and the rest of the assertion follows in a relative construction. This is a pragmatically marked format with the basic communica­ tive function to identify "that a certain focus item rather than other possible ones are correct" (Chafe 1976:35). Whereas topics are topics, the fronted

See now Willis (1998:102-133) for a more detailed discussion of verb-first constructions in Middle Welsh. There is no formal difference between P and Prel in Middle Welsh.

MIDDLE WELSH CONSTITUENT ORDER

47

constituents in the cleft-construction are marked comments (see (10) for an example of contrastive focus).14 (10)

Nyt ymywn ty yd oydynt, namyn ymywn palleu NEG A Prel V "it was not in a house where they were, but in tents". (PKM 31.11)

The contrast between (11a) with focus and (11b) with topic in sentenceinitial position is instructive with regard to the formal differences between the two constructions: (11) a. 'Miui a uyd gwassanaethwr hediw' C F o c u s R V3sg "(It's) I [and nobody else] (who) will be server today." (WLSD 7.30-31) b. 'Miui a gerdaf y gyt a chwi hyt y sened' $Topic P V 1 s g

"I shall go with you to the synod" (WLSD 9.31)

2.4 Left-dislocation In left-dislocations (LD) an element outside the syntactic frame of the clause is anaphorically taken up in the clause itself. The pragmatic function of such constructions is perhaps most adequately described with the words of Simon Dik (1981:13) as "a domain or universe of discourse with respect to which it is relevant to pronounce the following predication". Left-dislocated elements can be characterized as marked topics, which in Middle Welsh are typically followed by ordinary topics (see (12)). (12)

'Dy holl golledeu eiryoet, ' heb yr ynteu, 'o'ra wneuthum i ytti, mi a' e S P Opronl LD1 hennillafitt' (KLl 164-165) V "AH your losses, 'he said', which I have caused you, I will restore them to you".

3.

The synchronic stability of the pragmatic principle In the context of a comparison of sentences in Breuddwyd Rhonabwy which express the same or, at least, very similar information content, I have found that "variation in word-order and sentence types is remarkably infrequent" and I have suggested that this "seems to indicate that the selection of the fronted constituent does not depend on a purely stylistic principle of variatio" (Poppe 1990:458) - I would now reformulate that this situation is a result of the pragmatic principle's synchronic stability. One methodological See Poppe (1991:264-266) for a brief typology of focus-functions in Middle Welsh; for a general discussion of so-called it-clefts, cf., for example, Erdmann (1987) and Sornicola (1988).

48

ERICH POPPE

complication needs to be taken into account, however, that formulae and conscious parallelism/repetition are narrative devices which play an important part in the medieval Welsh narrator's technique (see Davies 1995:104-188) and Breuddwyd Rhonabwy is a particularly literate and structurally complex and polished text (see Slotkin 1989). Doris Edel's (1996:319-327) analyses of parallel scenes in Peredur yield similar results with regard to the use of word order and sentence patterns, but the same methodological reservations remain valid.15 The story of Lludd and Llefelys occurs in two different versions belonging to two different genres, the Brut-corpus and the Mabinogion-corpus, and the pragmatic principle is realized in both, i.e. the statistical dominance of pattern (3). There are some differences in the realization of individual pairs of sentences, however; the Brut-version prefers TA TS P V, whereas the Mabinogion-version prefers T A PV (S) (Poppe 1991a:173-176, see (13)). (13) a. Brut-version Ac gwedy henny wynt a kymerassant [...] A S PV "and afterwards they took [...]". (quoted Poppe 1991a:174) b. Mabinogion-version Ac odyna y kymerassant [...] A PV "and then they took [...]". ( 70)

4. Conclusion I have outlined here some aspects of the variation in the attested wordorder patterns in Middle Welsh, with regard to the ordering of the constituents S, O, V, A, and of the stability of what I term the pragmatic principle which governs the selection of the pre-verbal constituent(s)16 - in distinction from a notion of stylistic, and random, variatio or of an automatic selection for fronting of a constituent-type in specific syntactic contexts. Topic and focus are central parameters for the synchronic analysis and the integrated explanation of word-order phenomena in the specific literary register of Middle Welsh prose. I have not addressed the question of word-order changes in the history of Welsh here, only the principle which governs synchronic stability underlying surface variation within one subsystem of Middle Welsh. I think, however, that There is hardly any 'inexplicable' word-order variation in her examples, except perhaps in the case of the position of S in Pan doeth y dyd gyntaf kyfodi a oruc Peredur versus Tranoeth y bore Peredur a gyfodes y vynyd (Edel 1996:325), but see above, fn 5, for a brief reference to the question of the status of postverbal definite Ss. The ordering of post-verbal constituents is also susceptible to pragmatic principles, but this is beyond the scope of my discussion here.

MIDDLE WELSH CONSTITUENT ORDER

49

linguists need to have a firm idea about the syntactic situation which may have been the spring-board for further developments, 1 7 before they are in a position to discuss the direction in which syntactic patterns change over time.

ABBREVIATIONS BR = Richards (1948). KLl = Roberts (1975). PKM = Williams (1930). WLSD = Evans (1980). REFERENCES Chafe, Wallace L. 1976. "Givenness, Contrastiveness, Definiteness, Subjects, Topics, and Point of View". Subject and Topic ed. by  N. Li, 25-55. New York: Academic Press. Davies, Sioned. 1995. Crefft  Cyfarwydd. Astudiaeth  Dechnegau Narratif   Mabinogion. Caerdydd: Gwasg Prifysgol Cymru. Dik, Simon C. 1981. Functional Grammar. Dordrecht: Foris. Duszak, Anna. 1987. "Prototypes, Sentence Topics and Word Order Typologies". The Nordic Languages and Modern Linguistics, 6 ed. by Pirkko Lilius & Mirja Saari, 147-62. (= Proceedings of the Sixth International Conference of Nordic and General Linguistics.) Helsinki: Helsinki University Press. Edel, Doris 1996. "The 'Mabinogionfrage': Arthurian Literature between Orality and Literacy". (Re)Oralisierung ed. by L. C. H. Tristram, 311-333. Tübingen: Narr. Erdmann, Peter. 1987. It-Sätze im Englischen. Heidelberg: Carl Winter. Evans, D. Simon. 1980. The Welsh Life of St David. Cardiff: University of Wales Press. Fife, James. 1986. "The Semantics of gwneud Inversions". Bulletin of the Board of Celtic Studies 33.133-144. Fife, James. 1988. Functional Syntax. A Case Study in Middle Welsh. Lublin: Redakcja Wydawnictw Katolickiego Uniwersytetu Lebelkiego. Fife, James. 1996. "The Syntax of the Middle Welsh Mair o'r Aifft". The Legend of Mary of Egypt in Medieval Insular Hagiography ed. by E. Poppe & . Ross, 237254. Blackrock: Four Courts Press. Fife, James & Gareth King. 1991. "Focus and the Welsh 'Abnormal Sentence': A Cross-linguistic Perspective". Studies in Brythonic Word Order ed. by J. Fife & E. Poppe, 81-153. Amsterdam: John Benjamins. Isaac, Graham H. 1996. The Verb in the Book of Aneirin. Studies in Syntax, There is, of course, the additional question of registers, and it is not at all clear that Modern Welsh VSO continues the register in which Middle Welsh CT P V (S) (O) was dominant. For a discussion of problems relating to changes in word order between Middle and (Early) Modern Welsh, see Oliver Currie's contribution to this volume.

50

ERICH POPPE

Morphology and Etymology. Tübingen: Niemeyer. Kleiber, Georges. 1993. Prototypensemantik. Eine Einführung. Tübingen: Gunter Narr. Mallinson, Graham & Barry J. Blake. 1981. Language Typology. Cross-linguistic Studies in Syntax. Amsterdam: North Holland. Poppe, Erich. 1989. "Constituent-ordering in Breudwyt Maxen Wledic". Bulletin of the Board of Celtic Studies 36.43-63. Poppe, Erich. 1989/90. "The Position of Temporal Adverbials with nos as Core in Middle Welsh Sentences —A Functional Approach". Studia Celtica 24/25.117-129. Poppe, Erich. 1990. "Word-order Patterns in Breudwyt Ronabwy". Celtic Linguistics. Ieithyddiaeth Geltaidd. Readings in the Brythonic Languages. Festschrift for T. A. Watkins ed. by M. J. Ball, J. Fife, E. Poppe, J. Rowland, 445-460. Amsterdam: John Benjamins. Mittelkymrischen. Poppe, Erich. 1991. Untersuchungen zur Wortstellung im Temporalbestimmungen und funktionale Satzperspektive. Hamburg: Buske. Poppe, Erich. 1991a. "Word Order in Cyfranc Lludd a Llefelys: Notes on the Pragmatics of Constituent-ordering in MW Narrative Prose". Studies in Brythonic Word Order ed. by J. Fife & E. Poppe, 155-204. Amsterdam: John Benjamins. Poppe, Erich. 1993. "Word Order in Middle Welsh: The Case of Kedymdeithyas Amlyn ac Amic" . Bulletin of the Board of Celtic Studies 40.95-117. Reinhart, Tanya. 1983. Pragmatics and Linguistics. An Analysis of Sentence Topics. Bloomington: Indiana University Linguistics Club. Richards, Melville. 1938. Cystrawen y Frawddeg Gymraeg. Caerdydd: Gwasg Prifysgoî Cymru. Richards, Melville. 1948. Breudwyt Ronabwy. Caerdydd: Gwasg Prifysgol Cymru. Roberts, Brynley F. 1975. Cyfranc Lludd a Llefelys. Dublin: Dublin Institute for Advanced Studies. Shisha-Halevy, Ariel. 1995. "Structural Sketches of Middle Welsh Syntax (I)". Studia Celtica 29.127-223. Shisha-Halevy, Ariel. 1997. Structural Studies in Modern Welsh Syntax. Aspects of the Grammar of Kate Roberts. Münster: Nodus Publikationen. Slotkin, Edgar M. 1989. "The Fabula, Story, and Text of Breuddwyd Rhonabwy". Cambridge Medieval Celtic Studies 18.89-111. Sornicola, Rosanna. 1988. "It-Clefts and Wh-Clefts: Two Awkward Sentence Types". Journal of Linguistics 24.343-379. Sornicola, Rosanna. 1994. "On Word-Order Variability: A Study from a Corpus of Italian". Lingua e Stile 29.25-57. Watkins, T. Arwyn. 1977/78. "Trefn yn y Frawddeg Gymraeg". Studia Celtica 12/13.367-397. Watkins, T. Arwyn. 1983/84. "Trefn y Constitwentau Brawddegol yn Branwen". Studia Celtica 18/19.147-157. Watkins, T. Arwyn. 1987. "Constituent Order in the Old Welsh Verbal Sentence". Bulletin of the Board of Celtic Studies 34.51-60. Watkins, T. Arwyn. 1988. Constituent Order in the Positive Declarative Sentence in the Medieval Welsh Tale 'Kulhwch ac Olwen'. Innsbruck: Institut für Sprachwissenschaft. Watkins, T. Arwyn. 1990. "Constituent Order in the Negative Sentence in the White

MIDDLE WELSH CONSTITUENT ORDER

51

Book Version of Kulhwch ac Olwen". Celtic Language, Celtic Culture. A Festschrift for Eric P.  ed. by A. T. E. Matonis & D. F. Melia, 247-252. Van Nuys: Ford & Bailie. Watkins, T. Arwyn. 1993. "Constituent Order in Main/Simple Verb Clauses of Pwyll Pendeuic Dyuet".Language Sciences 15.115-139. Watkins, T. Arwyn. 1997. "Constituent Order in the Main/Simple Verbal Declarative Clause in Math uab Mathonwy". Studia Celtica 31.195-217. Williams, Ifor. 1930. Pedeir Keinc  Mabinogi. Caerdydd: Gwasg Prifysgol Cymru. Willis, David. 1997. "Clausal Coordination and the Loss of Verb-second in Welsh". Oxford Working Papers in Linguistics, Philology and Phonetics 2.151-172. Willis, David. 1998. Syntactic Change in Welsh. A Study of the Loss of Verb-Second. Oxford: Clarendon Press.

PART II STABILITY AND DIACHRONY

STABILITY AGAINST THE ODDS? THE SURVIVAL OF VERB FINAL ORDER IN AKKADIAN

GUY DEUTSCHER University of Cambridge

Why are the last ones always at the end?

0.

Introduction The question raised in this article is very simple, but one which is not often asked in historical linguistics. Whereas normally we try to explain why changes do occur, I ask why, perhaps contrary to expectation, hardly any changes happen over a period of two millennia. The Akkadian language emerges on the stage of history with word order patterns which, according to established WO typologies, could be regarded as very 'inconsistent'. Akkadian is verb final, but in most respects it behaves like VO languages. Given such inconsistency, we might expect that this situation would be unstable and quickly lead to changes. Nevertheless, throughout its long history, Akkadian shows little inclination for any changes in its word order. Even further 'incentives' for change, such as the disintegration of the case system, do not make Akkadian oblige, and the language remains stubbornly attached to its 'problematic' word order. In the following sections, after surveying the facts in a little more detail, I try to explain why the situation in Akkadian was actually less problematic than it may appear in theory. I conclude that it can be misleading to treat word order in isolation from other structural properties of the language. In short, I argue that there is more to life than word order and case systems. 1.

The facts: word order patterns in Babylonian Akkadian Akkadian is a Semitic language which was spoken in ancient Mesopota­ mia, probably from the beginning of the third millennium . It is attested in writing from ca. 2500 , and survives as a spoken language until ca. 500 , when it is replaced by Aramaic. There were two main dialects of Akkadian, Babylonian (spoken in southern Mesopotamia), and Assyrian (spoken in the

56

GUY DEUTSCHER

north). In this paper, only Babylonian is considered, although I sometimes use the term 'Akkadian' as well, especially where general properties of the language are concerned. (There do not seem to be fundamental differences in word order patterns between Babylonian and Assyrian.) The text corpus which forms the basis of the investigation is comprised mainly of one genre of prose texts, namely letters. Letters were written on clay tablets, and they are attested from all periods of Babylonian. They deal with a variety of topics: personal, commercial, administrative, legal, and diplomatic. They contain various fixed formulaic expressions (especially greeting formulae), but are often written in very simple and colloquial style. I also consider legal documents (law codes and court protocols), which are more formal in style. The examples quoted in this paper are taken mostly from the two most richly attested periods, conventionally known as Old Babylonian (~2000-1600 , glossed as "OB"), and Neo-Babylonian: (~1000-500 , glossed as "NB"). All the statements made about Akkadian word order in this article concern only prose texts. I do not consider poetic texts at all, because, as one would imagine, word order (including the position of the verb) in poetry is much more flexible than in prose, and is determined more by the rhetorical demands of poetry than by the conventions of the spoken language. Akkadian word order is interesting because it is highly 'inconsistent'. Akkadian comes from a good Semitic home, with VSO word order and very 'consistent' VO patterns. But because of centuries of intimate cohabitation with the strictly verb-final Sumerian, Akkadian went astray: from the earliest documents it already appears as a strictly verb-final language. However, inside the phrase, the order of elements remains largely unchanged, and displays Akkadian's Semitic inheritance. The resulting situation is a language with strict verb-final order, but which otherwise behaves like VO languages.1 The following examples demonstrate the main word order patterns in Babylonian in the early periods. The verb follows the subject, the object (as full nouns), PPs and adverbials. The strict verb-final rule is valid both for main clauses and subordinate clauses:

It is natural to ask why Akkadian borrowed the verb-final order from Sumerian, but not other features such as postpositions. Sumerian itself has postpositions, but N-G and N-Rel order. (Is this in turn Akkadian influence?) Sumerian and Akkadian show signs of Sprachbund conver­ gence, but it is difficult to speculate on individual issues relating to borrowing, because these mostly occurred in prehistoric times.

THE SURVIVAL OF VERB FINAL ORDER IN AKKADIAN

57

(1) OB

UGULA.MAR.TU dabāb-am suātu ana Sin-iqīšam šāpir general statement-c that .acc to Sin-iqīšam governor.of Suhi ublam Suhum he.took "The general took that statement to Sin-iqīšam the governor of Suhum"

(2) OB

inūna šāpir-ī ištu B ābili ana Sippar ana mahar when overseer-my from Babylon to Sippar to face.of Šamaš...Ulikam marşāku-ma Šamaš he.came I.ill.stative-P "When my overseer came from Babylon to Sippar, into the presence of (the God) Šamaš, I was ill."

(3) OB

summa awīl-um balum bēl kir-îm ina kirī if man-nom without owner.of orchard-gen in orchard.of awīl-im iş-am ikkis ½MA.NA kasp-am išaqqal man-gen tree-c he .cut down ½mina silver-acc he.weighs "If a man cut down a tree in (another) man's orchard, without the permission of the owner of the orchard, he shall pay ½ mina of silver."

The strict verb-final order has only two exceptions. The first is the object pronoun suffixes, which are a feature of the inherited Semitic structure: (4) OB

ana Jahruru-šaplT uštābil-šināti to Jahruru-šaplT Isent-them "I sent them to Jahruru-šaplī."

(5) OB

kasp-am lušābilak-kim-ma silver- let.me.send-to you(f.sg)-P "let me send you(f.sg) silver"

The only independent constituents which can follow the verb (initially only rarely) are finite sentential complements and some temporal adverbial phrases. These are also the only constituents which did undergo word order change during the history of the language, and they will be discussed in § 3.2 below. Apart from the two exceptions above, the verb-final order is very strict. No devices are ever used to bring the verb forward in the clause. Nevertheless when we look inside the phrase, the order of constituents is entirely the VO type. The following examples show the order Noun-Genitive, Noun-Relative, Noun-Adjective2 and some prepositions.

Dryer (1992) has shown that the order N- does not in fact have a statistically significant correlation with the order VO. Strictly speaking, therefore, N- should not be mentioned alongside N-G, N-Rel, and prepositions.

GUY DEUTSCHER

58 (6) OB (7) OB

šēp-am ahīt-am ana libbi āl-im la foot-c foreign-acc to heart.of town-gen not "don't let a foreign foot into the heart of the town"

tuserrebă you(pl).let.enter

ina ēribti girr-im mahrï-tim sa irrubam sTm-am asâmam-ma at arrival.of caravan-gen first-gen REL enters.3sg goods-c I.(shall).buy-P "at the arrival of the first caravan that comes in, I will buy the goods..."

These basic word order patterns remain stable throughout the history of the language.3 The following examples from Neo-Babylonian show the verb-final order, and the order of constituents within the phrase: (8) NB

Asarëdu sākin tēmi sa Kūti bīt abī-ja ana najjālu Asarëdu governor of Kutha house.of father-my to dispossessed inamdin he.(will).give "Asarëdu, the governor of Kutha, wants to give the house of my father to the dispossessed person"

(9) NB

25 GUR uttatu ana PN ana muhhika 25 kor barley to PN for you(m.sg) "I have paid 25 kor barley to PN for you."

(10) NB

ana muhhi ad-ê sa Bābili sa sarru bēl-ā ispur-a concerning oath-gen of Babylon REL king lord-my he.wrote-to me "concerning the oath of Babylon that the king my lord wrote to me (about)."

ēteţir Lpaid(perfect)

It is often claimed that such inconsistent combination of strict OV word order but otherwise VO patterns should be unstable. We could therefore expect that during the long history of Babylonian, word order changes would abound. (Sumerian, the original culprit responsible for the inconsistent word order, died out as a spoken language probably around 2000 , thus leaving Babylonian 1500 years to develop without any 'adversary' influence.) Moreover, from around 1000 , the case system started to disintegrate, a development which is often claimed to be a catalyst for changes in word order. But as mentioned above, Babylonian shows little signs of unease with its word order patterns, and they remain remarkably stable. I therefore return now to the opening question of the article. In the following sections, I try to explain why the apparently problematic word order in Babylonian did not result in instability and change. First, in § 2, I examine why the disintegration of the case system in an 'OV' language did not pose any serious difficulty. Then, in § 3,I try to explain why the 'inconsistent' word order was much less problematic than what we may be led to believe.

In the Neo-Babylonian period, as the language is dying, we can find a very small number of examples of verb-medial sentences in letters. These probably appear under Aramaic influence (von Soden 1995:227). In any case, their number is negligibly small.

THE SURVIVAL OF VERB FINAL ORDER IN AKKADIAN

59

2.

Change because of disintegration of the case system? It has often been claimed that one of the main reasons for change from SOV to SVO word order in a language is the disintegration of its case system. This claim, however, has already been refuted (cf. summary in Harris and Campbell 1995:216ff.). Babylonian supplies further evidence that a language does not need to change its word order to SVO when cases disappear. But Babylonian also shows that a language does not even have to fix its (preverbal) word order at all when cases disappear. In Babylonian, both before and after the disintegration of the case system, the verb is in final position, but the word order before the verb is relatively free. Both SOV and OSV orders are very common. The following examples show SOV and OSV order in Old Babylonian: (11) OB

SOV (Subject-PP-Object-Adverbial-V) Sin-iddinam ina Bābili 1½ GÍN kasp-am ana SE-e Sin-iddinam in Babylon 1½ shekel silver-c to barley-gen sâm-im... iddin-ma buy.inf-gen he.gave-P "Sin-iddinam gave I½ shekels of silver in Babylon (in order) to buy barley."

(12) OB

OSV (Object-PP-S-PP-V-Indirect object pronoun) 1 ŠEGUR... ina bTt-i awīl-um... mahrī-ka liddin-su 1 kor barley from house-gen gentleman-nom front-your let.him give-to him "Let the gentleman give him 1 Kor barley from the house in your presence."

In the Neo-Babylonian period, the case system disintegrates, and the distinction between nominative and accusative disappears on nouns. (The genitive is still sometimes distinct.) But Neo-Babylonian remains verb-final, and moreover, pre-verbal word order remains flexible. Both OSV and SOV are still very common: (13) NB

(14) NB

OSV mīnamma sa ümu 10 GUR uţţatu ina libbi arrabu  işşur ikkalū why REL daily 10 kor barley fromit 'mouse'and bird they.eat "why is it that 'mice' and birds eat 10 kor barley from it daily?" SOV adû sarru bēl-ā kīnūtu sa ardi-su itti bīt now king lord-my loyalty of servant-his with house.of bēli-šu immar lord.gen-his he.(will).see "Now the king my lord will see the loyalty of his servant to the house of his lord."

GUY DEUTSCHER

60

But why should a language adopt a fixed word order when its case system disintegrates? The usual arguments claim that free word order without case distinctions would cause ambiguity between subject and object, and thus lead to confusion. But Babylonian speakers evidently did not find the situation too confusing. The reason is that there were other means at their disposal which helped to distinguish between the subject and the object. I shall briefly list them here. First, even before any syntactic devices are used, pragmatic considerations can determine what is the subject and what is the object in most instances of natural discourse. In laboratory examples, the subject and object may be semantically interchangeable. But in common discourse, this is less often the case. Consider the OSV clause in (13) above: '[10 kor barley]o [the mice and birds]s [eat]v'. Even without any syntactic devices to distinguish the subject from the object, there is hardly any room for confusion about who does the eating and who is being eaten. After pragmatics, there are some syntactic features of the language which help to distinguish subject from object. In Babylonian, verbal agreement plays an important role in the identification of the subject. The verb agrees with its subject in person, number, and sometimes gender. For example, in (15) below, the singular third person verb determines that the "king" is the subject rather than the object. (As it happens, this interpretation is obvious from the pragmatic context anyway. It is the king who arrests people, not vice versa.) (15) NB

OSV PN bū-šu  ahhē-šu šarru uşşabbit PN father-his and brothers-his king seize(perfect).3sg "the king should seize PN, his father, and his brothers."

In fact, many clauses do not have an overt subject NP at all. Independent subject and object pronouns are hardly ever used. Pronominal subjects and objects are generally only coded on the verb, as for example in (4), (5), (6), (7), (9) above. Thus, clauses which contain both the subject and the object as full independent NPs are not very common in practice. Moreover, in addition to the obligatory subject agreement on the verb, resumptive object pronouns are often used as well. This amounts to optional 'object agreement', and it can help further to distinguish subject from object. The following two examples show how resumptive object pronouns are used in cases which otherwise would perhaps be confusing. First, in (16) below (from the Old Babylonian law code of Hammurapi), both the semantic content of the sentence and the word order may cause confusion. The subject ("obligation") and the object ("man") are in the 'wrong'

THE SURVIVAL OF VERB FINAL ORDER IN AKKADIAN

61

thematic slots (i.e. reverse from prototypical) in terms of their animacy. The natural tendency therefore would be to parse the "man" as subject (rather than object, as it is here), and "obligation" as object. The OSV word order may also be confusing, because the general pattern in the law code is for laws to start with 'if a mans...', where the man is the subject, not the object. In this particular example, the subject and object are already marked by their case ending. But the object resumptive pronoun 'him' is used here as well, and this gives a secondary identification of the 'man' as an object rather than the subject ('obligation' is feminine). (16) OB

summa awīl-am e'ilt-um işbas-şu-ma assas-su... man-c obligation-nom seized.3sg-him-P wife-his if ana kasp-im iddin for silver-gen he.gave "If a financial obligation 'seized' a man (i.e. the obligation is outstanding), and he sold his wife... "

In example (17) from the Neo-Babylonian period, the word order flexibility itself may cause confusion. Here, the resumptive pronoun "him" is used to smooth the passage from an SOV order in the first clause ("kings Şillajao - asked"), to an OSV order in the second clause ("Bel-eţir0 - kings should ask-him"). The resumptive pronoun used in the second clause (but not in the first) makes the transition from the SOV to the OSV structure less confusing. (17) NB

sarru ū-.. ana muhhi dibba agâ Şillăja iš'al king father-your concerning matter this Şillăja he.asked enna Bēl-ētir rēs-sarri sarru liš'al-šu-ma now Bēl-ētir advisor king let.him.ask-him-P "your father the king would ask Şillăja concerning this matter, now the king should ask Bēl-ētir, the king's advisor "

Resumptive object pronouns often come in conjunction with topicalization of the object, and this results in even stronger marking of the object: (18) NB

altemû umma şābe-ja sa halqu ahū-ā I.heard quotative: men-my REL lost.stative.pl brother-my iptatar-sunūtu he.ransomed-them "I have heard: 'my men who were lost, my brother ransomed them"'

62

GUY DEUTSCHER

Especially in the Old Babylonian period, we often find a high degree of over-determination of the object. In (19), for example, the combination of topicalization, resumptive pronoun, and a repetition of the full NP, may seem to us redundant: (19) aššum şuhări-ka sa taspur-am şuhăr-am suāti isahhurūniš-šu- OB about servant-your REL you.wrote-me servant-c this.acc they.(will)seek-him-P "Concerning your servant about whom you wrote me, this servantACC, they will seek him" (~ They will search for the servant about whom you wrote to me.)

We can therefore see that a combination of pragmatics and syntactic devices make the ambiguity between subject and object more an academic problem than a practical one. It is therefore misleading to isolate word order and case systems from other properties of the language. When we examine the whole range of relevant factors, it emerges that the need to fix the pre-verbal word order was not necessarily very pressing, even after the case system disintegrated. 3. Change because of inconsistency and parsing problems ? Perhaps a more serious reason why we could expect changes in Babylonian is its highly 'inconsistent' word order patterns. To understand why the situation again was less problematic in practice than in theory, it may be useful to refresh our memory on why we might expect inconsistent word order to be unstable. The explanations for why languages tend towards consistent word order patterns have been based on three main lines of argument, briefly summarized below: (I) Languages tend towards an abstract structural notion of consistency (or harmony) between the order of 'heads' and 'dependants' (also known as 'modifiers', 'operators' etc.). (II) Consistent word order patterns emerge as a result of repeated 'historical accidents'. For example, adpositions often emerge from serial verb constructions, and these would result in postpositions in OV languages, and prepositions in VO languages. Similarly, genitives often give rise to relatives so NG would give rise to NRel, etc. (III) Languages tend toward consistent word order because of processing considerations. Inconsistent patterns are more difficult to parse, and because they strain processing resources, languages tend to avoid them.

THE SURVIVAL OF VERB FINAL ORDER IN AKKADIAN

63

Explanations based on (I) have largely been discredited, and need not detain us here. The diachronic explanations in (II) (as for example suggested in Aristar 1991) may be responsible for many cross-linguistic correlations. But in the case of Akkadian, these explanations are overruled by borrowing. The original Semitic 'consistent' VO word order may well reflect the cumulative result of grammaticalization and reanalysis. But once Akkadian borrowed verbfinal word order, and thereby introduced 'inconsistency', arguments based on (II) do not imply that changes in the position of existing structures should occur in order to regain consistency. In the case of Akkadian, therefore, the only reason why one might expect a development towards greater consistency would be because of arguments based on (III). Inconsistent word order patterns should pose more problems for parsing and processing, and therefore are likely to be unstable. According to recent literature on processing explanations (e.g. Hawkins 1994), the most severe processing problems are caused by heavier sentential phrases in 'inconsistent' positions. To explain why the Babylonian situation was less problematic in practice than in theory, I shall therefore concentrate on such heavier sentential phrases, namely relative clauses, complements and adverbial clauses. In § 3.1,I first look at relative clauses, and show how Babylonian copes with the apparent difficulties in processing which arise from their 'inconsistent' position. In § 3.2, I then discuss the position of adverbials and complements, which is the only area where Babylonian seems to 'behave'. 3.1 OV and postposed relative clauses Postposed relative clauses in a verb final language result in structures of the type: VP{NP{N Rel} V}, and in particular, in the word order: N [... V]REL V. This structure is generally claimed to be problematic from the point of view of processing. Nevertheless, this is the standard structure in Babylonian. Hawkins (1994:266) acknowledges that such structures do exist (he gives the German example in (20)), but he asserts that they clearly strain processing resources, and are therefore extremely rare in natural discourse (1994:267). (20) Wie soll ich also deinem Vater [der wochenlang im Ausland war] erklären, daß... "How should I then to your father [who weeks abroad was] explain, that... "

64

GUY DEUTSCHER

Hawkins claims that even in languages whose word order patterns would naturally produce such structures, these are avoided in practice, by moving the verb forward. In German, for example, the head noun is often separated from the relative clause, and the relative moves to post-verbal position, giving structures such as (21): (21) Wie soll ich also deinem Vater erklären [der wochenlang im Ausland war] daß... "How should I then to your father explain [who weeks abroad was] that... "

But Babylonian does not seem to display any unease with the unmodified structures of the type N [...V]Rel V. They are very common and productive in older and younger periods: (22) OB

tuppa-ka ša tušābil-am ešme letter-your [REL you.sent-to me] I.heard "I heard (= was read aloud) the letter that you sent me."

(23) NB

uttatu  suluppū ša ina makkūru šaknū līmur barley and dates [REL in storeroom placed.stative 3pl] let.him.see "he should see the barley and the dates that there are in the storeroom"

(24) NB

pāni sipirtu sa PN u PN2 ana muhhi eleppăti anakâšunu face.of message [REL PN and PN2 concerning boats to you(m.pl) is-pu-┌ra┐ nidaggal they .sent] we.see "we are waiting for (=seeing the face of) the message that PN and PN2 sent to you concerning the boats"

Even multiple recursion of relative clauses in this way is found: (25) OB

assum eql-im ša PN ša šatamm-um ša aţrudaš-šu about field-gen of PNj [REL clerki-nom [REL I.sent-himj] ipluk-u-šum eqel-šu kinnā-ma hei.divided-sub-to himj] field-hisj you(pl) .assign .imperative-P "concerning the field of PNj which the clerkj whom I sent to you has divided (=marked off) for hiirij, assign hisj field."

If these structures are supposed to strain processing resources, why do they not create difficulties in Babylonian? Again, I think that isolating word order from other structural properties of the language can create a misleadingly simplified picture. In Babylonian, the apparent difficulties of processing are offset by a few syntactic features of relative clause formation. First, a special form of the verb is used in subordinate clauses, called the 'subordinative' (glossed here as "sub")- This forms helps prevent 'garden path' interpretations based on the confusion between the verb in the main clause and the subordinate clause. For example, in (25) above, the verb "divided" is in the

THE SURVIVAL OF VERB FINAL ORDER IN AKKADIAN

65

subordinative form, marked by the -u- in ipluk-u-sum. Moreover, because the verb is always in final position, the verb in the relative clause (especially when it is morphologically marked as subordinative)4 marks a very salient final boundary for the relative clause. This means that there is in fact a clear marking both for the beginning of the relative clause (the relativizer sa) and the end of the relative clause (verb in subordinative). The second factor which helps the processing of relative clauses is the use of resumptive pronouns. In Babylonian, relative clause formation is based on resumption rather than on deletion, and this strategy is known to increase the ease of processing (Keenan 1985). In (25), we can see this strategy in the clause 'the clerk whom I sent", which in Babylonian has the structure "the clerk that I sent-him'. The third factor, which I already mentioned in the previous section, is the frequent use of topicalization. The topicalized NP, with its heavy relative clause, is taken out of the main clause. It is then referred back to either with a pronoun, or by repetition of the head noun, as in (25) above ("concerning the field that..., assign his field!"). Finally, it is precisely the flexibility in pre-verbal word order which can help to prevent processing difficulties. The relativized NP, whether Subject, Object, or PP, can often appear earlier in the clause. Thus, the immediate preverbal position of the relative clause, which seems to be the most difficult factor in processing, in fact does not occur so frequently. It is especially common for the object to be in initial position when it is relativized, as we can see in (26) and (27) below: (26) OB

O+REL-S-V eql-am ša Kubatim sa ana bilt-im esh-u [field-c of Kubatum REL to rent-gen assigned ,stative-sub]o PN Tris PNs he.cultivated "PN cultivated a field in Kubatum which is assigned for field rent"

(27) OB

O + REL-S-V ŠE eql-im ša ištu MU.3.KAM şabtāku [barley .of field-gen REL for 3 years I.own.stative]o PN ul iddin-am PNS not he.gave-to me "PN did not give me the barley of the field which I have owned for three years."

The subordinative is not always morphologically distinct from the independent form of the verb. Thus, in the same example (24), the verb "sent", which is syntactically in the subordina­ tive, does not actually show the subordinative morphologically.

GUY DEUTSCHER

66

In (28) and (29), we can see variation in the positions of  and PP, with the relativized phrase standing in initial position: (28)

PP + REL-O-V

NB

ina nāri ša bīt ili ša niherra' --6 ittaši [from canal of temple REL we.dig] PP [water-our] o he.drew off (perfect) "he drew off our water from the temple canal which we are digging"

(29) NB

O + REL-PP-V

şāhē ša ana muhhi-šu illikūni ana kutallu [people REL against-him they.went.sub] o [to back] PP "he makes people who went against him return (to the back)"

usahhas he.makes return

In conclusion, we see again that what may appear to pose great difficulties when word order is considered in isolation in fact seems much less problematic when other structural factors are examined. 3.2 Complements and adverbial clauses The only area in which word order patterns undergo changes in Babylonian is the position of finite sentential complements and some adverbial clauses. In the Old Babylonian period, complements are overwhelmingly preverbal, as in (30). Post-verbal position is also found, as in (31), but much more rarely. (30) OB

kīma napišti māt-i eql-um-ma ul tīde COMP soul.of land-gen field-nom-P not you.know "Don't you know that the soul of the land is the field?"

(31) OB

ul tīde kTma kasap ekall-im  şibt-am not you.know COMP silver.of palace-gen and interest-acc ikaşşarū they .(will) .collect "don't you know that they will collect the silver of the palace and the interest?"

But in later periods, finite complements migrate to post-verbal position. From the Middle Babylonian period (ca. 1500 ), finite complements always have to come after the verb: (32) MB

bēl-ī īde kī ultu ēlâ diliptu mahratan-ni lord-my he.knows COMP since I.arrived trouble confronts.stative.fs-me "my lord knows that since I arrived, trouble has befallen me."

A similar development occurs with some finite adverbial phrases, mainly with the conjunction adi "until". In the older periods, adverbial phrases are almost always pre-verbal, as in (33). Only very rarely can adverbial clauses

THE SURVIVAL OF VERB FINAL ORDER IN AKKADIAN

67

(only with adi) come after the verb, as in (34). But in the Neo-Babylonian period, adverbial phrases with adi are most often post-verbal, as in (35). (33) OB

adi şuhār-um ša še'-am ubl-am iqbi-am until servant-nom REL barley-acc he.brought-to me he.said-to me masiāku I.forget.stative "until the servant who brought me the barley told me, I had forgotten (these things)"

(34) OB

alik šurib-šu adi kunukka-ka ubbalak-kum go.imp.sg bring.imp.sg-him until sealed document-your he.brings-to you "go, bring him in (to prison), until he brings you your sealed document."

(35) NB

sarru pursû-sunu la iparras adi PN1 ana pan king decision-their not he(will).decide until PN1 to face.of sarr-i bēli-ja illakam-ma king-gen lord.gen-my he.comes-P "the king will not decide their matter until PN1 comes before the king my lord... "

This seems to be the only area where Babylonian makes concessions to the predicted processing difficulties. Dryer (1980:130ff.) shows that many languages which are otherwise rigidly verb-final make an exception for sentential complements (e.g. Wappo, Hopi, Persian, Yaqui). Hawkins (1994:298ff.) claims that pre-verbal finite complements (with initial complementizer) cause serious processing difficulties even in consistent verbfinal languages. He argues that this is the reason why otherwise strict verb-final languages use various strategies to place finite complements after the verb. On this issue, Babylonian is no exception, and word order change does occur in the predicted direction. It is not possible to gauge precisely the level of difficulty that pre-verbal finite complements caused in the Old Babylonian period. It is probably suggestive, nevertheless, that finite complements are fairly rare in Old Babylonian, especially compared to the ubiquitous relative clauses. What we can see is that when they do appear, finite complements are almost always in sentence-initial position, as in (36), rather than embedded between the subject and the verb. Embedded position, like (37) is attested, but much more rarely. (36) kXma 30 puhādīnēmetta-kunu ana  ābili la tublānim PN...iqbi'-am OB [COMP 30 lambs payment-your.pl to Babylon not you(pl).brought] PN he.said-to me "PN told me that you did not bring the 30 lambs, your payment, to Babylon." (37) PN kīma puhād nëmetta-ka ana ekall-im la tublam iqbi'-am OB PN [COMP lambs payment-your.sg to palace-gen not you.brought] he.said-to me "PN told me that you did not bring the lambs, your payment, to the palace."

68

GUY DEUTSCHER

It was probably because of processing difficulties that finite complements migrated to post-verbal position in later periods. The change must have been relatively easy, because even in the older periods, the pre-verbal position of complements was not as rigid as that of other constituents. The synchronic variation in the older periods undoubtedly facilitated the diachronic shift in later periods. 4.

Conclusion Babylonian Akkadian demonstrates that languages have more to offer than word order and case systems. When word order is viewed in isolation (or at most in relation to the case system), the situation in Akkadian may seem problematic. But I suggested that the apparent difficulties are offset by various factors: pragmatics, verbal agreement, morphological distinction between verbs in main and subordinate clause, resumptive pronouns, and frequent use of topicalization. Thus, on closer inspection, the stability in Akkadian word order does not seem so unlikely after all. The Akkadian situation can also suggest why so many of the cross-linguistic word order universals are observed more in the breaking than in the keeping. At least in some cases, inconsistent word order patterns may only appear to cause difficulties because other structural factors of the language are ignored.

REFERENCES PRIMARY SOURCES Glossing Abbreviations: PN = Personal name. Sub = The subordinative mode of the verb. P = The Akkadian particle -ma, used for a variety of purposes, including emphasis and coordination. The examples quoted are from the following texts: (1) AbB 13:60 (2) AbB 1:18 (3) CH § 59 (4) AbB 1:7 (5) AbB 11:40 (6) AbB 7:50 (7) AbB 12:52 (8) Dietrich: 64 (9) CT 22: 105 (10) ABL: 202 (11) AbB 13:72 (12) AbB 13:72 (13) Streck, p.113 (YOS 3:137) (14) ABL: 521 (15) Streck, .133 (BIN 1: 93) (16) CH § 117 (17) Dietrich: 9 (18) Cole: 24 (19) AbB 9:18 (22) AbB 9:11 (23) Streck, p.91 (YOS 3:126) (24) Streck p. 89 (YOS 3,173:9), (25) AbB 11:174 (26) AbB 4:35 (27) AbB 4:93 (28) Streck, p.91 (BIN1: 44) (29) Dietrich: 48 (ABL 1255) (30) AbB 9:48 (31) AbB 13:65 (32) Aro, p.154 (33) AbB 2:93 (34) AbB2:106 (35) ABL 517 (36) AbB 2:75 (37) AbB 2:64.

THE SURVIVAL OF VERB FINAL ORDER IN AKKADIAN

69

AbB = Old Babylonian letters. Fritz Rudolf Kraus, ed. (1964-1994) Altbabylonische Briefe in Umschrift und Übersetzung. Volumes 1-13. Leiden: E.J. Brill. ABL = Neo-Babylonian letters. Cuneiform edition: R. Harper (1892-1914) Assyrian and Babylonian letters. University of Chicago Press. Aro = Middle Babylonian letter (BE 7(1):43), quoted in Jussi Aro (1955) Studien zur Mittelbabylonischen Grammatik. Helsinki: Studia Orientalia 20. CH = Old Babylonian Law Code of Hammurapi. Quoted from: Martha T. Roth (1997) Law Collections from Mesopotamia and Asia Minor. Atlanta: Scholars Press. Cole = Neo-Babylonian letters: Steven W. Cole (1996) Nippur TV, The Early NeoBabylonian Governor's Archive from Nippur. Oriental Institute of Chicago Publications. CT 22 = Neo-Babylonian letters. Transliteration and translation in Erich Ebeling (1949) Neubabylonische Briefe. München: Verlag der Bayerischen Akademie der Wissenschaften. Dietrich = Neo-Babylonian letters quoted from Manfried Dietrich (1970) Die Aramäer Südbaby loniens in der Sargonidenzeit. Neukirchen: Butzon & Bercker Kevelaer. Streck = Examples from Neo-Babylonian letters (BIN 1 and YOS 3), quoted from Michael P. Streck (1995) Zahl und Zeit, Grammatik der Numeralia und des Verbalsystems im Spätbabylonischen. Groningen: Styx.

SECONDARY SOURCES Aristar, Anthony R. 1991. "On Diachronic Sources and Synchronic Pattern: An Investigation into the Origin of Linguistic Universals".Language 67:1.1-33. Dryer, Matthew S. 1980. "The Positional Tendencies of Sentential Noun Phrases in Universal Grammar". Canadian Journal of Linguistics 25.123-195. Dryer, Matthew S. 1992. "The Greenbergian Word-Order Correlations". Language 68.81-138. Harris, Alice C. and Lyle Campbell 1995. Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Hawkins, John A. 1994. A Performance Theory of Order and Constituency. Cambridge: Cambridge University Press. Keenan, Edward L. 1985. "Relative Clauses". Language Typology and Syntactic Description ed. by Timothy Shopen, 141-170. Cambridge: Cambridge University Press. Soden, Wolfram von. 1995. Grundriss der Akkadischen Grammatik. Roma: Pontificio Istituto Biblico.

STABILITY IN CLAUSAL/PHRASAL PATTERN CONSTITUENT SEQUENCING: 4000 YEARS OF EGYPTIAN (WITH SOME THEORETICAL REFLECTIONS, ALSO ON CELTIC)

ARIEL SHISHA-HALEVY The Hebrew University, Jerusalem 1. Egyptian The linguistic study of Egyptian, fully deciphered only about 150 years ago, is a young discipline: modern Egyptian linguistics, dating more or less from the work of Hans-Jakob Polotsky, is much younger still: no more than about half a century old. Coptic, the final stage of Egyptian, dead as a spoken language at some point after the XIIIth century AD, had been scientifically known in the West from around the XVIIth century. It is a curious and somehow sobering thought that Champollion le Jeune probably got the brainwave and forward push to the final decipherment of the hieroglyphic script by a wholly and deeply erroneous idea about diachronic word order correspondence. He believed (or took for granted) that Coptic f-sôtm "he is hearing" (roughly, "he + hear") was the inversion of a 'pan-Egyptian' sdm.f ("hear-he"),1 which, he thought, had the same tense form, but which - we now know - is in fact a cluster of homographs, drastically differing, formally and functionally, in tense form and syntactical status from one phase of Egyptian to another and within one and the same phase.2 The idea was wholly misguided,3 By Egyptological convention an  vowel is inserted between consonants in Egyptian words, to make pronunciation possible: [sedjemef] (in the present text, I transliterate as a simple voiced dental, d, the transliteration of a hieroglyph probably representing a voiced and emphatic dental-alveolar affricate). For a recent overview of the grammatical systems of Egyptian as a whole, bridging between different theoretical approaches, see Loprieno 1995. 2 In Middle Egyptian, for instance, we have the so-called prospective sdm.f ("that he [should] hear", "may he hear"), the substantival misleadingly named 'emphatic' sdm.f ("that he hears", as in: I know that he hears and especially in the Cleft Sentence type It's because of me that he hears), and the syntactically adverbial converbal so-called circumstantial sdm.f "...he hearing". All are distinct, and distinguished, formally, by syntactic-slot occupancy privileges resolved in close textual analysis, correlated, conjoint and coupled with such consonantal morphological features as can be observed. In Demotic we find only the fossilized remnants of the prospective sdm.f and a single case of the 'perfectai' sdm.f. In Coptic, only a lexically closed-list class of

72

ARIEL SHISHA-HALEVY

yet the confidence it gave him, and his conviction that Coptic and Egyptian were two phases of the same language were not unjustified, and led him to eventual success. Today we have a reasonably good synchronic resolution and, paradoxically, a sometimes seemingly sharper diachronic resolution - of nearly four millennia of uninterrupted evolution of a language (or rather an ensemble of dialects and language varieties), made visible to us in the written documentation of five or six distinct broad linguistic systems (in the sense of la langue as well as norme and usage). Roughly, with some arbitrariness and considerable overlapping, Old Egyptian ("OE", 2800-2200 ), Middle Egyptian ("ME", 2200-1500 ), Late Egyptian (or Neo-Egyptian) ("LE", 1500-700 ); Demotic, from the VIIth-VIIIth century  to the Vth century AD, and finally Coptic, 'Christian Egyptian', written in customized graphemic systems based on the Greek graphemes and several Egyptian ones, from the IVth century AD on, until its death as a spoken language: Arabic entered Egypt in the VIIth century AD, but Coptic probably lingered on until the XVIIth century. (Incidentally, Coptic is formally differentiated as 'Egypto-Coptic' in the current International Linguistic Bibliography. Roughly since the Fifties, Coptic Studies have moved away from Egyptology, a separation unfortunate for both Egyptology and Coptic studies, which has all but wiped out Coptic linguistics as a discipline). Most phases, as we conveniently and simplistically delimit them (ignoring here the relationships, complicated in Egyptian, between language phase and script phase, as well as the religious-political implications of traditional archaizing use of earlier phases) have considerable overlapping or 'mutual leaking' with preceding ones, as well as transitory stages, and of course numerous diasystems of registers and other linguistic varieties which become clearer as detailed description progresses. Some phases extend up to a thousand years, which makes the need for a finer subperiodization obvious (Junge 1985). Generally speaking, we witness the uninterrupted evolution of a language on one and the same terrain, in its first attestation cradled in a Neolithic culture, before the end of its life-span a paraclassical language, part of a pious and totally Christian civilization: very little secular literature is attested in Coptic. quality-expressing lexemes (e.g. nanou-f, nane-prôme "good [nexus] he", "good [nexus] theman" = "he is good", "the man is good") maintains what is comparable to the old sdm.f pattern. In earlier Egyptian, there were other #VERB STEM + THEME (PRO)NOUN# cases, with the stem marked by distinctive affixes: -n-, -in-, -k3- etc. - these disappeared earlier. 3 We find it expressed as late as Steinthal and Misteli (1893:283-284) in a still valuable typological sketch of Egyptian: stm paï-k sn "(es) hört dein Bruder" with paï-k sn stm "dein Bruder hört".

STABILITY IN CLAUSAL/PHRASAL PATTERN

73

The full informed story of first the decipherment of the script, then the century-long, slowly improving resolution of the individual grammatical systems and subsystems of the respective phases of Egyptian, is truly breath­ taking and still remains to be told in full. Beside clashes, often vigorous, between schools of thought and grammatical method (most notably, at the beginning of the century, the Paris and the Berlin ones, of which the latter finally prevailed, leading to the Polotskyan Jerusalem School and currently to descriptive, 'functionalist' and evolved generative approaches), it is also the story of a Hugo-Schuchardt-style scholarly dissent and creative individualism, as well of a progressive detachment from the Eurocentric-Semitic methodo­ logical and theoretical bias and an unfolding realization of the special intricacy of Egyptian as formulated in its own internal, not genealogically biased terms. As some of its historical landmarks may be considered Adolf Erman's Ägyptische Grammatik of 1901, Kurt Sethe's Der Nominalsatz im Ägyptischen und Koptischen (1916); Battiscombe Gunn's Studies in Egyptian Syntax (1924), Alan Henderson Gardiner's Egyptian Grammar (1957 3); A. Erman's Neuägyptische Grammatik (1933); Frank Lloyd Griffith's Stories of the High Priests of Memphis (1900) and Catalogue of the Demotic Papyri in the John Rylands Library, Manchester (1909), and Wilhelm Spiegelberg's Demotische Grammatik (1925); for Coptic, Ludwig Stern's Koptische Grammatik (1880), Hans-Jakob Polotsky's Études de syntaxe copte (1944), his "Coptic Conjugation System" (1960) and his Grundlagen des koptischen Satzbaus (1987-1990) and most recently Layton (2000) (see Loprieno 1995 for a detailed bibliographical inventory). 2.

Some idiosyncrasies Several properties and aspects of Egyptian and Egyptian linguistics must be pointed out here, as being of the highest theoretical relevance to the issue in point. I am hardly going to mention the genealogical assignment of Egyptian, still controversial and mysterious: is Egyptian part - a crucial, if not defining part - of a hypothetical Hamito-Semitic language family (the implicitly or explicitly consensual view at present), or is it a substrate - (or otherwise) formed Mischsprache, almost an analogue (mutatis time-depth and other mutandis) of Amharic as Semitic Stoff and African Form?4 This has an immediate bearing on such a far-removed issue as the structural typological nonIndo-European affinities of Insular Celtic, a question recently redeemed after a long period of near-disrepute. Formulating the query in terms of a non-geographical (or hyper-geographical) Sprachbund (cf. Shisha-Halevy 1995:§ 7), I find Egyptian the most striking term of compari­ son, far more significant than the currently used Arabic (cf. Vennemann 1995, 1997; Borsley

74

ARIEL SHISHA-HALEVY

• First, no vowels are allowed for in any of the graphemic systems up to (and not including) Coptic.5 Since consonantal formal (morphology signifiant) distinctions are rare, we have little inflectional morphology in the IndoEuropean sense (consonantal affixal formation should really be considered syntagmatics or syntax). Analytic Grammar in Egyptian thus means 'syntax netto'. We are, in the study of Egyptian, largely spared what De Boer felici­ tously called "la tyrannie de la morphologie" and "la superstition de la forme"', the theoretical higher rank of syntax (a realization agreed in structuralist general linguistic discussion since the first decades of this century, but never really implemented in actual descriptive work) is in Egyptian linguistics de rigeur. What we have to deal with are subtextual slots only, and applying 'structural ID' procedure is inevitable and crucial: the decoded-analytic identity-definition of an element and entity of language is effected only by the conjoint coordinates of positional (syntagmatic) and commutational localiza­ tion; its 'role-name', its analytic identity or individual essential profile consists of its simultaneous syntagmatic and paradigmatic coordinates. This is, I believe, of the highest relevance to the issue in point here, namely "word order through time": for it indicates that we must resort to patterning, with sequencing but one of its constituent distinctive features (and probably hyperpattern, placement of certain discourse-referred elements, such as discourse signals). The specific individual grammatical category is the signifié of a substitution class in a given syntactical slot, without any 'morphological back­ ground noise', and ideally without the biasing by a pre-eminent conceptualterminological tradition based on the morphological form-entity (consider, for a striking instance, the category of tense). The grammatical form is consequently a precise formal-relational feature or signal, in a sense the ideal grammeme, and the subtextual unit the 'pattern', defined as "a bounded (i.e. 1995). But the ambition to rephrase the 'contact' in an historical scenario (e.g. as a substrate phenomenon) must come to terms with the fact that Egyptian, while affording a very considerable time depth - greater than that of 'Semitic' alone - is hardly a straightforward 'language on a family tree'. I personally, following thirty years' descriptive work on Egyptian grammar, favour the Mischsprache hypothesis, for numerous reasons outside the scope of the present article. Incidentally, and in the context of the present issue, the alleged 'VSO' property of 'Semitic' (as exemplified by Arabic) or of Egyptian (so Hawkins 1983:320) is ohne weiteres conventionally associated with the Celtic (e.g. Brythonic) VSO 'basic word order'. But all these components of a schematic typological comparison are either simplistic, or inexact, or even a petitio principii, and there are in fact affinities more unambiguous, sharp and valid. Coptic does have seven graphemes for approximately represented vowel phonemes, yet it is significant that efforts to reconstruct the pre-Coptic Egyptian vowel systems have so far shed little additional light on the respective grammatical systems (cf. Polotsky 1964).

STABILITY IN CLAUSAL/PHRASAL PATTERN

75

delimited) sequence of categories", which in their turn may themselves be pat­ terns, reducible to constituent categories, and so on and so forth. Heuristically and technically speaking, the first and most important single analytic step is the isolation of delimitations (boundary signals); the analytical procedure is necessarily descending, from text-scope downwards. • The remarkable time depth, which affords a unique opportunity and special laboratory conditions for observing features and factors of diachrony at work; so, for instance, the periodic-cyclical nature of many changes in features and subsystems. However, the danger of the 'optical illusion of historical continuity' is in the case of Egyptian very strong and ever present: on this further below. We thus witness for instance the familiar synthesis - analysis (e.g. periphrasis) - agglutination-resynthesis evolution of junctural structure typol­ ogy. From this point of view, Coptic resembles in many respects the oldest Egyptian and completes a millennia-long cycle, with LE an analytical peak (the phase prior to OE not documented, that posterior to Coptic sadly non-existent). Another specific and, for me, especially fascinating case is the history, familiar from a European isogloss in a Romance-Germanic West European Sprachbund or at least isogloss, of the recurring neutralization of the opposition perfect (the nunegocentric, i.e. speaker's hic-et-nun, past/present statal complex tense) vs. preterite, neutralizing always in favour of the marked term, the perfect, which is de-pertinented, representing the category as a whole while the preterite disappears and a new periphrastic perfect enters the picture. The full cycle is observable not less than three times in the course of the history of Egyptian.6 • For Egyptian, text-linguistics is the only feasible linguistics, and textscope grammar is perforce the only grammar. The descending analysis from full text-level down is unavoidable; no textual subdivision is usually given (the a priori or intuitive 'sentence' and 'word' units are far more dubious and

The evolution of future tenses in Egyptian is another instance of analysis-to-synthesis cyclicity. It is instructive to compare its stages to the familiar Romance ones (Fleischman 1982); in Egyptian, so far as we can see, the 'spatial' future ("I am to go") is earliest, whereas the modality factor starts to play a role in the same formation only very late, from Coptic on, and then only in certain dialects (temporal-to-modal, never the other way round). The analytic "go" futures appear around LE, and end up by being the only tense-form in Sahidic Coptic, one of two (a tempus instans, intentional, typically interlocutive-sphere, opposed to an 'absolute', uncircumstantial, unconditional, non-subjective one) in the Bohairic dialect. In any case, in Egyptian there is no direct association between the evolving forms and change in word order (cf. Fleischman 1982:50).

76

ARIEL SHISHA-HALEVY

obfuscating than in the European linguistic systems).7 'Narrative tenses' (to take one instance) and narrative building-blocks and texture units in general must be carefully isolated, as the only Tempuslehre possible; needless to say, this is immediately instructive and eye-opening for the well-trodden 'nostratic' systems. Another example, perhaps less banal and directly relevant to the word order issue, is the concept of syntactic (or rather syntaxic) autonomy: the notion and feature-complex of 'ordination' (as so strikingly epitomized in the Eurocentric logic/rhetoric-based dichotomy, still going strong, of 'main' vs. 'subordinate', with 'superordinate' a real improvement, albeit still rarely ap­ plied)8 cannot be simply applied to Egyptian, where, for example, 'non-initial' may be the distinctive feature of a clause (as for example in a special apodotic clause-form, attested in some form throughout the history of the language). • An invaluable metalinguistic or 'technical' feature of Egyptian linguis­ tics, due to the special circumstances, constraints and slowness of its evolution, is its relative resistance to Eurocentric terminology and careless taxonomic compartmentalization: there has been so far little uncritical en masse and en bloc application of Western terminology (probably with the exception of Greco-Latin terminology applied to Coptic in the XIXth century). This happy state is rapidly, and alarmingly, changing at present, in the pretext of 'modernizing' Egyptian linguistics, as if forcing the models, buzz-words and general terminology that are en vogue onto a non-European system still only partly understood can replace patient text-based categorial analysis. Conse­ quently, there still exists the opportunity for a leisurely and careful considera­ tion of applicability and suitability of concepts and names for concepts, and for evolving internally rational, not imported conceptual and terminological systems. Clearly, this too is potentially beneficial for a re-examination of wellstudied languages and the general-linguistic models based on them.

7

Note a recent general discussion of 'sentence' in Kindt 1994. Somehow, Satzbegriff sounds better (since probably less committed) than 'notion of sentence' and 'clause' in the English abstract (for one thing, there's no need to distinguish the grammatically and logically based terminology). Then, grammatikunabhängig is not at all the same as 'syntax-independent'. Occasionally, moreover, Kindt does not seem to distinguish Satzdefinition from Satzsegmentierung. Unfortunately, the discussion, very interesting and lucid per se, of Satz in spoken German, cannot be immediately applied in the case of a language such as Egyptian, written, dead, so very different culturally; and here, once again, Kindt's approach is incorrigi­ bly Eurocentric and verb-clause-oriented: consider for instance the Vollständigkeit test by the presence or absence of the verb "be" (cf. Kindt 1994:37ff.). Cf.Shisha-Halevy 1995.

STABILITY IN CLAUSAL/PHRASAL PATTERN

3.

77

Word order: stability and variation over time At this point, I must confess to several doubts and even misgivings about the terms in which our workshop theme has been phrased. "Variation". In the structuralist semiotic methodological framework within which I operate, every formal feature - every signifiant (and sequencing, when pertinent and not conditioned, is certainly a constituent or constituent-part tagmemic signifiant) has its signifié, until the relatively rare case of free variation has been proven. We have thus only three conceivable synchronic situations obtaining between two or more elements: (a) opposition and full or partial pertinence; (b) environmental neutralization alternation (as typically encountered in a complementary distribution situation); or (c) the always provisionally (and, for a structuralist, reluctantly) stated case of 'free variation', that is, formal differentiation not correlatable in some way with a functional one. Still, this last state of things is, more often than not, a preanalytical petitio principii. "Over time"; diachrony. It is crucial to bear in mind what is so obvious, has repeatedly been pointed out, yet is as a rule ignored in the actual practice of our thriving Historical Linguistics, that one written phase of the language does not necessarily or usually evolve out of a chronologically preceding one: Literary Modern Welsh (in itself consisting of several sub-phases) did certainly not evolve out of Mabinogion Middle Welsh, any more than Vulgar Latin or Romance texts evolved (as representative of états de langue with distinct grammatical and lexical systems) out of the Classical Latin literary corpus. The realization that the 'single line of development' is a convenient fiction - this self-evident, seemingly trivial truth, but too often unheeded - is no less than crucial when we study diachronic syntax, for evolutively juxtaposing and comparing grammatical patterning in an historical sequence of attestation is an easy trap (as is, to take a rather worn-out instance, the case of quod and the accusativus cum infinitivo as exponents of clause substantivation in Latin and Romance). In Egyptian, major successive phases match more or less known periods of extralinguistic (social and political) turbulence, and did not evolve successively, certainly not in a straight route and at a constant rate. Extensive systemic areas and many features in all phases are not documented in preceding ones; we know little of the systemic context and pragmatic or textual context of evolution. Of course, this 'optical illusion' of evolutive continuity may ideally, almost metaphorically, be taken as valid, yet only panoramically, for blueprint resolution and therefore, in my opinion, to little use prior to synchronic system delineation: so many details are lost, systemic factors are inevitably taken as negligible, underplayed or downright ignored that one finds

78

ARIEL SHISHA-HALEVY

oneself in a practically model-reshaped or meta-linguistic world. (In parenthe­ ses, let me point out that the historical perspective, as truly envisaged by Saussure - this is now being made more and more manifest by the new Saussurean philology - focusses on viewing a synchronic systemic fact, a feature in a système des valeurs, in diachronic light, with a holographic effect, rather than a cinematographic 'animation' by the rapid shifting of juxtaposed 'stills'. And, of course, the full systemic feature must be fully formulated first. Moreover, linguistic change concerns most intimately, stems from and is effected in microstructures and subsystems). Lastly, and most importantly, 'Word Order". Linguistic sequencing is a complex issue, not exhaustible in schematic generalizing, often sweeping statements of 'word order' as used in current linguistic inquiry, let alone in terms of the arguments of an Eurocentric (deductive, logic-based, not empiri­ cally founded) verbal-predication 'sentence', viz. S[ubject], V[erb], O[bject]. Indeed, I see this as a deplorable obfuscation, almost a vulgarization of an intricate vital issue. First of all, this assumes - and unqualifiedly builds on the universality of a prime, hierarchically dominant verbal predication : this is a deeply rooted Eurocentric verb-centered grammatical thought, almost equating 'clause' with 'verb clause', and a 'no-verb' construction with 'verbabsence' or worse.9 Moreover, a basic distinction of actual syntagmatic sequencing and structural (not necessarily of adjoining elements: 'discontinu­ ous') sequencing is essential;10 elements representative of categories often 'reside' in actual locations that are structurally, albeit not typologically, a matter of coincidence.11 And then, sequencing or placement may be pertinent (pattern-distinctive) or, like almost any linguistic signifiant element, condi­ tioned: such distributional statement can only be formulated on the basis of precise oppositional and environmental study.12 The dominance of the verb and the 'verb clause' is nowadays more a presupposition of current General (often in the sense of universalist-typological) Linguistics, stemming from the blend of Neo-Grammarian Indo-European-oriented Sprachwissenschaft with the logic-based normative tradition, still associated with various current schools transformed from the original generative models of the Sixties and Seventies, than of the individual branches of modern descriptive Indo-European linguistics, especially of Greek, Latin and Indo-Aryan. For instance, the valential structure of verb lexemes; or many cases of mutation in Celtic. So for instance the location of some Indo-European case-endings, structurally a discontinu­ ous constituent of the verb, in the noun phrase. 12

It is frankly difficult for me to understand word order ('SV/VS') functional opposition outside the synchronic 'transsection' or 'still shot' of an individual linguistic system, as a grammemic linguistic sign which may subsequently be compared or contrasted (typologically) with correspondent signs in other languages or (diachronically) with other phases of the same

STABILITY IN CLAUSAL/PHRASAL PATTERN

79

My contention is, in brief, that (a) constituent ordering is not a prime, independent, overruling, governing or motivating feature of syntax. It is one of several cumulatively and conjointly definitive features of any pattern, already defined here as a delimited sequence of categories. Constituent ordering does not 'exist' absolutely, on its own, in a given language, as one of its typological traits. Sequencing is a grammeme (a tagmeme), a signifiant which, as I see it, rules out the validity or feasibility of isolating SVO/VSO word order schemes independently of precise patterning; (b) the universal applicability of the 'SV' model is unwarranted or else trivial;13 (c) the universal existence and then typological predominance in a given language of a verbal-predication 'sentence' which can be stated in the same terms as traditionally used for IndoEuropean is questionable and cannot be taken for granted. Frankly, I find it difficult to see how current word order typologies, synchronic or diachronic, can be maintained in the face of these objections. (Needless to say, a specific world-order or sequencing may constitute part or the whole of a descriptive statement, always resulting from a process of analysis). A final word on 'subject'. It will be noted that I refrain from using this term, even for verb clauses, just as I reject the a priori distinction of 'syntactic' and 'pragmatic' word order, as if the syntactic features of la parole, of a text and of its texture were in some way transcendental, abstracted or independent of its contextual or situational functions. 'Subject' (like 'Predicate'), originally a functionally conceived term, has been irretrievably impaired by its rigid association with the European-brand verbal nexus, in what is probably the most striking case of terminological distortion in syntax. ('Theme' is the nonrhematic constituent in any nexal interdependency, including verbal nexus, and 'agent' the 'first actant' in the valency matrix of the verbal and associated lexe­ mes). As usually applied, 'subject' - or the flaringly ethnocentric and internally language. In the synchronic system, sequence would thus be pertinent, not conditioned. Consequently, I cannot accept the thesis in Hopper 1986, which, as it were, puts the typological and diachronic carts, in tandem, before the synchronic-internal horses. Even supposing that 'universal basic word order' is at all a scientifically viable concept, a presupposition cogently questioned in Mithun 1987. The inbuilt (Indo-) European ethnocentric squint in Western linguistics and ever present in Western linguistic discussion, is now a very real danger of forcing an alien model on language in general. Consider the new term 'Euroversals' (see Kortmann 1997:33ff.), which might at first be conceived of as a welcome qualified restriction, yet still has potentially comprehensive universalistic associations and evocations in an 'imperialistic' epistemology. Observe that it is not systematization as such that I am advocating here against, its inevitable abstractive and idealizing components notwithstanding; what I would point out, following Hugo Schuchardt, is the absolutely crucial need to distinguish between internal and external systems, and let the former - especially the easily obliterated micro-systems - take priority, as data for the latter.

80

ARIEL SHISHA-HALEVY

contradictory 'grammatical subject' - is spurious, and constitutes a prime instance of advanced terminological conditioning. Incidentally, the exiling from syntax to pragmatics of the issues of information structure and the highlevel signification of macrosyntactic grammar such as narrative grounding and perspective is a lamentable sign of the times, associated with the per se welcome surge in sociolinguistic awareness, discourse analysis and the study of spoken language. But - if I may be excused some slight exasperated irony written language - and dead written language, too - has been known to present grammatical systems worthy of study, and by no means second-hand or reflected or deficient.14 4. Excurse: word order in Celtic At this point and by way of an excursion, I wish to make three brief observations on Brythonic Celtic, à propos of Graham R. Isaac's recent syntax of Old Welsh Poetry (Isaacl996); they concern slips of method which I believe are instructive for the issue at point here. First, the periphrastic analytic verbal constructions of Welsh (as important as and strikingly similar to their Egyptian correspondents from OE on), which turn out to be a trap for the word order typologist. For on p. 21 Isaac states, off-handedly, that Mae Siôn wedi darllen "John (has) read", (approx. "John-exists/is after-reading") is a case of SV sequencing, as if the existential/statal nexus exponent auxiliary mae did not exist. In fact, this is either a case of discontinuous 'mae..+ converb' rheme, or of a 'there exists' statement (mae) preceding its existant theme. Second, the Cleft Sentences Siôn sy wedi darllen, wedi darllen  mae Siôn focalizing respectively the theme and converb (approx. "[It is ] John who-is after-reading" and "[It is] after-reading that-is John"): Isaac sees these as straightforward cases of SV and VS respectively, again as if the relative and substantive conversion forms of "be" did not exist or structurally matter. Last, the Middle Welsh unmarked narrative verbal clause-form, the so-called 'Abnormal Sentence', '(pro)noun + α-verb', the vexata quaestio of MW syntax, which is generally presented as the main argument and show-piece for the SV order of MW in contrast to VS in the modern language. Yet again, this involves totally ignoring the relative-converter element a- (apparently just because it is not simply and transparently reflected in translation). See Jamieson 1993 for a welcome manifesto-like call on behalf of dead-language syntax, pre-eminently applicable to Egyptian linguistics. My own personal research experience fully corroborates Visser's caveat (apud Jamieson, 219 n 1), viz. that hastily pronounced 'errors' constitute the data for most insights into a dead language's syntax: in Coptic, it is a teaching and research commonplace that the apparatus of critical editions by insensitive editors is the place to look for the most significant forms and constructions. On non-attestation in a dead language, see also Shisha-Halevy 1986, Introduction.

STABILITY IN CLAUSAL/PHRASAL PATTERN

81

5. Egyptian patterning In the main part of my exposition, I will now present and illustrate several striking patterning features of Egyptian throughout its history,15 with a focus on the evolution of sequencing. I would here draw attention to the remarkable stability of sequencing within the enveloping diachronic systems, in the framework of which the pattern in point is 'held' as evolution takes place: this is the dynamic, pluridimensional Structural (or Systemic) Diachronic Frame ("SDF"), which (as it were) contains the static synchronic syslérnes des valeurs, layered in chronological sequence. What is being followed over time is always the individual pattern, the key analytic unit isolated in a descending analysis from text-level; 'Word Order' is (as has been stressed) in-pattern category sequencing, a distinctive constituent feature of the pattern, and not a 'floating' placement property of individual elements: it is no more, no less 'important' (in the sense of 'essential' or 'basic') than other, categorial distinctive constituents of the pattern. Needless to say, 'stability' in this context has a relative applicability, but is no less remarkable for that. I must stress once again that in the following presentation there is considerable schematization and thus simplification, this being the main shortcoming of the typologicistical format of grammatical presentation. 5.1 Nexus types: no specific SDF The trait that I would choose as typologically most basic throughout the history of Egyptian is the predicative theme+rheme (i.e. predicative nexus) patterning. Three main predicative-nexus patterns are always in evidence, remaining practically unchanged in categorial constituency and in sequencing, from Old Egyptian through Coptic. (A fourth pattern eventually merges with the nominal-rheme one: see § 5.6 below). Note that patterns (I) and (II) below are not 'verb-less', but rather devoid of any formal association with the verbal nexus. From the earliest Egyptian onwards, pattern I supplies considerable parts of the verbal system by predicating converbs - subnexal finite or nonfinite adverbial verb-forms.

For Coptic, the only phase of Egyptian for which dialects are fully resolvable, I give Sahidic examples unless otherwise specified. Of all Coptic dialects, Bohairic (a northern dialect originally, by the Xth century extended all over Egypt) has most 'diaglosses' with Late Egyptian: cf. Shisha-Halevy 1981.

82

ARIEL SHISHA-HALEVY

(I) adverb-rheme nexus, adverbial and converbal rhemes, supplying the Present and (before Coptic and in Bohairic Coptic) non-modal Future tenses, from Late Egyptian/Demotic on a tense strikingly similar to the GrecoRomance Imperfect.16 In ME, and especially for substantival lexical themes, a prefixed formal thematic element jw characterizes an hic-et-nunc ('presentsituation'), discourse-syntactically unmarked status of the nexus, with a zero + nominal signalling its inclusion or otherwise discourse-syntactically marking. (ME) jw mtr.j m-hnw PRESENT-SITUATION- DEICTIC + testimony-1st sgl. [poss.] pron.+ in-+ residence "My testimony is in the Residence". jw hnw m sgr PRESENT-SITUATION-DEICTIC- residence + converb-marking preposition - being silent "The Residence is being silent." jw m pn n nsw hr m33 PRESENT-SITUATION- DEICTIC - army + demonstrative adj. + of (sgl.m.) - king + converbmarking preposition - watching "This army of the king is looking on." (LE) twj m-dj.k 1st sgl. - with - 2nd sgl. m. [poss.] pron. "I am with you." n3y.s g3b m mfkt plur. +3rd sgl. f. poss. article - leaf + predicate-marking preposition - malachite "Her leaves are of malachite.": incidental (i.e. non-essential circumstance-referred) nominal predication. (Dem.)p-t3 jr-hr-j m-qd wc.t cftj.t def. art. sgl. f. - earth + before - 1st sgl. [poss.] pron. + in-manner - zero-article - chest "The earth is before me like a chest." n3-hrd.w n-p3-lh m n-p3-hyr def. art. pl. - child + 'a-walking' (CONVERB) in - def. art. sgl. m. - street "The fool's children walk the street." (Copt.) f-mmau 3 sgl. m. them, pron-there "he is there"

16

In the component notation of Egyptian examples on the following pages , "+" indicates combination in general, while "-" marks close-juncture combination. Verbs of movement/posture and other intransitives have the "in" (m) converbs in the present tense; cf. the Modern Irish tå sé in-a- + verbal noun statal present conjugation. This is a typological trait of a considerably higher order than the alleged 'subject/verb' sequencing.

STABILITY IN CLAUSAL/PHRASAL PATTERN

83

f-sôtm 3 sgl. m. them. pron. - 'a-hearing' (CONVERB) p-rôme sôtm 18 def. sgl .m. art.-man + 'a-hearing' (CONVERB) "he /the man hears/is hearing".

(II) noun/pronoun-rheme nexus (Nominalsatz) - see the special discussion below : (ME) jnk nb.tn 1st sgl. them. pron. - lord - 2nd pi. [poss.] pron. "I am your lord". (Copt.) ang-petn-jôt 1st sgl. them. pron. - sgl. m. poss. art., 2nd pi. -father "I am your father".

(III) Verbal nexus is effected in a complex set of subpatterns. In ME, and to a large extent already in OE, the non-modal verbal system (as it is conveniently referred to) is almost entirely cast in the non-verbal adverbial nexus matrix (I): see Polotsky 1965, 1976: certain various sdm.f and sdm.n.f stem + pron./subst. theme subnexal constituents occupy the thematic (substantival, initial) and rhematic (adverbial, second) slots, preceded by deictic markers (of which the most striking is the formal-thematic here-and-now jw).19 (1) In rhematic status, the converbal (finite adverbial) sdm.f and sdm.n.f forms are (for transitives) opposed to, or (for intransitives, especially verbs of movement) neutralized with a peculiar statal finite converb (currently known as 'Stative', alias 'old Perfective' or Pseudopartizip), with personal thematic suffixes not occurring anywhere in the Egyptian system, but with clear formal correspondents in Accadian.20 The main forms of the ME verb clause may be presented as follows:21

The converb preposition is orthographically zeroed from Demotic (partly LE) onwards, and the converb is thus fully grammaticalized. 19 Sdm "hear" set of stems or bases (rhematic);-/"he" suffixed pronoun category (thematic). Another case where it is pointless to refer to a 'VS' typology. The Stative expresses perfect state in the present, with an added passive diathesis for transitive lexemes. 21 This is definitely not a full or even a summary account of the Egyptian verbal system, but a drastically abridged scheme, to give an idea of the verbal-nexus mechanism.

84

ARIEL SHISHA-HALEVY

hr sdm sdm.f

"he/my brother hears/ is hearing" (actual present) "he/my brother hears" (generic/atemporal 'present' or aorist)

sdm.n.f

"he/my brother (has) heard" (main narrative foreground carrier)

sdm.w

"he/my brother has been heard" (3rd sgl. m. Stative: 1st sgl. jw.j sd2m.kwj)

jw.w

"he/my brother is/has come" (Stative: 1st sgl. ink wj /jw-jjW'kwf)

sdmw.f

"he/my brother was heard"

(c) chc.n.f

sdm.n.f

"that he arose is..." [+ CONVERB "having heard"] = "then he heard..." (foreground delimitation)

(d) prr.f

ADVERBIAL

"[it is] ADVERBIAL/CONVERB FOCUS + that he comes/will come out"

jj-n-f

ADVERBIAL

"[it is] ADVERBIAL/CONVERB FOCUS + that he came/has come"

gm.n.tw.f

ADVERBIAL

"[it is] ADVERBIAL/CONVERB FOCUS+ that he was found"

(a)jw-.f/sn.j (b) m.k -sw/sn.j

Table 1: The main forms of the ME verb clause

Note: (a) and (b) are mutually opposed as predications in (respectively) unmarked and allocutive/interlocutively marked deictic perspectives; (c) the verb chc "stand, get up" is used as a periphrastic strongly delimitative narrative turn-of-event or foreground-spurt tense; this is a formalized Cleft Sentence, with the converbal ('circumstantial') clause in focus, and the verbal nexus predicating "stand" as grammaticalized topic: "It was having heard that he got up" = "He up and heard" = "...and then he (suddenly) heard"; (d) Cleft Sentences, with the adverbial or converb in focus, and a verbal nexus as topic. I hope to have made evident the insignificance, descriptively speaking, of the 'SV' typology in the face of pattern complexities. (2) Illustrating here only some of the tenses carrying the affirmative narrative basic (non-focussed) foregound, in opposition to report {Bericht, Mitteilung), with transitive-valency verb lexemes: 'Aorist' as used in Turkish or Coptic Tempuslehre. See also Shisha-Halevy 1995 (Middle Welsh).

STABILITY IN CLAUSAL/PHRASAL PATTERN

85

(OE) sdm.f vs. jw(.f) sdm.n.f 'hear-pret. - 3rd sgl. m.' = "he heard" vs. 'PRESENTSITUATION-DEICTIC (+ 3 sgl. m.) + POSSESSED {hear} of- 3rd sgl .m.' = "he has heard". (LE) sdm.f "He has heard" vs. the so-called Sequential jw.f) hr sdm "...and (then) he heard": ph wc n-rmtp3-sd t3y.k-srt "A man has reached your daughter's window", reported event following the narrative event information wn-jn p3-hrd hr jjt jw.f hr pwjt jw.fhrphp3-ssd n-t3-srjt... "...then the youth came and he jumped and he reached the window of the daughter...". (Dem.) sdm.f I jr.f sdm "he heard" vs. w3h.f sdm "he has heard" (auxiliary w3h "lay, rest"). (Copt.) a-f-sotm, dial, ha-f-sotm "he heard" vs. a-f-ouo e-f-sotm "he has (already) heard" (lit. he-rested/finished + CONVERB "he-hearing").

(3) In view of all this, I believe that it is less than useful to focus on the sequence diachrony of verb-nexus constituents alone. While 'VS' is the unanimous verdict of word order typologists regarding Egyptian, this can refer only to the synthetic, non-periphrastic verb form which, as we have seen, is in ME very different from a verb clause, and even then needs to be qualified: for these verbal formations are all cast in a rhematic matrix of a STEM + possessive-like PERSONAL SUFFIX, with the delocutive pronoun commutable with a noun. From ME on, and with a marked acceleration in LE, auxiliaries take on the verbal-categories load, in an analysis - agglutination - re-synthesis process that is all but complete in Coptic ("he [has] heard": a-f-sötm, neg. mpf-sötm, with the prefixal so-called conjugation bases only in part evolved out of auxiliaries, and certainly not a 'VS' case), and makes a 'VS' schematic view synchronically meaningless: in this sense, a synthetic form like Greek ÓKOUU "I hear" would be equally 'VS'. (4) The nominal-theme placement paradigm is in OE/ME a binary one (always taken schematically, with considerable simplification; sn "brother", -j T ' , -ƒ "he"jH>- "here-and-now" deictic thematic exponent): (a1) jw sn.j sdm.f "my brother hears" (usually, habitually, by nature, generically). (a2) sdm sn.j "may my brother hear" (sdm.f\, the so-called 'prospective' formal/functional category), "...my brother hearing" (sdm.f2, 'circumstantial' or converbal), "my brother (has) heard" (sd2m.f3, affirmatively only in OE, 'perfectal'). (b) (jw) sn.j sdm.f "my brother, he hears" I jr-hm nb r pn, ncq.n.f "as-for anyone who-does-not-know this spell, he cannot enter". While (a1) and (a2) seem to be unmarked terms, with a formal complementary distribution, (b) is an important topicalization construction, the discourse-

86

ARIEL SHISHA-HALEVY

functional details for which have not yet been conclusively studied: it is, however, suggested that the sdm.f nexus form itself is rhematic to the thematic topic, as a high-level case of the basic # NOMINAL + ADVERBIAL# adverbrheme nexus pattern. By Demotic and Coptic, however, the nominal-theme placement category becomes extremely complex. Affirmatively predicating a preterite (a-) auxiliary, expanded with the lexeme "hear" (sôtm), of "the man" (p-rôme, pronominally referred to by "he" [-ƒ-]), we encounter the following quaternary paradigm:23 (a) p-rôme a-f-sotm (b) a-pröme sôtm (c) a-próme a-f-sötm (d) a-f-sötm nci-prôme (nci is an obligatory element introducing the postposed theme-actor) As a matter of fact, what we know at present about the functional assignment of this paradigm (always allowing for dialectal and textemic, sometimes considerable differences) is very little. In some text-grammatical ('discourse') configurations, (a) is a topicalization construction; in others, it is a (mostly narrative) theme-switching delimitation; (b) is a somewhat rarer, perhaps theme-maintaining construction; (c) is a common placement, sometimes theme-switching (but differently - especially as regards the theme repertories - from [a]). Typically, it appears to be superordinating and nexushighlighting; (d), restricted to certain corpuses but in them fairly frequent, is still not well understood, but probably signals theme retrieval from an accumulated inventory, and thus some intermediate degree of theme continuity. While the category as a whole appears to be a gradient one of textual cohesion - linkage and delimitation, with reference to degrees of depth of theme 'storage' - it is still almost entirely uncharted. However, it is obvious that neither is (a) a case of 'SV', nor (c) a case of 'VS'. Their precise functional evaluation (currently undertaken by the present writer for a given corpus) must take into account at least the following: (1) the internal analytic interpretation of the individual constructions; (2) their possible correlation with formal (a) clause-internal and/or (b) text-grammatical (combinatory, narratological and narrative-grammar) parameters. Among the former, agens/theme extent, agens/theme specificity, internal Flexionsisolierung (grammeme/lexemeTypically contrasted with a binary paradigm: finite verb (verbal rheme-suffixed theme)+ nom. theme vs. nom. theme + finite verb (verbal rheme-suffixed theme) in the Greek Vorlage.

STABILITY IN CLAUSAL/PHRASAL PATTERN

87

separating structure, tense and negation of verb component etc.); among the latter, opposition or gradience of switching vs. persistence of the theme, thematicity of the verb clause, cohesiveness or linkage, disruptivity or delimitation, concatenability, narrative status, compatibility or incompatibility with discourse signalling etc.; (3) their possible correlation with narrativegrammar and narratological functions, such as information chunking, pacing; dramatic highlighting; narrative focussing, mode and so on; (4) scanning of markedness/un-markedness gradience or opposition; (5) examining the difference between theme-switching vs. theme (re)introduction; or between various hierarchical theme classes or 'folders' (narratologically more or less prominent or significant); or the possibility that the themes of all four constructions are drawn from different cumulative 'theme folders' (differing for instance in narrative depth or immediacy of presence), or from a single 'folder' but four different 'recall triggers' signalling the extent of rewinding required for decoding the reference. 5.2 The nucleus /expansion (determinatum /determinans) Grundrichtung in noun phrases. SDF: evolution of the nuclear determinators; loss of the syntactic category 'adjective' (a) Possessive-associative expansion of nominals: (Phase I) •

SUBSTANTIVE + SUBSTANTIVE, NOUN + NOUN ('loose composition'?) (OE, ME)Xs3-Y "X son of Y", hm-ntr "servant-of-God" = "priest", imy-r pr "overseer of- house" = "steward". (LE) restricted to fossilized phrases: hr-c .wy p3-ntr on -hands-dual + def. art. sgl. m.—god "in God's hands".

•

SUBSTANTIVE + nj- SUBSTANTIVE (nj- in partial concord with the nucleus: linking-relating adjective: ["the one] related to") (OE, ME) c3.w nj.w shy pn "the donkeys of this peasant", Nsw nj Kmt "the King of Egypt".

•

(Phase II) SUBSTANTlVEspec + n - non-zero article - NOUN (n- 'nota relationis') (LE) jbnnb.f heart + of- lord-poss. 3rd sgl. m. "his lord's heart"

ARIEL SHISHA-HALEVY

88

sb3yt n šct teaching +of-letter "epistolary teaching".

•

• •

(Phase III) suBSTANTlVEnon-spec + nte- non-zero specifying article - NOUN/PRONOUN (nte- relating/associating preposition): general/unmarked association or appurtenance; ARTICLEnon-phor - SUBSTANTIVEinalien + n- non-zero article - SUBSTANTIVE: inalienable personalized association (possession); ARTlCLEphor - SUBSTANTIVEalien + nte- non-zero article - SUBSTANTIVE / PRONOUN: depersonalized association, expanded class name. (Copt.)p-sôma m-pa-sêre def. art. sgl. m. - body + of -poss. art. m. 1st sgl. - son "my son's body". šêre nta-k zero art. -son + of- 2nd sgl. m. [poss.] pron. "a son of yours". (Boh. Copt.) t-jij m-pi-romi non-phoric def. art. sgl .f. - hand + of- phoric def. art. sgl. m. - man "the man's hand" but ni-ei nte-ti-baki phoric def.art.pl. - house + of - phoric def. art. sgl. f. - town "the houses of the town".

•

(b) Attributive expansion: Phase (I) SUBSTANTIVE + ADJECTIVE ('lexical adjective' - participle: nfr-0 "lovely-one masc",nfr-t "lovely-one fem", 'pronominal' adjective': quantifiers, demonstratives,nb "all, every", syntagmatic -j nucleus derived adjectives [nisbe "the one referring to/ associated with"]): (OE, ME, LE [Dem.]) shtj pn peasant + this adj. "this peasant, the above-mentioned peasant" hm.w.t nfr(w).t woman [PL.+FEM]. + lovely [PL.+FEM]

"lovely women/the lovely women".

Cf. Shisha-Halevy 1994a.

STABILITY IN CLAUSAL/PHRASAL PATTERN

•

89

Phase (II) SUBSTANTIVE + n- zero article - NOUN attribute ([Dem.] Copt.) snof nim n-dikaios blood + every + of - Ø-righteous "every righteous blood" ou-angelos n-ouoein an-angel+ of- 0-light "a shining angel" ou-sophos n-rôme a-wise-one+ of-man "a wise man".

(c) Determinated noun phrase: •

DETERMINATOR + NOUN (OE, ME)pr "house",p3-pr 'prox.dem.sgl.m. - house' = "this house",pr.j 'house - 1st sgl. [poss.] pron' = "la/una mia casa" (not specific?) (LE)p3-pr 'def. art. sgl. m. - house' = "the house", p3y.j pr 'sgl. m. poss. art. sgl. m. 1st sgl. ["the-of-me"] - house' = "my house" (specific) (Sahidic Copt.)p-êi 'def .art. sgl. m. - house' = "the house", pa-êi 'poss. art. sgl. m. 1st sgl. ["the-of-me"] - house' = "my house"; ou-ei "a house", hen-êi "(some) houses"; (Boh.Copt.) t-phe, ph-nobe "the sky", "Sin" (uniques and genus notion names);pi-rômi "the man", "man" (phoric specifier and discriminative intensive generic).

Note that Coptic (and later Egyptian generally) provides conclusive evidence for the nuclearity of the determinators (noted in the Sixties for European languages by H. Frei and H.-J. Seiler).25 The non-adjectival demonstratives v/ere always nuclear, and their evolution to articlehood involved no major syntagmatic change. While much is known of the evolution of the definite articles in Romance and Germanic (albeit the precise scenario is still controversial), only a few details of the phenomenon are as yet clear for Egyptian. LE 'emerges' with a fully operational definite article, and an incipient "one"-type indefinite article (we catch occasional glimpses of the former in colloquial language as early as the Old Kingdom), with marked differences in the different varieties or registers. The final phase of Egyptian teaches us caution: almost every dialect has its idiosyncratic determinator system, and some have more than one specifying articles (Bohairic Coptic, for instance has a phoric/generic1 and a See Shisha-Halevy 1986: Chapter 4.

90

ARIEL SHISHA-HALEVY

non-phoric/generic2/inalienable-possession 'definite' articles). It is evident that sophisticated discourse-level analysis and pragmatics of referentiality are essential for understanding and charting the phenomena involved; for instance, genericity/indefiniteness oppositions involve a complicated resolution depending on slots and the reduced paradigms therein. It seems still early days for this in Egyptian linguistics.26 5.3 Negators In Egyptian the resolution of negative forms and patterns is much higher than of affirmative ones: this typologically interesting lack of affirmative / negative symmetry is instructive in itself. Some highlights of negation are: (a) Preverbal nuclear negation: (OE/ME) n sdm.f"he did not hear/has not heard" (preterite/perfect); nn sdm.f "may he not hear" (prospective; existential negation, negative existence of "that"-form?). n sdm.n.f "he cannot hear" "it is not in his nature to hear" (generic; affirmative counterpart jw.f sdm.f). (LE) bw-pwy.f sdm "he did not hear/has not heard" (n-> bw-, aux.pwy.f "he has done in the past"); bw-jr f sdm "he cannot hear" "it is not in his nature to hear" (generic; n-> bw-, "that"-form sdm.n.f >j.jr.f sdm, with auxiliary jrj "do"); bn-jw.f r sdm "he won't hear" (future: affirmative jw.f r sdm, future - no longer presentbased directional adverbial). (Copt.) mp-f-sôtm "he did not hear/has not heard" (bw-pwy.f-> mp-f-); nne-f-sotm "he won't hear" (future: affirm e-f-e-sôtm; bn-jw-f > nne-f-); me-f-sôtm "he cannot hear" "it is not in his nature to hear" (generic; affirm: aoristÆ«-/sôtm, appearing in Demotic; bw-jr.f > me-f-).

26

There are many scenarios for the evolution of indefinite articles: the Problematik and definitions involved are very different from those of the definite article. If anything, the borderline between the pre-article and articular phases is more fuzzy, and the issues more complicated for the indefinite, with presentative, quantifying, specifying and generic, negative referentiality roles all involved in the transition, while non-referentiality is a key factor (see for instance Hopper & Martin 1987). However, the usual relative chronology of the indefinite evolving later than the definite agrees with the Egyptian facts. For some interesting theoretical considerations concerning the early stages of an evolving definite article, see Epstein 1993: 115,126ff.).

STABILITY IN CLAUSAL/PHRASAL PATTERN

91

We see here the same (structurally and in part 'etymologically') negative prefixes, for verbal nexus negating auxiliaries evolved from synthetic 'lexemic stem' forms, fused into nuclear negative bases, carrying, pro-verb like, all grammatical categories for the verb clause. (b) Nexal discontinuous ('embracing') negation: (OE, ME) n (js) negativing nominal and adverb-rheme nexus (LE) bn (jwn3) negativing nominal and averb-rheme nexus (Dem., Sah. Copt.) (n) an, (Boh. Copt.) (n)...(an) negativing nominal and adverbrheme nexus - not verbal nexus proper; local (nexus and clause-constituent) negation (with focussing associations). (ME) n hs js rn.k "your name is not 'excrement'". (LE) bn-rmt jwn3 "it isn't a human-being" ('not-/human-post-neg.', situational zero theme); bn sw m-sš "it is not in order" ('not-he-m order'). (Copt.) m-p-rome mmau / sôtm an "the man is not there/does not hear"; m-pa-son an pe "he is not my brother" (pe delocutive pron. theme).

Two etymologically unrelated postposed negation elements occur in the course of Egyptian diachrony. The first, js, is an enclitic, also occurring in focalizing and clause-including roles. The second, LE jwn3 to Coptic an, has no other functions than post-negation, but may also have been a focalizing exponent (cf. Winand 1997). Now it is evident that the so-called postverbal negation in Egyptian is unrelated to any word order change, or "shift from pragmatic to grammatical word order" (cf. Burbridge 1993:212ff.), nor is it associated with the verb clause as such (but rather with the converbal, i.e. adverbial verb in rheme status)27 - on the contrary, it is 'customized' for essentially non-verbal (adverbrheme and noun-rheme) nexus, or for negativing nexus constituents 'locally'. "Jespersen's Cycle" is only valid in Egyptian from Coptic onwards:28 in preCoptic Egyptian, from the earliest documentation on, the negation is 'embracing', i.e. discontinuous, with both constituents fully functional, and the negation not in effect post-verbal. 5.4 The causative conjugation Various syntagms containing forms of the auxiliary (r)dj, ti, t- , imperative jm(j), ma- "give, cause" as their grammatical nucleus evolve into what seem 'morphological complexes':

The association with the verbal predication and the 'SV' order is suggested also in Bernini 1984:31-32,1986. 28 Probably not fully in Bohairic Coptic: cf. Shisha-Halevy 1981.

ARIEL SHISHA-HALEVY

92 (ME) (infinitive) djt wnm.f to-cause-that-he-eat "to-make him /let him eat";

(1st sgl. post-imperatival sequelling) ... dj.j. wnm.f that-l-cause+ that-he-eat "...and I will make/let him eat"; (imperative) jm wnm.f cause-that-he-eat "make him eat!"; (imperative) jm dd.f hrt-jb.f cause- that-he-say + that-with-heart-his "let him tell what is in his heart". (Dem.) t-jr.f wm "make/let-him eat"; (1st sgl. post-imperatival sequelling)... t.j jr.f wm "...and I will make him eat"; (imperative) mj jr.fwm "make/let-him eat!". (Copt.) (infinitive) t-re-f-ouôm cause-that-he-do-eating "to-make him eat", "that he eat", "for him to eat", tmmo-f lexicalized "to-feed him" (metanalytical from t-mmo-f ' cause that-he-eat'); ('personally guaranteed' post-imperatival sequelling)... ta-re-f- ouôm "...and he shall eat"; sine ta-re-tn-cine "seek and ye shall find"; (imperative) ma-re-f-ouôm "make him eat" (causative imperative); "let him eat" (jussive).

This evolution - with a total stability of sequencing - is a striking instance of the 'syntax-to-morphology' agglutination process (cf. Fleischman 1982:126127). However, the underlying issue of juncture (Fleischman 1982:151ff.) is here crucial, since the evolutional presupposes a precise definition of such concepts as 'separate (words)' and 'fusion', that is, a precise and sophisticated synchronic account of junctural gradience. The main operative differences between the earlier and later Egyptian forms are: (1) the analysis by means of the auxiliary jr "do" as carrier of grammatical categories in later Egyptian (from LE on); (2) the differentiation of a (closed list) 'lexical' and 'grammatical' causative forms in Coptic; (3) the use of the causative imperative also as jussive (i.e. 3rd-personimperative), and the causative infinitive also as a personal infinitive and thatform in Coptic, roles attested from ME on. In any case, while a metanalytical process is in evidence, constituent ordering shift is not.

STABILITY IN CLAUSAL/PHRASAL PATTERN

93

5.5 The cleft sentence and other focussing constructions The Egyptian Cleft Sentence pattern set is very complex, both formally and functionally, and is of the highest typological importance (as has been often noted, it is strikingly similar to its Celtic correspondents, formally to a degree, functionally, to an extent distributionally and in its 'devaluation' grading). It is hardly possible to report in this limited context even on the most essential features of the formal repertory, role relationship and distribution of mise en relief constructions in the respective phases (see Polotsky 1944, 1965, 1976, 1990; Shisha-Halevy 1986: Chapter 2, 1987). Observe, however, that it is only at the last phase - in Coptic - that focus-initial adverbial-focus patterns, Cleft and non-Cleft, interrogative and declarative, make their appearance, a placement shift that is arguably also due to contact with Greek. Egyptian has always had two Cleft Sentence types, focus-initial and topic initial; nominal and pronominal foci are have been invariably initial in Egyptian Cleft Sentence patterns, but in Coptic we encounter (pro)nominal foci also in the topic-initial set.29 Up to Coptic, we find: • # that-VERB-FORM + ADVERBIAL FOCUS #, in complementary focusconstituence distribution with •

# (PRO)NOMINAL FOCUS + [FORMAL THEME] + RELATIVE/PARTICIPIAL FORMS30 #

•

# ADVERBIAL/(PRO)NOMINAL FOCUS +BASIC VERB / thatl-ing COnverb #

•

# that-VERB + (pro)nominal focus # ('autofocal', foc. a lexemic/adverbial rheme):

Coptic [especially Sahidic, the most innovative dialect] also:

(ME) jj.n.k hr sjj jsšt? that-you-have-come + because-of-what "What is it you have come for?" = "Why have you come?". (LE)j.jri k itt 3ht.j hr-ih that-you-have-stolen/are stealing my-things + because-of-what "Why is it that you have stolen/are stealing my things?" ntk p3-jr-shr you (sgl.m.) + the (sgl. m.)-who-do-counsel "It' s you who are counselling". Note that in Egyptian (as in Celtic) many interrogative pronominals and adverbials condition the Cleft Sentence construction, that is, are always marked for focality. Note that from OE through LE, that forms and relative forms are structurally identical, with the latter gender/number-commutable (and concording) allo-forms of the former (or of course the other way round).

94

ARIEL SHISHA-HALEVY

(DQm.)j.jrJjy rjr n.k t3y.s sbt that-I-have-come to-make for-you (sgl. m). + its-repayment "It is to repay you for it that I have come"; wc-wcb n Jmn pS-jjr-tSts a- (masc.)-priest of-Amun it is-who-did-stealing-it "It's a priest of Amun who stole it". (Copt.) etbe-ou a-k-ei because-of-what + you-did come "Why did you/have you come?"; n-as n-he e-ou-nobe na-n an pe in-what of-manner + that-a-sin + for-us neg. it-is "How is it that it isn't a sin for us?"; etbe-pai e-a-i-joos nê-tn because-this + tf^-did-I-say-it + to-you (pi.) "That's why I told you this"; p-noute p(e)-ent-a-f-aa-s the (sgl. m.)-god + it-is-w/zo-did-he-do-it "It's God who did it"; e-i-na-r-ou that-I-shall-do + what "What is it that I shall do?" "What shall I do?". 5.6 The r'hematic (predicative) adjective SDF: the disappearance of 'adjective' as a form-class and the emergence and evolution of the articles: • # ADJECTIVE + (PRO)NOMINAL THEME # (OE, ME, [LE]) nfr tw h j good - 2nd sgl. m. dependent pron. - with-lst sgl. [poss.] proa "You are well-off with me". • # article - NOUN + (PRO)NOMINAL THEME # (Copt.) oumepe "He is true" (ou- indef. article, me "truth" [dÅrfØeia], -pe delocutive thematic pronoun) "dArjOi vdg êcjriV' (Joh. 5:32). 5.7 Nominal-Sentence Patterning SDF: emergence and evolution of the determinators; disappearance of the adjective morpho-syntactic class (see § 5.2). This is probably the most widely discussed feature of Egyptian throughout its diachrony, and in fact, one of its main typological showpieces.31 The most striking diachronic processes encountered in this context (in particularly broad outlines and high simplification) are: See Sethe 1916, Shisha-Halevy 1987.

95

STABILITY IN CLAUSAL/PHRASAL PATTERN

(a) pronominal theme-to-copula grammaticalization (evident also in most Semitic languages); (b) topic-to-theme grammaticalization in Bohairic Coptic; (c) the virtual disappearance of the binominal predication pattern (replaced by pronominal-theme patterns, see below) and conjointly (d) the replacement of the noun-predication patterns by determinatorpronoun and personal-pronoun predication. The basic formal dichotomy is between delocutive (3rd-person) and interlocutive (lst-2nd-person) pronominal themes: (I)Delocutive themes: •# [TOPIC+] RHEME + EXPANSION = 'THEME') #,

PRONOMINAL

DELOCUTIVE

THEME

(+THEME

(ME) shtjpw n SktHm3 fa) peasant-mutiic/m. deloc. them. pron. + of-Field-of-Natron "He was a peasant of the Wadi Natrün"; phr.tpwn.t w.s mrh.t remedy (-fem.)-neutric/m. deloc. them, proa + for- body-3rd sgl fem.poss.pron-ointment "Ointment is the/a remedy for her body"; Hknw pf dd.n.kjntfbw pw wr n jw pn Hknw-oil + distal demonstrative + past relative stem-2nd sgl. m. pron. + infinitive " bring"-deloc. m pron. + abstract-marking lexeme-neutric/m. deloc. them. pron. + big + of island + proximal demonstr. adjective "That Hknw-oil you spoke of sending, it is a staple of this island". (LE)p3y.kjtp3y sgl. m poss. art. 2nd sgl. m.- father + neutric/m. them, proa "It's/he's your father"; jr nbw hc ntrw topic-marker-gold + body-[of] god-pl "Gold, [it's] the gods' body" (zeroed theme). (Dem.) p3-sp n3j.j m3c n3j sgl. m. def. ait - rest + pl. poss. art. 1st sgl. neutric/m-justification + plur. them, pron "The rest, these are my justifications"; bn-jw 3hy jn p3j p3-rmt PRENEGATOR - (a) reed + POSTNEGATOR + sgl. m them, proa + sgl. m. def. art. -(a) man "Man is not a reed". (Copt.) ou pe pek-hôb inanimate interr. pronoun- m. sgl. them proa + sgl. m. poss. art. 2nd sgl. m-work "What is your work?". (Boh. Copt.) Gen. 29:15 pek-bekhe ou pe sgl.m. poss. art. 2nd sgl. m -wages + interr.proa- ra them. pron. "What are your wages?";

96

ARIEL SHISHA-HALEVY

pa-nobi ou-nisti pe sgl. m. poss. art. 1st sgl.- sin + indef.art. -great- m. them. pron. "My sin is great" (Gen. 4:13).

•

(Il)Copular: #THEME - COPULA - RHEME # (OE)hnm.t.fpj J3t wet-nurse- deloc. m. poss. pron + copula +J3t "His wet-nurse is Iat". (ME) srwh.fpw hms.t treatment- 3rd. sgl. m. poss. pron + copula + sit (inf.) "His treatment is sitting". (Copt.)p-jok ebol m-p-nomos pe t-agapê sgl. m. def. art. -fulfilment + of- sgl. m. def. art -law + copula + sgl. f. def. art. -caritas "The fulfilment of the Law is caritas".

•

(Ill)Interlocutive: # PRON. THEME - RHEME # (ME)jnks3.f 1st sgl .thematic pron- son- deloc.m. poss. pron. "I'm his son" ( vs.jnkpw, I + deloc. them. pron. = "It's me"). (LE)ntkshj 2nd sgl. m. them. pron. -(a) dumb person "You're dumb". (Dem.)jnk-wcb 1st sgl. them, pron -(a) priest "I'm a priest". (Copt.) ang-ou-fnt 1st sgl. them, pron.) -sgl. indef. art. -worm "I'm a worm".

(IV) Wechselsatz

('balanced nexus')

(OE)jnkpw sy I + copula + she "I am she". phty [PN] phty Sth strength [of-] [PN] + strength [of-] Sth "The strength of X is the strength of Seth". (Dem.) p3j.j šr c3 p3j.t šr c3 sgl. m. poss. art. 1st sgl.- son + sgl. m. poss. art. 1st sgl.- son "My eldest son is your eldest son"; n3j.j mdt.w n3j.j mdt.w cn pi. poss. art. 1st sgl.- word + pl. poss. art. 1st sgl.- word + again (grammeme) "My words are my words", "my case rests".

STABILITY IN CLAUSAL/PHRASAL PATTERN

97

(Copt.) anon pe ntof we + copula + he "We are he", "We and he are one and the same".

Observe that there are considerable formal differences, largely maintained in the various phases of the language, between the patterns above as regards theme/rheme constituencies, constituent sequencing and prosodic/junctural contours; differences which correlate with type of predication as well as macrosyntactic integration. Perhaps the most striking change observable is from a predication of substantive lexeme to predication of pronominals - the determinators, evolving from LE on. Constituent ordering alone is not informative of pattern identity; theme-to-rheme sequencing is basic for an interlocutive pronominal rheme but marks cotextual boundness and/or thematic progression for nominal or demonstrative themes, whereas rheme-to-theme sequencing is unmarked for delocutive pronominals. 5.8 Sentence prosody - enclitic placement: cyclic shift of low-stress slotting. (I) Clause-initial sentence-initial focal tone in later Egyptian: The trigger to change from clause- or colon-final (phase II, LE to Demotic) to colon-second placement of enclitics (phase III, Demotic [?] and some Coptic dialects) might be attributed to Coptic/Greek adstratic contact32: 'Wackernagel's Law' (with E. Fraenkel's modifications) as applied to Coptic (the 'floating' elements, or enclitics: sentence particles; some pronominal forms of prepositions; augentia; some adverbials - see Shisha-Halevy 1981, 1986: Chapter 6). However, the earliest état de langue (phase I) seem to reveal the same colon-second placement (in the examples below; the boldface pronominals are coreferent; the translation aims at reflecting this coreference and the pronominal placement, and is thus ungrammatical as English): (II) Clause/colon-final: (LE)jw.j hr djt n.fp3y.f nty m-dj.j gr ntf "...and I gave him his, which was in my possession, he too"; m-jr dd n.j "wn.kjm"jnk "Don't tell me 'you were there', I".

(I, III) Clause/colon-second: (OE, ME) wnnjrf t3 pf mj-m m hmt.f "But what will this poor land be like without him?".

On the complicated associations of linguistic contact and evolution, see Goebel, Nelde et al 1996:534ff.

98

ARIEL SHISHA-HALEVY (Copt.) n-aš n-he anok e-i-na-eire n-ou-hôb ejô-k? "How, I, (is it) that I will do anything without you?"; n-t-he hôôn on tenou et-n-r-nobe, ... "As we too now, we sin...".

To conclude, with a single sentence: pattern-related sequencing in Egyptian is remarkably stable throughout the long history of the language, since the patterning itself is largely maintained unchanged (despite some basic systemic developments mainly associated with the noun phrase), and sequencing is a pattern constituent.

REFERENCES Borsley, Robert D. 1995. "On Some Similarities and Differences Between Welsh and Syrian Arabic". Linguistics 33.99-122. Bernini, Giuliano. 1984. "Vers une typologie aréale de la négation discontinue". Romanistique-germanistique: une confrontation. Actes du colloque de Strasbourg, 23-24 mars 1984 ed. by C. Bouridant, 28-32. De Boer, Charles. 1928. "Études de syntaxe francaise, V: Facteurs troublants dans les études de syntaxe francaise ".Revue de Linguistique Romane 4.301-310. Epstein, Richard. 1993. "The Definite Article: Early Stages of Development". Historical Linguistics 1991: Papers from the Tenth International Conference on Historical Lingustics ed. by J. Van Marie, 111-134. (= Amsterdam Studies in the Theory and History of the Linguistic Science, Series IV, vol. 107.) Amsterdam & Philadelpia: John Benjamins. Erman, Adolf. 1028-9.Aegyptische Grammatik. Berlin: Reuther & Reichard. . 1933. Neuaegyptische Grammatik. Leipzig: Engelmann. Fleischman, Suzanne. 1982. The Future in Thought and Language: Diachronic Evidence from Romance. (= Cambridge Studies in Linguistics, 36.) Cambridge: Cambridge University Press. Gardiner, Alan Henderson. 1975. Egyptian Grammar. (3rd ed.) Oxford: The Clarendon Press. Goebel, Hans. et al. eds. 1996. Kontaktlinguistik: ein internationaler Handbuch zeitgenössiger Forschung, 1 Halbband. Berlin & New York: De Gruyter. Griffith, Frank Lloyd. 1900 [1985]. Stories of the High Priests of Memphis. Osnabrueck: Zeiler. Griffith, Frank Lloyd. 1909. Catalogue of the Demotic Papyri in the John Rylands Library. Manchester: The John Rylands Library. Gunn, Battiscombe. 1924. Studies in Egyptian Syntax. Paris: Geuthner. Hawkins, John A. 1983. Word Order Universals. New York & London: Academic Press. Hickey, Raymond & Puppel, Stanislaw eds. 1997. Language History and Linguistic Modelling. A Festschrift for Jacek Fisiak on his Sixtieth birthday. Berlin: Mouton de Gruyter.

STABILITY IN CLAUSAL/PHRASAL PATTERN

99

Hopper, Paul J. 1986. "Discourse Function and Word Order Shift: a Typological Study of the VS/SV Alternation". Language Typology 1985: Papers from the Linguistic Typology Symposium, Moscow, 9-13 December 1985 ed. by Winfred P. Lehmann, 123-140. (= Amsterdam Studies in the Theory and History of Linguistics, Series IV, vol. 47.) Amsterdam & Philadelphia: John Benjamins. Hopper, Paul J. & Martin, Janice 1987. "Structuralism and Diachrony: the Development of the Indefinite Article in English". Papers from the Seventh International Conference on Historical Linguistics ed. by A. G. Ramat, O. Carruba & G. Bernini, 295-304. (= Amsterdam Studies in the Theory and History of the Linguistic Science, Series IV, vol. 48.) Amsterdam & Philadelphia: John Benjamins. Isaac, Graham R. 1996. The Verb in the Book of Aneirin: Studies in Syntax, Morphology and Etymology. (= Beihefte der Zeitschrift fur Celtische Philologie, 12.) Tübingen: Niemeyer. Jamieson, Stephanie W. 1993. "Determining the Synchronic Syntax of a Dead Language". Historical Linguistics 1989: Papers from the Ninth International Conference on Historical Lingustics ed. by H. Aertsen & R. J. Jeffers, 211-220. (= Amsterdam Studies in the Theory and History of the Linguistic Science, Series IV, vol. 106.) Amsterdam & Philadelpia: John Benjamins. Junge, Friedrich. 1985. "Sprachstufen und Sprachgeschichte". XXII Deutscher Orientalistentag ed. by W. Röllig, 17-32. Ausgewahlte Vorträge, Fachgruppe I: Ägyptologie. Stuttgart: Franz Steiner Verlag Wiesbaden. Kindt, Walther. 1994. "Satzbegriff und gesprochene Sprache". Lingua 94:1.25-48. Kortmann, Bernd. 1997. Adverbial Subordination. A Typology and History of Adverbial Subordination Based in European Language. Berlin & New York: De Gruyter. Layton, Bentley. 2000. Coptic Grammar. (The Sahidic Dialect) Wiesbaden: Harrassowitz. Loprieno, Antonio. 1988. "On the Typological Order of Constituents in Egyptian", Journal of Afro Asiatic Linguistics 1.26-57'. . 1991. "May Themes Follow on Rhemes, and Why Might They Do So?". Lingua Aegyptia 1.293-300. . 1995. Ancient Egyptian: A Linguistic Introduction. Cambridge: Cambridge University Press. Mithun, Marianne. 1987. "Is Basic Word Order Universal?". Tomlin 1987. 281-328. Polotsky, Hans-Jakob. 1944. Études de syntaxe copte. Cairo: Société d'Archéologie copte . 1960. "The Coptic Conjugation System". Orientalia 29.392-422. . 1964. "Ågyptische Verbalformen und ihre Vokalisation". Orientalia 33.267285. . 1965. Egyptian Tenses. Proceedings of the Israel Academy of Sciences and Humanities, II/5. Jerusalem. . 1971. Collected Papers. Jerusalem: the Magnes Press. . 1976. "Les transpositions du verbe en égyptien classique". Israel Oriental Studies 5.1-50.

100

ARIEL SHISHA-HALEVY

. 1987-1990. Grundlagen des koptischen Satzbaus, I—II. (= American Studies in Papyrology, 27, 29.) Decatur, Atlanta: Scholars Press. Ramat, P., Bernini, G. & Molinelli, P. 1986. "La sintassi della negazione germanica e romanica". Parallela 2, Aspetti della sintassi delVitaliano contemporaneo ed. by K. Lichem, E. Mara & S. Knaller, 237-270. Tübingen: Narr. Schenkel, Wolfgang. 1987. "Zur Struktur des dreigliedrigen Nominalsatzes mit der Satzfolge Subjekt-Prädikat im Ågyptischen". Studien zur Altaegyptischen Kultur 14.265-282. Sethe, Kurt. 1916. Der Nominalsatz im Ågyptischen und Koptischen. Leipzig: Teubner. Shisha-Halevy, Ariel. 1973. "Apodotic efsôtm: a Hitherto Unnoticed Late Coptic, Tripartite Conjugation Form and its Diachronic Perspective". Le Muséon 86.455466. . 1981. "Bohairic-Late Egyptian Diaglosses: a Contribution to the Typology of Egyptian" Studies Presented to HJ. Polotsky ed. by D. W. Young, 413-338. Beacon Hill: Pirtle & Poison. . 1986. Coptic Grammatical Categories: Structural Studies in the Syntax of Shenutean Coptic. Rome: Pontificium Institutum Biblicum. . 1987. "Grammatical Discovery Procedure and the Egyptian-Coptic Nominal Sentence". Orientalia 56.147-175. . 1994a. "Pluridimensional Oppositions: Case Studies from Scripture Bohairic". Coptology: Past, Present and Future: Festschrift R. Kasser ed. by S. Giversen, M. Krause & P. Nagel, 225-247. (= Orientalia Lovanensia Analecta, 61.) Leuven: Peeters. . 1994b. "Some Reflections on the Egyptian Conjunctive". Divitiae Aegypti: Koptologische und verwandte Studien zu Ehren Martin Krause ed. by C. Fluck et al, 300-314. Wiesbaden: Reichert. . 1995. "Structural Sketches of Middle Welsh Syntax I: the Converter Systems". Studia Celtica 29.127-223. . (Forthcoming, 2000). "Structural Sketches in Middle Welsh Syntax II: Nominal Predications and Presentatives". Studia Celtica 33. Spiegelberg, Wilhelm. 1925. Demotiche Grammatik. Heidelberg: Carl Winter. Steinthai, Haymann & Misteli, Franz. 1893. Charakter istik der hauptsachlichsten Typen des Sprachbaues. Berlin: Dümmler. Stern, Ludwig. 1880. Koptische Grammatik. Leipzig: Weigel Tomlin, Russell S., ed. 1987. Coherence and Grounding in Discourse: Outcome of a Symposium, Eugene, Oregon, June 1984. Amsterdam: John Benjamins. Vennemann (gen. Nierfeld), Theo. 1995. "Zur Erklärung des Baiern-Namens". Sprachwissenschaft 20.380-395. . 1997. "Some West Indo-European words of Uncertain Origin". Hickey & Puppel 1997.879-910. Winand, Jean. 1997. "La négation bn...iwn3 en néo-égyptien". Lingua Aegyptia 5.223-236.

STABILITY,VARIATION AND CHANGE IN WORD ORDER SOME EVIDENCE FROM THE ROMANCE LANGUAGES

ROSANNA SORNICOLA Università di Napoli Federico II

0.

Introduction In this paper, on the basis of a brief analysis of the structures of some Romance languages, a few ideas are presented on the general subject of the relationship between stability, variation and change in WO patterns, and on some of the methodological and theoretical problems it raises. 1.

Criteria for determining stability, variation and change in WO across time A prerequisite to the study of the relationship between continuity and discontinuity in patterns of WO over time is the selection of a homogeneous class of structures (for example, the sentence and those constituents of it which bear GFs, the NP and its head and modifying constituents, etc.) and to examine their WO properties, according to a perspective which we shall call 'longitudinal'. The structure must be defined in such a way as to restrict the range of variations in the patterns arising as a result of its properties. Thus, as far as the sentence is concerned, it is necessary to distinguish structures with transitive V from those with intransitive V; at a more detailed level of analysis it is important to differentiate full and pronominal NPs, and so on. The structure thus determined constitutes the unit of observation over time. A further requirement is the selection of a single language, or at most, a family of genetically related languages, according to a perspective which may be defined as 'microscopic'; this is oriented towards an examination of the properties of WO patterns which takes into account the specific, that is historical, conditions which may have brought them about. 'Longitudinal' and 'microscopic' approaches constitute methodological criteria different from classical approaches of typology. But traditional typological models have limited use in the diachronic domain, even though they may very often be useful, and in certain cases even indispensable,

102

ROSANNA SORNICOLA

especially when working on very remote periods in time. With this in mind, it may be appropriate to make a distinction between a typology which applies its tools directly to the study of historical development and a diachronic study which, with due care and attention, makes supplementary use of typological tools. The longitudinal and microscopic approaches require comparison between the properties of the class of structures chosen and those of other classes of structures to be carried out secondarily. In other words, traditional typological correlations between parameters, such as order of main constituents and types of head-dependency, should follow and not precede the study of a given phenomenon in the temporal dimension. The method just outlined is appropriate for a variety of reasons, some of which regard the very nature of the theoretical problem of stability, variation and change in order patterns. In fact, how to define these concepts and the way they relate to each other is a problem in itself. Here the assumption is made that variation may not necessarily bring about changes; on the contrary, it may well be related to 'stable' states over long time spans. This is true of both alternations in patterns of order and fluctuations in their frequency of occurrence (cf. § 2-3). A further assumption is that the integration of stability, variation and change into a unitary framework requires resorting to historical models of continuity and discontinuity. Another reason for the choice of the longitudinal and microscopic approaches concerns aims of research. Typological models have been set up with very different objectives from those with which studies on stability, variation and change are concerned. The aim in setting up typological models of various types has always been the classification of languages and language families on the basis of certain parameters, while the aim of a study of continuity and discontinuity is the understanding of the possibilities in patterns of order, in that examination of possible alterations allows light to be thrown on the nature of the patterns themselves and of the structure with which they are associated; this is made possible by establishing the conditions and restrictions they are subject to and which may be verified over time spans of greater or lesser duration. In other words, the aim of such a study is the understanding of the dynamics within the structure over time. There is a further difference between studies of continuity vs. discontinuity and most of the typological studies on diachronic development. The latter examine the development over time of correlations between patterns of order of different structures. Such correlations, drawn up on the basis of synchronic surveys carried out on a large scale (that is, of tens or hundreds of languages),

SOME EVIDENCE FROM THE ROMANCE LANGUAGES

103

are both too rigid and too general. They multiply the variables involved and are at a level of abstraction which makes it difficult to ascertain the dynamics of the structure examined. The main interest in research carried out in this vein is the description and explanation of change in WO patterns in a given language, understood as change in a set of correlations between parameters, that is, as the change of a system. On the whole, this constitutes a synchronic view applied sic et simpliciter to the diachronic domain. The former, on the other hand, require specific initial modelling, in order to represent the concepts of 'stability', 'variation' and 'change' and their inter-relationships. They also require consideration of the kind of particular conditions that play a central role in the domain defined as 'history of language'. The time factor also requires adequate modelling. On the methodological level, then, in the study of dynamics, procedures are preferred whereby individual structures are isolated in order to carry out broad analyses of the range of fluctuations in the patterns of order concerned. Closely related to the first is a second reason. A study with the objectives defined above requires what may be called dynamic models (cf. Sornicola 1994, Sornicola forthcoming), for which an integrated analysis of a multiplicity of interacting textual, pragmatic and cultural factors is indispensable. Finally, a not insignificant reason is that typological correlations are in themselves problematic. For example, it has been claimed in several independent accounts that it is not the basic order of constituents which determines constituent order at a lower hierarchical level. Furthermore, there are numerous diachronic studies in which the application of correlational generalizations gives inconsistent or circular results. The unit of observation over time and the set of WO patterns which it may assume, according to the pragmatic and historical conditions of the texts in which they occur, constitute the characteristic defined as flexibility of the structure. Clearly much importance is given in this approach not only to dominant order with respect to basic order, but also to orders which are only minority or marginal. Unlike some current approaches, where languages may be divided between those with pragmatic WO and those with syntactic WO, here it is assumed that there are no languages without flexibility of WO. If anything, languages differ from each other according to differing maps of flexibility in individual structures. This means that in each language the areas of flexibility vary in size and organization. Areas of flexibility are often, but not exclusively, connected to pragmatic functions (PF). The analysis of the flexibility of a given structure over time intervals of a sufficient length allows a better understanding of the

104

ROSANNA SORNICOLA

nature of WO diachronic processes. Note, however, that the length of such intervals is a problem in itself (cf. § 5). In addition to the preceding criteria, there are others which are usual in any research of a historical nature; these are: • precise choice of sources across time; • attention to cultural specificity in the sources. 2.

Flexibility over time in transitive structures in the Romance languages Romance languages provide an interesting testing ground for the study of continuity and discontinuity of WO. The vast amount of documentation allows for a historically based examination of their characteristics of flexibility. Of course, the use of literary and documentary texts requires all the caution that sources written in the past demand. One hardly needs reminding of the continuous influence that medieval Latin and, later, classical Latin have had on texts written in Romance, and more generally European languages. Hilty's observations concerning the language of the Strasbourg Oaths, that "die Zugrunde liegende Spontansprache schimmert nur in verhaltnismässig wenigen Erscheinungen durch die Lateinische Schrifttradition hindurch" (1966:228), could also apply to other Romance texts of successive centuries, although with certain differences. Awareness of the none too close correspondence between spontaneous phenomena and phenomena associated with the written language is particularly important with respect to WO patterns in sentential structures, since for many centuries WO was also significantly shaped by the classical and medieval rhetorical traditions (cf. Scaglione 1972). To the extent that examination of historical documents makes such a claim possible, SVO is consistently the dominant order in medieval and modern prose texts (cf. Sornicola, Barbato, Cesarano & Puolato 1994; Sornicola forthcoming; Salvi 1996 has an interesting analysis of WO patterns of the diachrony of Romance languages with special emphasis on the syntax of pronouns). In contrast to what is often assumed, a greater freedom of constituent order in Italian and Spanish than in French is unsubstantiated for modern phases (cf. § 3), as well as for medieval phases, where the literary languages show a significant degree of homogeneity in WO patterns (cf. Sornicola, Barbato, Cesarano & Puolato 1994). OVS and OSV types are minority orders with a higher frequency in the prose of medieval Romance texts. Nonetheless, they have a comparatively low frequency: the frequency of O in PI (sentence first position) or in P2 (sentence second position)is below 15% in XHIth and XlVth century texts. OVS order falls within the more general rule of S inversion when a constituent occupies

SOME EVIDENCE FROM THE ROMANCE LANGUAGES

105

the pre-verbal position (cf. Foulet 1919:387ff.); the second, more rare, occurs almost always when S is a pronoun (cf. also Price 1961: 29, 43, 45 for Froissart). These orders have one or other of two possible functions: marking the orientation of the structure towards the left context (more rarely right: cf. Ruelle 1966) or marked focalization of a constituent. In the first case O is a completely or partially thematic element (or, alternatively, anticipatory), as may be seen from the fact that in the majority of cases the constituent which realizes such a function is an anaphoric demonstrative (cf. here, examples (15)): (1)

"Hé! Douce terre pleinne de toutes beneürtez et en qui mes esperis et ma vie remaint outreement, beneoite soies tu de la bouche de celui qu'en apele Jhesucrist..". Itex paroles dist Lancelos quant il parti del roiaume de Logres (MRA, §124, 1)

(2)

...il li acreanta come rois que, ja plus tost la Pasque ne seroit passee, qu'il iroit a ost banie seur Lancelot et tant se traveilleroit... qu'il abatroit les fortereces de Banoïc et de Gaunes en tel maniere qu'il ne leroit en mur pierre seur autre. Ceste promesse fist li rois a monseigneur Gauvain (MRA, § 128, 13-14)

(3)

Dont j'en parlay a deux, qui m'ont compté qui furent cinq qui, par bonne compaignie devisant des adventures de celle cave, tous d'un vouloir entreprindrent d'aler jusques es portes de metal qui jour et nuyt batent... Et ceste chose avoient ilz entreprins, comme jeunesse fait souventeffois entreprendre les gens oyseux (PRS, 80-82, C)

(4)

ca pero non se mesauan nin se rascauan, nin dauan bozes, a todos semejaua quel quebraran por los coraçones, dando sospiros e llorando muy fuerte e poniendo las manos sobre los oios. E eso mesmo fazia el infante Roboan e toda la su gente (Zifar, 390)

(5)

Li vecchi l'insegnaro: Ragunerai il populo tuo, e con dolci parole dirai che tu li ami sì come te medesimo e ch'elli sono la corona tua, e che, se tuo padre fu loro aspro, che tu sarai loro umile e benigno... Queste parole l'insegnaro i savi vecchi del regno (Novellino, VIII, 74)

Note that in (3) the pattern O + Aux + S + Past Part occurs, which is typical of VP structures with Aux + Past Part. O in pre-verbal position is, although more rarely, an element introducing a TOPIC [+NEW] (cf. examples (6-7); for further examples see Ruelle 1966): (6)

Pour ce que, quant venoit Ie vendredi, après la mienuyt, sa compaigne se levoit d'emprès lui et s'en aloit à la royne, et toutes les autres de leans aussi. Et la estoient, en chambres et en autres lieux ad ce ordonnez, en estat de couleuvres et de serpens toutes ensemble; et ainsi estoient jusques après la mienuit du samedi, que chascune retournoit a sa compaignie; etl'endemain, sembloit estre plus belle que jamais n'avoit esté, Car elles jamais n'envieillissent, ne scevent que douleur est. Des vestemens ont il a leurs vouloirs; de viandes est chascun servi a l'appetit de son cuer; richesses ont ilz a planté (PRS, 97, C, r. 6-8)

106

ROSANNA SORNICOLA

(7)

Si dirent les gens que c'estoit une voix de paradis de la Sibille. Mais, quant a moy, je n'en croy riens; ains croy que feussent mes chevaulx qui au pié du mont estoient, combien qu'ilz feussent moult bas et loings de moy. Et nulle autre chose je ne vis né ne sçay, fors seulement ce que les gens du païs et de ladicte ville m'en ont dit (PRS, 80, , . 4-5)

Examples in which  in pre-verbal position has a PF of marked FOCUS (sometimes contrastive) are: (8)

Quant la nouvele fu espandue par la cité et l'en sot que li rois Artus fu ocis et tuit cil qui  lui estoient alé, grant duel firent et li povre et li riche por le roi Artu (MRA, § 136, r. 16-17)

(9)

E [el Cauallero Zifar] boluiose de rostro contra el señor de la hueste e puso la lança so el sobaco e dixo asy: "Cauallero defendetvos". "E quien eres tu", dixo el señor de la hueste, "que atanto te atreues?". " srtas", dixo el Cauallero Zifar, "agora lo veredes". E finco las espuelas al cauallo e fuelo ferir, e diole vna grant lançada por el costado quel paso las guarni iones, e metiose por el costado la lança bien dos palmos, e dio con el en tierra. La su gente commo yuan veniendo, yuan feriendo sobre el e trabajauanse mucho de lo poner en el cauallo. E entretanto el Cauallero Zifar tornose con su gente e pasaron el alcantariella en saluo. E mas mer ed fizo Dios al cavallero Zifar e a su gente; quel fijo del señor de la hueste, quando vio que el padre era derribado, finco las espuelas al cauallo e fue ferir vn cauallero de los de la villa; peroque lo non enpesçio, e metiose en la espesura de la gente e presieronle, e asy lo leuaron preso a la villa {Zifar, 115)

(10)

io non so quello che de' vostri pensieri voi v'intendete di fare: li miei lasciai io dentro dalla porta della città (Decameron, I, Introd., 28, § 93)

In main clauses, the SOV type is highly marginal throughout the entire history of the Romance languages (as to French, cf. Foulet 1919:316-318; Brunot 1966: 264, 268, 499; Brunot 1967:481-482; Marchello-Nizia 1979:331; Buridant 1987). Very rare in Xlllth and XlVth century texts, it is presumably a Latinism, as shown by its occurrence in authors primarily influenced by the classical tradition, such as Boccaccio or Antoine De la Sale: (11)

Per la qual cosa alla buona donna con cui dimorava interamente ogni suo accidente aperse e le disse sé disiderare d'andare a Tunisi, acciò che gli occhi saziasse di ciò che gli orecchi con le ricevute voci fatti gli aveano disiderosi. La quale il suo desiderio le lodò molto (Decameron, V, 2, § 38)

(12)

E dopo alquanti dì partitosi Melisso da Giosefo e tornato a casa sua, a alcun, che savio uomo era, disse ciò che da Salamone avuto avea; il quale gli disse: "Niuno più vero consiglio né migliore ti potea dare" (Decameron, IX, 9, § 34)

SOME EVIDENCE FROM THE ROMANCE LANGUAGES

107

(13)

La mattina dunque seguente, in su l'ora del mangiare, primieramente i quatro fratelli di Tedaldo, così vestiti di nero come erano, con alquanti loro amici vennero a casa Aldobrandino, che gli attendeva; e quivi, davanti a tutti coloro che a fare lor compagnia erano stati da Aldobrandino invitati, gittate l'armi in terra, nelle mani d'Aldobrandino si rimisero, perdonanza domandando di ciò che contro a lui avevano adoperato. Aldobrandino lagrimando pietosamente gli ricevette e tutti basciandogli in bocca, con poche parole spacciandosi, ogni ingiuria ricevuta rimise (Decameron, III, 7, § 86)

(14)

Lors, au saillir de l'ostel, veissiez chevaulz saillir et contourner, courre, recourre et en l'air tourner, estincelles de feu par l'air voller, crier, huer, ou chascun acouroit, que oncques de telle chose plus joyeuse ne fut. Et ainssy allerent, qu'ilz furent en la grant cour de Saint Pol. Lors chacun sa joye renforça; car bien savoient que le roy fust esveillé (PJS, 325)

(Cf. also Martin and Wilmet 1980:280). Most of the examples gathered show the SOV pattern in which  is a backgrounded constituent (O is almost always a [+GIVEN] constituent), but cases such as (12) may occur, where O, although [+GIVEN], is in FOCUS. It is not possible here to discuss the complex question of the V-2 nature of medieval Romance languages, which according to some would explain the rarity of the SOV type (while the SVO and OVS types fall completely within the rule originally described by Thurneysen). I merely observe that with regard to this I share the scepticism expressed by Marouzeau (1938:94) and Herman (1954:359,379). V in initial position with respect to S and  is a highly minority option in Romance texts of the Middle Ages, as it is in the modern languages and in SVO languages in general. VSO and VOS orders are very rare in French and Italian texts, and more generally in all the historical documentation handed down to us (the type (Adv) + VS occurs with considerable frequency with intransitive verbs: for an examination cf. Sornicola, Barbato, Cesarano & Puolato 1994). As is well known, in Old French initial V is found only in rhythmic prose or verse in the oldest documentation, while the type of structure in which V is preceded by an adverb is much more common. In any case, structures such as the following may still be found in Froissart: (15)

Enssi estoient adont li seigneur et les terres en le duché d'Acquittainne en grant variement et gueriiet de leurs voisins. Si ne savoient li pluiseur bonnement que faire. Or entendí messires Robers Canolles les nouvelles coumment li Franchois faisoient trefforte guerre au prinche (Chroniques, I, 1, § 757, r. 5)

108

ROSANNA SORNICOLA

(16)

Là leur remonstra il moult bellement et sagement eoumment en toutte honneur et en pais à son pooir il lez avoit maintenus et gouvrenés.... Tout chil baron et chevalier dessus nonmet li jurierent et se obligierent par foy et par sierement que ossi feraient il. Adont prist li prinches moult doucement congiet à yaux et se parti assés tost depuis de Bourdiaux {Chroniques, I, 1, § 816, r. 23-24)

In (15) and (16) the structure Adv + V + S +  signals a turning point in the narration; in the following example instead the structure has the function of expressing the thematic continuity of V and S, while  is in FOCUS: (17)

Se le rechupt li prinches à grant joie et le fist tantost cappittainne et souverain de touttes les Compaingnes qui estoient nouvellement venues de Normendie. Si l'envoiea li dis prinches et tourtes ces gens d'armes en le terre le comte d'Ermignac et le seigneur de Labrech, pour ardoir et destruire leur pays et faire y guerre car chil li estoient grant ennemit. Encorres envoiea li prinches son frere monsigneur Ammon et le jone comte de Pennebrucq à tout grant fuison de gens d'armes en le comté de Pieregorch {Chroniques, I, 3, § 750, r. 11-13)

(For inversion after the adverb in Old French cf. Foulet 1919:307ff; Brunot 1966: 268, 499; Herman 1954; for inversion after et cf. Bergh 1952, Price 1961:46: this is a stylistic device of Middle French.) The VSO type often appears in Italian prose of the XIIIth and XIVth centuries as well, with a stylistic value of 'turning point in the narration'; as in French, VS inversion is less frequent with transitive than with intransitives: (18)

Uno signore di Grecia, lo quale possedea grandissimo reame, e avea nome Aulix, avea uno suo giovane figliuolo, il quale facea nodrire e insegnarli le sette liberali arti, e faceali insegnare vita morale, cioè di be' costumi. Un giorno tolse questo re molto oro e diello a questo suo figliuolo {Novellino, VIII, 75)

(19)

A questo gran piacere di Pericone e di lei, non essendo la fortuna contenta d'averla di moglie d'un re fatta divenire amica d'un castellano, le si parò davanti più crudele amistà. Avea Pericone un fratello d'età di venticinque anni, bello e fresco come una rosa, il cui nome era Marato {Decameron, II, 7, § 32)

(20)

Il che come Giosefo ebbe udito, così si ricordò delle parole di Salamone e disse verso Melisso: "...". Quindi, dopo alquanti dì divenuti a Antiocia, ritenne Giosefo Melisso seco a riposarsi alcuni dì {Decameron, IX, 9, § 23)

But it is in Old Spanish texts that the patterns (X)VSO, (X)VOS seem to have the widest diffusion, although as minority orders. These structures sometimes have the usual textual function of turning point in the narration, as shown by the fact that they have the pragmatic property of being "all-inFOCUS" (note that both S and  are [+GIVEN] in the Italian and Spanish examples quoted):

SOME EVIDENCE FROM THE ROMANCE LANGUAGES

109

(21)

E fueron para alla, e en las primeras non fallaron recabdo ninguno; mas vno que estaua mas en cabo que todas, dixo que poco tienpo auia que leuaron mercaderos del tres quintales de tales poluos commo ellos dezian. E preguntaronle sy fincara alguna cosa ende, e el dixo que non sabia, e fizo commo que escrudiñaua sus arcas e sus sacos, e mostroles aquellos pocos de poluos quel auia dado el cauallero. E demandaronle que por quanto gelos daría, e el dixo que non menos de dies doblas. E el cauallero dixo que gelas diesen por ello, syquier por fazer la proeua, e dieronle dies doblas, e tomo los poluos el mayordomo e leuolos para el rey. (Zifar, 404)

(22)

e [la infante] dixole asy: "... E lo que vos he ha dezir, commoquier que lo digo con grant verguença, es esto: que sy el vuestro casamiento e el mio quesisese Dios allegar, que me plazeria mucho. E non he mas a dezir, ca a orne de buen entendimiento pocas palabras cunplen." Desy abaxo los ojos la infante e pusolos en tierra e non lo pudo catar con grant verguença que ouo de lo que auia dicho. (Zifar, 389)

More rarely VSO order is associated with a function of focalization of S: (23)

E estonçes el enperador mando quel vestiesen de otros paños muy nobles, e çiñole el espada e caualgaron e fueronse para casa del enperador, e el infante trayendo el espada desnuya en la vna mano e el pendon en la otra mano con la lança, e la guirnalda en la cabeça. E desque se asentaron a la mesa, tenia vn cauallero delante el espada desnuya, e el otro la lança con el pendon, fasta que comieron (Zifar, 399)

Elsewhere VSO order seems to occur when an adverbial constituent is in FOCUS in P1 (but the two examples differ in that in the first the prepositional phrase has anaphoric function, in the second cataphoric): (24)

E porque es el alma espritual e el cuerpo elemental, por eso ha el alma uirtud de guiar el cuerpo (Zifar, 270)

(25)

Ca Dios non fizo el orne commo las otras animalias mudas, a quien non dio razon nin entendimiento, e non saben nin entienden que fazen... E por eso dio Dios al orne entendimiento e razon, por que se podiese guardar del mal e fazer bien (Zifar, 269)

The frequently observed V-initial structures in Spanish have been ascribed to the influence of Arabic; but this hypothesis does not appear to have been sufficiently substantiated (cf. Lapesa 1980: 153-154 and fn 37). If, however, external influences are sought, Biblical Latin cannot be altogether ruled out: it is well-known, for example, that in a text such as the Peregrinatio Aegeriae there is a high incidence of V-initial structures, both with intransitive and transitive verbs, and that this reflects contemporary cultural models of the Latin of Biblical translations.

110

3.

ROSANNA SORNICOLA

Discontin uity or minor temporary fluctuations ? In trying to decide whether or not WO patterns of the Romance languages conform to a diachronic model of continuity or discontinuity, account should be taken of an inherent limitation of historical linguistics, that is, that the properties established concern texts and not necessarily languages. With respect to the picture drawn in § 2 modern Romance languages appear to have minor differences. SVO order has a very high frequency in the various contemporary languages, both written and spoken. This applies not only to French, but also to Italian and Spanish (cf. Sornicola, Barbato, Cesarano & Puolato 1994; Sornicola forthcoming). For example, in contemporary Italian the average frequency of such a pattern in a textually and stylistically differentiated corpus is 77.5%; OVS order, always marked and contrastive, has an average frequency below 10%, V initial orders have frequencies ranging between 1% and 4% (cf. Sornicola 1994). SOV and VSO patterns, which in medieval texts appear to be artificial and typical of the educated and literary language, are more or less non existent in both written and spoken registers of the modern languages. Where they occur, they are associated with prosodic structures in which the non in situ constituent has contrastive stress (cf. Sornicola forthcoming). The VSO pattern in particular does not have the function of turning point in the narrative, so typical in medieval texts. Although they have not entirely died out, OVS and OSV patterns are confined to pragmatically restricted contexts. They are only possible as marked orders in contemporary Italian and Spanish (O is a marked FOCUS, with relevant prosodic properties). In French, the OVS type has now disappeared, while the OSV type (with pronominal S), which was still frequent as a minority order in the XVIth century, is confined to a few marked expressions (cf. Ruelle 1966). In any case, these types do not these days have the function of using  to signal thematic continuity with the left context, as in examples (1-5). In French, this function was still found associated with the OVS pattern in Rabelais (cf. Ruelle 1966:310-311). Structures with left-dislocation of  generally appear with this function today, such as It. Maria non la posso soffrire "Maria I can't stand her", Fr. Cet élève je l'aime bien "This pupil I like very much". This is not a case of dramatic discontinuity, because this syntactic type is present uninterruptedly in Latin and Romance texts and in other IndoEuropean languages (cf. Sornicola 1984; Sornicola 1988-89). Generally speaking, on the structural level, the two most conspicuous differences of WO properties, in comparison with those found in texts of the XIIIth and XIVth centuries, appear to be: the fact that all the non-SVO orders

SOME EVIDENCE FROM THE ROMANCE LANGUAGES

111

have a marked FOCUS; absence of S-V inversion when  occupies pre-verbal position, a phenomenon, frequently observed and discussed in the Romance linguistics literature, which falls within the more general 'loss' of S-V inversion. But again, one may well ask to what extent these differences justify the hypothesis of a real 'change'. In both cases, in fact, we are merely conjecturing on a linguistic situation which to a large extent is accessible to us through the distortion of literary or documentary prose. In reality, it may well be claimed that not only the dominant pattern but also the flexibility of the structure have remained substantially unaltered in the various Romance languages. Of course, oscillations in the frequency of each of the patterns in corpora of different periods, and in certain cases in PFs or textual functions, may also be noted. They seem, however, to be connected to the influence of external factors rather than to spontaneous linguistic development. This is especially evident in the diachronic development of French (cf. Sornicola 1995). The apparent greater freedom of WO patterns in medieval Romance texts may therefore be only an optical illusion brought about by their literary character. Something which should be given due consideration is the very low frequency of V initial patterns in transitive structures, in all the available documentation (except for the stylistically determined type exemplified in (1520)). This seems to be the result of an important structural factor, since it allows us to make some more general hypotheses on the role of internal factors in the dynamics of the structure. These seem to operate in a passive rather than active way: they determine the limits and conditions beyond which flexibility cannot be pushed, that is conditions which prevent change, rather than conditions which actively favour it. As such, they are related to stability rather than to change. PFs for their part characterize flexibility of the structure, rather than processes of change. For example, in the Romance languages topicalization phenomena have always been a particular option, never grammaticalized in a dominant order, a phenomenon which is also found in other groups of the IndoEuropean family (for example in Celtic languages, cf. Poppe's contribution to this volume) and in other language families (for example in Hamito-Semitic, cf. Loprieno's and Shisha-Halevy's contributions to this volume). It is doubtful, or at any rate difficult to demonstrate, that the distribution of PFs can determine processes of change in WO. I would like to emphasize here the importance of the concept of 'stability' as opposed to 'change'. In the study of WO patterns according to the method

112

ROSANNA SORNICOLA

defined as 'longitudinal', stability rather than change emerges as the dominant structural property in conditions where external factors such as contact or the influence of prestigious patterns play no role. Although at first sight this may seem strange, flexibility seems to be directly connected to stability rather than to change. The more flexible a structure is, the more stable it is. In a certain sense, therefore, one could say that flexibility is what guarantees stability. 4.

How old is the SVO trend? The Romance historical linguist might perhaps be satisfied with formulating hypotheses on WO based on periods of time stretching from the earliest records of the Romance languages up to today, but not so the typologist. These two types of linguist do not work within the same time scale. The time scale of the typologist has a much longer span than that of the historical linguist. The former could object that a millennium or so is not a sufficient period to appreciate 'change' and that the basic stability which emerges through patterns of order in transitive structures in the Romance languages is in a certain sense foreseeable. From this point of view, the crucial point in a course of change may emerge only by taking into account a longer time span, which includes the Latin period, from the earliest records, and even further back to the earliest documentation of the Indo-European languages. Consideration of the difference in time scale is important in defining the problem of stability, variation and change, but it seems to be a premise rather than a solution. In fact, in the vast bibliography on what today we call the basic order of constituents in the sentence in Latin and in the ancient documentation of the Indo-European languages consensus is far from reached on the basic/dominant order of the various linguistic stages. This is partly due to the fact that the techniques of reconstructive analysis (to which classical typological ones belong) seem to give different results from philological techniques. As to Latin, it is true that the results of Linde's (1923) study suggest a sort of steady increase over time in SVO order compared to SOV, at least from the first century AD up to late Latin records. A fact which seems to me to have a certain interest, however, is that the majority of research carried out with essentially philological techniques questions not only the predominance of SOV order in Latin but also whether it is indeed the basic one (cf. Adams 1976, 1991; Panhuis 1985; Pinkster 1991). Once again the division between literary prose and spoken registers seems to be crucial. Once again we must bear in mind that spoken Latin is only indirectly reflected in written texts, and in a not unproblematic way (cf. Adams' observations on the letters of Claudius

SOME EVIDENCE FROM THE ROMANCE LANGUAGES

113

Terentianus [Adams 1977:66ff.]). Recall also Elise Richter's study (Richter 1903): although framed within a methodological perspective belonging to the beginning of the century, it contains a hypothesis which, in short, is not too far removed from that held by Adams in works of the late Seventies (for a critique of Richter from the Romance linguistics perspective cf. Herman 1954). The same problems arise in an even more complex manner when we try to widen the perspective further back in time, to documentation of the early IndoEuropean languages. Here, too, we find a sort of polarisation between the results found by researchers using mainly typological reconstructive techniques and the results of those who make use of historical techniques. Of course, the dominance of SOV order in surviving records has been given considerable attention least from the studies of Delbrück onwards. But is it enough to consider this order as the starting point of a linear development which runs from the ancient Indo-European languages up to modern ones? Researchers such as Gonda (cf. Gonda 1952) or Watkins (cf. Watkins 1964; Watkins 1976), who place more emphasis on stylistic variation in texts have, justifiably in my opinion, called attention to WO flexibility in ancient Indo-European languages since the earliest documentation.

REFERENCES PRIMARY SOURCES Chroniques = Jean Froissait, Chroniques. Livre I édité par G.T. Diller (tome I - IV), 4 vols. 1991-1993. Genève: Droz. Decameron = Giovanni Boccaccio, Decameron a cura di V. Branca. 1985. Milano: Mondadori. MRA = La mort le roi Artu édité par J. Frappier. 1954. Genève: Droz. Novellino = Novellino e conti del Duecento a cura di S. Lo Nigro. 1963. Torino: UTET. PJS = Antoine de la Sale, Le petit Jehan de Saintré, texte nouveau publié d'après le manuscrit de l'Auteur avec des variantes et une introduction par P.Champion et F. Desonay. 1926. Paris: Editions du Trianon. PRS = Antoine de la Sale, Le paradis de la royne Sibille in Antoine de la Sale, Oeuvres complètes. Tome I: La Salade. Édition critique par F. Desonay. 1935, 63114. Paris: Les belles Lettres. Zifar = Libro del Caballero Zifar. Edición de . Gonzales. 1983. Madrid: Catedra.

114

ROSANNA SORNICOLA

SECONDARY SOURCES Adams, Jim N. 1976. "A Typological Approach to Latin Word Order". Indogermanische Forschungen 81. 70-99. . 1977. The Vulgar Latin of the Letters of Claudius Terentianus. Manchester: Manchester University Press. . 1991. "Some Neglected Evidence for Latin Habeo with Infinitive: the Order of the Constituents". Transactions of the Philological Society 89:2.131-196. Bergh, Lars. 1952. "Quelques Réflexions sur l'inversion après la conjonction et en ancien et en moyen français". Mélanges de Philologie Romane offerts à M. Karl Michaëlsson par ses amis et ses élèves, 43-55. Göteborg: Bergendahls Boktryckeri. Brunot, Fernand. 1966. Histoire de la langue française des origines a nos jours. Tome I: de Tèpoqe latine à la Renaissance. Paris: Armand Colin. . 1967. Histoire de la langue française des origines a nos jours, tome II: Le XVIe siècle. Paris: Armand Colin. Buridant, Claude. 1987. "L'ancien français à la lumière de la typologie des langues: les résidus de l'ordre  V en ancien français et leur effacement en moyen français". Romania 108.20-65. Foulet, Lucien. 1919. [1930]. Petite syntaxe de l'ancien français. Paris: Champion. Gonda, Jan. 1952. Remarques sur la place du verbe dans la phrase active et moyenne en langue sanscrite. Utrecht: Oosthoek's Uitgevers-Mij. Herman, József. 1954. "Recherches sur l'ordre des mots dans les plus anciens textes français en prose". Acta Linguistica Academiae Scientiarum Hungaricae 4.69-94, 351-382. Hilty, Gerold. 1966. "Die Romani sierungen in den Strassburger Eiden". Vox Romanica 25.227-235. Lapesa, Rafael. 1980. Historia de la lengua española. Madrid: Gredos. Linde, P. 1923. "Die Stellung des Verbs in der lateinischen Prosa". Gioita 12.153-178. Marchello-Nizia, Christiane. 1979. Histoire de la langue française aux XIVe et Xve siècles. Paris: Bordas. Marouzeau, Jules. 1938. L'ordre des mots dans la phrase latine, II: Le verbe, Paris: Les Belles Lettres. Martin, Robert & Wilmet, Marc. 1980. Manuel du français du moyen age, 2: syntaxe du moyen français. Bordeaux: Sobodi. Panhuis, Dirk G. 1985. "Is Latin an S  V language? A Diachronic Perspective". Indogermanische Forschungen 89.140-159. Pinkster, Harm. 1991. "Evidence for S V  in Latin?". Latin and the Romance Languages in the Early Middle Age ed. by R. Wright, 69-82. London & New York: Routledge. Price, Glanville. 1961. "Aspects de l'ordre des mots dans les "Croniques" de Froissart". Zeitschrift fur romanische Philologie 77.15-48. Richter, Elise. 1903. Zur Entwicklung der romanischen Wortstellung aus der Lateinischen. Halle: Niemeyer.

SOME EVIDENCE FROM THE ROMANCE LANGUAGES

115

Ruelle, Pierre. 1966. "L'ordre complément direct - sujet - verbe dans la proposition énonciative indépendente". Mélanges de grammaire française offerts à M. Maurice Grevisse pour le trentième anniversaire du Bon Usage, 307-322. Gembloux: Duculot. Salvi, Gianpaolo. 1996. From Latin Weak Pronouns to Romance Clitics. Budapest: Institutum Linguisticum Academiae Scientiarum Hungaricae. Scaglione, Aldo. 1972. The Classical Theory of Composition from its Origins to the present. A Historical Survey. Chapel Hill: University of North Carolina Press. Sornicola, Rosanna. 1984. "Indeterminate Relations and the Notion of 'QuasiGovernment'". Folia Linguistica 18:3-4.379-408. . 1988-89. "Il relativo in irlandese antico. Una riconsiderazione". Lndogermanische Forschungen 93.124-167; 94.234-271. . 1994. "On Word-Order Variability: a Study from a Corpus of Italian". Lingua e Stile 29:1.25-57. . 1995. "Mutamenti di prospettiva culturale nelle lingue europee moderne: l'influenza del latino sulla sintassi". Kulturwandel im Spiegel des Sprachwandels ed. by K.-E. Lönne hrsg., 41-57. Tübingen & Basel: Francke. . forthcoming. "The Order of Main Constituents in a Pragmatic Perspective". The EUROTYP Survey of the Typology of the Languages of Europe ed. by G. Bernini vol. 9: Pragmatic Organization of Discourse. Berlin: Mouton-de Gruyter. Sornicola, R., Barbato, M., Cesarano, M. & Puolato, D. 1994. "Word-order in Old Romance Languages", paper presented to the meeting in Naples of the research group on "Pragmatic Organization of Discourse" of European Science Foundation's EUROTYP Project (November 11-13, 1994). Watkins, Calvert. 1964. "Preliminaries to the Reconstruction of Indo-European Sentence Structure". Proceedings of the IX International Congress of Linguists ed. by H. G. Lunt, 1035-1045. The Hague: Mouton. . 1976. "Towards Proto-Indo-European Syntax: Problems and PseudoProblems". PCLSM 12, Papers from the parasession on Diachronic Syntax (April 22, 1976), 305-326.

PART III REANALYSIS, GRAMMATICALIZATION AND CHANGE

REANALYSIS IN WORD ORDER STABILITY AND CHANGE

JAN TERJE FAARLUND University of Oslo 0. Introduction In this chapter I will examine two instances of syntactic change in North Germanic, both having to do with linear order and syntactic structure. I will claim that they can both be understood in the context of discourse functions and functional requirements on word order. The basic principle underlying discourse functional organization of the sentence, is that given information precedes new information, topic precedes comment, theme precedes rheme (these statements may for the present purpose be considered terminological variants of each other), and simple elements precede complex ones. One of the changes to be dealt with here involves word order change, bringing the new linear order more in accordance with general principles of information structure than the old order. The other change involves a structural change making it possible to maintain a functionally motivated linear order, which would otherwise be unavailable as a result of other syntactic changes. These two changes can both be interpreted as involving syntactic reanalysis, that is the assignment of a new structural description to an unaltered surface string. In § 1 of this chapter, I will offer a closer description of reanalysis and discuss certain basic concepts and tenets. The first change to be treated, the change from object-verb to verb-object order in North Germanic, is treated in § 2. In § 3, I discuss the introduction of ergative constructions in modern Scandinavian as a case of reanalysis that made it possible to maintain a desired word order. In § 4, I discuss a theoretical problem with reanalysis and its empirical basis, and propose a solution to it in terms of modularity.

This paper is dedicated to my good friend and colleague, Ekkehard König, on the occasion of his 60th birthday.

120

JAN TERJE FAARLUND

1. Reanalysis Although reanalysis does not in itself involve any change in the surface manifestation, it may eventually lead to such changes. And this is precisely what diachronic syntax is concerned with. "Reanalysis is a mechanism which changes the underlying structure of a syntactic pattern and which does not involve any modification of its surface manifestation" (Harris & Campbell 1995:50). In this quotation, the terms 'underlying structure' and 'surface manifestation' are used in a rather noncommittal sense. It can of course be understood as denoting levels in a multistratal theory, as deep structure and surface structure (or phonological form) in generative grammar. More generally, surface manifestation may be taken to mean the surface string of morphemes, while underlying structure is the structural description assigned to the string. The theoretical notion of reanalysis is based on the assumption that a given surface manifestation allows more than one syntactic analysis. Reanalysis takes place whenever a new generation of speakers assigns a structure to an input string, which is different from the structure assigned by the grammar of the previous generation. It follows from this that the locus of reanalysis is language acquisition, and that this type of syntactic change is implemented by first language learners. Reanalysis takes place if the choice made by a significant number of speakers of the new generation is different from the underlying structure which is part of the internalized grammar of the older generation. 2. Word order change Those of the world's languages that exhibit a preferred or unmarked word order where the verb and its object are adjacent can be divided into OV lan­ guages and VO languages. In the former, the verb follows its arguments, and in the more consistent members of the group, a governing word generally follows its complement. In VO languages it is the other way around. (For discussions of these patterns and their implications and interpretations in various frameworks, see, among others, Greenberg 1963:76-80; Vennemann 1975; Hawkins 1979, 1983; Chomsky 1986; Lightfoot 1991:42-43; Haegeman 1994:13-16.) In many languages the two orders of verb and complement are both used frequently, sometimes to the extent that it may be difficult to determine which order is the basic or underlying one. In other languages, the basic order may be

REANALYSIS IN WORD ORDER STABILITY AND CHANGE

121

violated under certain circumstances. Thus in German, which is basically OV, (la)1, an object follows the verb if it is a clause, as in (lb). (1) a.

b.

Peter hat ein Buch Peter has a book "Peter has bought a book"

gekauft bought

als Peter versuchte, sein Haus when Peter tried his house "when Peter tried to sell his house!"

zu to

verkaufen sell

In French, a VO language, an unstressed pronominal object precedes the verb: (2) a. b.

Pierre "Peter

a acheté has bought

un livre a book"

Pierre V a acheté Peter it has bought "Peter has bought it"

Generally, it seems that deviations from the basic and unmarked order in many languages is conditioned by certain pragmatic factors, such as weight or discourse function. Direct objects, consisting of an entire phrase, are more variable in terms of complexity than the verb. They can be anything from a clitic to a subordinate sentence. It is also more common for the object to carry new information. Semantically, too, verbs are less variable, and less 'heavy'. Typical and frequent verbs in transitive sentences tend to express abstract relations such as "have", "get", "receive", "own", operations such as "buy", "sell", "lend", "borrow", non-descript events such as "find", "discover", "observe", "hide", or very general, trivial activities such as "eat", "drink", "read" etc. With such verbs, the new information is carried by the direct object rather than by the verb. An exchange such as (3) is therefore much more meaningful than the one in (4). (3)

- What kind of food did you find? - I found some fruit.

(4)

- What did you do to the fruit? - #I found it.

When a German main clause contains a main verb alone, this verb is moved to second position, and therefore it automatically precedes its complement (except when the latter is topicalized). Peter kaufte gestern ein Buch Peter bought yesterday a book "Peter bought a book yesterday"

122

JAN TERJE FAARLUND

In several language families and groups, a change from OV to VO can be documented or reconstructed (Leumann et al 1965; Lehmann 1974; Hopper 1975; Hyman 1975; Braunmüller 1982; Faarlund 1985,1990).2 One such group is North Germanic (along with many other Indo-European languages). The modern representatives of North Germanic, the Nordic languages, are all VO languages. The oldest records in North Germanic, the Ancient Nordic runic inscriptions, have mostly OV order, cf. (5), but very soon we find inscriptions with VO order, (6), as described in Faarlund (forthcoming). (The placename and the figure after each example indicate the place of the find and the assumed year AD of the inscription.) (5) a.

godagastiz runo faihido Godagasti rune painted "Godagasti painted the rune" (Einang 350-400)

b.

ekhlewagastiz holtijaz horna tawido I Hlewagasti Holt-ing horn made "I, Hlewagasti of Holt, made the horn" (Gallehus 400)

c.

hagiradaz Hagirada "Hagirada

(6) a.

i tawide in made inscribed (it)" (Garbølle 400)

ekhrazaz satido staina ana... I Hraza set stone on "I, Hraza, set the stone on ..." (Rö 400)

b.

ek hagustadaz hlaaiwido magu minino I Hagustada buried son mine "I, Hagustada, buried my son" (Kjølevik 450)

c.

eka sigimArAz AfsAkA raisido-ka stAinA I Sigimara off-guilt raised-I stone "I, Sigimara, the acquitted one, raised the stone" (Ellestad 550-600)

In Old Norse (IXth - XVth century) the VO order is the unmarked one. On the basis of other branches of Indo-European, OV can also be reconstructed for Proto-Indo-European (Lehmann 1974). As shown in Faarlund (1985:366-7), any change from OV to VO has to pass through a stage where both are possible. There are thus two separate events to be accounted for: the appearance of the new form (in this case VO),

The opposite change, on the other hand, seems to be almost unattested, except in clear contact situations, where a novel OV pattern can be ascribed to borrowing.

REANALYSIS IN WORD ORDER STABILITY AND CHANGE

123

and the disappearance of the old one (OV). The course of events may plausibly be described as follows: Some OV languages allow a phrase to be extraposed to the right of the verb if it is heavy or focussed. This is an expressive and marked form. As with other expressive linguistic forms, such as lexical items, the conditions on its use will be weakened over time, consider how the word terribly has come to be used with a very neutral meaning in expressions such as terribly nice. As the conditions on the use of the VO order is relaxed, the frequency increases to the point where new learners of the language interpret VO as the basic form. From then on, the old OV structure is doomed, since there will be little motivation for moving the verb behind the object. The mechanism by which this change happens is reanalysis, taking place at the point where the new form has reached a certain frequency. At the time when the VO order in Nordic was conditioned by certain functional factors, it was a result of movement, as illustrated in (7a), where the preverbal NP (the object) is moved to a position behind the VP. This structure was then reanalyzed as in (7b), where the NP is generated behind the V in the deep structure. The sentences in (6) may therefore have been generated either by (7a) or by (7b). After reanalysis has taken place, there is no possibility for a movement in the opposite direction, and the OV order has disappeared completely in the Nordic languages, except as relic forms in certain fixed expressions. (7) a. v'[ti V] NPi b. v'[V NP]

In French and other Romance languages, however, the OV order was maintained when the object was a pronoun. This, then, would require another reanalysis. The process indicated in (7) also took place in the Romance languages, which are now all VO. However, a pronominal object is generally a proclitic on the verb.3 In order to maintain this order after the reanalysis, another reanalysis took place. The order [pr V] was no longer analyzed as underlying OV, but as VO with movement, as in (8). (8)

v'[ NP pr i v'[Vt i ]]

The details about the placement of pronominal clitics vary somewhat among the Romance languages. We are here only concerned with the general picture as represented by French.

124

JAN TERJE FAARLUND

Note that this is reanalysis of a different kind: this time a movement rule is introduced and a new structure created in order to maintain a desirable order. In the former case, a movement rule was removed as a result of reanalysis, thereby creating a desirable underlying order. In the next section, we will look at a case of reanalysis of the latter kind in Nordic, one that has had fundamental effects on the syntax of the modern Scandinavian languages. 3. Structural change 3.1 The Old Norse subject Old Norse, the ancestor of Modern Norwegian and Icelandic, has a rather free word order, case marking of nominal categories, and verb agreement. The subject of the sentence can be defined as a nominative NP. Old Norse being a verb second language, the unmarked position of the subject is either first or following the finite verb, as in (9a-b). (The subject is in bold face.) (9) a. Hálfdan hvítbeinn var Halfdan Whiteleg was "H. W. was a powerful king"

konungr king

ríkr powerful

b. pátók Loki upp stein then took Loki up stone "Then Loki picked up a stone"

However, the NP in the nominative case carrying the subject role may come at the end of the sentence, in extraposition (Christoffersen 1993; Kristoffersen 1998; Haugan 1998): (10) a. par myndi vera kominn Óláf-r there might be come Olaf-N "Olaf Tryggvason might have come there"

Tryggvason Tryggvi's son

b. Penna tíma réð fyrir Noregi Eirík-r jarl Hákonarson this time ruled for Norway Eirik-N earl Hakon's son "At this time Earl Eirik Hakonsson ruled Norway" c. sner-u-st til hlyöni við hann all-ir turned-PL-REFL to obedience with him all-PL.N í Noregi menn men.PL.N in Norway "All men in Norway declared their obedience to him" d. pávár-u lokin sund then were-PL closed channel.N.PL "Then the channels were closed" e. pat sverð hafði átt Ketil-l that sword-A had owned Ketil-N "Ketil Salmon had owned that sword"

hœng-r salmon-N

REANALYSIS IN WORD ORDER STABILITY AND CHANGE

f. ok urð-u honum fegnir frænd-r ok and became-PL him.D happy kinsmen-PL.N and "And kinsmen and friends were happy to see him"

125

vin-ir friends-PL.N

In (10a) there is a prototypical ergative verb; in (10b) an intransitive, nonergative verb; in (10c) a reflexive verb; in (lOd) a passive verb; in (10e) a transitive verb with an object; and in (lOf) a copula verb with a predicate complement. The nominative NPs in the sentences in (10) carry new informa­ tion. Therefore their position is in accordance with functional requirements. Sentence (11a) (a simplified version of (10a)) has the structure in (l1b), with the NP with the subject role adjoined to VP. This option, made available by the grammar of Old Norse, is exploited as a means to bring the linear order in accordance with the information structure of the sentence. (11) a. Er kominn Óláfr Tryggvason "Olaf Tryggvason has come"

3.2 Ergative constructions Modern Norwegian and the other Scandinavian languages (Swedish and Danish) use a certain kind of ergative construction to obtain a similar informa­ tion structure as the one illustrated for Old Norse. Ergative constructions of this

126

JAN TERJE FAARLUND

kind have an expletive subject, an intransitive verb, an indefinite NP in the object (postverbal) position, and possibly a local adverbial. The expletive subject is det "it" in standard Swedish and standard Norwegian and der "there" in Danish, southern Swedish dialects and southeastern Norwegian dialects. Examples of ergative sentences in Norwegian are given in (12). (12) a. Da hadde det komme ein then had it come a "Then a stranger had arrived"

framandkar stranger

b. Det hadde herska ein brutal diktator it had ruled a cruel dictator "A cruel dictator had ruled the country" c. Det har vendt ses mange it has turned RËFL many "Many people have turned against him" d. Det vart stengt it was closed "A door was closed"

ei a

mot against

i in

land-et country-DEF

han him

dør door

Sentences with a definite post-verbal NP are now ungrammatical, (13a), as are usually sentences with an extraposed subject NP further to the right than the canonical object position, (13b). (13) a. *Der kunne there might b. *Det har it have

det ha it have vendt turned

komme come

seg mot REFL against

Olav Tryggvason Olav Tryggvason han him

mange gamle venner many old friends

While the sentences in (12) have the same functional properties as those in (10), and while both have the same relative order of verb and NP, they have different structures. The structure of (14a) (a simplified version of (12a)) can be given as (14b). (14) a. Det har komme it has come "A stranger has come"

ein a

framandkar stranger

REANALYSIS IN WORD ORDER STABILITY AND CHANGE

127

The change from the Old Norse to the modern Scandinavian pattern can be described as a case of reanalysis whereby the extraposed subject NP was reanalysed as a complement of the verb. Another important change that cleared the way for this reanalysis was the loss of case marking. The reason that structures like the ones in (10) with extraposed nominative NPs fell out of use might be because they ceased to have a distinct nominative case marking. But the same lack of morphological case made it possible for them to be reanalyzed as 'objects' in ergative constructions. Norwegian ergative constructions are not limited to typical ergative or unaccusative verbs. Thus an intransitive verb with an agent can also be used in ergative constructions, as in (12b). Even some transitive verbs, when used intransitively, can form ergative constructions, as in (15). (15)

Det har studert mange studentar på it has studied many students at "Many students have studied in our department"

vårt our

institutt department

These verbs have alternative argument structures with external arguments, and can therefore also passivize (16). (16) a. Det vart herska over land-et på it was ruled over country-DEF on "The country was ruled in a cruel way"

ein brutal måte a cruel way

128

JAN TERJE FAARLUND

b. Det vart eti godt på denne it was eaten well on this "One/we/they ate well in this restaurant"

restauranten restaurant

While in Old Norse we find postverbal subjects in a variety of sentence types and with many different verbs, as illustrated in (10), the kinds of verbs that are found with an expletive subject and a postverbal NP in Modern Norwegian are ergative (12a), intransitive (12b), or reflexive (12c). In addition, transitive verbs may be used in impersonal passives, as in (12d). Active transitive verbs or the copula with a predicate complement do not allow it: (17) a. *Det har ått ein annan mann it has owned an other man "Another man has owned the boat before" b. *Det vart glade mange it became happy many "Many people were happy"

båt-en boat-DEF

før before

menneske people

It turns out, then, that only patterns with one postverbal NP were reanalyzed. For an old construction with a postverbal subject to be reanalyzed in this way, it must be possible to analyze the subject as a direct object. Patterns where a complement position was not vacant, could not be reanalyzed, and that sentence type disappeared from the language. Therefore the sentence types in (17) do not exist in modern Norwegian.4 The result of this process of reanalysis is a sentence type in modern Scandinavian with an expletive subject, an ergative, intransitive, reflexive or passive verb, and an NP in object position. This NP of course carries the semantic role of the subject, and with intransitive verbs it can even be an agent, as in (12b). But as we have seen, they clearly occupy the object position. 4. A problem 4.1 The output of reanalysis It follows from the scenario sketched in § 1 that the options available to new generations of language learners are structures defined by categories and Haugan (1998) calls sentences like (lOe) "passive-like", and suggests that subsequent change would yield a passive sentence with the topical NP as the surface subject, and the postverbal, or extraposed, original subject as an adjoined agent phrase. Sentence (lOe) would then correspond to modern Norwegian Det sverdet hadde vore ått av Ketil "That sword had been owned by Ketil" Although the change implied here cannot be ruled out a priori, it cannot be understood in terms of reanalysis alone.

REANALYSIS IN WORD ORDER STABILITY AND CHANGE

129

principles that already exist in the language or at least in universal grammar. It is a priori very unlikely, or perhaps even ruled out on principle, that reanalysis would introduce novel categories or structures. For example, if a subject is reanalyzed as something else, this something else must be an already existing category. The change described in § 3 and the reanalysis postulated to account for it, seems to be a case of subjects becoming objects. The postverbal NPs in (12) are structurally objects. Besides their surface position, this can also be seen from other properties. Facts about VP pronominalization, coordination, and topicalization show that the verb plus the NP make up one constituent. Compare (18a) with a transitive verb and a regular direct object, to the ergative sentence (18b), in both cases the VP in question is represented by the neuter pronoun det in the answer. (This det must not be confused with the expletive subject det, in (18b) marked with the subscript e for clarity.) (18) a. -

b. -

Har du Have you Ja, deti yes that "Yes, I have"

[sett ein seen an har eg have I

elg]i? elk?

Har dete [komme ein has it come an Ja, deti har dete yes that has it "Yes, there has"

elg]i? elk

The verb + NP can be coordinated as a consituent. (19)

Det [kom inn ei ung kvinne] og it came in a young lady and ut ein gammal mann] out an old man "A young lady entered and an old man left"

[gjekk went

The verb + NP can be topicalized like any other verb phrase, compare (20a) and (20b). (20) a. [Sett ein elg]i har seen an elk have "We have not seen an elk" b. [Komme ein elg] come an elk "An elk has not come"

vi we har has

ikkje ti not det it

ikkje t¡ not

Modern Norwegian has case marking only of pronouns, and pronouns would violate the definiteness constraint on postverbal NPs. Therefore the syntactic

130

JAN TERJE FAARLUND

function of the postverbal NP can usually not be determined by its case form. There is, however, one instance where this is possible: a pronoun combined with the adverb berre "only, no one but" has indefinite reference and can occur in existential sentences. The pronoun is then in the accusative, as expected for direct objects, (21). (21)

Det manglar berre it lacks only "Only she is missing"

henne herA

The postverbal NP also shares with regular direct objects the fact that it has narrow scope. Observe the difference between the real subject in (22a) and the postverbal NP in (22b). In (22a) the NP "ten cows" is outside the scope of the negation, in (22b) it is inside. (22) a. Ti kyr kom ikkje tilbake til gard-en ten cows came not back to farm-DEF "Ten of the cows did not return to the farm" (but stayed in the field) b. Det kom ikkje ti kyr tilbake til garden it came not ten cows back to farm-DEF "There were not ten cows (but twelve) that came to the farm"

So far so good. It seems here that we have a clear case of reanalysis of a subject as an object. But on closer inspection it turns out that these NPs still retain some subject properties. In coordinated sentences, the subject of the latter sentence can be deleted under identity with the postverbal NP of the former, (23a). This is what we usually find with real subjects, as in (23b), but not with real objects, (23c). (23) a. Det kom ein elg¡ og _i begynte å eta it came an elk and started to eat "An elk came and started to eat my apples" b.

Elg-ni elk-DEF

c. *Eg såg I saw

kom came

og and

_i

ein an

elgi elk

og and

begynte started _i

å to

epl-a mine apples-DEF mine eta eat

begynte å started to

epl-a mine apples-DEF mine eta epl-a mine eat apples-DEF mine

The postverbal NP of an ergative construction can bind an anaphor: (24)

Det sat mange turistari i bila-ne it sat many tourists in cars-DEF "There were many tourists sitting in their cars"

sinei their.REFL

From this evidence it seems that the reanalysed postverbal NPs are partly subjects and partly objects, unless they belong to a third, novel category. Both of these solutions seem unacceptable on a principled basis. As long as

REANALYSIS IN WORD ORDER STABILITY AND CHANGE

131

functional categories are defined in terms of structural positions, we would expect a given phrase to belong either to one or the other category. And we do not expect reanalysis to create new categories. 4.2 Modularity and change In order to resolve this dilemma, it is necessary to distinguish between two separate modules: a purely syntactic module, where structural relations such as dominance and precedence are defined, and a semantic, or referential, module, where reference and binding relations are defined. This is modularity in the sense of Autolexical Syntax (Sadock 1991), where modules are seen as independent components with their own rules and principles. The object properties of the postverbal NPs examined in this chapter are all clearly structural, having to do with position and phrasal hierarchies. The subject properties have to do with reference and binding relations, besides of course the semantic role. Note that the starting point of the change described, the Old Norse extraposed subjects, also represents a mismatch: a phrase with a subject role and subjectlike morphology (nominative case, verb agreement) is not expected at the end of the sentence. By attributing object properties to such a phrase, the mismatch is partly resolved. The resolution of a mismatch is probably a very common type of diachronic change. 5.

Conclusion In this chapter I have presented two types of reanalysis: one that changes the order and not the structure, and one that changes the structure, maintaining the order. Both of these can of course be given a teleological interpretation, but teleology is at best a post hoc explanation, or a metaphor for other mechanisms and factors. In the two cases examined here, teleology can be taken as a metaphor for functional factors connected to language used in communication, combined with a mechanism of reanalysis of syntactic structures connected to language acquisition.

REFERENCES Braunmüller, Kurt. 1982. Syntaxtypologische Studien zum Germanischen. Tübingen: Günter Narr. Chomsky, Noam. 1986. Knowledge of Language. Its Nature, Origin, and Use. New York: Praeger. Christoffersen, Marit. 1993. Setning og sammenheng. Syntaktiske studier i Magnus Lagabøters landslov. Kristiansand: Agder Distriktshøgskole.

132

JAN TERJE FAARLUND

Faarlund, Jan T. 1985. "Pragmatics in diachronic syntax". Studies in Language 9.361-393. . 1990. Syntactic Change. Toward a Theory of Historical Syntax. Berlin & New York: Mouton de Gruyter. . (forthcoming). "Early Northwest Germanic". Encyclopedia of the World's Ancient Languages ed. by Roger D. Woodard. Cambridge: Cambridge University Press. Greenberg, Joseph H. 1963. "Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements". Universals of Language ed. by Joseph H. Greenberg, 73-113. Cambridge, Mass. & London: MIT Press. Haegeman, Liliane. 1994. Introduction to Government and Binding Theory. 2nd ed. Oxford: Blackwell. Harris, Alice C. & Lyle Campbell. 1995. Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Haugan, Jens. 1998. "Right Dislocated 'Subjects' in Old Norse". Working Papers in Scandinavian Syntax 62.37-60. Hawkins, John A. 1979. "Implicational Universals as Predictors of Word Order Change".Language 55.618-648. . 1983. Word Order Universals. New York: Academic Press. Hopper, Paul J. 1975. The Syntax of the Simple Sentence in Proto-Germanic. The Hague & Paris: Mouton. Hyman, Larry M. 1975. "On the Change from SOV to SVO: Evidence from NigerCongo". Word Order and Word Order Change ed. by Charles N. Li, 113-148. Austin & London: University of Texas Press. Kristoffersen, Kristian E. 1998. "Norrøn frasestruktur - ein LFG-analyse". MONS 7. Innlegg på Møte om Norsk Språk i Trondheim 1997 ed. by Jan T. Faarlund, Brit Mæhlum og Torbjørn Nordgård. Oslo: Novus. Lehmann, Winfred P. 1974. Proto-Indo-European Syntax. Austin & London: University of Texas Press. Leumann, Manu, J. . Hofmann & Anton Szantyr. 1965. Lateinische Syntax und Stylistik. Lateinische Grammatik, vol 2. Munich: Beck. Lightfoot, David W. 1991. How to Set Parameters: Arguments from Language Change. Cambridge, Mass. & London: MIT Press. Sadock, Jerrold M. 1991. Autolexical Syntax. A Theory of Parallel Grammatical Representations. Chicago & London: The University of Chicago Press. Vennemann, Theo. 1975. "An Explanation of Drift". Word Order and Word Order Change ed. by Charles N. Li, 269-305. Austin & London: University of Texas Press.

WORD ORDER HARMONIES AND WORD ORDER CHANGE IN GEORGIAN* ALICE C. HARRIS Vanderbilt University 0. Introduction The word order universals discovered by Greenberg in 1963 and further studied and refined by many others, especially Hawkins (1983) and Dryer (1992), present a special challenge to linguistics. There is said to be a high degree of correlation among the orders listed in Table 1. Here "O" is (direct) object, "V" verb, "Po" postposition, "Pr" preposition, "A" adjective, "N" noun, "G" genitive, "Rel" relative clause, "Aux" auxiliary, "St" standard of comparison, and "Adj" comparative adjective.1 ov

vo

Po Pr AN NA GN NG RelN NRel Vaux AuxV StAdj AdjSt Table 1 : Some commonly recognized word order correlations

A number of linguists have shown that word order harmonies can be both established and maintained through a process referred to here as construction reanalysis (cf. Greenberg 1963:99; Givón 1971, 1974; Mallinson & Blake 1981:395; Heine & Reh 1984:242; Aristar 1991 and others). According to this view, a particular dyadic configuration, such as GN (genitive-noun), may be *

Some of the material presented here is based upon research supported in part by the National Science Foundation under Grant No. BNS-8419143 and in part by a Fellowship for University Teachers from the National Endowment for the Humanities. I am grateful also to Reza Ordoubadian and to Michael Dobrovolsky for helping me to understand the Azerbaijaini and Turkish comparative construction and to John Grappin for help with Armenian. 1 But see Dryer (1992) on AN/NA and further note 11 below.

134

ALICE C. HARRIS

reanalyzed, for example as NPo (noun-postposition) constructions. Specifi­ cally, it has been proposed that the relative order of an innovative construction may result from construction reanalysis of a source construction involving the following dyads: relative order of noun and adposition from that of the genitive and noun (e.g. Greenberg 1963:99) or from that of object and verb (e.g. Andersen 1979; Lord 1973, 1993); relative order of object and verb from that of genitive and noun (Heine & Reh 1984: 103-104, 240-244); relative order of genitive and noun, as well as that of adjective and noun, from that of relative clause and noun (e.g. Haider & Zwanziger 1984; Aristar 1991). This scenario thus elegantly explains the existence of at least some word order harmonies. Harris and Campbell (1995, Chapter 8) agree that this is an important way of both initiating new harmonies and preserving old ones, but they show that extension also provides a way of establishing and maintaining such harmonies. The purpose of the present paper is twofold. (i) It provides a general account of a thoroughgoing change of word order in the transition from Old to Modern Georgian, a language of the Kartvelian family attested for a millennium and a half. Only changes in the historical period are considered. (ii) It proposes that construction reanalysis may be responsible for the relative orders of verb and auxiliary, and for that of comparative adjective and the standard of comparison. As far as I am aware, neither of these has been proposed before. The latter is argued on the basis of facts in Georgian, and the former on data from Laz, a sister of Georgian. In § 1, the word order of Old Georgian is described, limiting the inquiry to the dyads included in Table 1 for the sake of brevity. The word order of Modern Georgian is described in § 2. In § 3 I propose construction reanalysis as the origin of the change in the word order of the comparative adjective construction in Georgian, and as the origin of verb-auxiliary order in Laz. 1. Old Georgian Old Georgian dates from the fifth century AD, but the oldest complete manuscripts that have been preserved date from the eighth. Many of the early sources are translations of religious texts; since the order of major constituents was very free, their order in Old Georgian sometimes reflects that of the source language. It is therefore difficult, if not impossible, to be certain of the basic order of major constituents (subject, object, and verb) in Old Georgian; and for this reason, order of major constituents is not here discussed for Old Georgian. This problem does not carry over to the order of other constituents, which was more fixed and less susceptible to influence during translation.

WORD ORDER IN GEORGIAN

135

The order of constituents in Old Georgian can be summarized as Table 2, where capital letters represent the unmarked order, small letters represent the marked order, those in parentheses barely occur, and those omitted do not occur. Orders to the left of the slash are those that are harmonious with OV order, while those to the right of the slash are harmonious with VO order. Orders of individual dyads are discussed in greater detail below. Po/pr an/NA gn/NG /NRel /AdjSt V2Aux/auxv Table 2: Order of Constituents in Old Georgian

While most adpositions in the language occur only as postpositions, Old Georgian did have some 'ambipositions', adpositions that could either precede or follow the nouns with which they were associated.3 (1) gives examples of four of these as postpositions, their usual position in Old Georgian; (2) shows them as prepositions. Adpositions are highlighted; Georgian orthographic con­ vention requires monosyllabic adpositions to be written with the head, polysyllabic ones to be written as separate words. (Many of the examples in this section are cited by Imnaisvili 1957: 322-325, 444-447, 548-549, 553-554, 705-709.) (1) a.

kadagebdit mas erdo-eb-sa zeda preach-IMP-PL it roof-COL-DAT on "preach it on the housetops" (Mt 10:27Ad)

b.

saq'uedreli cerni k'ac-ta soris reproach my man-PL.OBL among "my reproach among men" (L 1:25)

c.

romel-ta sen-da-mi hrc'ams which-PL.OBL you.SG-ADV-thither they.believe.it "who believe in you" (Wisdom of Solomon 16:26)

d.

q'o cem-da-mo did-did-ni he .do me-ADV-hither big-big-PL.NOM "he did great things to me" (L 1:49Ad)

My study of word order in Old Georgian is based primarily on the gospels, using materials in Blake and Brière (1950), Brière(1955), and Blake (1974 and 1976). I have also drawn on studies by Pätsch (1971), based on the Sat'berdi manuscript, and by Imnaisvili (1957), based broadly on all of the Old Georgian literature. Some of the ambipositions, such as zeda "on, over", are derived from adverbs and are cognate with preverbs, derived from the same source. In this way they are similar to the Indo-European adverbs that developed into preverbs and ambipositions, the latter becoming eventually prepositions or postpositions. However, some Georgian ambipositions, such as soris "among, between", are derived from adverbs but are not associated with cognate preverbs, while others, such as c'iag "across, on the other side of", appear to be derived from a noun (in this instance c'iag "bosom").

136

ALICE C. HARRIS

(2) a.

romeli iq'o zeda tav-sa cemsa which it.be on head-DAT my "which was on my head" (Genesis 40:17)

b.

romelime davarda šoris ek'al-ta some it .fall among thorn-PL.OBL "some fell among thorns" (L 8:7)

c.

mi-mun-dγe-d-mde romel-sa ševida nove thither-there-day-ADV-until which-DAT he .enter Noah "until the day when Noah entered the ark" (Mt 24:38Ad)

d.

dγe-ta mat gan iohanes natlismcemel-isa-ta mo-aka-mde day-PL.OBL the from John Baptist-GEN-PL.OBL hither-here-until "from the days of John the Baptist until now" (Mt 11:12Ad)

k'idobn-ad ark-ADV

A count of adposition order in the first eight chapters of Matthew in the Adisi manuscript reveals a total of 173 adpositional phrases. Of these, 6 involve a circumposition (two adpositions with a combined meaning, one immediately preceding and one immediately following an NP), 24 involve prepositions (alone), and 143 involve postpositions (alone). These numbers seem to be typical for texts in Old Georgian. There are several indications that the prepositions are relics of an older, dominant Preposition-NP order (cf. Šani3e 1942:371). We may note first that, while Imnaisvili (1957:324-325) cites twenty-eight occurrences of mi- "thither" as a preposition, in every instance it is used with the postposition mde (or mdis) "until, to", as in (2c). The preposition mo- "hither" likewise combines frequently with this postposition; but it provides an additional indication that it is not productive, namely, most of its occurrences are in one of a handful of expressions, including mo-aka-mde, illustrated in (2d). In unmarked order, adjectives follow head nouns, which are highlighted here. (On Paetsch's analysis, please see the appendix to this chapter.) (3) a.

siq'mil-i did-i famine-NOM great (L 4:25AB)

b.

tapl-sa velur-sa honey wild (Mk 1:6)

c.

sul-ita c'mid-ita spirit-INST pure-NST (Mt3:ll)

Participles that modify nouns follow those nouns in unmarked order. (4) a.

santel-i aγntebul-i candle-NOM light .PTCPL-NOM "a lit candle" (J5:35A)

137

WORD ORDER IN GEORGIAN

b.

santel-i sačino-y candle-NOM shine .PTCPL-NOM "a shining candle" (J 5:35B)

c.

cecxl-sa mas daušret'el-sa fire-DAT the quench .NEG.PTCPL-DAT "to the unquenched fire" (Mk 9:43Ad)

d.

sakme-ta mat tvis damt'k'icebul-ta čuen-da establish.PTCPL US-ADV thing-PL.OBL the for "about the things established among us" (L 1:1 Ad)

momart among

Adjectives and modifying participles may precede their heads, but in Old Georgian this order is marked and infrequent. (5) a.

dac'eril-i igi c'ign-eb-i write .PTCPL-NOM the book-COL-NOM "the books written " (J 21:25Ad)

b.

gansat 'evebel-i c 'ign-i leave.PTCPL-NOM book-NOM "[a] writ of divorcement" (Mt 5:31 Ad)

In unmarked order genitives follow the nouns they modify, whether the genitive is a noun or a pronoun. (6) a.

3e-sa

k'ac-isa-sa

son-DAT man-GEN-DAT

"[the] Son of man" (Mt 8:20) b.

''-ni sasupevel-isa ca-ta-ysa-ni key-PL.NOM kingdom-GEN heaven-PL.OBL-GEN-PL.NOM "keys of [the] Kingdom of [the] Heavens" (Mt 16:19)

c.

qorci cerni flesh my "my flesh" (J 6:51)h

d.

gul-sa tv-is-sa heart-DAT self-GEN-DAT "in his own heart" (Mk 5:30Ad)

It may be noticed in the preceding examples that modifiers that follow the head noun bear its case, even when they also bear their own; in this way in (6d) we get the genitive in tv-is-sa marking possession of "heart", while the dative marks agreement with the head noun.4 While other modifiers sometimes 4

This case marking peculiarity of Old Georgian is one of the reasons we can be certain that the NG order, and hence probably other N-modifier orders, were not influenced by the language from which a translation was being made. Apparently no other language of this era in this area used the multiple case marking (Plank 1990). (There is disagreement about the language from which the

138

ALICE C. HARRIS

precede the head noun, genitives do so considerably less frequently; (7) gives an example of this highly marked order. p'ur-i oras-isa drahk'n-isa two.hundred-GEN drahkna-GEN bread-NOM "two hundred drahknas' bread" (J 6:7A)

(7)

While pre-head order is highly marked for simple genitives such as that in (7), (8) illustrates the genitive objects of deverbal nouns, where the genitive precedes its head relatively more often. (8) a.

γvin-isa wine-GEN

msumel-i drink .PTCPL-NOM

"a drinker of wine" (Mt ll:19Ad) b.

c'q'l-isa mis suma-y water-GEN the.GEN drink.MAS-NOM "to drink the water" (II Samuel 23:16)

c.

k'eisar-sa xark'-isa micemad Caesar-DAT tax-GEN give .INF "to give (pay) tax to Caesar" (L 20:22B)

The examples in (8a-c) illustrate a present participle, a masdar, and an infinitive, respectively. The same kinds of deverbal nouns are shown in (9a-c), where the arguments follow the deverbal nouns. (9) a.

msumel-i γvin-isa-y (Mt 11:19AB) drinker.PTCPL-NOM wine-GEN-NOM "a drinker of wine"

b.

štasaval-sa mas mt-isa mis zetis-xil-ta-ysa descend .MAS-DAT the .DAT mount-GEN the.GEN oil-fruit-PL.OBL-GEN "the descent of the Mount of Olives" (L 19:37Ad)

c.

miq'vanebad mariam-isa col-isa sen-isa take .INF Mary-GEN wife-GEN your-GEN "to take (i.e. marry) Mary, your wife" (Mt 1:20)

With deverbal nouns, both orders are frequent in Old Georgian, but the NG type is more common than the GN. In a study of non-finite verb forms in the first eight chapters of the Adisi manuscript of the Gospel of Matthew, 25 examples of NG order were found,

Gospels were translated, but Greek and Armenian are the most likely candidates.) If the language(s) from which translations were being made had influenced Georgian into using an NG order to which it was not accustomed, there would be no explanation for the existence of this unusual multiple case marking, which is ordinarily limited to modifiers that follow the head.

139

and 10 examples of GN. In comparison, among examples with non-derived noun heads in the same text, 272 had NG order, and only 2 had GN order.5 Old Georgian has two kinds of relative clause, and both follow the head. The first type contains a relative pronoun and a finite verb form; it follows the head, apparently without variation. The head nouns are highlighted. (10) a. mravali bork'il-i da ač'v-i, roml-ita šek'rian many fetter-NOM and chain-NOM which-INST they.bind.him "many fetters and chains with which they bound him" (Mk 5:4Ad) b. k'ac-sa, romel-i č'esmarit'-sa man-DAT which-NOM truth-DAT "[a] man which tells the truth to you"

igi him

get'q'vi[s] tkuen he.tell.you.it you (J 8:40AB)

c. marcvial-sa mdogv-isa-sa, romei rnoiγo seed-DAT mustard-GEN-DAT which.NOM he.take.it "a mustard seed which a man took..." (Mt 13:31)

k'ac-man... man-NAR

The second construction that may be considered a relative clause contains future or past passive participles, illustrated in (4 a-d) above. We can summarize the preceding discussion of the positions of constituents by combining the unmarked orders of constituents, as in Table 3. Of course, all these constituents probably never cooccur in a single example. N Po Gen Adj Rel Table 3: Unmarked Sequence of Constituents of NP in Old Georgian.

In the comparative construction, the comparative adjective precedes the standard of comparison, whether the dative is used, as in (11), or the genitive, as in (12).6 (11) a. ravden tkuen umjobes xart mprinvel-ta-sa how .much you .PL better you.PL.be bird-PL.OBL-DAT "how much better are you than the birds" (L 12:24Ad) b. xolo šemdgomad cemsa momavali ulie[reys] but after mecoming-NOM mightier "but the one that comes after me is mightier than I" (Mt 3:11)

cemsa ars me-DAT he.be

(12) a. raoden me umjobes ars k'ac-i cxovr-isa how.much better he.be man-NOM animal-GEN "how much better is a man than an animal" (Mt 12:12Ad) Compounds involving genitives are not included in these numbers; see Harris & Campbell (1995, Chapter 8) on the order of morphemes in compounds in Georgian. According to Imnaišvili (1957:441-452), the dative construction in (11) is older than the genitive construction in (12).

140

ALICE C. HARRIS

b. suli upro[ys] ars sazrdel-isa da qorc-ni samosl-isa soul more it.be food-GEN and flesh-PL.NOM clothing-GEN "The soul is more than food, and the flesh [is more] than clothing."(L 12:23Ad) c. iq'os uk'uanaysk'neli sacturi usures p'irvel-isa it.be last error worse first-GEN "the last error will be worse than the first" (Mt 27:64)

Another way of expressing the comparative involves the use of a complementizer, vidre (ara) "than"; this construction usually involves whole clauses, as in (13). (13)

umjobe[s] ars senda seslvad cxovrebasa mk'elobelisay gina better it.be you .ALL to .enter life lame or qelni da orni perqni tu uqeloysay, vidre orni without.hands than two hands and two feet if gesxnes da štahvarde cecxlsa mas sauk'unesa you .have.them and you .fall fire the eternal "it is better for you to enter life lame or without hands than [that] you have two hands and two feet and fall into eternal fire" (Mt 18:8AB)

In all three comparative constructions of Old Georgian the comparative adjective precedes the standard of comparison, apparently without variation. Auxiliaries follow main verbs in Old Georgian; auxiliaries are highlighted in the examples. (14) a. uk'uetu ara mocemul ars misa zec-it if not give.PTCPLit.be him.ALL heaven-INST "unless it is given to him from heaven" (J 3:27B) b. iohane... aydgomil ars mk'udret-it John rise.PTCPL he.be place.of .dead-INST "John had risen from the dead" (L 9:7B) c. mokceul vart sam gz-is return.PTCPL we.be three way-GEN "we would have returned three times" (Genesis 43:10) d. je k'ac-isa-y micemad ars mydelt-mojγuar-ta... son.NOM man-GEN-NOM give.MAS.ADV he.be priest-teacher-PL.OBL "The Son of man is to be turned over to the priests..." (Mk 10:33Ad)

The verbal complexes in (14) represent the analytic passive, the analytic evidential, and an analytic future. Sometimes the auxiliary precedes the main verb; (15) gives an example. (15)

uk'uetu ara ars mocemul misa if not it.be give .PTCPL him .ALL "unless it is given to him from heaven" (J 3:27A)

zec-it heaven-INST

WORD ORDER IN GEORGIAN

141

Most of the unmarked orders exhibited by Old Georgian are those harmonious with VO order. As Table 2 shows, there are two main exceptions to this generalization: Po and VAux. 2. Modern Georgian Word order in Georgian has undergone considerable change in the millennium and a half during which the language has been documented. Major constituents now occur in the order SOV; this is confirmed by text studies reported in Pocxua (1962) and in Apridoni3e (1986).7 Order of other constitu­ ents is summarized in Table 4. Po/(pr) AN/(na) GN/ng RelN/NRel StAdj/AdjSt VAux/auxv Table 4: Order of Constituents in Modern Georgian.

The adpositions of Modern Georgian are basically postpositions, as illustrated in (16); however, a handful may be considered ambipositions or prepositions (Sani3e 1973:599; Harris and Campbell 1995:215). The postpositions are highlighted in these examples; note that (16c) uses a complex postposition. (16) a. t'q'e-si forest-in(to) "in(to) the forest" b. megobar-tan friend-at "at a friend's place" c. megobar-tan friend-at "with a friend"

ertad together

d. zmn-eb-is šesaxeb verb-PL-GEN about "about verbs"

Adjectives precede nouns, as shown in (17).

It has been reported in the literature that Modern Georgian has SVO order. This incorrect characterization is evidently based upon Kita Tschenkéli's (1958) grammar of Georgian. Intended as a teach-yourself grammar and written in German, Tschenkéli's grammar uses German word order where possible; the exercises in the workbook section use word order that is comfortable for a German-speaking readership. All adpositions included in the grammar are postpositions, but SVO order is predominant in the exercises. Even the first story in the chrestomathy, however, reflects a natural Georgian word order, SOV. See Fähnrich (1986: 154-156) for an accurate and accessible statement of word order for the modern language.

142

ALICE C. HARRIS

(17) a. didi bina big apartment "big apartment" b.

čents lurj 'abas my blue dress-DAT "my blue dress"

Genitives precede the nouns they modify. (18) a. seni d-is disert'acia your sister-GEN dissertation "your sister's dissertation" b. zurab-is važišvili Zurab-GEN son "Zurab's son" c. x-is saxl-s wood-GEN house-DAT "(to a) house of wood, wooden house"

It is clear that in unmarked order modifiers precede the deverbal nominals, as in (19). (19) a. mizn-is garemoebiti damok'idebuli c'inadadeb-is purpose-GEN adverbial dependent clause-GEN sinaars-is gadmomcemi (3i3iguri 1973:468) content-GEN transmit .PTCPL "a transmitter of the content of the purpose adverbial dependent clause" b. nazmnari saxel-is mier saxel-is mortva brunvasi deverbal noun-GEN by noun-GEN government .MAS case-in udur ena-si (title of Pančvi3e 1960) Udi language-in "Government of nouns in a case by a deverbal noun in the Udi language" c. ori c'inadad-eb-is mier movlena-ta an mokmedeba-ta two sentence-PL-GEN by phenomenon-PL.OBL or action-PL.OBL dap 'irisp 'ireb-is gamosaxat'avad (Basilaia 1974:25) opposition-gen express .INF "for two clauses to express an opposition of phenomena or of actions" d. da ilagebda madlob-is gadauxdelad, xm-is amouγeblad and she .straighten thanks-GEN count .NEG .PTCPL sound-GEN raise .NEG .PTCPL "and she straightened [things] up without counting thanks, without making a sound" (Važa, cited in Academy Dictionary 2:329)

Although the participles in this example are passive, the translation in this particular case works more smoothly with active voice.

WORD ORDER IN GEORGIAN

143

The examples in (19a-c) illustrate the modern equivalents of the non-finite verb forms which were exemplified for Old Georgian in (8) and (9)9, and (19d) illustrates the negative participle. As these examples show, arguments precede the non-finite verb forms that govern them; this generalization includes negative participles. Ordinarily the arguments of a non-finite verb form in Modern Georgian are postposed only to avoid ambiguity, to avoid the stacked modifier interpretation, as in (20a-b), or because they are heavy, as in (20c-e). (20) a. damok'idebuli c'inadadeb-is dasak'avšireblad mtavar-tan dependent clause-GEN connect INF main-with "to connect the dependent clause with the main one" (Basilaia 1974:39) b. p'arat'aksul-i šexameba-ni c'inadadeba-ta paratactic-NOM juxtaposition, MAS-PL.NOM sentence-PL.OBL "paratactic juxtapositions of clauses" (Si^iguri 1973:441) c. moq'vana msgavsi magalit-eb-isa, sadac uaxloesi obiekt'i adduce .MAS similar example-PL-GEN where closest object mxolod III p'iri-a only 3rd person-it.be "adducing of similar examples, where the closest [i.e. direct] object is only third person" (Šerozia 1980:121) d. k'argi magaliti-a imis gasatvalisc'ineblad, tu rogori good example-it.be that GEN examine.INF COMP what.kind šedegebi šeijleba miviγot ucxo met'q'velebis zegavlenis sedegad results possible we.get.it foreign speech influence result "[a certain dialect] is a good example for examining what kind of results it is possible to obtain as a result of the influence of foreign speech "(Čikobava 1938:7) e. axsna-ganmart'ebas an dasabutebas imis, rac explain.MAS or document.MAS that.GEN what.REL natkvamia mtavar c'inadadeba-si (Basilaia 1974:39) say.PTCPL main clause-in "explanation or documentation of that which is said in the main clause"

In (20a) one modifier is postposed to avoid the stacked modifier interpretation; that is, if mtavar-tan immediately preceded the head, damok'idebuli c'inadadebis would be interpreted as modifying what immedi­ ately followed it, namely mtavar-; this is not the intended meaning. In (20c), 9

. . . .

The infinitive of Modern Georgian is historically a future participle in the adverbial case, while the infinitive of Old Georgian was historically a masdar in the adverbial case. The functions of the two are also partly different. Therefore, the example of (19c) is not exactly comparable to those of (8c) and (9c).

144

ALICE C. HARRIS

the genitive modifier, magalitebisa, has its own relative clause, which makes this modifier heavy and leads the author to postpose it. Both (20d) and (20e) also have a heavy modifier, a dependent clause. Both use a correlative, imis "it, that", with the non-finite verb form in the function of "pointing to" the dependent clause, but in (20d) the correlative precedes the non-finite form, while in (20e) it follows. That is, one could consider (20c) to be Extraposition and (20d) to be Heavy NP Shift. Splitting of the correlative pronoun and the subordinate clause is not dependent upon the type of non-finite form used; for all non-finite forms, a correlative in this function may precede (i.e. occupy the unmarked position) or follow, together with its dependent clause. In Modern Georgian most departures from the unmarked order are caused by one or another of these considerations. Modern Georgian has several types of relative clause. Participial relatives generally precede the head, which is highlighted here. (21) a. g. k'art'ozia-s mier dadast'urebuli iseti magalit-eb-i G. K'art'ozia-GEN by confirm.PTCPL such example-PL-NOM "such examples, which are confirmed by G. K'art'ozia"(Cxubianisvili 1972:132) b. met'q'veleb-is, azr-is, ...aγmnišvnel-i semant'ik'-is zmn-eb-i speech-GEN opinion indicate.PTCPL-NOM semantic-GEN verb-PL-NOM "verbs of (with) semantics that indicate speech, opinion,.." (Enuki3e 1985:311)

In (21a), g. k'art'ozias mier dadast'urebuli "confirmed by G. K'art'ozia" is a participial relative clause on the head magalitebi "examples". In (21b), semant 'ik'- "semantics" is the head of the participial relative clause met'q'velebis, azris, ...aymnisvneli "indicating speech, opinion...", and this, in turn modifies the head zmnebi "Verbs". Relative clauses formed with a relative pronoun follow the head, while gapped relatives may follow or precede the head. Head-internal (nonreduction) types also occur. All of these types contain finite verb forms, and "examples" of each may be found in Harris (1994). Among the types with finite verb forms, the gapping type may follow the head, as in (22), or precede it, as in (23). (22)

nek'liudov-i, mere rom senat'-is p'rok'uror-ad da minist'r-is N.-NOM later COMP Senate-GEN procurer-ADV and minister amxanag-ad iq'o (Ak'ak'i 460, cited by 3isiguri 1973:256) buddy-ADV he.be "Nekiudov, who was later Senate Procurer and buddy of the minister"

(23)

minda, betaniasi rom k'olmeurnoba-a, is vnaxo I.want Betania.in that collective iarm-it.be it I.see "I want to see the collective farm [that is in Betania]." (cited by Vogt 1971:51)

WORD ORDER IN GEORGIAN

145

Both gapped relatives and relative pronoun relatives may be extraposed, and gapped relatives may precede the entire main clause. Both the type in (21) and that in (22) are very common in speech. Modern Georgian has two constructions for the expression of comparison, with opposing word orders. (24)

upro mayali-a more tall-she.be "she is taller than you"

vidre than

(25)

sen-ze mayali-a you-on high-she.be "she is taller than you"

sen you

The first construction is inherited from the Old Georgian construction illustrated in (13), but has undergone changes. In Old Georgian a comparative form of the adjective was made with the circumfix u-es(i) "-er", but in Modern Georgian this is superlative (or elative), not comparative. Now upro "more" forms the comparative analytically, as in (26). (26)

upro didi otaxi room more big "I want a bigger room"

minda I.want.it

A second change that the construction in (24) has undergone is that vidre is now used with a standard of comparison that may consist of a single word, as in this example, rather than being essentially limited to a clausal standard of comparison, as in the Old Georgian example (13) (cf. Andersen's 1983 analysis). The construction in (25) is new in Georgian and is discussed in detail in §3.2. Auxiliaries follow main verbs in unmarked order, but the opposite order is not unusual. (27) illustrates both orders in the analytic passive, first with the reduced form of the auxiliary aris "he/she/it is", then with the full form. (28) gives examples of both orders in the analytic perfect. (27)

samecniero lit'erat'ura-ši gamotkmuli-a mosazreba, titkos express.PTCPL-it.be opinion as.tho scholarly literature-in svanur ena-si a-s redukci-it y aris miyebuli. it.be derive, PTCPL Svan language-in a-GEN reduction-INST  "In the scholarly literature the opinion is expressed that in the Svan language y is derived by reduction of a." (Kaldani 1959:91)

(28) a. p'rop. s. q'auxčišvil-s šenišnuli akvs... Prof. S.Q'auxčišvili-DAT note.PTCPL he.have.it "Prof. S. Q'auxcisvili has noted [that]..." (Abesa^e 1963:18)

146

ALICE C. HARRIS

b. es ponet'ik'ur movlena senisnuli akvs i q'ipši&e-s this phonetic phenomenon note.PTCPL he.have.it I. Q'ipsi3e-DAT "I. Q'ipsi3e has noted this phonetic phenomenon." (Nadareisvili 1962:181)  is pilmi makvs that film I.have.it "I have seen that film."

nanaxi see.PTCPL

In Modern Georgian, if the orders that harmonize with VO (and are shown to the right of the slash) occur, usually it is for stylistic reasons; this is true of the unusual noun-adjective order and of noun-genitive order. When relative clauses follow their heads it is often, but not always, because they are heavy. The AdjSt construction illustrated in (24) sounds somewhat formal or literary in the modern language.10 We can summarize this section and the previous one with Table 5, which compares the word orders at these two stages of the language. OGa: Po/pr an/NA gn/NG /NRel /AdjSt VAux/auxv MGa: Po/(pr) AN/(na) GN/ng RelN/NRel StAdj/AdjSt VAux/auxv Table 5: Comparison of Order of Constituents in Old and Modern Georgian.

Modern Georgian has SOV order, and all of the harmonies of that order are unmarked in speech. In every instance, the OV harmonies occur at least as frequently in speech as the non-harmonious orders. In Old Georgian, most of the unmarked orders of paired constituents are those that are harmonious with VO order.11 The trigger of the marked order auxiliary - main verb in Modern Georgian is less clear, but it may be conditioned by position relative to the end of the sentence. That is, when the unmarked order, VAux, would make the auxiliary fall at the end of the sentence the order may be inverted. If the auxiliary is aris "he/she/it is", the full form seems to be used in this situation to avoid occurrence at the end of the sentence, since the reduced form is always enclitic. However, I have examples of the order auxiliary - main verb also in other contexts, and I suspect that it is possible for the auxiliary to fall at the end of a sentence. This matter requires further investigation, but seems to be a stylistic tendency only. The description above is limited to the dyads listed in Table 1, which seem to be the most cited. However, Dryer (1992) lists adjective + noun as a "noncorrelation pair", meaning that no correlation was found between the order of this dyad and the order of others in his set of 625 languages. Another dyad which is on Dryer's "Complete list of correlation pairs" (1992:108) and which has been studied in Old Georgian is article + noun. We could, therefore, substitute it for adjective + noun in an overview of the change in word order from Old Georgian to Modern Georgian. In Old Georgian the demonstrative tends to precede the head, while the definite article, derived from and identical to the demonstrative, usually follows (Imnaisvili 1957:563, Mart'irosovi 1979:130-131, among others). In the examples below, I have glossed and translated the determiners in this way. In its function as a definite article, its post-nominal position is harmonic with the order of other constituents of the NP. In its function as a demonstrative

WORD ORDER IN GEORGIAN

147

3. Construction reanalysis 3.1 Auxiliary and verb Although it has apparently not been pointed out before, auxiliary and comparative constructions can also originate in ways that naturally account for the correlation of their word order with other orders in the language. In some SOV languages, such as Japanese, complement clauses occur in the position occupied by  in a simple sentence, as in (29), an example of the structure shown in (30). (29)

(30)

Watakusi wa nihongo ga muzukasii koto o mananda I Japanese difficult COMP learned "I learned that Japanese is difficult." (Kuno 1973:213) [S1[S2O2V2]oV1]

The object clause is likely to occur in the position of the  in a VO language as well, producing the order in (31), where SVO is used as an example of VO order. pronoun, it precedes the noun; demonstrative + noun is on Dryer's "noncorrelation" list (1992:108). Head nouns are highlighted. (i)

a. b.

k'ac-isa magis gan man-GEN the from "from the man" (Mk5:8AB) q'ovel-i kalak-i igi (Mt 21:10Ad) all-NOM

city-NOM

the

"all the city" (ii) a. magis k'ac-isa gan(Mk5:8Ad) that man-GEN from "from that man" b. mas am-sa vhrkua (Mt 7:23AB) that time-DAT I.say "at that time I said" In Modern Georgian, demonstratives still precede nouns: (iii) a. im k'ac'is deda that man-GEN mother "that man's mother" b. es tetri supra this white tablecloth While Old Georgian distinguished definite articles from demonstratives by position, Modern Georgian has no articles. With regard to this constituent, the change from postnominal to prenominai placement was carried out in part by the loss of the definite article function. In the 12-13th centuries, the definite article category, and with it the NDet order, was lost (Mart'irosovi 1979:137).

148

ALICE C. HARRIS

(31)

I

(32)

learned that Japanese is difficult. [SlV1[S2V202]0]

Universally, biclausal structures of the sort in (30) and (32) frequently are fused into monoclausal ones, with the verb of the matrix clause becoming an auxiliary, and the verb of the embedded clause the lexical verb (Harris and Campbell 1995, Chapter 7). In the sort of SOV language illustrated by Japanese, this would result in VAux structures (with V from V2 and Aux from V1 in (30)), naturally correlating with the OV order from which it developed. In VO languages of the type illustrated in (31), fusion produces AuxV order, correlated with the VO order. An example of an auxiliary originating from a structure such as that in (30) is provided by the  dialect of Laz, a Kartvelian language spoken primarily in Turkey. In Laz the V1 (verb of the matrix clause) "want" was reanalyzed as an auxiliary of the future tense (see Cikobava 1936:148-149; Holisky 1991:430-431). (33) provides an example of the matrix verb "want" (second person subject form) with a noun object, illustrating its earlier meaning. (34) shows the usual position of a dependent clause (here with first person subject form). This is one of the regular positions for object clauses in Laz (Holisky 1991:455-456). (35) provides an example of the future tense auxiliary (third person subject form). (33)

makvali-a va gi-non NEG you-want it egg-QUOT "You don't want an egg (he said)." (Asatiani 1974:5)

(34) a.

man oc'k'omu do osumu var mi-non-ya I eating and drinking NEG I-want.it-QUOT "I do not want to eat and drink (he said)." (K'art'ozia 1968:156, 74)

b. [ S [ V 2 a n d V 2 ] 0 V 1 ] (35) a. miti akolen var ida-su-non-an no.one from.here NEG go-he he-FUT-PL 12 "No one will go from here." (Asatiani 1974:4) b. [ S V A u x ]

While the auxiliary has a sense of obligation in this example, in others it may retain a sense of intention or prediction. In the dialect illustrated here () the main verb is in the optative and inflected for person, while the auxiliary is inflected for both person and number. In the speech of Čxal, on the other hand, the auxiliary is invariant and the main verb is inflected for both person and number (Cikobava 1936:148, citing Q'ipše).

WORD ORDER IN GEORGIAN

149

In Laz, the negative particle must occur before the verb it negates; in the un-reanalyzed (34a) it occurs before the verb "want", in the reanalyzed (35a) before the main verb, not the auxiliary from "want". The VAux order of (35) may be attributed to the fact that the auxiliary is diachronically derived from the verb of the matrix clause in a structure like that in (30). 3.2 The comparative construction It is likely that comparative constructions likewise develop out of other constructions whose order they may reflect. In this area, however, there has been less systematic study of the universals, and the discussion is therefore necessarily more speculative. In his study of the universals of comparative constructions, Andersen (1983:118) distinguishes four basic types, as in (36).13 (36) a. Juxtaposition: b. Adpositional: c. Case: d. Particle :

A is big,  is small/not (big). A is big(ger) FROM . A is big(ger) -FROM. A is bigger THAN .

The word order here within the clause is intended to illustrate only one possibility; that is, (36b) is intended to include both prepositions and postpositions, and (36b) and (36c) are intended to include both the AdjSt order given, and the StAdj order. Similarly, "from" in (36b) and (36c) is intended only as an example (perhaps the most common) of the adposition or case that may be used. Type (36a) is not further discussed here, and a fifth type ("A surpasses  (in size)") has been omitted altogether. Andersen (1983:121-127) points out that the word order in constructions of types (36b) and (36c) in a given language is just a special case of adjective + complement constructions in that language. Adjective + complement construc­ tions include phrases such as "experienced in work", "faithful to a friend", "dear to me". While Andersen makes the connection between general adjective + complement constructions and comparative constructions of types (36b) and (36c), he stops short of suggesting that the latter develops out of the former diachronically, though that seems to be what he has in mind. We can document Stassen (1985) presents a different typology and an interesting analysis, linking the comparative construction that a language uses to the way in which it expresses consecutive and simultaneous actions, and this in turn to word order. Unfortunately, this analysis predicts (Stassen 1985:106) from the so-called fixed case comparative of Georgian (the innovative comparative), that it will be a so-called deranking language. On the contrary, it falls into Stassen's balancing" type. For this reason his typology proved to be less useful to me than Andersen's.

150

ALICE C. HARRIS

here such an example. In Georgian, adjective + complement constructions have the unmarked order Complement-Adjective, as in (37).14 The adjectives are highlighted. (37) a. -tvis me.GEN-for "good for me"

k'argi good

b. am bič'is-tvis this boy.GEN-for "difficult for this boy"

teli

difficult

c. es noxi cerni this rug my "This rug is dear to my family."

o ixis-tvis family .GEN-for

, virpasi-a dear-is

The same order is unmarked in those adjective + complement constructions that contain modifiers that are formally participles, such as those in (38). (38) a. -ze damok'idebuli me.DAT-on dependent-PTCPL "dependent on me" b. mosk'ovis-gan damouk'idebeli Moscow .GEN-from independent-PTCPL "independent of Moscow" c. sakartvelo-ši camosuli Georgia.DAT-in arrived-PTCPL "arrived in Georgia, come to Georgia" d. amit k'maq'opili it.INST satisfied-PTCPL "satisfied with it" savse e. c'q'l-it water-INST full-PTCPL "full of water"

When a new comparative construction developed in historical times, it devel­ oped as an adjective + complement construction: Complement - Adjective, in particular, Standard - Adjective; this is illustrated in (39).

WORD ORDER IN GEORGIAN

(39) a. -ze k'argi me.DAT-on15 good better than I" b. am p'roblema-ze iteli this problem .DAT-on difficult "more difficult than this problem" c. es noxi im bec'ed-ze virpasi-a this rug that ring .DAT-on dear-is "This rug is more valuable than that ring."

The order of adjective + complement constructions has not been the focus of intensive study, and it is not yet known whether it is universally correlated with the order of other constructions.16 In Georgian, at any rate, the innovative Standard-Adjective order in the comparative construction seems to have developed out of the general Complement-Adjective order. 3.3 Conclusions The approach taken here, which is also the approach taken in other works cited above, is to argue: • that certain constructions may develop out of certain others by reanalysis; • that when this occurs the order of constituents in a phrase may reflect the order of the constituents in the construction from which it developed; • that in instances observed here the order of the derived construction is correlated with that of the source construction. I have presented evidence that the order of Verb and Auxiliary may develop as a correlation of the order of embedded clause and verb of the main clause. Similarly, there is evidence that the order of the Standard of Comparison and the Comparative Adjective is in some instances a reflection of the order of Complement and Adjective. This adds two dyads to those known to reflect the order of the elements out of which they often develop. Historically the meaning of this postposition included the meaning "over", probably the meaning that is relevant to its use in the comparative construction; today, outside this construction, it means "on", not usually "over". The case form to which it is attached was historically dative, but because the marker of the dative has been lost before this postposition in nouns, an argument could be made that it is attached to the stem form synchronically. Andersen (1983:125) reaches the opposite conclusion, based on the same data, suggesting, for type (36 b) only, a correlation of VO order with CompAdj, and OV with AdjComp. The difference in our conclusions is due in part to the additional Georgian data, but also to the fact that all of the languages he cites happen to have mixed characteristics, and of course to the lack of a systematic study.

152

ALICE C. HARRIS

4. Borrowing as an alternative explanation Any discussion of the explanation of change would be incomplete without considering the possibility of borrowing. Two Georgian dialects, Mesxian and Ac'arian, have been heavily influenced by Turkish, which has consistent SOV order. However, there has not been a great deal of contact between Turkey and mainstream Georgians. In particular, the literary dialect, which is traced in this paper, is based on the Kartlian and K'axetian dialects, which are not strongly influenced by Turkish or by Azerbaijani (Azeri), the Turkic language neighbor­ ing to the east. Historically it is chiefly the religious difference between Christian Georgians and Muslim Turks and Azerbaijanis that has kept these neighbors apart. (However, Laz, described in § 3.1, is spoken primarily in Turkey, and Turkish influence on that language is well documented [Jikia 1964, 1967,1974,1975]). Greek was in a position to exert a greater influence, since much of the early Old Georgian literature was translated from Greek. Watkins (1964:1039) has noted that Greek had a very free word order, and Friedrich (1975) states that the OV and the VO characteristics are found about equally in Greek, apart from a preference for AN order.17 It is possible that AN order influenced the Georgian change, but this does not seem to have been among the first OV harmonies put into effect in Georgian. Armenian is a near neighbor and another Christian nation, and was therefore in a position to exert some influence. According to Friedrich (1975:40-42), Old Armenian had (1) both AN and NA, with a slight predomi­ nance of the former, (2) GN and NG, with a strong predominance of the latter, (3) four comparative constructions, all with the order AdjSt, (4) most frequently NRel, (5) prepositions, with rare postposing, (6) basic SVO, with VSO as an alternative. When we compare these characteristics of Old Armenian with those of Old and Modern Georgian, we see that Old Armenian has no predominant order that is the same as the dominant order of the same elements in Modern Georgian, with the single weak exception of AN. Therefore it seems that Old Armenian order could not have provided the model for the dyadic changes undergone in historical times, with the exception of AN. Armenian did undergo changes from predominant NG order to GN order, and from almost exclusively prepositions to predominantly postpositions, thus Watkins' and Friedrich's statements are based primarily on the earliest Greek, while it was Christian writings that were translated into Old Georgian. But the order that Greek developed into, according to Aitchison 1979 included VO, AN, NG, AdjSt, Pr, and NRel. Apart from AN, Georgian was developing away from these at about the same period. It therefore seems unlikely that Greek had a major influence in shaping the Georgian change.

WORD ORDER IN GEORGIAN

153

undergoing some of the same changes undergone by Georgian. It appears, however, that in the change from prepositions to postpositions, at least, it was Georgian that made the change first and perhaps influenced Armenian, not the reverse. A third ethnic group with which Georgians interacted a great deal in ancient times was the so-called Caucasian Albanians (unrelated to European Albanians), who spoke a language of the North East Caucasian family, an ancestor of modern Udi. Caucasian Albanian was a written language, but so little of it is attested (Mouraviev 1981, 1990), that we must reconstruct its characteristics from modern Udi and from its sister languages in the Lezgian subgroup of North East Caucasian. All of the sister languages and Udi are strongly post-positional, with AN, GN, and RelN orders. In Udi, the order StAdj is used (Harris, ms. 1998b). Udi and most of its sisters have dominant SOV order, with other orders permitted, especially SVO. Udi provides strong evidence of an earlier VAux order (Harris, ms. 1998a), and a number of the sister languages have this order as well. Thus it is likely that Caucasian Albanian had all of the orders that Modern Georgian eventually acquired. Did Georgian borrow SOV and its harmonies from Caucasian Albanian? Although the Caucasian Albanians were much more numerous than their Udi descendants, it appears that they were considerably fewer in number than the Georgians. Their alphabet appears to be derived from those of Armenian and Georgian, and generally it appears that these two exerted much more cultural influence on Caucasians Albanians than the latter did on the Georgians. Thus, it appears unlikely that these word orders would have been borrowed outright from the Albanians. In general it is difficult to prove presence or absence of influence in a particular syntactic change, but in this instance it would seem unlikely that Georgian could completely escape the influence of so many neighbors with SOV order. In the Georgian changes of most word order dyads from the VO harmonies to the OV harmonies, one cannot find evidence of specific influence or of lack of it. Since every language has dyads composed, for example, of adjectives and nouns, how could one, even in theory, show that AN order resulted from the influence of another language? Those who would claim that the Georgian word order change must be the result of the influence of other languages do so not on the basis of specific evidence, but on the basis of a general conviction that word order changes are due to the influence of other These statements are based on descriptions of the order in Modern West Armenian, from Kogian (1949: 48,136ff, 159,213); and Feydit (1948:122).

154

ALICE C. HARRIS

languages. I have argued elsewhere that we must not merely assume influence from another language but should base any claim of borrowing on specific evidence (Harris & Campbell 1995, Chaper 6). There is, however, some evidence that Georgian word order was not borrowed outright from Turkic languages, from Greek or Armenian, nor from other Caucasian languages. Two new constructions that helped to establish the new OV harmonies were introduced; these are complex enough and have characteristics specific enough to Georgian, that we can look in neighboring languages for their prototypes. The two constructions at issue are the pre­ nominai relative clause, using the gapping strategy (see (23)), and the comparative construction in the StAdj order (see (25)). The Georgian pre-nominal gapped relative is unlike relative clauses in Turkish (as described in Underhill 1976:273-296), in Greek (as described in Aissen 1972), or in Caucasian languages. Turkish relatives are pre-nominal participles, as are those in most languages of the Caucasus. Old Georgian already possessed participial relatives (see § 3);19 to make these conform to the OV order, Georgian simply moved the participial clauses to pre-nominal position. Greek has post-nominal relatives formed with relative pronouns; these do not provide a prototype for an OV harmony. The innovation of the prenominal gapped relative was one of the changes that increased conformity to OV harmonies, but this relative clause structure does not seem to have a preexisting model in the region.20 The pre-nominal gapped relative developed from the Middle Georgian post-nominal gapped relative (Harris 1993, 1994); the post-nominal gapped relative simply moved to the other side of the head noun, just as other modifiers did. If the pre-nominal gapped relative is entirely a Georgian development as the evidence indicates, this suggests that the introduction of this construction was entirely language-internal and conseGeorgian participles are unlike Turkish relatives in that in Turkish the object of the participle bears the object case (as in (iv)) and the subject bears the genitive (Underhill 1976), while in Georgian the direct object of the participle bears the genitive, the subject of an intransitive participle bears the genitive, and the subject of a transitive participle is marked with the postposition mier "by" (see Harris 1981:156-160 and examples above), (iv) pencere-yi kiran taş window-OBJ broke.PTCPL stone "the stone that broke the window" (Underhill 1976:277) At least two of Georgian's sister languages, Laz (Harris 1988:88-89) and Mingrelian (Harris 1991:384) do have pre-nominal gapped relatives similar in structure to those in Modern Georgian; it is not known when these developed. If we suggest that Georgian borrowed its pre-nominal gapped relative from one of its sisters, we are only displacing the problem by one step. That is, we still must ask where Laz and Mingrelian got their pre-nominal gapped relatives.

WORD ORDER IN GEORGIAN

155

quently that it was not borrowed. If Turkish, Greek, or another Caucasian language were responsible for the change, rather than language-internal pressures, we would expect to find the innovative relative clause calqued on the foreign model. The comparative construction in Turkish, as described in Underhill (1976:224-225), does have the order StAdj, found also in the innovative Georgian construction. (40) a. Hasan, Halil'-den daha hos bir insan-dir Hasan Halil-ABL more pleasant one person-EMPH "Hasan is a more pleasant person than Halil." (Underhill 1976:224) b. Hasan, Halil'-den (daha) tembel Hasan Halil-ABL more lazy "Hasan is lazier than Halil" (Underhill 1976:225)

The Turkish comparative construction has the following additional characteristics, none of which is shared by Georgian: (i) Turkish uses the ablative suffix to mark the standard of comparison; Georgian uses the postposition -ze "on", not -gan, -dan "from" which would be comparable to the Turkish. (ii) The Turkish comparative uses the form daha "more" with the adjective, although this can be omitted, as in (40b). Georgian possesses such a form; upro "more" can be combined with adjectives, as in upro didi "bigger", upro jvirpasi "more valuable". However, upro "more" is not used in the new StAdj construction, only in the AdjSt construction inherited from Old Georgian.21 Like Turkish, Azerbaijani uses the ablative suffix on the standard of comparison, an adjective with no suffix, and the StAdj word order. Azerbaijani offers the possibility of use of the emphatic suffix -rax on the adjective (Householder & Lotfi 1965:195-196). Literary Georgian does not use its ablative here and has no comparable emphatic comparative suffix. Thus, although the innovative comparative construction of Georgian has the same word order that the Turkic construction has, it does not have the characteristics we would expect if the Georgian had been borrowed outright from the Turkish or from the Azerbaijani.22 Tschenkéli (1958:224) cites another construction: es kali (upro) lamazi-a im kal-ze this lady (more) pretty-she.be that Iady-on "this lady is prettier than that lady" It is also possible to use upro "more" in this AdjSt construction. The comparative in Laz, a Kartvelian language spoken mostly in Turkey, does use the ablative case to mark the standard of comparison; it does not seem to have an equivalent of Turkish daha "more". Mingrelian, which is further from the sphere of Turkish influence, nevertheless has a

156

ALICE C. HARRIS

One comparative in Modern West Armenian is structurally similar to the Turkish. (41) a. mezmë aweli erjanik us .ABL more happy "happier than we" (Kogian 1949:174) b.

mezmē US.ABL

erjanik happy

"happier than we" (Kogian 1949:174) (42)

Sahak-ēn (aweli) gitown ër Mastoc' [StAdj] scholar was Maštoc' Sahak-ABL (COMPAR) "Maštoc' was more scholarly than Sahak." (Kogian 1949:173)

Like both the Turkish and the innovative Georgian construction, this comparative in Modern West Armenian has the order StAdj.23 Like Turkish and unlike Georgian, the Armenian StAdj construction uses the ablative and a word meaning "more". Armenian illustrates the kinds of effects we would expect if borrowing or direct influence had been responsible for the Georgian construc­ tion. Since these effects are not found in Georgian, there is thus little reason to think that the Georgian construction was strongly influenced by either Turkish or Armenian. Although one Greek comparative is similar to the Old Georgian compara­ tive in (13), it is not similar to the innovative Georgian StAdj construction and therefore does not provide a potential model for it. Schmidt (1979) has surveyed comparative constructions in the Caucasus and finds several with the StAdj word order. Among these, the Ossetian (IndoEuropean, located in part in Georgia), somewhat like the Turkish, uses the ablative case and an adjective marked for the comparative degree. Xinalug and the southern dialect of Tabassaran (both of the Lezgian subgroup of the North East Caucasian family and both unrelated to Georgian) have this word order and both mark the standard of comparison with a special suffix of comparison and leave the adjective uninflected. Though Schmidt did not discuss every language of the region, he discusses no comparative that would be a prototype for the innovative Georgian construction using the postposition "on" to mark construction that seems to be exactly parallel to that in Turkish: te cira ti cira-še umosi skvami re this girl that girl-ABL more pretty she.be "this girl is prettier than that girl" (Kipšidze 1914:034) Since Old Armenian had constructions with the order AdjSt (Friedrich 1975:41-42), I assume that the StAdj construction of Modern West Armenian is an innovation. (In both Georgian and Armenian constructions with the AdjSt order survive).

WORD ORDER IN GEORGIAN

157

the standard of comparison and leaving the adjective uninflected and without "more". Udi, the modern descendant of Caucasian Albanian, speakers of which Old Georgian speakers interacted with, has a comparative construction similar to that of its sister languages in the Lezgian subgroup. (43) a.sa sunsun-axo šavat'-t'e one RECIPROCAL-ABL beautiful-3SG "Each is more beautiful than the next." (Dirr 1928,62:6) b. me xod t'e xod-axo kala-ne this tree that tree-ABL big-3.SG "This tree is bigger than that tree." (Pancvi3e 1974:79,125)

As shown by (43), the Udi comparative is characterized by a standard of comparison in the ablative and an adjective without marking. The differences between this and the innovative Georgian comparative suggest that the comparative construction illustrated in (39) is a purely Georgian creation. Because there are numerous languages with OV order in the region, outside influence cannot be ruled out in the whole change of VO harmonies to OV harmonies. Nevertheless, no specific evidence has ever been presented, as far as I am aware, to support the view that OV order was borrowed in Georgian. As shown here, at least two of the changes show clear evidence of having been internally motivated, rather than borrowed outright. Certainly the burden of proof is on those who would argue that Georgian word order is borrowed. 5. Conclusion In this paper I have documented the changes from the word order of Old Georgian to Modern Georgian, over nearly a millennium and a half. Although we cannot be really certain about the basic word order of major constituents (S, O, and V) in Old Georgian, the order of dyads is secure. I have shown here that the unmarked order of several important dyads changed during this period, with NA being replaced by AN, NG by GN, and with NRel coming to coexist with RelN, and AdjSt with StAdj. This is consistent with the fact that it appears that in Old Georgian postpositions, though already unmarked, had recently replaced old prepositions, which in fact still lingered. I have proposed that construction reanalysis is in some instances responsi­ ble for the relative order of V and Aux and of St and Adj, and I have adduced examples of each. I have argued that these particular innovative constructions were not borrowed, but rather that they are the result of internally motivated changes. This, of course, does not mean that overall word order is not influenced by contact languages.

158

ALICE C. HARRIS

APPENDIX Pätsch (1971:256-257) claims that for adjectives (qualities) the postnominal position is emphatic. This cannot be correct for the codices studied here, given that adjectives follow their heads far more often than they precede them. Pätsch's evidence (1971:257) is of two types. First, in the expression k'urneba-ni da sasc'aul-ni healing-PL.NOM and miracle-PL.NOM "great healings and miracles"

did-did-ni (151) big-big-PL.NOM

the reduplicated adjective is emphatic and is postposed. According to Pätsch this shows that the post-nominal position is emphatic. However, were this so, the emphatic form would not be needed; the ordinary form, did-ni, would be made emphatic by being in emphatic position. Thus, this example suggests the opposite, that the postposed position is not emphatic. The second type of evidence Pätsch introduces is the order in certain fixed expressions, citing p'at'iosani juari "honorable cross", c'miday nino "Saint Nino", net'ari nino "blessed Nino", and as an exception, suet'i cxoveli "living column". These examples are of different sorts. The first example, p'at'iosani juari "honorable cross", in the order AN, and the last, suet'i cxoveli in the order NA (where the modifier is a participle), genuinely involve modifiers and can be compared with fixed expressions involving modifiers in the Gospels. In the Adisi, Opiza, and Tbet'i Gospels, "Holy Ghost" is likewise a fixed expression and has the modifier ("holy") in post-nominal position: eg. sulisa c'd-isa (Mt 12:31AB),suli c'miday (Mk 13:11; L 1:35, L3:22, L 12:12Ad; J 7:39AB, J 14:26),sulita c'midita (Mt 3:11; L 1:15, L 3:16, L 4:1; J 1:33),sulman c'midaman (L 12:12AB). "My beloved Son" is another fixed expression with a participle and occurs in the order "son""my"-"beloved": jey čemi saq'uareli (Mt 3:17; Mk 1:11 AB; L3:22AB, L9:35), je čemi saq'uareli (Mt 17:5, Mk 9:7), jeey čemi saq'uareli (Mk l:llAd; L 3:22Ad). We have here two fixed expressions in the order AN (p 'at'iosani juari, net'ari nino, both from the Sat'berdi manuscript) and three examples having the order NA. In the expression 'miday nino "Saint Nino" the first word is a title. This particular title, "Saint", generally does precede the name in Old Georgian, but others follow. For example, caesar was a title similar to "king", and we find (omitting now cases and other variations) t'iberia k'eisari "Caesar Tiberias" (L 3:1), agust'os k'eisari "Caesar Augustus" (L 2:1). "King" is another title that follows: davit mepe "King David" (Mt 1:6), herode mepe "King Herod" (Mt 2:1), and this usage was continued at least into the Middle Ages, with tamar mepe "Queen Tamar". The Sat'berdi manuscript itself contains the fixed expression helene dedopali "Queen Helen" (Pätsch 1971:256). Thus c'miday differs from other Georgian titles in this respect. In any case, the universals of the word order of titles have been less studied. The last of Pätsch's examples, net'ari nino "Blessed Nino", could be analyzed in several ways. It is possible that net'ari is intended as a modifier, but it also may be either a title or an appositive. Fixed appositive expressions in the Gospels generally have the order name-descriptor, as in "John the Baptist", always in the order John {iohane or iovane)-Bzptist: iohane natlismcemeli (Mt 3:1Ad), iovane natlismcemeli (Mt 3:1AB, Mt

WORD ORDER IN GEORGIAN

159

14:2AB; Mk l:4Ad, Mk 6:14), iohanesa natlismcemelisa (Mt 11:11Ad), iohane natlismcemelisa (Mt 11:1 LAB), iohanes natlismcemelisata (Mt ll:12Ad), iohane natlismcemelisait (Mt 11:12AB), iovane natlismcemelisay (Mt 14:8AdA, Mk 6:24), iohanesi natlismcemelisay (Mt 14:8B), iovane natlismcemelman (L 7:20). net'ari nino does not follow the usual order for modifiers, for titles, or for appositives. Thus, three of Pätsch's examplesp 'at'iosani Juari, c'miday nino, and net'ari nino are not typical of comparable expressions found in the Gospels at about the same time and a bit earlier. At leastp 'at 'iosani juari "honorable cross" is probably best viewed as an advance wave of the order that was just beginning, with the later codices of the Gospels following the order that had already been established for these fixed expressions in earlier Old Georgian codices. "King", being a old traditional Georgian title, may have remained conservative and may reflect an older state of affairs (following a name) than does the title c'miday "Saint", which precedes a name. Modern Georgian gives us little help in this respect, as the titles bat'oni, kalbat'oni are used before names (eg. kalbat'oni luiza "Miss Louise"), but kinship titles follow the name (eg. luiza deida "Aunt Louise"). ABBREVIATIONS USED FOR GLOSSING EXAMPLES ABL = ablative; ADV = adverbial case; ALL = allative; COL = collective; COMP = complementizer; DAT = dative; EMPH = emphatic; GEN = genitive; IMP = imperative; INST = instrumental; MAS = masdar; NAR = narrative case; NEG = negative; NOM = nominative; OBL = oblique; PL = plural; PTCPL = participle; QUOT = quotative; SG = singular Georgian does not distinguish masculine from feminine; third person singulars are glossed with masculine pronouns if they refer to masculine beings in the context from which they are taken; otherwise they are glossed with feminine pronouns. Many examples of Modern Georgian are taken from the texts of the linguistic literature. Old Georgian sources are abbreviated as follows: Mt = Matthew (Blake 1976); Mk - Mark (Blake 1974); L = Luke (Brière 1955); J = John (Blake and Brière 1950). The codices are abbreviated as Ad = Adisi Gospels (A.D. 897); A = Opiza Gospels (A.D. 913);  = Tbet'i Gospels (A.D. 995).

REFERENCES Abesa3e, Nia. 1963. "Rom k'avsiri kartvelur enebsi". [The conjunction rom in the Kartvelian languages.] TUS 96.11-20. Academy Dictionary (KEGL)=Čikobava, et al 1950-64. Aissen, Judith. 1972. "Where Do Relative Clauses Come From?". Syntax and Semantics, 1 ed. by John P. Kimball, 187-198. New York: Seminar Press. Aitchison, Jean. 1979. "The Order of Word Order Change". Transactions of the Philological Society 43-65.

160

ALICE C. HARRIS

Andersen, Paul Kent. 1979. "Word Order Typology and Prepositions in Old Indic". Studies in Diachronic, Synchronic, and Typological Linguistics: Festschrift for Oswald Szemerényi on the Occasion of his 65th Birthday ed. by Bela Brogyanyi, part I, 23-34. Amsterdam: John Benjamins. . 1983. Word Order Typology and Comparative Constructions. Amsterdam: John Benjamins. Apridoni3e, Sukia. 1986. Sit'q'vatganlageba axal kartulsi. [Word order in Modern Georgian.] Tbilisi: Mecniereba. Aristar, Anthony Rodrigues. 1991. "On Diachronie Sources and Synchronic Pattern: An Investigation into the Origin of Linguistic Universals". Language 67.1-33. Asatiani, Irine. 1974. C'anuri (lazuri) t'ekst'ebl [Can (laz) texts.] Tbilisi: Mecniereba. Basilaia, N. 1974. Uk'avsiro rtuli c'inadadeba. [The asyndetic complex sentence.] Tbilisi: Universit'et'i. Blake, Robert P. 1974. "The Old Georgian Version of the Gospel of Mark". Patrologia Orientalis 20:3 (1st ed. 1928). . 1976. "The Old Georgian Version of the Gospel of Matthew". Patrologia Orientalis 24:1 (1st ed. 1933). Blake, Robert P. & Maurice Brière. 1950. "The Old Georgian Version of the Gospel of John". Patrologia Orientalis 26:4. Brière, Maurice. 1955. "La version géorgienne ancienne de L'évangile de L u c " . Patrologia Orientalis 27:3. Cikobava, Arnold. 1936. C'anuris gramat'ik'uli analizi. [A grammatical analysis of Laz.] T'pilisi: Mecnierebata Ak'ademiis Sakartvelos Pilialis Gamocema. . 1938. Č'anur-megrul-kartuli sedarebiti leksik'oni. [A Laz-MingrelianGeorgian comparative dictionary.] T'pilisi: Mecnierebata Ak'ademiis Sakartvelos Pilialis Gamocema. Cikobava, Arnold, G. Axvlediani, Sim. , Vikt'. ''r, Tamar Lomouri, V. Topuria, G. C'ereteli, eds. 1950-64. Kartuli enis ganmart'ebiti leksik'oni. [Explanatory dictionary of the Georgian language.] Tbilisi: Ak'ademia. Cxubianisvili, D. 1972. Inpinit'ivis sak'itxisatvis jvel kartulsi. [On the question of the infinitive in Old Georgian.] Tbilisi: Mecniereba. Dirr, Adolph. 1928. "Udische Texte". Caucasica 5.60-72. Dryer, Matthew S. 1992. "The Greenbergian Word Order Correlations". Language 68.81-138. Enuki3e, L. 1985. K'onperencia 't'ip'ologiuri metodebi sxvadasxva sist'emis enata sint'akssi'. [Conference 'Typological Methods in the Syntax of Various Systems of Language'.] Iberiul-K'avk'asiuriEnatmecniereba 23.308-318. Fähnrich, Heinz. 1986. Kurze Grammatik der georgischen Sprache. Leipzig: Verlag Enzyklopädie. Feydit, Frederic. 1948. Manuel de langue arménienne (Arménien occidental moderne). Paris: C. Klincksieck. Friedrich, Paul. 1975. Proto-Indo-European syntax. Journal of Indo-European studies, Monograph 1. Butte: Montana College of Mineral Science and Technology. Givón, Talmy. 1971. "Historical Syntax and Synchronic Morphology: An Archaeologist's Field Trip". Chicago Linguistic Society 7.394-415.

WORD ORDER IN GEORGIAN

161

. 1974. "Verb Complements and Relative Clauses: a Diachronic Case Study in Biblical Hebrew". Afroasiatic Linguistics 1.4:1-22. Greenberg, Joseph H. 1963. [1966.] "Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements". Universals of Language ed. by J. H. Greenberg, 73-113. Cambridge, Mass: MIT Press. Haider, Hubert, & Ronald Zwanziger. 1984. "Relatively Attributive: The 'ezâfe'Construction from Old Iranian to Modern Persian". Historical syntax ed. by Jacek Fisiak, 137-172. Berlin: Mouton. Harris, Alice  1981. Georgian Syntax: A Study in Relational Grammar. Cambridge: Cambridge University Press. . 1988. "On Hypotaxis in Laz". C'elic'deuli 15.87-103. _. 1991. "Mingrelian". The Indigenous Languages of the Caucasus. The Kartvelian Languages ed. by Alice C. Harris, vol. I, 313-394. Delmar, New York: Caravan Press. . 1993. "Changes in Relativization Strategies: Georgian and Language Universals". Caucasologie et mythologie comparée ed. by Catherine Paris, 391403. Paris: Peeters. . 1994. "On the History of Relative Clauses in Georgian". Non-Slavic Languages of the USSR: Papers from the Fourth Conference ed. by Howard I. Aronson, 130-142. Columbus, Ohio: Slavica. . ms. 1998a. Endoclitics and the Origins of Udi Morphosyntax. Vanderbilt University. . ms. 1998b. Udi. The Languages of the Caucasus ed. by Alice C. Harris & Rieks Smeets, Curzon Press. Harris, Alice C , & Lyle Campbell. 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: Cambridge University Press. Hawkins, John A. 1983. Word Order Universals. New York: Academic Press. Heine, Bernd & Mechtild Reh. 1984. Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Holisky, Dee Ann. 1991. "Laz". The Indigenous Languages of the Caucasus. The Kartvelian Languages ed. by Alice  Harris, vol. I, 395-472. Delmar, New York: Caravan Press. Householder, Fred W., Jr. & Mansour Lotfi. 1965. Basic Course in Azerbaijani. [Uralic and Altaic Series, 45]. Bloomington: Indiana University Publications. Imnaisvili, I. 1957. Saxelta bruneba da brunvata punkciebi jvel kartulsi. [The declension of nouns and the function of cases in Old Georgian.] Tbilisi: Universit'et'i. Jikia, Sergi. 1964. "Turkul-lazuri enobrivi urtiertobidan: Turkuli leksik'uri k'alk'ebi lazursi". [On the Turkish-Laz linguistic relationship: Turkish lexical calques in Laz.] Tbilisis Universit'et'is Šromebi 108, Aymosavletmcodneobis seria, 4, 253269. . 1967. "Turkul-lazuri enobrivi urtiertobidan: 2. Turkuli sint'aksuri k'alk'ebi lazursi". [On the Turkish-Laz linguistic relationship: 2. Syntactic calques of Turkish in Laz.] Orioni: Ak'ak'i Sani3es, 367-77. Tbilisi: Universit'et'i. . 1974. "Turkul-lazuri enobrivi urtiertobidan: 3. Turkul bgerata gadmocemisatvis lazursi". [On the Turkish-Laz linguistic relationship: The transfer

162

ALICE C. HARRIS

of Turkish sounds into Laz.] Tbilisis Universit'et'is Šromebi  8-9 (155-156), 115128. . 1975. Turkul-lazuri enobrivi urtiertobidan: 4. Turkuli leksik'uri element'ebisatvis lazursi. [On the Turkish-Laz linguistic relationship: On Turkish lexical elements in Laz.] Tbilisis Universit'et'is Sromebi 164.59-72. Kaldani, Maxime. 1959. "Y-s genezisisatvis svanursi". [On the origin of [y] in Svan.] Kartvelur enata st'rukt'uris sak'itxebi 1.91-99. K'art'ozia, Guram. 1968. "Masalebi lazuri zep'irsit'q'vierebisatvis". [Materials on Laz traditional oral literature.] Kartuli lit'erat'uris sak'itxebi, 132-78. Tbilisi: Mecniereba. Kipsid3e, I. 1914. Grammatika mingrel'skogo (iverskago) jazyka. S.-Peterburg: Akademija Nauk. Kogian, Fr. S.L. 1949. Armenian Grammar (West dialect). Vienna, Austria: Mechitharist Press. Kuno, Susumu. 1973. The Structure of the Japanese Language. Cambridge, Mass.: MIT Press. Lord, Carol. 1973. "Serial Verbs in Transition". Studies in African Lingusitics 4.26996. . 1993. Historical Change in Serial Verb Constructions. Amsterdam: John Benjamins. Mallinson, Graham & Barry J. Blake. 1981. Language Typology: Cross-Linguistic Studies in Syntax. [North-Holland Linguistic Series, 46.] Amsterdam: NorthHolland. Mart'irosovi, Ar. 1979. "Saxelis gansazγvruloba-ganusazγvrelobis k'at'egoria 3vel kartulsi". [The definite-indefinite category of nouns in Old Georgian.] Arnold Cikobavas ed. by Sota 3i3iguri, 125-138. Tbilisi: Mecniereba. [Mouraviev] Murav'ev, Serge N. 1981. "Tri ètjuda o kavkazsko-albanskoj pis'mennosti". C'elic'deuli 8.222-331. Mouraviev, Serge N. 1990. "The Aluanian (Old Udi) month names". The Annual of the Society for the Study of Caucasia 2.43-49. Nadareisvili, L. 1962. "Mimyeoba zanursi". [The participle in Zan.] IberiulK'avk'asiuri Enatmecniereba 13.177-190. Pancvi3e, VI. 1960. "Nazmnari saxelis mier saxelis martva brunvasi udur enasi". [Government of a noun in a case by a deverbal noun in the Udi language.] IberiulK'avk'asiuri Enatmecniereba 12.405-409. . 1974. Uduri enis gramat'ik'uli analizi [Grammatical analysis of the Udi language]. Tbilisi: Mecniereba. Pätsch, Gertrud. 1971. "Zur Entwicklung der attributiven Wortfolge im Georgischen". Bedi Kartlisa 28.253-261. Plank, Frans. 1990. "Suffix Copying as a Mirror-Image Phenomenon". Linguistics 28.1039-1045. Pocxua, B. 1962. "Sit'q'vatganlagebisatvis kartulsi". [On word order in Georgian.] Iberiul-K'avk'asiuri Enatmecniereba 13.109-122. Sanige, Ak'ak'i. 1942. "Mde tandebulis genezisisatvis". [On the origin of the adposition mde.] Sakartvelos SSR Mecnierebata Ak'ademiis Moambe 3:4.367-371.

WORD ORDER IN GEORGIAN

163

. 1973. Kartuli enis gramaVik'is sapu vlebi. [Fundamentals of the grammar of the Georgian language.] Tbilisi: Universit'et'i. Schmidt, Karl Horst. 1979. "Probleme der Komparation". Saenatmecniero k'rebuli, 224-230. Tbilisi: Mecniereba. Serozia, Revaz. 1980. "P'ofencialisis k'at'egoria da mastan dak'avsirebuli zogi sak'itxi kartvelur enebsi". [The category of potential and some questions related to it in the Kartvelian languages]. Nark'vevebi iberiul-k'avk'asiur enata morpologiidan ed. by G. Bedosvili and B. Jorbena^e, 119-126, Tbilisi: Mecniereba. Stassen, Leon. 1985. Comparison and Universal Grammar. Oxford: Basil Blackwell. Tschenkéli, Kita. 1958. Einführung in die georgische Sprache. Zürich:Amirani Verlag. Underhill, Robert. 1976. Turkish Grammar. Cambridge, Mass.: MIT Press. Vogt, Hans. 1971. Grammaire de la langue Géorgienne. Oslo: Universitetsforlaget. Watkins, Calvert. 1964. "Preliminaries to the Reconstruction of Indo-European Sentence Structure". Proceedings of the Ninth International Congress of Linguists ed. b y H . G. Lunt, 1035-45. The Hague: Mouton. 3i3iguri, Sota. 1913.K'avsirebi kartul enasi. [Conjunctions in the Georgian language.] Tbilisi: Universit'et'i.

WORD ORDER AND THE FIRST PERSON IMPERATIVE CAROL F. JUSTUS University of Texas at Austin 0. Introduction Older Indo-European languages usually encode directive categories as suf­ fixes on the verb. As suffixes, these markers follow patterns usually associated with verb final word order, as one might expect of verbal qualifiers (Lehmann 1973:49-54). Compare the imperative paradigm in Hittite, the oldest attested Indo-European language (Friedrich 1960: 93,96,99). (1)

1s 2s 3s

-(l)lu -lit, -lut -i, -t, zero -u, -tu

peske-llu, segg-a-llu, iya-llu eslit, eslut pesk-i, saak, iya, huisnu-t peski-ddu, sak-du, iya-du [peski-tu, sak-tu, iya-tu]

"Let me give, know, do" "Let me be" "give, know, do, save" "Let him / her give, know, do"

In English similar first and third person imperative categories are expressed by the particle let(s) (Quirk et al. 1986 [1985]:829-830). (2) a. b.

Let ME sit here, Let's all work hard Let NO ONE think that a teacher's life is easy

(first person imperatives) (third person imperative)

While second person imperatives ("Sit here, Give me that!") are universal in language, third person imperatives are less frequent, and first persons are often thought to be non-existent. In fact first person imperatives are associated with a speaker beneficiary meaning, whether expressed by a verbal suffix on the main verb or by a sentence-initial particle. Grammaticalization of this category in Indo-European involves both word order change and the rise of new auxiliaries. Toshiko Orita initially pointed me to Japanese kudasai "please" and lexical kudasaru "give", and, with the help of a small research grant from San Jose State University's College of Humanities, began work on related issues. Discussion with Jaan Terje Faarlund later pointed me to the Norwegian data, while comments from Bill Darden led to careful rethinking of the idea of a first person imperative, and Yukiko Alam's data confirmed its crosslinguistic validity. Facilities at the Linguistics Research Center and those of the University of Texas made available to me by the Classics Department and the Center for Middle Eastern Studies made completion of this study possible.

166

CAROL F. JUSTUS

1. Person-differentiated imperatives and thefirstperson imperative Although first person plural imperatives, termed hortatory, are frequent enough, a first person singular imperative was termed 'voluntative' when it was found in Hittite (Friedrich 1960:139). Quirk et al. (1986 [1985]:830) call let of the English imperative a particle distinct from the transitive verb let meaning "permit, allow", but they do not discuss its origin as an imperative marker nor the categorial status of the first person imperative. This section focuses on crosslinguistic data comparable to English grammaticalization of let beginning with (1.1) an account of the grammaticalization of let(s); (1.2) examines the implied dative argument of similar erstwhile lexical verbs; and (1.3) adduces Hittite person-differentiated imperatives as they illustrate nuances proper to the first person imperative. 1.1 English let as a person-neutral imperative marker While Quirk et al. distinguish the imperative particle let from the transitive lexical verb meaning "permit", Hopper & Traugott (1994:10-14) analyze meaning changes by which lexical verb let gradually came to be a grammatical marker of the imperative in English. They argue that second person imperative let ("Let [allow] us [to] go, Let yourself in, Let Bill go") and a contracted form of cohortative let us ("Let's go to the circus") are restricted uses that resulted in a particle lets with a more general directive function (Hopper & Traugott's 1994:11 with examples). (3) a. b. c.

Lets you and I take 'em on. Lets you go first. Lets you and him fight.

(complex predicate "lets ... take") (complex predicate "lets ...go") (complex predicate "lets ... fight")

The grammaticalized use of lets spread so that second person imperative let became reanalyzed as a person-neutral directive marker as in "Let him speak now or forever hold his peace" (Hopper & Traugott 1994:13). The proc­ ess started from a restricted use of lexical let as an imperative and spread. Once lexical let underwent the semantic change that reanalyzed the final s of lets as part of a new word; it was a grammatical marker. 1.2 Dative-argument verbs with speaker-bene)active uses A similar process seems to be occurring in modern colloquial Norwegian, but with the verb få "get, receive", a verb that has an implied dative argument, while Japanese verbs meaning "give" already have auxiliary functions.

167

WORD ORDER AND THE FIRST PERSON IMPERATIVE

1.2.1 Imperative uses of Norwegian få "receive, get". In modern colloquial Norwegian, the imperative få "get, give" has undergone a shift in meaning in some contexts. Whereas the present indicative få means "receive, get", imperative få adds an agentive force to an action benefiting a first person in simple and complex predicate constructions as in (4) below (cf. Faarlund 1985:151-152). (4) a.

Simple Predicate får "I get, receive" Eg FÅR smøret. "I GET the butter."

Imperative få "get, give": FÅ smøreu "GIVE me the butter."

The non-agentive indicative "receive, get", as an imperative, comes to have an agentive meaning, "make me receive, give me, get me" or "do getting for my benefit". This semantic shift continues when, in complex predicates indicative "get go", as imperative, means "let me go", "get see" becomes "show me", "get hear" becomes "tell me", and "get borrow", "lend me". Nonagentive indicatives får gå "get to go", får sjå "get to see", får høra "get to hear", and får låna "get to borrow" thus contrast with agentive imperatives with få in få gå "let go", få sjå "show", få høra "tell", and få låna 'lend". (4)b.

Eg I

FÅR GÅ GET (TO) GO

tidleg. early.

c.

Eg I

FÅR SJÅ GET(TO)SEE

kva what

FÅ SJÅ SHOW (me)

kva what

FÅR HØRA GET (TO) HEAR

kva som skeddje. "what happened."

d.

Eg I

FÅ HØRA TELL (ME)

e.

Eg I

FÅGÅ LET (me) GO

tidleg! early.

du you

har have (brought)

med. with.

du you

har have (brought)

med\ with.

va som skeddje! what happened!"

får låna get (to) borrow

sykkelen din. "your bike."

Få låna Let (me) borrow = lend (me)

sykkelen din.' "your bike!"

The agentive force, whether causative or permissive, varies with the meaning of the main verb ("Let me go, Make me see = Show me!"). What does not vary is the recoverability of the speaker as beneficiary. As Faarlund (1985:150-151) noted, imperatives are characteristically formed only from verbs that have a semantic controller, someone who can intentionally instigate the action. The imperative is thus syntactically a (second

168

CAROL F. JUSTUS

person) agent deletion process. Transitive verbs like "defend, cut, throw" with agentive subjects naturally form imperatives ("Defend yourself! Throw the ball! Cut the cake!"), while non agentive intransitive verbs such as "be an­ noyed, occur, receive" form quite peculiar imperatives, if at all: *Be annoyed! *Occur! *Receive me the book! Subjects of "annoy, occur, receive" do not intentionally instigate the action, hence the semantic difficulty. Norwegian få "receive" is semantically a verb that does not form an imperative. An imperative form with an understood controlling second person agent thus means "give", not "receive". Making få imperative shifts its meaning from non-agentive "receive" to agentive "give". This hereby restricted imperative context, now with its dative argument covert, implicitly fills the dative slot with the speaker resulting in an implicit added speaker beneficiary meaning. Imperative auxiliary få, as a new grammatical marker, thus encodes two covert discourse persons, both the speaker as the dative/beneficiary argument of "receive" and the second person as agent argument with an imperative. Like English "let(s)", få is an imperative that has become the grammaticalized head of a complex predicate, but with added discourse meaning. Lexical verb får "let" comes to encode, in its limited use as the imperative få of request, the category, speaker beneficiary. Auxiliary uses of "give" in Japanese offer well-studied examples of a verb of giving encoding a speaker beneficiary. 1.2.2 Japanese KUDASAI "give, please". There are no less than seven distinct Japanese verbs of giving with both lexical and grammatical functions (Kuno 1976 [1973]:127-135). Among them kudasaru and kureru indicate that the speaker (or someone associated with the speaker) benefits from the action, each encoding different honorific relations between speaker and hearer (Kuno 1976 [1973]:130-131; Nakau 1976:479-480). An example of the verb of giving, kudasaru, when it has an auxiliary function, comes from Kuno (1976 [1973] :131 (15a)). (5) a.

Sensei ga teacher subject

BOKUNIhon  me dat book object

onde KUDASATTA reading gave (the favor of)

"The teacher GAVE ME (THE FAVOR OF) reading a book",

or, with Ishiguro (1991:234): "The teacher KINDLY read ME a book."

A different verb of giving, yaru, indicates that someone other than the speaker benefits from the action (Kuno 1976 [1973]:129 (7b)).

WORD ORDER AND THE FIRST PERSON IMPERATIVE

(5)b.

169

John ga

MARY NI

kono hon o

YATTA

John subject

Mary dative

this book object

gave (the favor of)

"John KINDLY GAVE MARY this b o o k . "

Alam (1994) defined the grammatical feature that Japanese encodes with kudasaru as benefactive deixis, a deixis that contrasts with the locative deixis encoded in English motion verbs "come" and "go". While an overt speaker recipient marked with postpositional dative ni is characteristic of lexical uses, with grammatical uses the beneficiary is the (covert) speaker, and an overt dative is ungrammatical. In fact the deictic properties of such verbs make person marking (which Japanese does not do) unnecessary as in (4d-e); (cf. Alam 1994 (19a-b); see also note 8 below). c.

Soko-e itte-kureru? there-to going-give? "(Can you) go there (FOR ME)?"

d.

Soko-e itte-yaru. there-to going-give. "(I'll) go there (FOR YOU)."

Japanese verbs of giving in fact lexically encode "a culture-bound notion of reciprocity" in which social 'debt' and 'credit' favors are noted (Alam 1994:9-10). When kudasaru (as an auxiliary) and yaru (as event verb) co-occur, the speaker is asking as a personal favor an action that benefits both the speaker (kudasai) and a non-speaker (yatte): Kuno (1976 [1973]:135 (26c)). (5) e.

Sign  sign object

site doing

YATTE KUDASAI giving (SOMEONE favor of) give (ME favor of)-IMP

"GIVE (ME the favor) of GIVING HIM/HER (the favor) of signing",

i.e., "Please (for ME: kudasai) kindly sign (for HIM/HER:yatte )."

In English polite "please" corresponds to Japanese kudasai. For a Japanese speaker grammatical kudasai "please" is as psychologically separate from lexical kudasaru "give (as a favor to the speaker)" as English "have" of "I have fed the dog" is from "have" of "1 have a dog". Like English 'let(s)" and Norwegian få "get, give", the restricted use of Japanese kudasai as an imperative heading a complex predicate with the preceding gerund has undergone a semantic shift. The semantic path from lexical "give (as a favor to the speaker)" to grammaticalized imperative kudasai "please" in complex predicates deserves more study, as do paths from English lexical verb "please" or German bitten "ask" to polite particles "please" and Bitte "please". What we

170

CAROL F. JUSTUS

can, however, conclude is that Japanese kudasai "give" has a directive function associated with a speaker beneficiary, as does Norwegian få "get/receive". Alam's work on Japanese parallels Faarlund's on Norwegian. In English, too, there is an implicit speaker benefit associated with the first person imperative with let as in Let me sit here, Let me open the door. 1.3 The suffixal first person imperative of oldest Indo-European Hittite The Hittite imperative paradigm corresponds in function to that of the English paradigm with let. Both express the same imperative categories but with quite different formal strategies. While SVO English uses the sentenceinitial particle let(s) with personal pronouns, Hittite, a language with verb-final order, uses instead person-differentiated suffixes on the sentence-final main verb. Recall forms with 'give' and 'know'. (6) a.

Hittite

"give"

"know"

1s 2s 3s

peske-llu "Let me give" pesk-i "Give!" peski-ddu "Let him/her give,"

segga-llu "Let me know." sak(-ø) "Know!" sak-du "Let him/her know."

Hittite prayer Requests illustrate the nuanced range of meanings associated with person-differentiated imperatives. The Hittite king, Muwatalli (early XIIIth century ), in an attempt to motivate the Stormgod to aid him, prayed using the first person imperative. (6)b.

ispanduzd-ya kuin peskimi n=an-ta [dusg]arauwanza peskellu libation=and which-acc give-Is ptc=it=you [g]ladly-nom give-let-me "The libation that I usually give1, LET ME CONTINUE  GIVE you gladly." (CTH 381 A III 65-66 =  IV 33-35, Lebrun 1980:268,282.)

Here the king addresses this first person imperative as a request to the Stormgod, implying that the Stormgod can make the libation gifts worthwhile, presumably by a response that shows the efficacy of such gifts. The king, in requesting that he be happy about his sacrifice to the Stormgod, thus implicitly asks that his act be in fulfillment of part of the divine-human covenant, i.e., in his own (speaker's) interest. And it will be, if the Stormgod fulfills his part. The Hittite first person imperative, peskellu, is morphologically part of a multi-person imperative paradigm. Friedrich (1960:139) termed it a "voluntative", third person imperative an "optative", and first person plural imperative a "cohortative". Hittite peskellu is built on the iterative-durative stem pe-sk(i)- from the the verb pai- Ipiya"give".

171

WORD ORDER AND THE FIRST PERSON IMPERATIVE

Muwatalli's father, Mursili (II), in a prayer for relief from a plague, used all three persons of the imperative, demonstrating the nuanced differences among them (CTH 378 = Goetze 1927:214, lines 5-6; Lebrun 1980:207, lines 19-24). (6) .

 SISKUR SHA íd[Mala] iyallu ptc offering of (river)Mala make-let-me n =at=kan asnullu ptc=it=ptc establish-let-me 'The offering for the Mala River LET ME MAKE , LET ME ESTABLISH it (as a rite)."

The first person imperative indicates the king's intent that the offering to be established be a gift that will make the king want to have made it. The king's desire to make this offering is motivated by future benefits that he will get for it from the deity. This becomes clearer from the second and third person passages that follow. (6) d.

iyami=ya=at=za make-ls=and=it=ptc

kuedani which-d/1

uddani matter-d/1

hingani plague-dA

ser for

ginzu datten nu=mu DINGIR mesh BELUmesh=JA ptc=me god(plural) lord(plural)=my mercy take-2p-IMP "and because of which matter, because of the plague, I do this, O gods, my lords, (so) TAKE MERCY on

e.

me;"

nu- INA SHÀ KUR uruHatti hingan lazziyattaru ptc=ptc in heart land (city)Hatti plague subside-let-it "(in that you) LET the plague in the heart of the Hatti land SUBSIDE."

The force of these imperatives emerges naturally from the divine-human contract on which the older IE prayer was predicated: "let me do (first singular imperative), so that you do (second person imperative), so that X happens (third singular imperative)." All imperatives have elements of voluntatives to the extent that they express actions undertaken in the speaker's interest, and all make requests that appeal to a hearer who may, in responding with empathy for the speaker, requite the speaker's acts of devotion (sacrifice for the Mala River) and therefore bring about the desired (third person) effect. The speaker's acts and requests overtly invoke the deity on behalf of the speaker (first person imperative: 6c), then suggest an appropriate hearer's (psychological) response (second person imperative: 6d), before finally specifying the concrete deeds needed (third person imperative: 6e), here in terms of the effect on the (speaker's) Hatti land.

172

CAROL F. JUSTUS

In business letters the first person imperative seggallu "let me know" indicates what the speaker wants, his end goal (XXII 88 Vs. 5f. CTH 175; = Justus 1981:24 example 38: 6-7 and KBo XVIII 2 Rs. 8ff., = Justus 1981:37,86). (6)f.

nu=mu=za ptc=me=ptc

man when

SHESH=JA brother=my

kuwapi pi[esta] whether gfave]

nu=mu Hatrai nu seggallu ptc=me write-2s-IMP ptc know-let-me "WRITE to me whether (and) when my brother g[ave]; let me KNOW." namma =mu further=me man =mu ptc=me

man whether kuiski someone

uppai sends

UL kuiski not someone

uppai sends

nu=mu uppa[i kuit ] hatratten nu sikkallu ptc=me send[swhat] write-2p-IMP ptc know-ls-IMP "Further, whether no one sends to me, whether someone sends to me, [what he] send[s] me, write! LET ME KNOW!"

In the letter genre, first person imperatives of "know" state the way in which the speaker will benefit. An imperative of non-agentive "know" might more properly be interpreted as "let knowing happen for me" parallel to Norwegian "let receiving, getting happen for me", or "give knowing on my behalf", "let offering happen in my interest". 2.

Older Indo-European prayer requests Second person imperatives in Homeric Greek and Classical Latin prayer are comparable to those in Hittite and elsewhere in form and meaning. Third person imperatives range in form from the older Indo-European third person imperative suffix cognate with the Hittite third person form -tu l -du to the innovating optative mood in Greek and the subjunctive in Latin (Justus 1993:141-155). In addition, both Greek and Latin use a second person imperative form of "give" as a quasi-auxiliary. English recalls a similar construction in archaic prayer structures with "grant" ("Grant to me,  Lord, to know what I ought to know" Appleton 1988 [1985] :92, nr. 284; "Grant us,  Lord, not to mind earthly things, but to love heavenly things" Appleton 1988 [1985]:164,nr.544).

WORD ORDER AND THE FIRST PERSON IMPERATIVE

173

2.1 Suffixing strategies for prayer requests The Lord's Prayer recalls second person imperative request forms (7a) that still contrast with third person imperative requests (7b). (7) a.

King James

"GIVE US this day

FORGIVE US ...,

DELIVER US..."

our daily bread,...

b.

Good News

"GIVE US ...,

FORGIVE US ...,

KEEP US SAFE ..."

Latin Greek

DA nobis,..., DOS hëmon ...

DEMITTE nobis..., APHÈS hêmôn ...

LIBERA nos... HRúSAI hemàs ...

Gothic Latin King James Good News Greek

QIMAI thiudinassus þeins, ADUENIAT regnum tuum, Thy kingdom COME, MAY your kingdom COME, elthâtōhēbasileia sou

WAIRAI wilja peins opt. FIAT uoluntas tua subj. Thy will BE DONE subj. MAY your will be DONE aux. genēthetōtòthélëmà sou imp.

While second person imperatives have formally predictable endings (or lack of them), third person endings range from the old suffix cognate with the Hittite third person imperative (Greek) to a new modal auxiliary (English). Gothic uses the optative mood for a third person imperative, while Latin and King James' English use a subjunctive. The English of the Goods News version then makes the subjunctive more explicit by adding the modal auxiliary may. Greek, with the old third person imperative (-tō) cognate with Hittite -tufdu, continues the Homeric third person imperative form in -tō (Justus 1993: 140-141). The Homeric third person prayer imperative, however, is preserved only in the formula "let Zeus witness" (Iliad 10.329 and elsewhere). (8)

istö nun Zeùs know-let-him now Zeus "LET Zeus himself now KNOW."

autós himself

The more usual Homeric form for expressing this third person directive is the optative mood. With the verb "pour", the third person singular optative has the suffix -oi: réoi / hréoi "may it pour" (The Greek army prays to Zeus, Iliad 3.300), while the third person plural (aorist) optative suffix on "pay" is -eia(n): tiseian "let them pay" (Chryses prays to Apollo, Iliad 1.42). (9) a.

b.

hōdé sph'enképhalos khamádis hréoi hōshóde oinos thus their brain(s) earth-ward pour-let-it as this wine "LET their brain POUR out upon the ground like this wine." tiseian Danaoi ema dákrua soisi bélessin PAY-LET-THEM (FOR) Danaeans my tears with your with arrows "LET the Danaeans PAY for my tears with (the damage from) your arrows."

174

CAROL F. JUSTUS

Other optative prayer requests include passages in the Odyssey (20.61ff.; 20.80; and 21.200-201; Justus 1993:135-136). In Latin, which has no optative, the subjunctive mood (Justus 1993:147148) replaces older third person imperatives (cf. ā-subjunctive cern-â-t "let him see" in Aeneid 10.462): (10)

CERNĀT semineci sibi me rapere arma cruenta "LET HIM SEE me strip him, half-dead, of his bloody armor."

Other Latin subjunctives of prayer request include passages from the Aeneid (4.689-691; 10.460-463) and Cato's agricultural prayers (e.g., de agri cultura 141.2; Justus 1993:137-138). Latin quite regularly used third person subjunc­ tives, where in Homeric prayer the third person optative was frequent (Justus 1993:143-151). Inflectional third person imperatives alternating with equally inflectional optative or subjunctive suffixes persisted in IE third person prayer requests for a long time. In the Lord's Prayer, koine Greek still used a third person imperative, but Latin, although it still had cognate third person imperative {-to) forms (Justus 1993:138-139), used its subjunctive, and Gothic its optative for what is translated by the King James' subjunctive and the Good News version with auxiliary "may" plus subjunctive. When English translations of the Lord's Prayer differ in the extent to which they do or do not continue a third person imperative, the range of non-imperative forms attesting a multiplicity of language-specific strategies points to innovation on the earlier third person imperative mechanism, the auxiliary strategy being the most recent. If we reconstruct an imperative suffix as original (Justus 1993:152), then it is not surprising to find the later Gothic, Latin, and English prayer forms with such diverse strategies. While English let(s) now encodes all three persons of the imperative,2 between the old Hittite first person verb-final suffix and the use of English let as a sentence-initial first person speaker beneficiary particle lie both suffixal and non-suffixai strategies. A non-suffixai strategy prominent in Homeric Greek and Classical Latin prayers parallels the Norwegian and Japanese first person imperative auxiliary strategy and the English "grant" of archaic prayers. The imperative "give" as a directive marker in Greek, Latin, English, however, appears with a new word order patterning comparable to English let(s). While Quirk et al. (1985:829-830) find lets as the directive particle with first and third persons, Hopper & Traugott (1994:10-14) additionally find examples like "Lets wash your hands" or "Lets eat our liver now, Betty" when spoken to children as second person imperatives.

WORD ORDER AND THE FIRST PERSON IMPERATIVE

175

2.2 Lexical and (quasi-)auxiliary "give" In Homeric Greek and epic Latin prayer requests, second person imperatives of verbs of giving have both lexical and restricted auxiliary uses. They thus take both simple abstract noun objects of request and entire propositions. Simple noun phrase objects of "give" are comparable to those with other main verbs expressing a prayer request. (11) a. Hittite n =an kez inanza HUISNUT 3 b. Greek PHULÁSSETE d ' hórkia pista c. Latin DA deinde auxilium

"RESTORE her from this illness! " "PROTECT then the faithful oaths ! " (Iliad 3.276-277) "Give finally aid!" (Aeneid 2.689)

Constructions in (11) are similar to the Lord's Prayer example, "GIVE us this day OUR DAILY BREAD" in which the request content is the noun phrase Object of the verb. (11) d. VERB-imperative + Request [NP OBJECT]

The abstract noun expanded to a noun clause after the imperative of "give" has the form of an accusative-infinitive construction but is otherwise comparable to the unexpanded noun. (11) e. GIVE-imperative + Request [NP] (11a-c above) f. GIVE-imperative + Request [NP-acc + INFINITIVE (+ NP)] (12-17 below)

In Greek and Latin prayers the accusative-infinitive construction with "give" forms a complex predicate: "give to become, give to settle, give to escape". Expansion of the Request content then as an accusative-infinitive construction dependent on second person imperative "give" serves as the restricted context in which "give" undergoes the semantic shift that ultimately separates it from lexical "give". 2.2.1 Latin complex predicates with "give". Typical Latin prayer petitions with quasi-auxiliary 'give' include the accusative-infinitive construction in (12: Aeneas to Jupiter, Aeneid 5.691). (12)

DA flammam EVADERE GIVE-s/ag + REQUEST [ [N-acc INF "Grant (us) that the flame pass by the fleet!"

classi [N-abl] ]

(CTH 380 Obv. 16'-17' Tischler 1981:12); inanza huisnut are written logographically with phonetic complements: GIG-za TI-nut.

176

CAROL F. JUSTUS

The Latin verb dare "give" normally takes as Subject (S) an Agent (AG) of the giving, a Dative or Beneficiary, and an Object of the giving. Here the speaker is the covert dative / Beneficiary. The vocative pater "father" is the Agent/Hearer to whom the imperative is addressed, and the accusativeinfinitive construction is the Object. In the accusative-infinitive unit accusative (singular) flammam "flame" is the subject of the infinitive evadere "to pass by" [N-acc INF], and ablative classi "from (the) fleet" is an ablative object of the infinitive where the prefixed e(x) governs the ablative case. Aeneas here is asking on behalf of himself (and his group) that Zeus save their ships from being burned. The speaker, as dative argument of "give", however, is covert. Where we might have expected a nobis "to us", we must recover it from the semantics of "give". An unexpressed dative argument with this restricted use of "give" refers to the speaker. The speaker as the (benefiting) dative argument has been incorporated in the meaning of this auxiliary use of "give": "give (me)". In Japanese overt expression of the dative speaker with kudasai "please" (give) is ungrammatical because it is implied in the new auxiliary meaning. Speaker deixis, as part of the grammaticalized meaning of kudasai, has already filled the dative slot, hence the ungrammaticality of an overt dative (Alam, personal communication). Lack of an overt dative with Latin imperative uses of "give" in complex predicates suggests that the speaker as beneficiary here too is implicit enough to render any other overt dative ungrammatical. The difference between Latin examples with da and Japanese kudasai lies primarily in word order. In (13) Arruns (Aeneid 11.789), as he lies in ambush, asks Apollo to help him kill Camilla, who has been successful in battle against his side. Accusative hos ... dedecus, the subject of the accusative-infinitive Request, is expressed with characteristically nonlinear Latin constituent order. (13)

DA, pater, give-2s-IMP father-voc give-s/ag S / AG -voc "GIVE (US), father, (+ REQUEST): hos nostris ABOLERI dedecus armis these-acc our-dat to-be-abolished dishonors-c arms-dat [ [N-acc [MOD-dat] PASS INF N-acc [N-dat] ] these dishonors to our arms TO BE ABOLISHED", or "Let the power of our arms abolish these dishonors (for us)."

The accusative subject of passive infinitive aboleri "to be abolished" is discontinuous hos ... dedecus "these dishonors" and equally discontinuous

WORD ORDER AND THE FIRST PERSON IMPERATIVE

177

nostris ... armis "to our arms" is governed by dedecus "dishonors". The dative speaker relation implicit in "give" is again covert. In (14) the second person singular da "give" governs the accusative / infinitive construction with the content verb "settle" (Aeneas prays, Aeneid 6.65-67). (14)

o sanctissima vates, ... DA ... S /AG-voc give-s/ag "O holiest of prophetesses,... (+ REQUEST): Latio... CONSIDERE ... Teucros [ [N-dat/loc] INF [N-acc] ] GRANT the Teucros (TO) SETTLE in Latium", or "LET the Teucres (for the sake of me, a Teucrian) SETTLE in Latium."

The infinitive considere heads an event structure or proposition which is the Request, while the imperative da expresses the modality of the proposition, a modality which includes a directive involving both discourse persons, speaker and hearer. The imperative allows recovery of the second person agent / hearer and the semantics of grammaticalized "give" recovers the first person (speaker) as beneficiary. In (12-14) "give" governs the Request content as if it were a grammatical Object, and the accusative-infinitive construction behaves as an abstract (accusative) noun. Da "give" takes the proposition expressed formally as an accusative-infinitive construction, as Object: flammam evadere ..., hos ... aboleri, and ... considere Teucros. Infinitives evadere, aboleri, and considere, although they head the Request events, as infinitives, are dependent for tense, mood, and aspect on the grammar of da "give". Semantically, da "give" and the infinitive thus form a complex event structure in which da encodes tense, mood, aspect, and discourse participant roles and infinitives, as main event verbs, govern the noun arguments of the Request events. 2.2.2 Greek "give" of prayer petition. As in Latin, imperative "give" in Homeric prayer introduces the prayer Request as the second part of a complex predicate. Using a form of "give" cognate with Latin da (PIE *dö-), Homeric prayers usually address deities with dós (second person singular imperative) or dóte (second person plural imperative). Homeric prayer petitions, however, also use imperatives of krainō "accomplish, bring to pass, grant". With formulaic (epi)krēēnon "grant, bring to pass" the dative speaker who benefits is overt {moi "me"), and the Object noun, tόd' ... eéldör "this wish", cataphorically points to the accusative-infinitive Request proposition that stands in apposition to it (Chryses prays to Apollo (Iliad A 1.455-456).

178

CAROL F. JUSTUS

(15)

... moi

tód'

EPIKRÉENON

[me-dat] [this-nom/acc] grant-2s-IMP "GRANT this wish for me (+ REQUEST):

eéldör,

[wish-nom/acc]

dē ũ Danaolsin aeikéa loigòn ÂMUNON [ particles N-dat [ADJ-acc N-acc] ward-off-2s-IMP ] Now too WARD the unseemly plague OFF (from) the Danaeans."

Greek epikrëenon "grant", as a variant of "give", introduces a Request proposition, but, unlike with "give", both the dative beneficiary and an Object are overt. The elements grammaticalized with dós, dóte "give" are explicit with kraìnō. Other such examples with krainö include (Iliad 1.37-42, 1.445-456, 8.242-244,1.502-510,514-516, 8.236; Odyssey 17.240-246, 20.112-119). Greek dós "give", like Latin da, takes an accusative-infinitive construction as Object, but without the cataphoric noun and with or without an overt dative speaker "me". In example (16) Homeric Greek accentual constraints affect the word order so that the unaccented imperative dòs "give" in second position is enclitic on stressed tón "him", the accusative subject of the infinitive dũnai "to enter". As a result, the Request Object linearly surrounds the auxiliary use of "give" (Agamemnon swears and prays, Iliad 3.323). (16)

ton DÒS apophthímenon DŨNAI dómon Aidos eisō [[him-acc]] GIVE-s/ag [dying-acc TO-ENTER [house-c [N-gen] into] ] "GIVE (that) him dying (he) ENTER the house of Hades", i.e. "(For my sake) LET him die and ENTER Hades."

Request verb, dünai "enter (into)", governs two noun phrases, its accusative subject, tòn ... apophthimenon "him dying", and the locative object, dómon eisö "into (the) house". The postpositional phrase "house into" is further modified by the genitive of Hades ("into the house of Hades") much as the accusative participle, "dying", modifies "him"). While dòs 'give' encodes the modality of the construction, expressing tense, mood, aspect, and discourse relations, the infinitive, dünai "to enter" governs the arguments of the proposi­ tion. Together "to enter" and the imperative of "give" form a complex Request event made in the interest of the speaker. In (17) second person plural imperative dóte "give" governs the Request Object "him to become distinguished among the Trojans" (Hector prays for his son,Iliad Z 6.476-4787). (17)

DÓTE de kal give-2p-IMP particles '%ut GIVE (me) indeed (+ REQUEST): tón=de GENESTHAI paĩd' emón, ... ariprepéa Troöessin [ [3s-acc] TO-BECOME [N-acc my-acc] ... [ADJ-acc] [N-dat] ] my child,  BECOME, ... distinguished among the Trojans", i.e. "For me LET my child BECOME distinguished among the Trojans."

WORD ORDER AND THE FIRST PERSON IMPERATIVE

179

Accusative paĩd' emón "my child" belongs with tón(de) "him, this one" as "this my child". The predicate adjective ariprepéa "distinguished", with its dependent dative Trōessin "among the Trojans", is the complement of "become". Here dote encodes tense, mood, aspect, and the speaker as benefici­ ary of the proposition, while the infinitive genésthai "to become" governs its accusative subject "this my child" and its adjectival complement "distinguished among the Trojans". Other uses of Homeric dòs in prayers include five from Iliad (3.320-324, 3.351-354, 3.365-370, 5.115-120, 6.305-310, 7.202-205) and four from the Odyssey (10.278-282, 16.514-526, 17.645-646, 6.323-327). The imperative auxiliary "give" in (12-17) has shifted from its lexical meaning to a grammatical modal meaning "may, let on behalf of, in the interest of the speaker". As a result, both the speaker as beneficiary and the addressee as agent are implicit in grammaticalized imperative "give", which now heads a complex predicate with an accusative-infinitive Object. Latin imperative auxiliary "give" (12-14) similarly grammaticalized the meaning "on behalf of the speaker". The prayer requests of (12-14) are speaker deictic "give ME": "give (me)" da (mihi)flammamevadere (12), da ... [mihi] hos ...aboleri dedecus 'give [me to have] this dishonor abolished' (13), da [nobis Teucri] considere Teucros "give [us Teucrians to have] the Teucres settle Latium" (14). More idiomatic translations include: "Do me the favor of (letting the flame pass by the fleet", "letting this dishonor to our side be abolished", "letting the Teucres settle in Latium"). Homeric Greek similarly encoded both an agentive hearer and a benefiting speaker in the semantics of imperative auxiliary "give". Formulaic examples like (15) moi tód' epikre non ... "fulfill for me this ...", i.e., "On my behalf, fulfill this wish ..." overtly specify that the speaker is the dative / beneficiary and that there will be an appositionally extended Request Object. In (16) Agamemnon prayed that a person who had violated the oaths of the gods should, in recompense for his [the speaker's] good faith, enter Hades: tòn dòs (moi) ... dũnai dómon Áidos "give (me) for him to enter the house of Hades", i.e., "(For me who observes the oath), let him (the oath breaker) go to Hades". Similarly, the speaker's (Hector's) interest is recoverable from the semantics of "give" and, redundantly too, from the first person possessive "my": dóte (MOI) d kal tón=de genésthai paĩd' EMÓN "give (ME) indeed too him, MY child, to become...", i.e., "For me let my child become distinguished ...". The more usual examples with reflexes of PIE *dö- have grammaticalized the pragmatic speaker-hearer relation. As in Japanese, the society of older Indo-European prayer was also a society of social debits and credits. Prayer was the form that asked the deity to

180

CAROL F. JUSTUS

make good on the debit incurred when a human being, through sacrifice, earned a credit (Benveniste 1973: 53-100, 481-486, 501-503; Mauss 1925: 31ff; 50-60, 53, 59).4 Use of a first person imperative (expressed in Hittite by a suffix) or its equivalent (Greek and Latin auxiliary "give") grammatically encoded this debit-credit relation. The fact that Norwegian imperative fă "receive, get" and Japanese imperative kudasai independently grammaticalize the speaker as an implied beneficiary offers crosslinguistic evidence for speaker deixis as a category with the imperative. 3. Summary and conclusions Imperatives belong to the category of modality, the category that operates on a proposition and usually includes tense, mood, and aspect. With the imperative mood, tense is characteristically neutralized as present tense, and the person is characteristically the second person, as directives are addressed to a hearer. Examples here have shown, however, that there are first (and third person imperatives) as well. When a directive has the purpose of benefiting the speaker, a language may grammaticalize the speaker as beneficiary of the imperative, above and beyond the usual second person implied in the imperative, by using a special imperative marking with speaker deixis. Forms for marking the directive modality with speaker as beneficiary vary chronologically in the Indo-European languages examined here and crosslinguistically in non-Indo-European Japanese. (18) a. Hittite

4

b. Japanese

[ispanduzzi(n) ... pesk e] -llu "Offering giving let me do." [Request Proposition] Speaker Deixis Modality [sign-o sitte ] kudasai. "Sign doing let me do." [Request Proposition] Speaker Deixis Modality

 Homeric Greek

dóte "Let (for me) Speaker Deixis Modality

[... genésthai paĩd' emón ariprepéa] my son become distinguished." [Request Proposition]

d. Classical Latin

da "Let (happen for me) Speaker Deixis Modality

[Latio considere Teurcros.] the Teucreans to settle Latium." [ Request Proposition ]

e. Norwegian

Få "Let (me get) Speaker Deixis Modality

[ låna sykkelen din. ] borrowing of your bike." [ Request Proposition ]

Since this study was done, Newman's (1998) crosslinguistic study on the grammar of "give" has appeared, too late for consideration here.

WORD ORDER AND THE FIRST PERSON IMPERATIVE

f. English

"Let (for me) Speaker Deixis Modality

181

[ me sit here."] [ Request Proposition ]

Examples ranging from the second millennium  Hittite to contemporary English over a period of nearly 4000 years of Indo-European languages5 change not only word order patterning but the mechanisms for expressing a first person imperative, a category also expressed in the genetically unrelated Japanese. Old verb-final Hittite (second millennium ) encodes this modality using a suffix on the event verb of the Request proposition. When verb-final Japanese also encodes the speaker-benefiting imperative at the end of the Request proposition, it uses a special speaker-deictic auxiliary verb of giving. It is perhaps not typologically surprising then to find the parallel grammaticalization of a similar modal meaning with a verb meaning "give" in Greek and Latin. Norwegian by contrast uses "receive, get" as the source for its restricted auxiliary use as a speaker deictic imperative. The data here suggest, with Alam, that there is a speaker deictic category in language associated with directives. It also shows that this category may renew itself over time in new auxiliary verbs. Much more investigation of Indo-European verbs of giving and of the kinds of verbs that come to function as speaker-deictic imperatives now remains to be done.6 Did older Indo-European ever have multiple verbs of giving related to discourse person functions? This needs further study. Ramat (1982) and others have shown how "have" as a new auxiliary in Latin and the Romance languages grammaticalized categories of tense and aspect. Heine, Claudi, & Hiinnemeyer (1991:187-189) and others show that one semantic path of grammaticalization as languages develop new case markers is the path from activity (verbs) to discourse metaphor. In Ewe ná "give" comes to encode benefactive, purpose, and dative case functions, while other verbs are the source of other case markers, but in Marathi, finite forms of "give" as a 'vector verb' come to express an aspectual modality (Hook 1991:66-74). The data here, while involving a path from the activity "give" over discourse metaphor, do not arrive at new case markings or verbal aspect, but at a speaker deictic directive category. The data here thus supplement data To be precise, Hittite, an Anatolian language that flourished in the second millennium , has no modern descendants, while there are no Germanic ancestors of Norwegian and English attested before the first millennium AD. Modern Greek and the Romance languages more directly continue Classical Greek and Latin, so that changes here are indicative, not direct, but further study would be expected to confirm general directions. Newman's (1998) crosslinguistic study of grammatical structures with "give" only begins to do this.

182

CAROL F. JUSTUS

from other languages on the paths that verbs of giving may take in (re)creating grammatical categories. Lightfoot's (1979) study of the rise of English modal verbs included a set of abstract rules that led to creation of a new verbal category in a new position in the sentence. Formulation of rules underlying such changes depended on identification of crosslinguistic correspondences attested over centuries of data in widely differing expressions of a category that included differences both in word order and in grammatical encoding strategies. Identification of the category, speaker deictic directive, here emerged in the context of genreconditioned pragmatic sameness. The request function in the prayer served as a heuristic for plotting the range of category mutation in sample Indo-European languages (Justus 1993). Noteworthy in Greek, Latin, and Norwegian is the placement of the new auxiliaries at the beginning of the Request proposition rather than at the end. If we hypothesize that older Indo-European word order patterns were more like Japanese (Lehmann 1974) by contrast with modern western Indo-European languages such as English with its sentence-initial "let", then new sentenceinitial auxiliaries in Greek, Latin, and Norwegian reflect the movement away from verb-final to verb-initial patterns. If we consider that English auxiliaries, including the modal auxiliaries, similarly moved the expression of tense, mood, and aspect of Old English away from sentence-final verb suffixes, this pattern of word order change is not novel. It is, however, new to see an imperative category encoding speaker deixis involved in this process. The lexical verbs that we see coming to be restricted in grammatical auxiliary functions for expressing speaker deictic directives include verbs of giving, getting, and allowing, all verbs that govern a dative case. We may now ask: Is governing a dative argument a requisite of such verbs? Why would one language path be more likely to chose a "give" source, a "get" source, or a "permit, allow" source? Identifying what is the same across changing and vari­ able linguistic expressions of pragmatic purposes, however, is fundamental to working out the details of semantic paths and the frequency with which they repeat themselves. Not until such equivalencies are established is it meaningful to ask either about semantic paths or about formal rules (or parameters) underlying linguistic change and variation. Knowing that the category, speaker deictic (first person) imperative, resurfaces over the millennia, it now makes sense to further explore its range of expression and semantic relations as it mutates over time and space.

WORD ORDER AND THE FIRST PERSON IMPERATIVE

183

REFERENCES Alam, Yukiko Sasaki. 1994 ms. "Deixis and Implicit Argument of Japanese Verbs of Giving". To appear in Proceedings of the 4th International Conference on Southeast Asian Linguistics (SEALS IV). Appleton, George, general ed. 1992 [1988]. The Oxford Book of Prayer. Oxford & New York: Oxford University Press. Benveniste, Emile. 1973 [1969]. Indo-European Language and Society. (= Miami Linguistics Series, 12). Coral Gables: University of Miami Press. Faarlund, Jan Terje. 1985. "Imperative Control: First Person Imperatives in Norwegian". Nordic Journal of Linguistics 8.149-160. Friedrich, Johannes. 1960. Hethitisches Elementarbuch, vol. I. Heidelberg: Carl Winter. Goetze, Albrecht. 1927. "Die Pestgebete des Mursilis". Kleinasiatisches Forschungen 1:2.161-251. Hook, Peter Edwin. 1991. "The Emergence of Perfective in Indo-Aryan Languages". Approaches to Grammaticalization ed. by Elizabeth C. Traugott & Bernd Heine, vol. II, 59-89. Amsterdam & Philadelphia: John Benjamins. Hopper, Paul J. & Elizabeth Closs Traugott. 1994. Grammaticalization. Cambridge: Cambridge University Press. Ishiguro, Teruhiro. 1991. "On Japanese Benefactive Accessory Verbs". Doshisna Studies in English 52-53.224-237. Justus, Carol F. 1993. "Mood Correspondences in Older Indo-European Prayer Petition". General Linguistics 33:3.129-161. . 1981. sak(k)- / sek(k)- "Know". (= Materialen zu einem hethitischen Thesaurus, 7, Lieferung, 10.) Heidelberg: Carl Winter. Kuno, Susumu. 1976 [1973]. The Structure of the Japanese Language. 2nd printing. Cambridge, Mass.: MIT Press. Lehmann, Winfred P. 1973. "A Structural Principle of Language and its Implications". Language 49.47-66. . 1974. Proto-Indo-European Syntax. Austin: University of Texas Press. Lebrun, Réné. 1980. Hymnes et prières hittites. (= Homo Religiosus, 4.) Louvain-LaNeuve: Centre d'Histoire des Religions. Lightfoot, David W. 1979 Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Mauss, Marcel. 1925 [1923-1924] "Essai sur le don. Forme et raison de l'échange dans les sociétés archaïques". L'Anηée Sociologique 1.31-186. Nakau, Minoru. 1976. "Tense, Aspect and Modality". Syntax and Semantics ed. by Shibatani, vol. 5, 421-482. New York: Academic Press. Newman, John, ed. 1998. The Linguistics of Giving. (Typological Studies in Language, 36). Amsterdam & Philadelphia: John Benjamins. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1986 [1985]. A Comprehensive Grammar of the English Language. 4th Impression. London & New York: Longman.

184

CAROL F. JUSTUS

Tischler, Johann. 1981. Das hethitische Gebet der Gassulijawija. (= Innsbrucker Beiträge zur Sprachwissenschaft, 37.) Innsbruck: Institut für Sprachwissenschaft der Universität.

PART IV VARIATION AND CHANGE

VARIANT ORDER OF SURFACE SEGMENTABLES ON THE BORDER BETWEEN MORPHOLOGY AND SYNTAX THE CASE OF PRERADICAL VERBAL MORPHOLOGY IN KARTVELIAN*

MARCELLO CHERCHI The University of Chicago

0.

Introduction Language analysis has traditionally been divided into the domains of phonetics, phonology, morphology and syntax. While these divisions have helped scholars gain considerable insight into how language works, it is well known that many linguistic phenomena do not fit comfortably into only a single one of these domains.1 From the standpoint of synchronic linguistics, various analytical 'solutions' to such phenomena have been suggested (viz. the phenomenon in question 'really' only belongs to one domain,2 or it straddles

The author wishes to thank the participants of the "Workshop on stability, variation and change in word-order patterns over time" at the XIIIth International Conference on Historical Linguistics (August 14, 1997, Heinrich-Heine-Universität, Düsseldorf) for their questions and comments at the presentation of this paper. Detailed discussions with Professors Erich Poppe, Ariel Shisha-Halevy and Rosanna Sornicola have been especially informative. The paper has also benefited from the comments of an anonymous reviewer. Any errors or inaccuracies are solely the responsibility of the author. Naturally, this begs the question of whether the difficulty of classifying these phenomena is merely an artifact of the analysis itself. If we were to propose an analytical framework consisting of a continuous gradation from phonetics through syntax, any given linguistic phenomenon would be representable by a point (or, perhaps, by a set of points) along that continuum, and the problem of 'not fitting into a particular domain' would never arise. 2

This is a not uncommon occurrence in the history of the discipline of linguistics, perhaps because specialists of one particular area of grammar are often wont to encroach upon the material and phenomena treated by specialists of other areas. For example, Paul Kiparsky's "Revised Alternation Condition" in Lexical Phonology made reference to "derived environments" which is essentially referring to morphology (cf. Spencer 1991:105-107). Another good example is Stephen Anderson's (1984) use of a Word-and-Process morphologi­ cal framework to construe Georgian case marking as a strictly morphologically-controlled phenomenon rather than as a (prototypically) morphosyntactic one. A more recent example is Rochelle Lieber's (1992) attempt to design a framework in which morphology as such does not exist, for all of its functions purportedly can be taken over by syntax.

188

MARCELLO CHERCHI

two or more domains,3 or it demands its own entirely new domain4). From the standpoint of diachronic linguistics, such phenomena have often been recognized as being 'in transition' from one domain to another, and this has provided material for the growing discipline of grammaticization. Clitics (in Romance, South Slavic and elsewhere)5 constitute a good example of a phenomenon which has attracted considerable attention in recent literature both from the synchronic and diachronic perspectives. In the present paper we shall examine another such phenomenon which has received less attention, probably due to the fact that the language family involved has been less studied in the West, namely the Kartvelian family, comprising Georgian, Mingrelian, Laz and Svan. We will be interested in the synchronic variation of certain morphological elements preceding the verbal root in these languages. It is probably accurate to say that we will be discussing a set of related phenomena, since the variation in each language seems to be of slightly different types, and is arguably due to slightly different reasons. 1.

Overview of Kartvelian The languages of the Kartvelian family are spoken in the Caucasus mountains in the present-day Republic of Georgia, and to some extent in northeastern Turkey. The genetic relationship among these languages shown in (1) was proposed by Deeters (1930:2) and has basically been accepted by most scholars.6

3

For example, in Jerrold Sadock's (1991) theory of Autolexical Syntax, any linguistic item is described by, and participates in, several separate 'modules' of grammar simultaneously (viz. phonology, morphology, syntax and others). For example, the 'morphophonemic' level invoked by the post-Bloomfieldian structuralists (cf. Spencer 1991:52-57, Anderson 1985:301ff.), or Nikolaj Trubetzkoy's 'morphophonology' (Anderson 1985:1 12ff.; Singh [ed.] 1996). See Spencer (1991:ch. 9) for case studies and discussion. Deeters (1930:3) explains the particular arrangement of the lines in the genetic tree as follows: "zwar besitzt das Lasisch-Mingrelische einige Gemeinsamkeiten mit dem Swanischen, sie fallen aber nicht ins Gewicht gegenüber den zahlreichen gemeinsamen Neuerungen, die es mit dem Georgischen teilt."

VARIANT ORDER OF MORPHEMES

(1)

189

The Kartvelian family (adapted from Deeters 1930:2)

Georgian is the only language of the family which has a written history dating from the Vth century C.E. It was not until the XIXth century that linguists, ethnographers and scholars of folklore began collecting texts of the other languages.7 Although the verbal morphology of these languages is fairly elaborate, we shall here be concerned only with the morphology preceding the verbal root. These languages contain the morphemes shown in (2), and their standard linear order is that which is shown there. (2)

Standard linear order of morphemes preceding the verbal root in Kartvelian 1. 2. 3. 4. 5. 6.

Status, including affirmation and negation (only in Mingrelian) Preverb(s) Imperfectivizer (only in Mingrelian) Tmetic particle(s) (only in Old Georgian) Person marker(s) Version vowel

Some work on Kartvelian languages other than Georgian was conducted even in the XVIIIth and early XlXth century by scholars such as Johann Anton Güldenstädt (1745-1781) and Julius von Klaproth (1783-1835). However, most of the work from this period consists primarily of comparative word lists (cf. e.g. Güldenstädt's [1834:189-246] "Wörtersammlung zur vergleichung der im Kaukasus gangbaren Sprachen," or Klaproth's [1823:509-544] word list). It was not really until the 19th century that more substantial data became available through the work of such scholars as Marie Ivanovich Brosset (1802-1880), Roderich von Erckert (18211900) (cf. e.g. Erckert 1895), Aleksandre Cagareli (1844-1929) (cf. e.g. Cagareli 1880) and Georg Rosen (1820-1891) (cf. e.g. Rosen 1845, 1847).

190

MARCELLO CHERCHI

Most of the time the morphemes do indeed appear in this order. However, there are attested instances of the morphemes occurring in different orders, and it is these unusual orderings which we shall examine in § 2. The possible reasons for the alternative orderings will also be discussed. 2.

Examples of variant morpheme ordering In this section we shall examine some alternative orderings of preradicai morphology in Old and Early Middle Georgian, Mingrelian and Svan. (Research limitations have prevented us from taking Laz into consideration in the present paper.) In order to interpret these data properly, it is important to keep three points in mind. First, we do not claim to have found all possible alternative orderings. Second, although a greater number of data could be presented, space considerations have required us to select only a few of the clearest examples to illustrate our points in each section. Third, and perhaps most significant, it must be understood that we shall be examining morpheme ordering patterns which are unusual, or in some sense 'marked' in their respective language contexts.8 The motivation for this is that investigation of these unusual patterns provides some insight into the limits and types of flexibility in morpheme ordering, and this constitutes a necessary first step towards any proposals concerning the parameters underlying that variability. Clearly, if one were to examine only the canonical patterns, there would be no variation to consider. 2.1 Old and Early Middle Georgian The primary source of ordering variation in verbal morphology involves various morphemes (nouns, pronouns, adverbs and others) which are inserted

VARIANT ORDER OF MORPHEMES

191

into the verb by tmesis.9 Tmesis, by definition, occurs when material gets inserted into a word which, under normal circumstances, is not split up. 10 The material which gets inserted we shall refer to as tmetic particles.11 Although verbal tmesis is not extremely common in Old and Early Middle Georgian,12 it is sufficiently attested to determine that the standard placement of tmetic particles is following the preverb(s) and preceding the person marker(s), if any, and the version vowels, if any. A typical example of tmesis is given in (3).13 The "split box" indicates the word which is broken up by the tmetic particle. (3)

še -ra- vida igi kaparnaum-d... Pvb-when-3sg.goes.AOR 3sg.NOM Capernaum-ADV "...when He came to Capernaum..." (Matthew 8.5D, cited by Imnaishvili 1986:456)

Since tmesis itself is rather uncommon, its occurrence may already be viewed as a deviation from the normal order of morphemes in the verbal complex. Even more striking is the fact that these tmetic particles sometimes appear in other locations in the verbal complex. For instance, in the datum in (4) we have the tmetic particle -re- splitting up the compound preverb ca-mo (which is not normally divided by any morphology). (4)

kvla again

mepe-man king-ERG

taqvanissca, 3sg.worships.3sg.AOR

9

Verbal tmesis in Old and Early Middle Georgian is examined at length in Cherchi (1994, 1997a). Other pertinent articles which address the topic directly are Boeder (1994), Chincharauli (1969) and Schmidt (1969). Although the nature of tmesis varies from language to language, our point of departure is a very general description of the phenomenon: "Die Sprengung der Einheit durch Zwischenschaltung materieller Bestandteile heißt man tmesis" (Lausberg 1960,vol.1:253). For comparisons of the nature of tmesis in Kartvelian (mostly Old Georgian) and Indo-European see Boeder (1994) and Schmidt (1969). The meanings and functions of some of the tmetic particles are not entirely clear. See Cherchi (1994, 1997a) for discussion. 12

Most of the translations of the Bible fall in the Old Georgian period (Vth-XIth centuries). As a paradigm example of Early Middle Georgian (XIIth and XIIIth centuries) we have drawn upon the epic poem The Knight in the Tiger's Skin by Shota Rustaveli. The interlinear glossing abbreviations employed for Georgian include: adv=adverbial case; aor=aorist screeve; conj=conjunctive screeve; dat=dative case; erg=ergative case; imptv= imperative; pres=present screeve; Pvb=preverb.

192

MARCELLO CHERCHI

ça -re- mo-vlo, çadga çina... Pvb-(?) 14- Pvb-3sg.goes.AOR 3sg.stands.before.AOR before "Then the king did homage to him, he came forward, he stood before him." (The Knight in the Tiger's Skin, 1528.3; 1725.3 HJKOT; 1504.3) 15

A similar example, though with a different tmetic particle, is given in (5). (5)

...cxen-sa ševje, horse-DAT lsg.mounts.3sg.A0R ça -ca-  -ve Pvb-and-Pvb-1sg.departs.AOR "...I mounted my horse and departed..." (The Knight in the Tiger's Skin, 558.4; 704.4; 542.4)

The datum in (6) exhibits another unusual ordering in which the tmetic particle -ca- follows the first person marker -m~ (rather than preceding it, which would be the standard order in such an instance of tmesis). (6)

da-m-ca-vedra: Pvb-lsg.-and-3sg.entreats.AOR utxar, 2sg.says.to.3sg.AOR[=IMPTV] "She entreated me to tell the Amirbar Skin, 383.4; 516.4 LZ; 369.4)

amirbar-sa Amirbar-DAT asre namusobdes. thus 3sg.behaves.C0NJ.PRES so to behave." (The Knight in the Tiger's

A similar example is found in (7), in which the tmetic particle -ca- follows the third person marker -h- (rather than preceding it). (7)

usen gakvirda, Usen 3sg.marvels.A0R ga-h-ça-krta, ra Pvb-3sg.-and-3sg.causes.to.tremble.AOR when suk-n-i naxna mz-isa-n-i. rays-PL-NOM 3sg.saw.3pl.AOR sun-GEN-PL-NOM "Usen marvelled, he even quaked when he saw the sunbeams." (The Knight in the Tiger's Skin, 1155.2; 1340.2 D ; 1133.2)

The reason (or reasons) for such unusual orderings of preradicai morphemes in Old and Early Middle Georgian is not entirely clear. Boeder 14

The meaning of this particular tmetic particle is not clear from the available data. When citing from Shota Rustaveli's The Knight in the Tiger's Skin, the first number is the quatrain as found in the standard text (Rustaveli 1957), the second number (and letters) indicate the textual variants in Qubineishvili et al. (1960-1963). The third number refers to the quatrain in the English translation by Wardrop in Rustaveli (1939).

VARIANT ORDER OF MORPHEMES

193

(1994) has suggested that the placement of tmetic particles was governed by Wackernagel's law (a law originally formulated concerning clitic placement in certain Indo-European languages). Cherchi (forthcoming) presents data contradicting that hypothesis, and suggests instead that even if verbal tmesis was active in some earlier (pre-attested) stage in Georgian, by the time of attested texts it was already moribund and was probably employed in conscious imitation of foreign models, i.e. by this point in the history of the language verbal tmesis was an obsolescent, register-related phenomenon. In any event, tmesis eventually died out in Georgian. Cherchi (1997b) suggests that a possible reason was the fact that most of the (relatively rare) constructions involving verbal tmesis could be replaced by (much more common) alternative constructions which did not employ tmesis. We should add here a possible typological consideration. In Old and Early Middle Georgian, aspectual distinctions (perfective-imperfective) were con­ veyed by inflecting verbs in different screeves.16 However, by later Georgian, the opposition of perfective-imperfective was generally conveyed through the presence vs. absence of preverbs (much as in Russian). This new structuralfunctional burden on the preverbs may have contributed to solidifying their position in the verbal complex, and may additionally have been (at least partially) responsible for the decline and ultimate demise of verbal tmesis. Therefore we suggest that a typological factor (viz. the new manner in which aspectual distinctions were conveyed by the language) may have been in part responsible for creating a more rigid order of morphemes in the verbal complex. 2.2 Mingrelian Among its preradicai verbal morphology, Mingrelian has a few morphemes beyond that exhibited by Georgian, as shown in (2). It has various status markers (e.g. the affirmative marker ko- and the negative marker va-), it has a more elaborate set of preverbs (both a wider inventory and greater flexibility in their combination), and it has an imperfectivizing particle, tm(V)-. These (and the other) preradicai morphemes in the verbal complex of Mingrelian show some degree of variation in their linear ordering. For instance, Kipshidze (1914: 090, § 101) gives examples of the second person marker breaking up compound preverbs. In the form gegə-r-ço-lapu "it A screeve is a set of verb forms varying only in person and number (and thus, in a rough sense, might be said to correspond to the "tenses" of Indo-European languages as presented in traditional grammatical analyses). The various screeves in Georgian convey different opposi­ tions of aspect, temporal reference, modality, witnessed vs. unwitnessed, and others.

194

MARCELLO CHERCHI

fell out for you" the second person marker -r- has split up the compound preverb gegə-co (the standard form would be gegəco-r-lapu). In the form ki-rsino-lebue "you were in the position (to do something), it was possible for you" the second person marker -r- has split up the compound preverb kisino (the standard form would be kišino-r-lebue). Kipshidze (1914: 0106, § 111) also gives examples of the first person marker breaking up simple preverbs. For instance, in the form i-p-še-elenk "I will ascend" the first person marker (-v-) appears within the (synchronically) simple preverb ise- (from esa- or isa-),17 whereas the standard form would be ise-e-v-lenk (where the first person marker is -v). In the form ke-v-le-ebi šuri "I drew breath" the first person marker (-v-) appears between the preverbs ke(from ko-) and le- (from la-), whereas the standard form would be kele-e-v-bi. Another unusual ordering involves the attachment of the preverb to the root (discussed in Kipshidze 1914: 0106-0107, § 112), in which case the person markers occur before the preverb. Kipshidze (ibid.) cites the example of b-sa-qarank "I gather it", in which the first person marker b- occurs before the preverb sa- (the standard form would be sa-p-qarank). Sometimes when preverbs are compounded, the ordering of those preverbs or of elements within those preverbs can be rearranged. Kipshidze (1914:0108, § 115al) cites the following examples: ge-miço-ciis "they jumped" rather than the standard ge-çimo-ciis; ge-mçu-vuçkurine "I put (AORIST) it in front for him" rather than the standard ge-çimo-vuçkurine. In some instances the person markers appear within the verb root, as discussed by Kipshidze (1914: 055-056, § 67d). He cites the following examples: ba-b-rgank "I weed it", in which the first person marker -b- has interrupted the root barg-; iva-b-rank "I refuse, decline, reject", in which the first person marker has interrupted the root var-; mu-b-senk or mu-v-senk "I am working", in which the first person marker (-b- or -v-) has interrupted the root musa-; --toə "I pinch it" (root: puton-); pi-b/v-rkenk "I am thinking" (root: pirk-); de-e-ma-m-nji "I will become calm, reassured" in which the first person marker (-m-) interrupts the root many. Kipshidze (1914: 074, § 87b) also cites dixa gaməno-pu-b-cxoni "I cast/threw earth" in which the first person marker (-b-) has broken up the root pucxon- (the standard form would be gaməno-m-pucxoni).

The various vocalic changes in the preradical verbal morphology are due to processes of vowel harmony and syncopy discussed in Harris (1991b: 321-322, § 1.4.1) and Kipshidze (1914: 108-113, § 115).

VARIANT ORDER OF MORPHEMES

195

Kipshidze (1914: 058, § 69b) also gives examples in which the first person marker (-b-) follows rather than precedes the version vowel: i-b-črə "I write it for myself' rather than the standard v-i-carunk (in which the first person marker is v- rather than b-); i-b-cqank "I begin it" rather than v-i-çqank. When both the first person marker (-v-) and the third person marker (-u-) occur in a verb form they generally occur in the order v-u-. However, Kipshidze (1914: 075, § 90) mentions that these markers are occasionally interchanged, and cites the following forms: din-u-v-bari "I blew down into it"; kiməšə-u-v-bogi "I built a bridge in the middle for him". Harris (1991b:360) notes that the negative status marker (va-) occasionally follows the preverb rather than precedes it. She does not cite examples thereof. Almost all of the preradicai verbal morphemes (viz. status markers, preverbs, person markers and preradicai vowels) are involved in the ordering variations discussed above. The only morpheme for which we have not found any evidence of positional variation is the imperfectivizer tm(V)-, and it would not be surprising if this, too, exhibited variation. In any case, this variation suggests that the organization of the preradicai verbal morphology is far from strict. It is difficult without further research to offer definitive reasons for what is apparently the relatively loose organization of the preradicai verbal morphol­ ogy, though we can suggest two possibilities. First, somewhat similarly to the typological reasons mentioned for Georgian in § 2.1, it appears that some of the preradicai verbal morphemes in Mingrelian are undergoing a shift in their structural-functional role (though this shift is not yet complete, which is why the ordering is not as strict as in Modern Georgian). In particular, the so-called 'affirmative' marker, ko-, seems to be acquiring a perfectivizing function (Harris [1991b:342], Deeters [1930:14, § 20]). Thus, although this 'status marker' is not a preverb in the strict sense (or at least not in the same sense as the directional and locational preverbs), it may be beginning to resemble those preverbs (which can also serve to perfectivize). The fact that this shift in function is not yet complete may contribute to instability in the relative ordering of it and adjacent morphemes. Second, since "Mingrelian has... a richer system of preverbs... than any other Kartvelian language" (Harris 1991b:341), this larger inventory of morphemes and their combinatoric possibilities may in some sense demand greater synchronic flexibility, permitting interruption (by person markers) and reordering.

196

MARCELLO CHERCHI

2.3 Svan Svan has two 'layers' of preverbs, the outermost of which can be separated from the innermost by intervening clitics and words, and which can even appear after the verb (Deeters [1930:16-17, § 24], Hewitt [1985:9-10], Tuite [1997: § 3.3.2]). Hewitt describes this as follows: Regarding the prev(erbs), Svan's complement of eight may be divided into two groups. As in G[eorgian], the basic role of prev[erb]s in the modern language is to indicate perfective...aspect, though the four members of Group A have the following directional force: zi- "up(wards)", cu- "down(wards)", sga- "in (to)", ka- "out (from)". In Group  la- generally has only an aspectual role, whilst the other three mark orientation, thus: an- 'to the speaker/hearer', ad- or es- 'away from the speaker/hearer'. If a verb takes two prev[erb]s, the first will be from Group A, the second from Group B, though the Group A prev[erb] may not only be split off from the rest of the verbal complex, it may even follow it. (Hewitt 1985:9-10)

This unusual ordering behavior of the outermost layer of preverbs is not very common. Concerning the frequency of this phenomenon, Tuite says: Nine times of ten [the preverbs belonging to the outermost layer] precede the verb, though other particles often intervene... On occasion the preverb follows the verb... (Tuite [1997:§ 3.3.2.1])

As an example of the preverb being separated from the verb by intervening particles, Tuite (1997:§ 3.3.2.1) cites the datum in (8).19 Notice that the preverb ka is separated from the verb ispi by two intervening particles. The 'broken box' is meant to indicate that usually the preverb would immediately precede the verb. (8)

ijri:njœ:l i ka de:m-te squirm[A].3sg.APR and out cannot.anywhere-to mos ispi. at.all turn[P].3sg.AOR "She squirmed, but could not turn in any direction." (Upper Bal dialect, from Shanidze et al [1978:161-162, text #183], also parsed in Tuite [1993  : "Upper Bal texts," p. 40, line 19])

18

I am greatly indebted to Kevin Tuite for data and information concerning Svan. All data has been drawn from Tuite (1993 ms.), which is a superb line-by-line parsing of and commentary upon selected Svan texts from Davitiani et al. (1957), Shanidze et al. (1939) and Shanidze et al. (1978). 19

We present the Svan data in accordance with the transcriptional system used in this paper. In addition to the interlinear glossing abbreviations already mentioned for Georgian, the following are employed for the Svan data: [A]=active, [P]=passive, [SV]=subjective version, impf=imperfect screeve, opt=optative particle, quot=quotative particle.

VARIANT ORDER OF MORPHEMES

197

Similar examples, though with the other preverbs, are given in the data in (9) through (11). (9)

œnkid woseb-d, zi gar 3sg.takes.3sg.[A-SV].AOR Woseb-ERG up only œnçkir γulœ-te-zi. 3sg.budges.3sg.[A-SV].AOR knee-to-up "Woseb picked it up, but barely lifted it up to his knees." (Upper Bal dialect, from Shanidze et al [1939:64-67], parsed in Tuite [1993 ms.: "Upper Bal texts," p. 11, line 87])

(10)

I u -œ nœnkd isgwriwx mine-mine-zi... down-also separately 3pl.sit.IMPF each-each-at "They also sat [at the tables] separately..." (Lower Bal dialect, from Davitiani et al. [1957:294-295, text #152], parsed in Tuite [1993 ms.: "Lower Bal Holidays," p. 3, line. 18])

(11)

ešwi txwim isg 20 eser-ow one.NOM head.NOM in QUOT-OPT laxcwir... 3sg.leaves.3sg.[A-OV].AOR[=IMPT] "Leave me one head..." (Lower Bal dialect, from Davitiani et al. [1957: text #179], parsed in Tuite [1993 ms.: "Lower Bal texts," p. 10, line 80])

As an example of the preverb following (rather than preceding) the verb, Tuite (1997: § 3.3.2.1) cites the datum in (12). The shape of the "broken box" is meant to indicate that normally the preverb would be prefixed to the verb (rather than following it). (12)

la-xaţulex ka ... Pvb.call.3pl.AOR out "They called out..." (Lentex dialect, from Shanidze et al. [1978:326-327, text #350])

Similar examples, though with the other preverbs, are given in the data in (13) through (15). (13)

eckas dœw anəgan zi ... then ogre.NOM 3s.stands.AOR up "Then the ogre stood up..." (Lower Bal dialect, from Davitiani et al. [1957: text #179], parsed in Tuite [1993 ms.: "Lower Bal texts," p. 7, line 50])

20

This appears to be a Lower Bal variant of the preverb sga (although the latter also appears in Lower Bal).

198

MARCELLO CHERCHI

(14)

amecu ser rih-d liswbi-liγrœ:l i here indeed dawn-ADV dancing-singing.nom and liçšxœš ma:ma kwecni round.dancing.NOM not 3sg.ceases.[P].PRES ču . down "Here the singing, dancing and round-dancing does not cease until dawn." (Upper Bal dialect, from Shanidze et al. [1939:15, text #13], parsed in Tuite [1993 ms.: "Upper Bal texts," p. 59, line 33])

(15)

...ej-zi çalœ-te adkwœrx isga . that-way river-to 3pl.throws.3sg.[A].AOR in "...they threw it [=a sack] into the river." (Lower Bal dialect, from Davitiani et al. [1957:323-325, text #178], parsed in Tuite [1993 ms.: "Lower Bal texts," p. 22, line 88])

What we are probably witnessing here is that the inner layer of preverbs became fused to the verb at an earlier stage of the language (acquiring perfectivizing force, etc.) while the outer layer of preverbs, which Deeters (1930:16) describes as "adverbialen Präverben", can still be considered "durchaus selbständige Wörter" (ibid.) and have thus not acquired a solid position in the inventory of Svan's preradicai verbal morphology. Thus we have an inner layer of preverbs which has proceeded further along the path of grammaticization than the outer layer of preverbs. 3.

Summary and conclusions In the Table in (16) we summarize the various unusual ordering patterns and possible reasons discussed in § 2. In a sense the morpheme ordering phenomena discussed in § 2 are both diverse and similar. They are certainly diverse in their details (e.g. Old and Early Middle Georgian exhibit verbal tmetic particles unique to that language; Mingrelian's status markers are unique to that language; the behavior of Svan's outer layer of preverbs is unique to that language). Yet at a more general level one can perceive similarities, such as the fact that the ordering of preverbs is an important source of variation in linear ordering in each language (viz. in Georgian the preverb can be split by a tmetic particle or separated from the verb; in Mingrelian the preverb can be split by person markers or have its elements rearranged or become fused to the verb root; in Svan the outermost layer of preverbs can be separated from the verb by tmesis or can even follow the verb).

VARIANT ORDER OF MORPHEMES

199

Language

Variation in ordering

Possible source

Georgian (V-XIII centuries)

Tmesis interrupts sequence of verb morphemes. Tmetic particles can break up compound preverbs. Order of tmetic par­ ticle and person markers seems somewhat flexible.

At time of attestation, tmesis seems moribund. Its use may be a register-related phenomenon.

Mingrelian

The position of person markers seems quite flexible; a person marker can interrupt simple or compound preverbs, can interrupt a verb root, can exchange places with a version vowel, and some person markers can exchange posi­ tions with each other. The constituent elements of a com­ pound preverb can be reor­ dered. The negative status marking prefix can follow the preverb rather than precede it.

The preradicai morpheme (ko-) seems to be acquiring perfectivizing force. This may be indicative of a broader shift in the functional role(s) of preverbs. Mingrelian's very large inventory of preverbs may demand greater combinatoric and sequential flexibility which extends to other preradicai morphemes.

Svan

The outermost layer of preverbs may be separated from the rest of the verb either through tmetically interposed particles, or by being placed after the verb.

The outermost layer of preverbs (which are also homophonous with postpositions) have not been as thoroughly integrated into the verbal complex as the innermost layer of preverbs. Perhaps they do not (yet) need to serve the structural-functional role of perfectivization, since the innermost layer of preverbs already serve that purpose.

(16) Synopsis of unusual ordering patterns of Kartvelian preradicai verbal morphology and possible sources

Similarly, the possible reasons behind the ordering phenomena are both diverse and similar. Their diversity stems in large part to the diverse inventory of preradicai verbal morphemes in each language and to their different structural-functional roles. Yet, again, certain similarities can be perceived. For instance, at a general level, it appears that the variant orders occur at least in part due to the shifting in structural-functional roles of certain of the morphemes and due to the incompletely grammaticized functions of those morphemes.

200

MARCELLO CHERCHI

Although we have been discussing morphemes in the verbal complex, the relevance to syntax is clear: it is highly likely that what are currently preradicai verbal morphemes in the daughter Kartvelian languages used to be independent words at an earlier (pre-attested) stage of Kartvelian (a development which is widely attested cross-linguistically). Although some broad or impressionistic generalizations concerning ordering variability can be offered (viz. Georgian seems to be the least flexible and Mingrelian the most), it is not possible to "locate" the languages along any single "scale of degree of grammaticization" (with regard to the preradicai verbal morphology) since we are dealing with phenomena governed by multiple (and different) grammatical categories. However, what the data for each language provide is an informative "snapshot" of the state of affairs in each language in what appears to be a relatively late phase in the syntax-to-morphology path of grammaticization.

REFERENCES PRIMARY SOURCES Davitiani, A. (Davitiani), V. Topuria (Topuria) and M. Kaldani (Kaldani), eds. 1957. Svanuri prozauli ţeksţebi 2: balskvemouri kilo [Svan prose texts 2: Lower Bal dialect]. Tbilisi: Mecniereba. Imnaishvili, Ivane (Imnaisvili). 1986. Kartuli otxtavis simponia-leksikoni [Concordance to the Georgian Gospels]. Tbilisi: Tbilisi University Publisher. Matthew=Blake, P., ed. 1933. "The Old Georgian Version of the Gospel of Matthew". Patrologia Orientalis, 24.1. Paris: Firmin-Didot & Co. Publishers. Qubineishvili (Qubineisvili), Kekelidze (Kekeli3e), Gugushvili (Gugusvili) and Lolashvili (Lolasvili). 1960-1963. Vepxisţqaosnis xelnaçerta varianţebi [Variants of the manuscripts of The Knight in the Tiger's Skin], in Rustvelologiuri jiebani 1 [Rustavelological research 1], parts 1-4. Tbilisi: Georgian S.S.R. Scientific Academy Publishers. Rustaveli, Shota (Rustaveli, Šota). 1957. Vepxisţqaosani [The Knight in the Tiger's Skin]. Tbilisi, State Publishing Co. .1939. The Knight in the Tiger's Skin. Translated by Wardrop, Marjory Scott. New York: International Publishers. Shanidze, . (Šni) & V Topuria (Topuria), eds. 1939. Svanuri prozauli ţeksţebi 1: balszemourikilo [Svan prose texts 1: Upper Bal dialect]. Tbilisi: Mecniereba. . M. Kaldani (Kaldani) & Z. Chumburidze (umburi), eds. 1978. Svanuri enis kresţomatia [A chrestomathy of the Svan language]. Tbilisi: Tbilisi State University Press. Tuite, Kevin. 1993 ms. Svan reference grammar with texts.

VARIANT ORDER OF MORPHEMES

201

SECONDARY SOURCES Anderson, Stephen. 1984. "On Representations in Morphology: Case, Agreement and Inversion in Georgian". Natural Language and Linguistic Theory 2.157-218. .1985. Phonology in the Twentieth Century. Chicago: The University of Chicago Press. Boeder, Winfried. 1994. "Kartvelische und indogermanische Syntax: Die altgeorgischen Klitika". Indogermanica et Caucasica ed. by Roland Bielmeier & Reinhard Stempel, 447-471. Berlin & New York: Walter de Gruyter. Cagareli, Aleksandre. 1880. Mingre skie ètjudy [Mingrelian studies]. Part 1: MingreVskie teksty s perevodom i ob jasneniem [Mingrelian texts with translation and explanation]. Part 2: Opyt fonetiki mingreVskogo jazyka [Towards a phonetic study of the Mingrelian language]. Saint. Petersburg. Cherchi, Marcello. 1994. "Verbal Tmesis in Georgian," AION: Annali del Dipartimento di Studi del Mondo Classico e del Mediterraneo Antico, Sezione Linguistica ed. by Domenico Silvestri, Part I, 16.33-115. Napoli: Istituto Universitario Orientale. . 1997a. "Verbal Tmesis in Georgian," AION: Annali del Dipartimento di Studi del Mondo Classico e del Mediterraneo Antico, Sezione Linguistica ed. by Domenico Silvestri, part II, 19.63-137. Napoli: Istituto Universitario Orientale. . Forthcoming. "Georgian Verbal Tmesis and Wackernagel's Law" Non-Slavic Languages of the Soviet Union ed. by Howard Aronson, vol. 9. Chicago: The University of Chicago Linguistics Department. . 1997b. "Über das Nichtvorhandensein von Morphologisierung". Georgica ed. by Heinz Fähnrich, 20.70-82. Konstanz: Universitätsverlag Konstanz. Chincharauli, Al. (Çinçarauli). 1969. "Ţmesi da misgan gamoçveuli zogi gaugebroba Vepxisţqaosanis ţeksţsi [Tmesis and some confusions due to it in the text of The Knight in the Tiger's Skin.Y ed. by Dzidziguri et al, 193-197. Deeters, Gerhard. 1930 Das kharthwelische Verbum: vergleichende Darstellung des Verbalbaus der südkaukasischen Sprachen. Leipzig: Kommissionsverlag von Markert & Petters. Dzidziguri, Shota (3i3iguri, Šota) et al., eds. 1969. Giorgi Axvlediani. Tbilisi: Tbilisi University Press. Erckert, Roderich von. 1895. Die Sprachen kaukasischen Stammes. Vienna: Alfred Holder. Güldenstädt, Johann Anton. 1834. Dr. J. A. Güldenstädts Beschreibung der kaukasischen Länder. Aus seinen Papieren gänzlich umgearbeitet, verbessert herausgegeben und mit erklärenden Anmerkungen begleitet von Julius Klaproth. Berlin: Stuhr. Harris, Alice C , ed. 1991a. The Indigenous Languages of the Caucasus. Volume 1: The Kartvelian Languages. Delmar & New York: Caravan Books. . 1991b. "Mingrelian". Harris 1991a.315-394. Hewitt, B.G.. 1985. "A Svan (Lasx) Text: Part 2".Revue des études géorgiennes et caucasiennes 1.9-34. Paris: Association des études géorgiennes et caucasiennes.

202

MARCELLO CHERCHI

Kipshidze, I. (Qipsš3e). 1914. Grammatika mingre skago (iverskago) jazyka [Grammar of the Mingrelian language]. Materialy po jafeticeskomu jazykoznaniju [Materials for the study of Japhetic linguistics], 7. St. Petersburg: Nauk. Klaproth, Jules. 1823. Voyage au mont Caucase et en Georgie. Paris: Librarie de Charles Gosselin, Librarie Classique-Élémentaire. Lausberg, Heinrich. 1960. Handbuch der literarischen Rhetorik, 2 vols. Munich: Max Hueber Verlag. Lieber, Rochelle. 1992. Deconstructing Morphology: Word Formation in Syntactic Theory. Chicago: The University of Chicago Press. Rosen, Georg. 1845. "Über die Sprache der Lazen". Abhandlungen der Königlichen Akademie der Wissenschaften zu Berlin aus dem Jahre 1843, Philologische und historische Abhandlungen, 1-38. Berlin: Druckerei der Königlichen Akademie der Wissenschaften. . 1847. "Über das Mingrelische, Suanische und Abchasische". Abhandlungen der Königlichen Akademie der Wissenschaften zu Berlin aus dem Jahre 1845, Philologische und historische Abhandlungen, 405-444. Berlin: Druckerei der Königlichen Akademie der Wissenschaften. Sadock, Jerrold. 1991. Autolexical Syntax: A Theory of Parallel Grammatical Representations. Chicago: The University of Chicago Press. Schmidt, Karl Horst. 1969. "Zur Tmesis in den Kartvelsprachen und ihren typologischen Parallelen in indogermanischen Sprachen". Dzidziguri et al., eds. 1969. 96-105. Singh, Rajendra, ed. 1996. Trubetzkoy's Orphan: Proceedings of the Montréal Roundtable "Morphonology: Contemporary Responses, " Montreal, September 30October 2, 1994. Amsterdam: John Benjamins. Spencer, Andrew. 1991. Morphological Theory. Oxford: Blackwell Publishers. Tuite, Kevin. 1997. Svan, in Languages of the World/Materials 139. Munich & Newcastle: LINCOM EUROPA.

WORD ORDER STABILITY AND CHANGE FROM A SOCIOLINGUISTIC PERSPECTIVE THE CASE OF EARLY MODERN WELSH

OLIVER CURRIE St John's College, University of Cambridge

1.

Sociolinguistic variation and word order theory The notion of stability is central to grammatical description, since any categorization of a language as having a certain structural property takes for granted that this property is a stable one, regularly and consistently reflected in discourse in that language. Yet a precise, unambiguous interpretation of stability cannot itself be taken for granted, since it remains an open question how much variability can be allowed in the realization of a 'stable' linguistic property. On the one hand, it is possible to associate stability with invariance, on the other hand, sociolinguistic studies have shown that there can also be 'stable' patterns of variation, and indeed have argued that such variation termed "structured heterogeneity" - is actually the norm in language (Weinreich, Labov and Herzog 1968:101). The aim of this article is to explore the implications of this central finding of sociolinguistic research for the specific area of word order theory, looking in particular at how we understand the notions of word order stability and word order change. We face three broad problems in examining word order stability and change from a sociolinguistic perspective. First, sociolinguistic theory has been devel-oped largely for the investigation of phonological variation, where it was taken as a basic premise that variant forms are equivalent in meaning - "the variants are identical in truth, but opposed in their social and/or stylistic significance" (Lavandera 1978:174); variant word orders, however, usually differ in meaning and pragmatic function at least to some extent. Second, it has been demon-strated that a multiplicity of factors contribute to word order variation - syntactic factors (e.g. type of clause, transitivity), pragmatic factors (e.g. topic continuity, focus), textual factors (e.g. genre), lexical and idiomatic factors - as well as sociolinguistic and stylistic factors. So in investigating sociolinguistic and stylistic variation in word order, we do not simply have a

204

OLIVER CURRIE

two-dimensional puzzle to tackle (i.e. the relationship between form and social/stylistic signifi-cance), but a multidimensional one, in studying the interaction of sociolinguistic and stylistic variation with several other kinds of variation. Third, mainstream word order theory has tended to exclude sociolinguistic factors a priori from its analysis, and does not seem adequately equipped to deal with sociolinguistic variation. This article also aims to show that the importance of sociolinguistic and stylistic variation has been underestimated in word order theory. Empirical evidence of this will be presented from a corpus-based study of Early Modern Welsh, where there is particularly striking stylistic variation in word order. 2. 'Top-down' vs. 'bottom up ' approaches to word order description The tendency within mainstream syntactic theory to marginalize sociolinguistic and stylistic variation stems from what we call - after Langacker (1988) - the 'top-down' nature of its conception of grammar. Typically in a 'top down' approach to grammar, the starting point of analysis is a maximally general description or rule, and priority is given to generalizations in the form of rules over the description of (reputedly) less systematic features. A 'top-down' approach to word order would typically model linearization in terms of a series of interacting abstract rules or principles, which specify what permutations of the major sentence constituents are grammatical, and, in certain cases too, specify under what circumstances particular word order patterns might be used. 'Top-down', rule or principle based approaches to linearization are found in both generative syntactic frameworks, such as Principles and Parameters, and in functional ones, especially Functional Grammar (Siewierska 1988). In contrast to 'top-down' approach, a 'bottom-up' one would not represent word order as being determined globally by very general linearization rules but rather as a series of choices between various pre­ existing but flexible patterns or constructions. The basic unit of analysis, then, is the construction, one which is both less general in scope and less abstract than the notion of the 'syntactic rule' (cf. Construction Grammar - Fillmore et al. 1988). 'Bottom-up' and 'top-down' models of grammar necessarily entail very different approaches to generalization in word order description. A 'topdown' approach starts from the general description or rule and then derives individual constructions from this; indeed, in the most rigorously 'top-down' approaches, such as minimalism, the notion of the construction itself is not just secondary but epiphenomenal, as Chomsky (1995:170) states: "the notion of grammatical construction is eliminated, and with it, construction particular rules". A 'bottom-up' approach contrasts with this in that it starts from

STABILITY AND CHANGE IN EARLY MODERN WELSH

205

individual constructions and builds up more general descriptions on the basis of what they can be shown to have in common. It must be stressed, though, that adopting a 'bottom-up' approach does not mean rejecting syntactic generaliza­ tions above the level of the construction. The point is that the interrelationship between constructions must be defined in empirical terms consistent with speakers' usage, rather than in abstract theory-internal terms. The significance of this distinction between 'top-down' and 'bottom-up' ap-proaches for the investigation of word order variation and change is threefold: (I) The units of analysis used in 'top-down' approaches - word order rules, the notion of the word order type (e.g. a VSO language), or grammar (e.g. a 'V2 grammar' in Principles and Parameters theory) - are too general and abstract to describe many actual instances of word order variation, which seem to occur on a more 'localised' level, for example between individual construc­ tions. In 'bottom-up' approaches, on the other hand, the units of analysis are small enough to capture such 'fine-grained' variation. (II) A 'top-down' word order description - e.g. a typological classification of language's word order - may quite commonly conceal considerable variation in word order between different discourse types, different media (e.g. prose vs. poetry) or different speakers, thus reducing the informativeness of the description. A 'bottom-up' approach need not assume that the same abstract grammatical rules necessarily underlie the word order patterns in different texts or authors. (III) Often word order change happens in a piecemeal fashion, extending or restricting the use of individual constructions, through such processes as syntactic reanalysis, grammaticalization, stylistic competition, rather than on a 'global' level, whereby a language abruptly changes from one word order type to another (e.g. SOV to SVO). While 'top-down' approaches can formulate elegantly the results of long term word order change, they cannot, in contrast to 'bottom-up' ones, model well the intricacies of the process. 3. Case study: absolute-initial verb order in Early Modern Welsh 3.1 Historical overview The data is drawn from a corpus of Early Modern Welsh manuscript and printed texts, spanning approximately two centuries from the mid XVIth century to the mid XVIIIth and covering almost the full range of texts available in Welsh during this period, both original and translated. This includes drama (morality plays and interludes in verse), autograph manuscript sermons, the XVIth century Bible translations, religious treatises of a popular and scholarly

206

OLIVER CURRIE

nature, narrative prose texts, letters and slander case records. In total, the corpus contains nearly 10000 positive declarative clauses ("PDMCs"), which have been chosen as object of study, since they, as opposed to negative or subordinate clauses, show the greatest variation and change in their word order patterns from Middle to Modern Welsh. While Modern Welsh is well known as verb-initial (VSO) language, the situation was quite different in Middle Welsh. Here, the order in PDMCs was predominantly verb-medial in prose, with at least one constituent - subject, object, verbal noun object or adverbial phrase - preceding the verb, and finite verb-initial order was rare. In Middle Welsh poetry, on the other hand, the pattern was more complex, since alongside the verb-medial constructions attested in the prose, finite verb-initial order was considerably more frequent. Further, Middle Welsh differs from Modern Welsh not just in the usual position of the verb, but also in the relativeflexibilityof the word order. There does not seem to have been any single statistically predominant, basic word order in PDMCs in Middle Welsh prose; the choice of the preverbal constituent, which Poppe (1991 & this volume) analyses as a topic constituent, seems to have been largely determined by pragmatic factors. In Modern Welsh, in contrast, verb-initial order has been grammaticalized as the basic word order. 3.2 Word order variation in Early Modern Welsh The most salient feature of the word order during period ca.1550-ca.1750 is the variation, particularly in the use of one verb-initial construction, absolute-initial verb order (1) and (2), where a finite verb comes in absoluteinitial position in a PDMC. (1)

Lladdodd Meurig eu brenin hwy Kill-3sg-PAST Meurig their king 3pl "Meurig killed their king in battle" (FfC 152)

mewn in

(2)

aethum yn fwy fwy gwrthryfelgar yn erbyn Go-lsg-PAST PRED more more rebellious against "I became more and more rebellious against God" (Hel II 4)

brwydr battle Duw God

AIV order was extremely rare in MW prose - with no examples at all in several texts and only one or two when the construction is attested - yet is common in MW poetry. In Gwasanaeth Meir, for example, a late XlVth century liturgical text translated from Latin, which contains both prose and poetry sections, there are frequent examples of AIV order in the poetry (48 out of 179 PDMCs, i.e. 26.8%), though none in the prose (out of 48 PDMCs). To give an idea of the frequency with which AIV order could occur in MW poetry,

STABILITY AND CHANGE IN EARLY MODERN WELSH

207

out of 100 PDMCs in the cywyddau (poems 100-126 in Parry ed. 1979) of Dafydd ap Gwilym (1320-1380), there are 37 examples of AIV order. The variation in the use of AIV order becomes even more striking in Early Modern Welsh, since we start to find the kind of variation only attested between poetry and prose in Middle Welsh, within the medium of prose itself. In the sixteenth and early XVIIth century part of the Early Modern Welsh corpus, we still find several prose texts with either no examples at all of AIV order in the sections analysed - e.g. Y Marchog Crwydrad (MS c.1575), Roland Puleston's (1583) Llefr o'r Eglwys Crhistnogedd - or with only one or two examples, e.g. Richard Davies' foreword to the 1567 Welsh New Testament Epistol at y Cembru (1 example of AIV order out of 130 PDMCs), Robert Gwyn's (1580) Gwssanaeth y Gwyr Newydd (2/198), the sermons of William Griffith (late 16th/early XVIIth centuries) (2/192), the c.1600 Welsh translation of the Gesta Romanorum (2/148) and the sermons of Evan Morgan (1610) (1/154). We also find prose texts from this period with a more frequent but still sporadic use of the construction, e.g. William Salesbury's translation of the gospel of Mark (1567) with a 2.3% frequency of AIV order, William Morgan's revised translation of Mark (1588) 6.5%, William Morgan's transla­ tion of the book of Esther (1588) 9.9%, and Maurice Kyffin's (1595) Deffyniad Ffydd Eglwys Loegr 3.5%. Besides these texts, however, there are others - all parts of the 1567 and 1588 Bible translations - with a significantly more frequent use of AIV order: William Salesbury's Psalms in the 1567 Welsh Book of Common Prayer 20.1%, William Morgan's translation of Isaiah in the 1588 Bible 24.8%, and above all his revised translation of the Psalms in the same Bible, with AIV order being used in 40.4% of PDMCs. This pattern of variation between prose texts becomes even more marked in the XVIIth and XVIIIth centuries, for while the frequency of use becomes higher in individual texts (over 50% of PDMCs in some cases), and while the number of texts with a more than sporadic use of AIV order (i.e. 10% or more of PDMCs) also increases, there are still several texts with a very low use of AIV order or with no original examples at all (that is excluding quotations from the Bible). To give an indication of the extreme range of variation between more or less contemporary texts, we have in 1675 Rondi Davies' Profiad yr Ysprydion with no original examples of AIV order at all, and two years later, Charles' Edwards Y Ffydd Ddi-ffvant, with 62.1% AIV order. Moreover, significant variation is found across all types of texts, original and translated, narrative prose and expository prose, and within the drama section of the corpus too, all of which is in verse. These findings are presented in Tables 1-6.

208

Psal I Mark I Mark II Esther Isaiah Psal II

OLIVER CURRIE

AIV 1 20.1 23 65 9.9 24.7 40.4

AdV 0.7 0 0 0 0 0

AdyV 15.3 22.8 18.9 18.8 21.2 15.2

DU 1.1 4.1 0.9 2.6 0 0

PSV 19.8 32.8 39.6 14.1 15.5 14.5

AdPS 0.7 1.4 4.1 5.2 0 0

NSV 25.0 18.3 22.1 29.7 26.5 24.5

AdNS 0 0 1.4 18.2 4.0 0.7

Oth 15.0 17.8 4.6 1.6 8.0 4.7

Table 1: Word orderfrequencies(%) in the XVIth century Bible translations AIV Crwy Gwel HI Prof  II Faf

AdV

0 35.7 9.4 43 27.7 93

1.2 22.2 8.0 0.7 18.4 4.7

AdNS Oth AdPS NSV AdDU PSV AdyV DU 6.2 4.9 21.0 35.8 21.0 0 7.4 3.7 5.9 0.5 1.9 12.4 16.2 0 4.9 1.1 10.1 1.4 5.0 1.4 21.7 24.6 13.0 5.1 9.4 13.8 18.1 2.9 0.7 37.7 4.3 8.0 6.4 12.1 1.4 3.5 0.7 10.6 16.3 2.8 6.2 3.1 0 24.8 6.2 24.8 2.3 18.6

Table 2: Word orderfrequencies(%) in the 1st person narrative texts AIV L1EC GR FfC  HBA Pel MF

0 1.4 563 54.9 24.1 21.8 3.7

AdV AdyV 2.4 14.5 0 16.9 23.4 6.3 18.3 0.3 10.2 11.5 7.1 6.1 6.2 6.2

DU AdDU 3.4 1.9 3.4 8.1 0 0 1.0 0.3 7.3 0.8 9.6 0.5 11.1 2.5

AdPS PSV 15.9 28.1 43.9 10.8 0.8 0 2.4 5.1 10.0 6.5 4.1 21.8 37.0 13.6

NSV 22.7 14.2 8.6 9.5 25.7 21.8 16.0

AdNS 4.8 1.4 0.8 2.7 1.5 4.1 1.2

Oth 6.3 0 6.2 5.4 2.3 3.0 2.5

Table 3: Word orderfrequencies(%) in the 3rd person narrative prose texts (excluding Bible translations - Mark I & II and Esther) Epist Gws Deff Gors Edif Car LITA Yspr Princ Rhin Gwy YBM Bed

AIV AdV AdyV DU AdDU PSV Oth AdPS NSV AdNS 0.8 0 13.1 40.8 2.3 0.8 1.5 2.3 15.1 22.3 1.0 0 10.1 13.1 8.6 5.1 46.5 9.1 2.0 5.0 0 3.5 15.7 4.7 29.1 8.4 3.5 7.0 23.8 2.9 3.3 25.7 6.5 16.4 5.6 1.9 5.1 1.4 30.8 3.3 0 0 23.0 2.7 2.0 29.7 4.7 25.7 6.1 6.1 3.2 16.8 18.4 24.0 5.6 0.8 16.8 6.4 3.2 4.8 05 0 5.0 18.8 53.2 4.5 5.0 10.6 0.8 1.6 0 0 18.9 7.4 2.9 18.5 4.1 3.7 38.7 4.9 0 0 19.6 22.5 3.6 27.5 5.8 13.8 2.9 4.3 68.8 8.3 9.2 0 0.9 0 0 11.0 0 1.8 40.0 7.0 17.5 1.0 0 8.5 2.5 19.0 2.5 2.0 22.4 3.3 15.4 22.4 3.0 22.7 3.3 4.7 0.7 2.0 9.4 465 8.9 5.9 14.4 10.4 0.5 1.5 2.5 1.5

Table 4: Word orderfrequencies(%) in the expository prose texts Abbreviations for order types: AIV = Absolute-initial verb order; AdV = Adverbial phrase + verb + (S); AdyV - Adverbial phrase + particle y + verb + (S); DU = dummy subjectƒe+ verb + (S); AdDU = Adverbial phrase + dummy subject fe + verb + (S); PSV = Personal pronoun subject + verb;AdvPS = Adverbial phrase + personal pronoun subject + verb; NSV= nominal subject + verb; AdvNS = Adverbial phrase + nominal subject + verb; Oth = other constructions.

STABILITY AND CHANGE IN EARLY MODERN WELSH

WG EM WW Piers C2219 NLW3  95 C2226 B362

AIV 1.4 0.6 1.0 10.1 3.6 5.6 0 0.9 22.0

AdV 0 0 4.0 0 1.5 3.4 0 0 5.1

AdyV 22.9 7.8 19.2 25.3 31.3 22.6 9.1 11.2 11.9

DU AdDU 5.2 9.9 24.0 5.2 1.0 10.1 2.8 1.1 2.9 0 1.1 7.3 31.8 2.3 23.4 10.3 1.7 0

AdPS PSV 34.9 8.3 33.8 3.9 43.4 7.1 2.2 14.6 17.5 3.6 25.4 2.8 34.1 11.1 26.2 3.7 29.7 4.2

NSV 8.3 19.5 11.1 37.6 32.8 28.2 6.3 14.0 23.7

209

AdNS 0.5 0 1.0 2.2 4.4 0.6 1.1 4.7 1.7

Oth 7.3 4.2 2.0 3.9 2.2 3.2 2.9 5.6 0

Table 5: Word order frequencies (%) in the sermons (allfromautograph manuscripts) Diodd GK I GK II Rhyfel Brutus FAfrad Cyndr Ffrew

AIV 13.5 38.4 25.6 3.8 18.8 20.9 25.6 3.8

AdV AdyV 4.0 10.3 3.5 8.3 0 7.7 2.5 2.5 10.1 0.8 1.8 2.7 5.1 3.2 0.5 1.6

DU AdDU 0.8 0.8 1.2 10.5 1.7 11.1 9.8 0.9 6.4 1.1 8.4 0.6 0.7 6.9 14.2 2.2

AdPS PSV 36.5 13.5 24.4 5.8 40.2 3.4 65.3 6.9 8.4 40.3 48.7 5.1 44.4 3.2 68.3 3.3

NSV 7.9 9.5 6.0 5.1 7.3 5.7 6.5 3.8

AdNS 0 0 0 0 1.4 0.3 0.7 1.6

Oth 9.5 2.3 4.3 3.2 5.3 6.0 3.6 0.5!

Table 6: Word order frequencies (%) in the drama

The range of variation in the use of AIV order is noticeably lower in the case of the sermons (only one sermon has a relatively high use of AIV order, above 20%) and in the drama, compared with the narrative and expository prose texts. Nevertheless, there is still significant variation in all sections of the corpus represented in Tables 1-6, between very low use (or even complete avoidance) of AIV order and where AIV order makes up at least a fifth of all PDMCs. Independent confirmation of this pattern can also be found in the existence of 'parallel' Welsh texts with significantly different usage of AIV order, and, consistent with the overall pattern, these texts also differ radically in genre and discourse type. There are four such cases: (I) Two different XVIth century manuscript versions of the morality play Y Gwr Cadarn, one (GK I) in Cardiff MS. 2.83 (copied c.1550) with 38.4% AIV order, the other (GK II) in Peniarth 65 (copied  late XVIth century) with 25.6% AIV order. GK II does not appear to a copy of GK I but rather an independent version. (II) William Salesbury's translation of the Psalms in 1567 with 20.1% AIV order, and William Morgan's revised translation in 1588 with 40.4% AIV order. (III) Two Welsh translations of Thomas Gouge's Principles of the Christian Religion, a popular religious treatise written in dialogue form, designed to instruct ordinary people in the fundamentals of the Christian faith; the first edition, translated by William Jones in 1676 as Principlauu 

210

OLIVER CURRIE

Grefydd Ghristianogol has no examples of AIV order, and the second, hugely expanded edition by Charles Edwards in 1679 as Gwyddorion y Grefydd Gristianogol has 40% AIV order. (IV) Two Welsh translations of John Bunyan's Grace Abounding (Helaethrwydd o Râs), an autobiographical account of how he found faith, narrated in the first person; the first translation in 1737 by John Einnion has 9.4% AIV order, the revised translation in 1763 by Rhys Thomas has 27.7% AIV order. 3.3 Accounting for the variation: competition between AIV order and other constructions As a note of caution, it should be stressed that variation in the crude percentage frequency of AIV order need not in itself imply an absence of underlying regularities in word order, even between apparently very divergent texts. An interesting and problematic example from within the corpus is the Bible translations (Table 1), where the use of AIV order is considerably higher in the two versions of the Psalms than in the other books translated by each author. William Salesbury uses AIV order over eight times more frequently in his Psalms than in his translation of Mark; in William Morgan's Psalms, the percentage of AIV order is over four times higher than in his translation of Esther, and six times higher than his translation of Mark. To an extent, the higher incidence of AIV order in William Morgan's Psalms compared to the narrative texts Esther and Mark could possibly be explained by differences in the discourse structure of the texts. Esther and Mark 1-5 are overwhelmingly written in the third person - 1st and 2nd person verbs making up less than 5% of the total. In the Psalms, though, 1st and 2nd person verbs make up 40% of the total, and more than half of the instances of AIV order in this text involve 1st and 2nd person verbs. In contrast to this, in 3rd person verbal constructions with a full nominal subject, William Morgan shows a very clear tendency across all four texts to use nominal subject + verb rather than AIV order: Mark II Esther Isaiah Psal II %NSV order 653 76.7 61.1 80.5 % AIV order 12.8 4.2 15.0 9.2 Table 7: Comparison of the frequency of NSV and AIV orders in PDMCs with a full nominal subject in parts of William Morgan's Bible translations

The discourse structure of the Psalms could be argued to be naturally more favourable to AIV order in another respect too. Because of their monologic and lyric form - expressing thanksgiving, praise and supplication to God - we do not find in the Psalms the frequent changes of perspective and protagonist, nor

STABILITY AND CHANGE IN EARLY MODERN WELSH

211

the multiplicity of potentially competing topics that we do in narrative texts like Esther and Mark. The two main personae of the Psalms, the author, in the first person, and God, either addressed directly in the second person or referred to in the third, are continuously present; equally, the dominant themes of praise, deliverance and thanksgiving constantly recur throughout the Psalms. Since a common use cross-linguistically of verb-initial order with zero anaphora is in contexts of topic and thematic continuity (Givón 1984:403, 1990:913-917), the Psalms provide, arguably, considerable opportunity to use this construction. Discourse-pragmatic factors do not seem, however, to account fully for the differences between, the Psalms (and to a lesser extent Isaiah), on the one hand, and Mark and Esther on the other. While Morgan continued Salesbury's pattern of using AIV order significantly more frequently in the Psalms than in Mark, he still decided, in revising the translation of the Psalms to use AIV order twice as often as Salesbury. It seems that Morgan could have used AIV order considera-bly less frequently in the Psalms, and Salesbury considerably more so. Authorial choice appears, then, to be a decisive factor in determining the frequency of use of AIV order; even in contexts favourable for its use - e.g. topic continuity and subjectless constructions - AIV order could be used more or less frequently, or, as the testimony of other texts suggests, even avoided altogether. This observation is borne out both by the range of variation in the use of AIV order across texts of the same genre and discourse type - whether 1st or 3rd person narrative prose, expository prose or verse interludes - and by more detailed analysis of the relationship between AIV order and the other competing constructions. Here, the patterns of usage suggest that Welsh writers of the period may have interpreted AIV order as functionally equivalent to and effectively interchangeable with two other constructions: the dummy subject construction and personal pronoun subject + verb order. The dummy subject construction was a verb-initial one like AIV order, the crucial difference being that a particle fe, etymologically derived from the third singular masculine personal pronoun, precedes the verb in clause-initial position. This construction is used not only with subjectless and impersonal verbs, (3), but also with transitive and intransitive ones, (4) and (5), which have full nominal subjects. (3)

Fe ddywedir fod Duw yn yspryd o berffeithrwydd it say-PASS-PRES/FUT be-VN God PRED spirit of perfection "It is said that God is a spirit of perfection" (Princ 3)

(4)

Fe gymmerth Mab Duw natur ddynol iddo ei hun it take-3sg-PAST son God nature human to-him his self "The Son of God took a human nature upon himself" (Gwydd 99)

212

OLIVER CURRIE

(5)

Fe fu gwraig Caleb Jones yn y gwely gida Llewelyn it be-3s-PAST wife Caleb Jones in the bed with Llewelyn Beavan tra fu Mary fy merch i yn godro Beavan while be-3s-PAST Mary my daughter 1s PROG milk-VN "Caleb Jones's wife was in bed with Llewelyn Beavan while Mary my daughter was a milking" (Slander: Llandaff, 323; 1730)

The personal pronoun subject + verb construction is exemplified in (6). The verb agrees with the preceding personal pronoun in person and number except in two cases: in contrastive focus constructions where the verb is always 3rd singular, and in synthetic passive constructions where there is only a single, impersonal passive inflection. Because of the rich verbal morphology, neutral use of preverbal personal pronouns could be seen as redundant, and so as broadly equivalent to AIV order. In addition to the possibility of suspending subject-verb agreement to highlight a personal pronoun subject, Early Modern Welsh (like Middle Welsh and Modern Welsh) also possessed a very rich pro­ nominal system with three paradigms of independent personal pronouns, which could be used as preverbal subjects to express various nuances of meaning. For example, beside the simple first singular form mi used in (6), there was also a reduplicated form myfi or y fi (7), and a conjunctive form minnau (Evans 1976 49-53; Thomas 1996: 243, 249-252). (6)

Mi a gysulltais a phobl ysbrydol o weinidogion Crist I PART mix with-1s-PAST with people spiritual of servants Christ "I kept the company of godly people who were Christ's servants" (Profiad 19)

The clearest evidence for the potential interchangeability of the dummy subject and personal pronoun subject + verb constructions, on the one hand, and AIV order, on the other, comes from the pairs of 'parallel texts' mentioned in 3.2. We have numerous directly comparable PDMCs in the two versions of the morality play Y Gwr Cadarn, e.g. (7), of the two XVIth century translations of the Psalms (8), and the two Welsh versions of Grace Abounding (9), with personal pronoun subject + verb order in one version and AIV order in the other. (7) a.

Y Gwr Cadarn, Cardiff 2.83, 71 (GK I) - AIV order gadawa fi hi yn sykutor om holl goweth am trysor leave-1s-FUT I her PRED executor of my whole wealth and my treasure

b.

Υ Gwr Cadarn, Peniarth 65, 56 (GK II) - Personal pronoun subject + verb order Myfi ai gadawa hi yn secktor Ar fy holl olyd I PART-her leave-Is FUT her PRED executor on my whole wealth "I shall leave her as executor of all my wealth and treasure"

STABILITY AND CHANGE IN EARLY MODERN WELSH

(8) a.

b.

(9) a.

b.

213

Psalms 5: 6, William Salesbury 1567 - Personal pronoun subject + verb order Ti ddestrywy y rei y ddywedant gelwydd You-2s destroy-2s-PRES/FUT the ones who say-3p-PRES/FUT lie Psalms 5: 6, William Morgan 1588 - AIV order Difethi y rhai a ddywedant Destroy-2s-PRES/FUT the ones who say-3p-PRES "You destroy those whole tell lies"

gelwydd lie

H I: 18, John Einnion 1737 - Personal pronoun subject + verb order mi a ddichwelais yn ffromwyllt at fy chwariaeth drachefn I PRT return-1sg-PAST ADV rash to my play back "I returned rashly to my play" Η II: 8, Rhys Thomas 1763 - AIV order dychwelais at fy Ynfydrwydd return-1sg-PAST to my folly "I returned to my folly"

drachefn back

The pattern of variation between each of these pairs of texts is consistent: the text with the higher frequency of AIV order has several examples with AIV order corresponding to personal pronoun subject + verb order in the other version of the text, and with very few the other way round. In the Psalms, for example there are 20 cases where William Morgan has changed personal pronoun subject + verb order to AIV order, but none where he has changed AIV order in Salesbury's version to personal pronoun subject + verb order in his own. In Helaethrwydd o Rås, Rhys Thomas has changed 10 instances of personal pronoun subject + verb order in John Einnion's version to AIV order, with only one change in the other direction. Finally, in Gwr Kadarn, there are seven cases where AIV order in the Cardiff 2.83 version corresponds to personal pronoun subject + verb order in Peniarth 65, but none the other way round. The most spectacular divergence of word order, though, especially in the use of AIV order in relation the dummy subject construction, is between William Jones' and Charles Edwards' versions of Thomas Gouge's Principles of the Christian Religion. William Jones and Charles Edwards show diametrically opposed usage: there are no examples at all of AIV order in William Jones' version but 37 examples of the dummy subject construction, whereas in Charles Edwards' translation there are 193 examples of absolute-initial verb order used in syntac-tic environments where the dummy subject construction could be used and only 32 examples of the dummy subject construction - that is, an 85.8% proportion of AIV order. Thomas Gouge's revisions were such that we do not have closely comparable parallel sentences differing only in the

214

OLIVER CURRIE

presence of absence of AIV order, but we do have several instances where William Jones uses certain verbs in partially idiomatized expressions with the dummy subject construction, which Charles Edwards then uses regularly with AIV order (Table 8). In (10) two directly comparable PDMCs from Helaethrwydd o Ras are cited, where AIV order in one version contrasts with the dummy subject construction in the other. William Jones Charles Edwards fe ddywedir dywedir "It is said (that)" fe ddywed [...] dywed [...] "[...] says" fe all [...] gall[...] "[...] is able" fe ellir [...] gellir[...] "[...] can be..." (PASS) fe ddylai [...] dylai [...] "[...] ought" fe gyfrifir [...] cyfrifir [...] "[...] is reckoned" Table 8: Comparison of William Jones' and Charles Edwards' usage (10) a. H I: 1, John Einnion 1737 - Dummy subject + Verb fe ryngodd bodd i Dduw i osod yn eu it please-3s-PAST to God to set-VN in their im dodi i mewn Yscol i ddysgu darllen to my put-VN into school to learn-VN read-VN

Calonnau nhwy hearts 3p a Scrifennu and write

b. Η II: 8, Rhys Thomas 1763 - AIV order rhyngodd bodd i'r ARGLWYDD i roi yn eu Calonnau please-3s-PAST to the Lord to put-VN in their hearts i'm dodi mewn Yscol i ddyscu darllen a 'scrifennu to my put-VN into school to learn-VN read-VN and write-VN "It pleased God to set it in their hearts to put me in school to learn to read and write"

This pattern of variation between AIV order and the dummy subject and personal pronoun subject + verb constructions in clause-initial position is also mirrored in clause-second position after a 'fronted' adverbial phrase (cf. Tables 1-6). Several constructions are possible after a clause-initial adverbial phrase: placing the verb directly after the adverbial phrase (Adverb + verb order); placing a particle y between the verb and adverbial phrase (Adverb + y + verb order); using a dummy subject construction where appropriate in the context (Adverb + dummy subject) or using a subject-verb (SV) construction, either Adverb + personal pronoun subject + verb order or Adverb + nominal subject + verb order. In XVIIth and XVIIIth century texts where there is a high frequency of AIV order, there tends also to be a relatively high frequency of Adverb + verb order, compared to Adverb + personal pronoun subject + verb order or Adverb + dummy subject order. Conversely, in texts with a high frequency of personal pronoun subject + verb and/or dummy subject

STABILITY AND CHANGE IN EARLY MODERN WELSH

215

constructions, these constructions also tend to be used frequently after clauseinitial adverbial phrases, instead of Adverb + verb order. The two Welsh translations of Grace Abounding reflect this pattern very clearly (see (11) and Table 9): (11) a. H I, John Einnion 1737 - Adverb + personal pronoun + verb order amser arall mi syrthiais or Bad i time other I fall-1sg-PAST from the boat into Afon Bedford river Bedford b. Η II, Rhys Thomas 1763 - Adverb + Verb order Amser arall syrthiais o'r Bâd i afon Bedford time other fall-1sg-PAST from the boat into river Bedford "Another time I fell from a boat into the Bedford river" John Einnion 1737 (H I) % AIV order (excl. Adv + XP) % PSV order (excl. Adv + XP) % of Adv + XP with AdvVorder % of Adv+XP with AdvPSV order

18.6 42.9 16.2 50.0

Rhys Thomas 1763 (H II) % AIV order (excl. Adv + XP) % PSV order (excl. Adv + XP) % of Adv+XP with AdvV order % of Adv+XP with AdvPSV order

48.8 18.8 42.3 14.8

Η Ι:Η II 1 :2.6 2.3: 1 1 :2.6 3.4 : 1

Table 9: Correlation between frequency use (%) ofAIV order and AdvV order in Η I and Η II

3.4 The role of stylistic choice in the extension of AIV order The pattern of variation - one of competition between functionally similar constructions - may also shed light on the diachronic development of AIV order. Perhaps the most puzzling aspect of the construction's history is how, from being rare and marginal in medieval and in much XVIth century Welsh prose, it became, by the second half of the XVIIth century, the dominant order in certain prose texts. A narrative prose text like James Owen's Trugaredd a Barn (1687) with 54.9% AIV order is, moreover, comparable in discourse type, if not in its content and cultural milieu, to many of the MW narrative prose texts analysed by Poppe (1991 & this volume). The dramatic rise of AIV order in Early Modern Welsh prose could be accounted for if, after being perceived as interchangeable with the dummy subject construction and personal pronoun subject + verb order, it then came to expand into their respective syntactic domains. Between them, the dummy subject and personal pronoun subject + verb constructions could be used in a very wide range of syntactic contexts - including all persons of the verb and all kinds of subject, nominal and pronominal (except, that is, focussed ones). This meant that both constructions could potentially be used extremely frequently, with some of the corpus texts - two prose texts (Morgan Llwyd's Llyfr y Tri Aderyn and John Griffith's sermons in Bangor 95) and two interludes (Y Rhyfel

216

OLIVER CURRIE

Cartrefol and Ffrewyll y Methodistiaid) - showing a combined frequency for the two constructions of over 67%. If AIV order could be used in exactly the same syntactic environments, we could also expect, potentially, a very high use of this construction too, and indeed this is what we find in two XVIIth century prose writers, Charles Edwards and James Owen, both of whom use it in over 50% of PDMCs in texts in the corpus. Moreover, they maintain a frequent use of AIV order in both their expository and their narrative prose texts, indicating that a frequent use of the construction was a characteristic feature of their individual prose styles. Two texts by each author are included in the corpus. For Charles Edwards these are: Υ Ffydd Ddi-ffvant (1677), a original history and description of the Christian religion, which contains sections of descriptive, narrative and argumentative prose, and Gwyddorion y Grefydd Gristianogol (1679), a didactic religious text translated from English. The two texts by James Owen are Trugaredd a Barn (1687), a narrative text consisting of series of anecdotes (mostly derived from English sources) about God's punishment of the wicked and mercy towards the righteous, and Bedydd Plant o'r Nefoedd (1693), an original treatise on baptism. Three sections of Charles Edwards' Υ Ffydd Ddi-ffvant were analysed: a descriptive section at the beginning (chapters 1-3) giving a religious survey of the various parts of the world (FfH) with 62.3% AIV order, a narrative section (chapter 16) relating the early history of Christianity in Wales with 56.3% AIV order, and an argumentative section (Rhinwedd y Ffydd Gristianogol chapter 22) expounding the virtues of the Christian faith with 68.8% AIV order. Remarkably, out of a combined total of 391 PDMCs, there are no examples of the dummy subject construction and only two examples of personal pronoun subject + verb order. The frequency of AIV order in Gwyddorion y Grefydd Gristianogol is lower, but still relatively high at 40%. Similarly, despite the very different genres of James Owen's texts, they both show a high proportion of AIV order: 54.9% for Trugaredd a Barn and 46.5% for Bedydd Plant o 'r Nefoedd. The perceived interchangeability of AIV order with the dummy subject and personal pronoun subject + verb constructions provided a syntactic and stylis­ tic resource which writers - like Charles Edwards and James Owen - could use to extend radically the range of AIV order, and so develop a new prose style. It need not necessarily be used to this end, however. It was equally possible to avoid AIV order in prose, as authors like Morgan Llwyd, Rondl Davies and John Griffith did, and this usage was well motivated in turn by being the traditional one. The fluidity and freedom of choice in usage was such that any style in between the systematic use of AIV order, on the one hand, and its systematic avoidance, on the other also seemed to be possible. Writers like

STABILITY AND CHANGE IN EARLY MODERN WELSH

217

Thomas Williams in Ymadroddion Bucheddol Ynghylch Marwolaeth (1691), and Simon Thomas in Hanes y Byd a'r Amseroedd (1718) and Histori yr Heretic Pelagius (1735), seem to come halfway in between, using both AIV order and the dummy subject construction relatively frequently. Thomas Williams uses the two constructions apparently interchangeably in many instances; there are 15 examples where he uses the dummy subject with the auxiliary verb gallu "to be able" in Ymadroddion Bucheddol Ynghylch Marwolaeth, and 11 comparable examples where he uses AIV order instead, e.g. (12) and (13): (12)

(13)

YBM 293,11; Thomas Williams 1691 - AIV order geill dynion fod mewn cyflwr Can-3s-PRES men be-VN in condition "Men can be in a righteous condition, [...]"

cyfiawnol righteous

YBM 249; Thomas Williams 1691 - Dummy subject construction feeill dynion fod mor gariadus, it can-3s-PRES men be-VN as loving ag ymae 'r gyfraith yn gofyn as is the law PROG ask-VN "Men can be as kind as the law demands [...]"

Simon Thomas, on the other hand, in both Hanes y Byd a 'r Amseroedd and Histori yr Heretic Pelagius, alternates between the two constructions, showing quite specific preferences for AIV order with certain idioms, and for the dummy subject construction with other idioms. So, for example, with the 3 s past of bod 'to be', Simon Thomas consistently uses the AIV form bu rather than the dummy subject form fe fu; for the 3sg future of bod, on the other hand, he consistently uses the dummy subject form fejydd, instead of the AIV form bydd. In the impersonal construction with the verb darfod 'to happen', Simon Thomas consistently uses the AIV form darfu, instead of the dummy subject Ïormfe ddarfu. A further possibility available to authors is style shifting in the use of AIV order. The higher use of AIV order in both Salesbury's and Morgan's Psalms compared to other parts of their Bible translations is most likely an example of this (cf. Currie 1999 for a discussion of the factors motivating such style shifting in the Psalms). 4. Implications of the data for sociolinguistic theory 4.1 The sociolinguistic nature of the variation It has been shown that the word order variation in Early Modern Welsh is not predictable in syntactic or discourse-pragmatic factors, though these do play a contributory role; instead, it has been argued that the variation is

218

OLIVER CURRIE

sociolinguistic in nature, with the single most important parameter of variation being stylistic choice on the part of individual authors. Here, sociolinguistic is understood not in relation to any specific variationist paradigm (such as those of Labov or the Milroys), but in the broader sense of denoting a speaker as opposed to a structure-oriented approach to the study of language. Indeed, from an orthodox variationist perspective, which makes correlations between social background and social group membership, on the one hand, and language use, on the other, the centrepiece of its analysis and explanation of language variation (Romaine 1984; Cameron 1990), the syntactic variation in Early Modern Welsh occurs precisely where we would least expect it. We find significant stylistic variation here between texts of a very similar nature (in terms of genre, content and discourse type), and written by authors (all men) of a similar social status and of similar social and educational backgrounds. Even if we possessed more biographical information about the various authors, we would still be hard pushed to correlate this linguistic variation with differences in social identity and social group membership. Yet, the existence of such individual stylistic variation as described here for Early Modern Welsh word order should not be seen in any way as surprising. Individual stylistic variation has been studied in other contexts (Eckert 1995; Johnstone 1996); moreover, from a theoretical perspective, as Romaine (1984) has convincingly argued, correlations between language and social group structures do not in themselves explain individual speakers' use of language, but are in fact secondary, being derived from the collective observation of individual speakers' linguistic behav-iour. Mainstream linguistics, including sociolinguistic theory, has tended to underestimate the role of individual speakers as agents of language variation and change, a role Romaine (1984:33-4) sums up as follows: by virtue of the fact that speakers can monitor and control their performance to varying degrees and act self-consciously in conveying meaning, they can vary and change language (where language is understood as linguistic behaviour).

4.2 Word order variation and pragmatic meaning A crucial factor in enabling the kind and degree of word order variation we see in Early Modern Welsh seems to be the perceived functional similarity between AIV order and two other constructions - the dummy subject construction and personal pronoun subject + verb order. The variation could, then, be a manifestation of speakers consciously or unconsciously exploiting a structural redundancy in the language. If this analysis is correct, it would appear to lend credence to the argument that sociolinguistic variation in syntax can be treated as analogous to sociolinguistic variation in phonology, where

STABILITY AND CHANGE IN EARLY MODERN WELSH

219

variation is modelled as occurring between variants 'identical in truth but opposed in their social and/or stylistic significance'. On closer analysis, though, it emerges that while this pattern is possibly the single most significant one, it is only part of a more complex picture. The variation is not, in fact, confined to AIV order, on the one hand, and the functionally similar dummy subject and personal pronoun subject + verb constructions, on the other, but also involves (though on a lesser scale) AIV order and nominal subject + verb order (e.g. 14), where there is a potentially much greater functional difference, since pre- or postverbal positioning of full nominal subjects is used crosslinguistically to distinguish variously between topic shift and topic continuity (Quakenbusch 1992), be tween (discourse) paragraph-initial and paragraphmedial clauses (Cooreman 1992), or between temporally sequenced and unsequenced clauses (Myhill 1992). (14) a. He I: 15, John Einnion, 1737 - Nominal subject + Verb order Yn ol fy mod i fel hyn dros amser after my be-VN Is like this over time meddwl arall a ddaeth im bwriad thought other PRT come-3s-PAST to my purpose "After being like this for some time, another thought came to my mind" b. Η II: 6, Rhys Thomas, 1763 - Adverb + Verb order Yn ol bod fel hyn tros Amser After be-VN like this over time daeth peth arall i'm meddwl come-3s-PAST thing other to my mind "After being like this for some time, another thing came to my mind"

Moreover, there is also significant variation between different authors involving constructions other than AIV order, in particular the dummy subject and nominal subject + verb constructions, and this variation has implications both for assessing the diachronic significance of the variation and for judging whether the Early Modern Welsh word order patterns can be described as 'stable' or not. A plausible explanation for the rise of AIV order in construc­ tions with nominal subjects is that AIV order simply began to take over the functions of the dummy subject construction, which is also a verb-initial construction, after its range had been extended at an earlier period. The dummy subject construction had been less common in Middle Welsh prose (Willis 1996:183-6), generally being restricted to subjectless, impersonal and presentative constructions, but later came to be extended to all kinds of constructions with nominal subjects, transitive and intransitive, with the exception of focus and topic-shift, where SV order continued to be preferred. Therefore, underly-

220

OLIVER CURRIE

ing the variation in the use of AIV order in nominal subject construc-tions, there could still be a stable pragmatic principle determining whether the nominal subject is pre- or post-verbal. If we look at the word order data for nominal subject constructions from the interlude section of the corpus (Table 10), this hypothesis would seem to be borne out, since there is considerably less variation in the percentage of nominal subject constructions with SV order than there is in the percentage of AIV order or the dummy subject constructions, whether specifically within nominal subject constructions or overall. This is not the pattern we find in the sermon section of the corpus (Table 11), however, where there is a range of variation in the percentage of nominal subject constructions with SV order from 19.1% to 81.1%, and a range of variation from 2.0% to 66.2% in the corresponding percentage of dummy subject constructions. It may be that the interludes are a more stable genre in discourse-pragmatic terms; this is prima facie a plausible hypothesis, but given the unpredictable nature of the variation overall, we cannot exclude the possibility that a different sample of interludes may have yielded a different pattern. NSV DU AIV AdV AdyV Other

Rhyfel 29.6 51.9 9.3 1.8. 5.6 1.8

Brutus 32.0 22.7 28.9 13.4 1.0 2.0

Afrad 31.2 39.1 18.8 4.7 3.1 3.1

Cyndr 29.4 17.6 35.3 10.3 4.4 3.0

Ffrew 26.3 68.4 5.3 0 0 0

Table 10: Word order frequencies (%) in PDMCs with nominal subjects in the interludes NSV DU AIV AdyV Other

WG 25.6 33.3 0 34.6 6.5

EM 41.7 47.2 0 5.6 5.5

WW 40.0 30.0 0 26.7 3.3

Piers 70.3 2.0 4.0 21.8 1.9

C2219 NLW3 65.3 60.7 2.4 10.3 3.8 3.6 16.7 29.8 3.9 3.5

Β 95 19.1 66.2 0 11.8 2.9

C2226 33.3 46.7 1.7 15.0 3.3

B362 81.1 2.7 5.4 10.8 0

Table 11: Word order frequencies (%) in PDMCs with nominal subjects in the sermons

The more complex pattern of variation that emerges on closer inspection suggests an important theoretical point. On the one hand, there are regular patterns of usage across otherwise widely divergent texts, in that we find the same constructions put to similar uses, for example the nominal subject + verb construction is typically used for focus, topic shift as well as in discourse paragraph-initial PDMCs, and AIV order in subjectless constructions and in contexts of topic continuity. Yet, on the other hand, as far as frequency of use is concerned, there is considerable scope for competition between constructions (cf. Barlow and Kemmer 1994:25), so that where an author shows a stylistic

221

STABILITY AND CHANGE IN EARLY MODERN WELSH

preference for a particular construction - AIV order in the case of Charles Edwards, the dummy subject construction in John Griffith's sermons, the nominal subject + verb construction in Rondl Davies' Profiad yr Ysprydion the use of the construction can be extended; the pragmatic or syntactic regularities by no means constrain this. 5.

Implications for word order description and word order change In several respects a 'top-down' approach could not cope with the variation we find in Early Modern Welsh. The most obvious shortcoming is that the extent of variation between VS and SV order - in the corpus texts the propor-tion of VS order ranges from 22% to 90%, while that of SV order ranges from 10% to 78% (Tables 12-13) - means that no straightforward typological classification of Early Modern Welsh word order in terms of the relative position of verb and subject seems to be possible. SV:V(S) SV:V(S)

Yspr 65:35 Gwel 35:65

NLW3 57:43 B362,5 59:41

Prin

Ffydd

50:50 HBA 44:56

10:90 Pel 52:48

Gwy 33:67 HI 58:42

TB 20:80 Prof 73:27

YBM 31:68 MF 68:32

Bed 29:71 Η II 31:69

C2226 49:51 Faf 59:41

Table 12: Variation in the frequency (%) of SV and VS orders in prose texts 1675-1772

SV : V(S)

Rhyfel 78:22

Brutus 57:43

Afrad 60:40

Cyndr 55:45

Ffrew 77:23

Table 13: Variation in the frequency (%) of SV and VS orders in interludes, late XVIIth c-1745

Moreover, as this variation is stylistic in nature and not predictable in pragmatic or syntactic terms, an adequate description of Early Modern Welsh word order has to incorporate a sociolinguistic dimension, typically excluded from mainstream word order description.A sociolinguistic dimension cannot simply be appended to a 'top-down' model, though, since the stylistic variation here occurs on a more 'localised' level than is usually included in 'top-down' descriptions. The variation can, in fact, only be described in terms of the inter­ relationship between individual constructions and their interaction with other (i.e. pragmatic, syntactic, idiomatic and stylistic) factors. Further, while some authors' usage of AIV order can be described in quite general terms - Charles Edwards and James Owen show an across-the-board preference for AIV order - for others such as Simon Thomas, we need to use a more localised level of analysis still, specifying particular idiomatic and lexical preferences for where AIV order is used. This does not rule out more abstract word order descriptions of a more general scope, since these can be derived from the detailed construction-based analysis, but these too are contingent on stylistic variation.

222

OLIVER CURRIE

VSO patterns appear to have been generalized in certain authors through stylistic preferences for #V(S)(O) and #AdvV(S)(0) constructions, while #PronSV(0) and #ParticleV(S)(O) patterns appear in other authors. At every level of analysis, then, we need to recognize the contingent nature of the word order variation - contingent in the sense that it is the consequence of individual writers exercising their (considerable though not unlimited) free choice in using their language, and not predictable in a deterministic manner. A 'bottomup' analysis does not aim to achieve the same formal elegance or concision as a 'top-down' one, but can justifiably claim to capture more detail and to be more informative about actual usage. When the word order patterns of Early Modern Welsh as presented here are compared with those of Middle Welsh prose, which Poppe (this volume) has described as showing a stable pragmatic (i.e. topic-based) verb-medial system, the question arises - is Early Modern Welsh word order radically different from Middle Welsh prose? If so, has there been an abrupt change in word order? The answers to these questions also hinge on whether a 'topdown' or 'bottom-up' approach to word order description is adopted. If a language's having pragmatic or topic-based word order is conceived in a 'topdown' sense, as governed, for instance, by an abstract overarching principle of grammar, then it would seem logical that changes which affect the pragmatic nature of word order in the language would also occur at this abstract and very general level. This could involve an abrupt loss of pragmatic word order. Such is the position adopted by Willis (1996), who analyses Middle Welsh prose word order within a Principles and Parameters framework as a verb-second (V2) system, where the positive setting of the parameter also rules out unmarked verb-initial order, specifically AIV order. He then posits a resetting of this parameter, which results in the loss of V2 and at the same time licenses unmarked VI: We have posited the V2-requirement to be a parameter of Universal Grammar. A language with a positive setting of this parameter will not allow core verb-initial structures. The introduction of verb-initial structures therefore requires a change in parameter setting, which in turn entails a sudden change in the grammaticality of verb-initial orders as unmarked structures. (Willis 1996:113)

The non-grammaticality of unmarked verb-initial order in Middle Welsh as well as its sudden re-emergence following the parameter resetting are required a priori by Willis' Principles and Parameters analysis of Middle Welsh prose, but this is not the only possible analysis of the Middle Welsh data. The key issues here are: (a) how do we define the pragmatic nature of Middle Welsh prose word order? and (b) does the strong tendency towards verb-medial word

STABILITY AND CHANGE IN EARLY MODERN WELSH

223

order in PDMCs in Middle Welsh prose necessarily mean that (unmarked) AIV order should be regarded as ungrammatical there? In Poppe's analysis, the five main verb-medial constructions in PDMCs in Middle Welsh prose - Nominal subject + particle a + verb, Pronominal subject + particle a + verb, Object + particle a + verb, Verbal noun object + particle a + auxiliary verb gwneuthur "to do", and Adverbial phrase + particle y(d) + verb - all conform to a more general abstract schema - (Topic2), Topici PARTICLE Verb (S) (O) - with no single construction being statistically predominant. Now, while there is clear formal parallelism between these five constructions (all being instances of 'fronting' in Poppe's terminology), it is not clear that there is a strictly comparable functional parallelism between them. In other words, the degree to which the choice of the preverbal constituent is determined by pragmatic factors seems to vary depending on the syntactic context. In narrative contexts where there is a nominal subject, for instance, there seems to be a greater likelihood that the choice of word order is determined pragmatically. When the verb is 1st or 2nd person in dialogue, however, there seems to be a strong tendency for personal pronoun subject + verb order to be used, approaching perhaps grammaticalization of the construction2. There are, as Poppe (1989:54) has noted, pragmatic motivations for such usage, though this does not lessen the case for distinguishing between the nature of topichood in these two syntactic contexts. Indeed, there are general theoretical grounds for doing so: the degree of flexibility of word order in a given syntactic context, and in turn the scope for pragmatically determined word order choice, depends on the range of alternative constructions available. As this varies from context to context, so the degree of word order flexibility varies. It is possible, then, to redefine pragmatic word order in 'bottom-up' terms, where the scope for pragmatically determined word order is seen as partly contingent on individual constructions and syntactic contexts. Using such an approach, we can also model gradual and piecemeal change whereby pragmatic word order phenomena is restricted or extended in certain syntactic contexts but not in others. So while overall, Early Modern Welsh appears to show much more stylistically rather than pragmatically determined word order variation in PDMCs compared to Middle Welsh prose, in certain specific syntactic contexts, especially in determining whether nominal subjects occur This can be measured empirically by calculating the proportion of actual occurrences of personal pronoun subject + verb order out of the total number of possible occurrences (in this case, the number of PDMCs without nominal subject or subjectless constructions). The interludes in the Early Modern Welsh corpus, which are dialogic in nature, do in fact show a very high proportion of personal pronoun subject + verb order, ranging from 63% to 90%,

224

OLIVER CURRIE

pre- or postverbally, pragmatic factors do still play a significant role. Equally, stylistic word order variation does also occur in Middle Welsh prose, including an example (though the only one I have found) of variation between AIV order and the dummy subject construction in two different manuscript versions of Mabinogi Iessu Grist (17). (17)

Version I {dummy subject) - Peniarth 14, Part. II; (Williams 1912:193) ef a damweiniawd digwydaw un or meibion gan it PART happen-3sg-PAST fall-VN one of the boys by y uwrw o un arall or lloffi yr llawr his striking-VN by one other from the loft to the ground yny uu uarw so thatbe-3s-PAST dead "One of the boys, after being shoved by another, fell from the loft to the ground and died."

(18)

Version II (AIV order) - Peniarth 5; (Williams 1912: 230) Damweinaud hagen gythyau o vn or meibion Happen-3sg-PAST however push-VN by one of the boys y Hall yr ar y lloffi yr llaur yny vu varw the other from on the loft to the ground so that be-3s-PAST dead "One of the boys pushed the other out of the loft onto the ground so that he died"

There are no a priori theoretical grounds within a 'bottom-up' framework for regarding AIV order in Middle Welsh prose as ungrammatical because it does not conform to the dominant verb-medial pattern; it seems to have been one of many constructions available in Middle Welsh, but one which happened to have been only rarely used in prose. Why it was so rare in Middle Welsh prose - while being common in poetry - and why it then re-emerged in Early Modern Welsh prose is a complex historical problem which, for reasons of space, it is not possible to address here (cf. Evans 1968; Willis 1996; Currie 1999). One dimension of the problem that must be considered, though, together with all the relevant syntactic and discourse-pragmatic factors, is the sociolinguistic one. 6. Conclusion The main aim of this article has been to show that the investigation of sociolinguistic and stylistic variation should be made an integral part of word order description - contrary to the tendency for mainstream word order theory to disregard to such variation - and further, that it is easier to describe and understand such variation using a 'bottom-up' methodology. The sociolinguistic perspective adopted here also poses a curious paradox with regard to how we define word order 'stability'. If the word order patterns in Early Modern

STABILITY AND CHANGE IN EARLY MODERN WELSH

225

Welsh can legitimately be defined as 'stable' on the grounds that stylistic variation is a natural part of language use and that language change is also a natural consequence and integral part of language use (Coseriu 1973), then are there any word order patterns - except perhaps in language death situations (e.g. Dyirbal; Schmidt 1991) - which are not stable? Who should be the final arbiter of this, the linguist with a set of predetermined criteria, or native speakers in a more subjective manner? Ultimately, there is no single answer to these questions, precisely because there are different conceptions of 'stability'. This should not necessarily be seen as a problem, though, since these different conceptions of the notion of 'stability' are all valid in themselves, and can shed light on different aspects of the functioning of word order in language. In the case study, we find word order patterns without significant variation (e.g. the use of SV order in nominal subject constructions in the interludes, 4.2) as well as patterns with considerable variation (involving, in particular, the use of AIV order), both of which can be categorized as 'stable' in different ways. The interest of stability as a concept is perhaps best seen, then, not as an end-point in word order description (i.e. is X 'stable' or not?), but as providing a starting point for investigating word order variation and change from a range of different perspectives.

ABBREVIATIONS USED FOR GLOSSING EXAMPLES ADV = adverbial particle; PRT = preverbal particle; PRED = predicative particle; PROG = progressive particle; REL = relative particle; VN = verbal noun.

REFERENCES PRIMARY SOURCES Middle Welsh Texts Parry, Thomas ed. 1979. Gwaith Dafydd ap Gwilym, 3rd ed. Caerdydd: Gwasg Prifysgol Cymru. Roberts, Brinley F. ed. 1961. Gwassanaeth Meir. Caerdydd: Gwasg Prifysgol Cymru. Williams, Mary. 1912. 'Llyma Vabinogi Iessu Grist'. Revue Celtique 33.184-248. XVIth Century Bible Translations Esther Morgan, William. 1588. "Llyfr Esther". Jones, Thomas ed. 1988. Rhyddiaith Gymraeg. Yr Ail Gyfrol. Detholion o Lawysgrifau a Llyfrau Printedig 15471618. 2nd ed. Caerdydd: Gwasg Prifysgol Cymru. 85-104.

226

OLIVER CURRIE

Isaiah Morgan, William. 1588. "Llyfr Prophvvydoliaeth Esay". Llyfrgell Genedlaethol Cymru 1987. Υ Beibl Cyssegr-lan 1588. Aberystwyth. 261a264a. Mark I Salesbury, William. 1567. "Llyma Cyssecrsanct Euangel Iesu Christ yn ol Marc". Parry, Thomas ed. 1967. Detholion o Destament Newydd 1567. Caerdydd: Gwasg Prifysgol Cymru. 1-12 MarkllMorgan, William. 1588. "Efengyl Iesu Grist yn 61 S. Marc". Llyfrgell Genedlaethol Cymru 1987. Υ Beibl Cyssegr-lan 1588. Aberystwyth. 454b457a. Psal I Salesbury, William. 1567. "Psalmae Dauid". Lliver Gweddi Gyffredin 1567. ii-xii. Psal II Morgan, William. 1588. "Psalmau Dafydd". Llyfrgell Genedlaethol Cymru 1987. Υ Beibl Cyssegr-lan 1588. Aberystwyth. 218a-221a Drama Afrad Υ Mab Afradlon, Gwenogvryn Evans MS 5 in National Library of Wales, copied 1750. Attributed to Huw Morys. Ed. in Jenkins, David 1948. Bywyd a Gwaith Huw Morys (Pont-y-Meibion). Traethawd M.A., Prifysgol Cymru, Aberystwyth. 491-519. Brutus Enterliwt Ynghylch Cronicl y Cymry er Amser Brutus hyd at Sior y 3ydd, MS, Cwrtmawr 211 A, 1-57. Copied 1734/5. Attributed to Mathew Owen. Cyndr Richard Parry 1737. Enterlute neu chwaryddiaeth Ar Destun Odiaethol yn dangospa Drigolion a fu'n Preswulo yn y Deyrnas hon cyn dyfod Cymru na Saeson erioed iw Meddiannu o wnaethuriad R. P. NLW 833B, 5-69. Diodd Υ Dioddefaint, BM Add. 14986. Jones, Gwenan ed. 1938. Three Welsh Religious Plays. Bala: R. Evans & Son. 152-237. Ffrew Roberts, William. 1745. Ffrewyll y Methodistiaid neu Buttein-glwm Siencyn ac Ynfydog GK I Y Gwr Cadarn. Cardiff 2.83, 59-78. Copied c. 1550 GK II Υ Gwr Cadarn/ Yr ymddiddan afy Ryng yr effeiriad ar gwr bonheddig. Peniarth 65, 40-72. Copied end C16th. Rhyfel Rhyfel Cartrefol, Cwrtmawr 42, attributed to Huw Morys, C18th manuscript. Expository Prose Bed Owen, James. 1693. Bedydd Plant o 'r Nefoedd. Llundain. 1-126. Car Ballinger, John, ed. 1930. Carwr y Cymry. First published 1631. Caerdydd: Gwasg Prifysgol Cymru. Deff Prichard Williams, W., ed. 1908. Deffynniad Ffydd Eglwys Loegr a gyfieithwyd i'r Gymraeg, o Ladin Esgob Jewel yn y flwyddyn 1595, gan Maurice Kyffin. Bangor: Jarvis & Foster. vi-xix, 1-105. Edif Dent, Arthur. 1629. Pregeth dduwiol yn traethu am iawn ddull, ac agwedd gwir edifeirwch. Translated by Robert Lloyd. Llundain. Epist Davies, Richard. 1567. "Epistol Episcop Menew at y Cembru". Rhagymadroddion 1547-1659 ed. by Garfield H. Hughes, 17-43. Caerdydd: Gwasg Prifysgol Cymru.

STABILITY AND CHANGE IN EARLY MODERN WELSH

227

FfH

"Hanes y Ffydd Cristianogol er dechreuad y byd hyd yr oes hon". Edwards, Charles 1677. Ffydd Ddi-ffvant, sef Hanes y Ffydd Gristianogol, a'i Rhinwedd. 3rd ed. Rhydychen. 1-21. Gors Parry, Thomas, ed. 1930. Theater du Mond (Gorsedd y Byd) gan Rhosier Smyth. Caerdydd: Gwasd Prifysgol Cymru. 1-110. Gws Robert Gwyn 1580. "Gwssanaeth y Gwyr Newydd". Gwssanaeth y Gwyr Newydd Robert Gwyn 1580 ed. by Bowen, Geraint ed. 1970. Caerdydd: Gwasg Prifysgol Cymru. 4-39. Gwy Gouge, Thomas. 1679.Gwyddorion y Grefydd Gristianogol. Translated by Charles Edwards. Llundain. 1-53. LITA Morgan Llwyd. 1653. "Llyfr y Tri Aderyn". Gweithiau Morgan Llwyd o Wynedd ed. by Thomas E. Ellis, 157-266. Bangor: Jarvis & Foster. Princ Gouge, Thomas. 1676. Priniplau neu Bennau y Grefydd Ghristianogol. Translated by William Jones. Llundain: A. Maxwell. Rhin "Rhinwedd y Ffydd Gristianogol". In Charles Edwards 1677. Ffydd Ddiffvant, sef Hanes y Ffydd Gristianogol, a'i Rhinwedd. 3rd ed. Rhydychen. 259-280. YBM Sherlock, William. 1691. Ymadroddion Bucheddol Ynghylch Marvvolaeth. Translated by Thomas Williams. Rhydychen: Thomas Jones. 1-84, 200-359. Yspr Davies, Rondl. 1675. Profiad yr Ysprydion. Rhydychen. 1st Person Narrative Prose Crwyd "Y Marchog Crwydrad". In Parry-Williams, T.H. ed. 1988. Rhyddiaith Gymraeg. Υ Gyfrol Gyntaf. Detholion o Lawysgrifau 1488-1609. Caerdydd: Gwasg Prifysgol Cymru. 96-105. Faf Powel, Vavasor. 1772. Hanes Bywyd a Marwolaeth y Parchedig Mr. Fafasor Powel. Translated by D. Risiart. Caerfyrddin. 6-21. Gwel Wynne, Ellis. 1948. Gweledigaethau y Bardd Cwsc. Caerdydd: Gwasg Prifysgol Cymru. First published 1703. 5-49. Hel I Bunyan, John. 1737. Helaethrwydd o Ras i'r Gwaelaf o Bechaduriaid. Translated by John Einnion. Caerfyrddin. 3-30. Hel II Bunyan, John. 1763. Helaethrwydd o ras i'r pennaf o bechaduriaid. Translated by Rhys Thomas. Caerfyrddin. 1-17. Prof Lloyd, Henry. 1750. Profiad Tufewnol o Nefoedd ag Uffern. Brista. 11-32. 3rd Person Narrative Prose FfC "Hanes y ffydd ymmhlith y cymru". Edwards, Charles 1677. Ffydd Ddi-ffvant, sef Hanes y Ffydd Gristianogol, a'i Rhinwedd. 3rd ed. Rhydychen. 150-162. GR Gesta Romanorum. Story 42, "Mab y Fforestwr". In Parry-Williams, ed. 1988: 122-130 and Story 10, NLW 13076, ff. 16a-18a. HBA Thomas, Simon. 1718. Hanes y Byd a 'r Amseroedd. Υ Mwythig. 1-48, 59-67, 91-103, 140-9. L1EC Puleston, Roland. 1583. Llefr o'r Eglwys Crhistnogedd, NLW 716B, la-25a, 149a-153b, 161a-168a. MF Defoe, Daniel. 1750. Hanes y ffortyn ar anffortyn, a fu i Mal Flanders. Υ Mwythig.

228

Pel TB

OLIVER CURRIE

Thomas, Simon 1735. Histori yr Heretic Pelagius. Υ Mwythig. 13-73, 137141. Owen, James (1687) Trugaredd a Barn. Llundain. 1-30.

Sermons B 95 John Griffith. 9 sermons. Bangor 95. Preached 1683-5: Lianelian. B362 Robert Wynne (?). 1 sermon. Bangor 362 (5). Preached 1717: Llanddeiniolen. C2.219 Anon. 2 sermons. Cardiff 2.219, 39-76. Preached 1668-1683: Llanwothyn, Caerwys, Wrexham. C2.226 Anon. 3 sermons. Cardiff 2.226. Preached late C17th. EM Evan Morgan. 4 sermons. 1610. Edited in Morgan, Glyn 1969. Pregethau Cymraeg William Griffith (?1566-1612) ac Evan Morgan (c. 1574-1623). Traethawd M.A., Prifysgol Cymru, Bangor. 578-646. NLW3 Anon. 6 sermons. NLW 3B, 23-97. Preached 1675-6. Piers John Piers. 5 sermons. mid C17th. NLW 12205A. WG William Griffith. 3 sermons. C16th/C17th. Edited in Morgan 1969. 70-91, 144-153,314-331. WW William Williams. 2 sermons. NLW 73B, Preached 1629-33, in Llanafan & Llanfihangel. Edited in Jones, Ruth Elizabeth 1980. Υ Bregeth Gymraeg. Traethawd M.A., Prifysgol Cymru, Aberystwyth. 28-90. Slander Case Records Slander Suggett, Richard. 1983. Early Modern Welsh Defamation Suits. SSRC Final Report (HR 6979). SECONDARY SOURCES Barlow, Micheal & Suzanne Kemmer. 1994. "A schema-based approach to grammatical description". The Reality of Linguistic Rules ed. by Susan D. Lima, Roberta L. Corrigan & Gregory K. Iverson, 19-42. Amsterdam & Philadelphia: John Benjamins. Cameron, Deborah. 1990. "Demythologizing Sociolinguistics: Why Language Does not Reflect Society". Ideologies of Language ed. by John E. Joseph & Talbot J. Taylor, 79-93. London: Routledge. Chomsky, Noam. 1995. The Minimalist Program. Cambridge& MA: MIT Press. Cooreman, Anne. 1992. "The Pragmatics of Word Order Variation in Chamorro Narrative Text". Pragmatics of Word Order Flexibility ed. by Doris L. Payne, 243278. Amsterdam & Philadelphia: John Benjamins. Coseriu, Eugenio. 1973. Sincronía, Diacronía e Historia. 2nd ed. Madrid: Editorial Gredos, S.A. Currie, Oliver. 1999. The Development of Verb-initial Word Order in Early Modern Welsh. Ph.D. thesis, University of Cambridge. Eckert, Penelope. 1995. '(ay) Goes to the City: Exploring the Expressive use of Variation'. Towards a Social Science of Language: Papers in honour of William

STABILITY AND CHANGE IN EARLY MODERN WELSH

229

Labov ed. by Gregory R. Guy, Crawford Feagin, Deborah Schiffrin & John Baugh, 47-68. Amsterdam & Philadelphia: John Benjamins. Evans, D. Simon. 1968. "The Sentence in Early Modern Welsh". Bulletin of the Board of Celtic Studies 22.311 -337. . 1976. A Grammar of Middle Welsh. Dublin: Dublin Institute of Advanced Studies. Fillmore, Charles J., Paul Kay, & Catherine O'Connor. 1988. "Regularity and Idiomaticity in Grammatical Constructions: the Case of let alone". Language 64.501-538. Givón, Talmy. 1984. Syntax: a Functional-Typological introduction. Vol. I. Amsterdam & Philadelphia: John Benjamins. . 1990. Syntax: a Functional-Typological Introduction. Volume II. Amsterdam & Philadelphia: John Benjamins. Johnstone, Barbara. 1996. The Linguistic Individual: Self-Expression in Language and Linguistics. (Oxford Studies in Linguistics). Oxford & New York: Oxford University Press. Labov, William. 1978. Sociolinguistic Patterns. Oxford: Basil Blackwell. Langacker, Ronald W. 1988. "A Usage-Based Model". Topics in Cognitive Linguistics ed. by Brygida Rudzka-Ostyn, 127-161. Amsterdam & Philadelphia: John Benjamins. Lavandera, Beatriz. 1978. "Where Does the Sociolinguistic Variable Stop?". Language in Society 7.171-182. Milroy, James. 1992. Linguistic Variation and Change. (= Language in Society, 19.) Oxford: Basil Blackwell. Milroy, Lesley. 1987. Observing & Analysing Natural Language: A Critical Account of Sociolinguistic Method. (= Language in Society, 12.) Oxford: Basil Blackwell. Myhill, John. 1992. "Word Order and Temporal Sequencing". Pragmatics of Word Order Flexibility ed. by Doris L. Payne, 265-278. Amsterdam & Philadelphia: John Benjamins. Poppe, Erich. 1989. "Constituent Ordering in Breudwyt Maxen Wledig". Bulletin of the Board of Celtic Studies 36.43-63. . 1991. "Word Order in Cyfranc Lludd a Llefelys: Notes on the Pragmatics of Constituent-Ordering in MW Narrative Prose". Studies in Brythonic Word Order ed. by James Fife and Erich Poppe, 155-204. Amsterdam & Philadelphia: John Benjamins. Quakenbusch, Stephen. 1992 "Word Order and Discourse Type: an Austronesian Example". Pragmatics of Word Order Flexibility ed. by Doris L. Payne, 279-303. Amsterdam & Philadelphia: John Benjamins. Romaine, Suzanne. 1984. "The Status of Sociological Models and Categories in Explaining Language Variation". Linguïstische Berichte 90.25-38. Schmidt, Annette. 1991. "Language Attrition in Boumaa Fijian and Dyirbal". First Language Attrition ed. by Herbert W. Seliger & Robert M. Vago, 113-124. Cambridge: Cambridge University Press. Siewierska, Anna. 1988. Word Order Rules. London & New York: Croom Helm. Thomas, Peter Wynn. 1996. Gramadeg y Gymraeg. Caerdydd: Gwasg Prifysgol Cymru.

230

OLIVER CURRIE

Weinreich, Uriel, William Labov, & Marvin I. Herzog. 1968. "Empirical Foundations for a Theory of Language Change". Directions for Historical Linguistics ed. by Winfred P. Lehman & Yakov Malkiel, 97-195. Austin: University of Texas Press. Willis, David W. E. 1996. The Loss of Verb-second in Welsh: A Study of Syntactic Change. D.Phil, Oxford.

CONVERGENCE AND DIVERGENCE IN THE DEVELOPMENT OF THE GREEK AND LATIN CLITIC PRONOUNS

MARK JANSE Bibliographie Linguistique (The Hague) and University of Ghent

0.

Introduction It is common knowledge that in many languages, clitic pronouns (and, for that matter, clitics in general) are "subject to special positioning rules only loosely, if at all, relating to their grammatical relation" (Comrie 1989: 89). 1 In his classic essay "On Clitics", Zwicky captured this observation by distinguishing between "special" and "simple" clitics, the latter having "ordinary syntax", the former "special syntax" (1977: 6). Following Zwicky's definition, the English clitic pronouns would be simple clitics, as in (lb), the French special, as in (2b): (1) a. I love my wife b. I love her (2) a. J'aime ma femme b. Je l'aime

It turns out, however, that within one and the same language, the position of clitic pronouns sometimes varies across dialects, as in the case of Modern Greek where the clitic pronouns obligatorily procliticize to finite verbs, as in (3 a), whereas they generally encliticize to finite verbs in such dialects as Cappadocian, as in (3b):2 1 This material is based upon work supported by the Fund for Scientific Research-Flanders (Belgium) under Grant 1.5.416.97. 2 Clitic pronouns are referred to by number and person: 1SG = first person singular, 1PL = first person plural, etc. Other abbreviations: ADV = adverb, ART = article, cc = clitic coordinator, CL = clitic, co = coordinator, CP = clitic particle, DEM = demonstrative pronoun, INTER = interrogative pronoun, NEG = negative, PRT = particle, REL = relative pronoun, SUB = subordinator. Source material: Mt = Matthew, Mk = Mark, Lk = Luke, Jn = John. The equals sign = marks the attachment of clitics, a simple hyphen - the attachment of affixes. Ancient and Medieval Greek quotations are transliterated, Modern Greek quotations transcribed.

232

MARK JANSE

(3) a. b.

m= 1SG aγapás you-love

aγapás? you-love =me? 1SG

The same phenomenon occurs within the Romance language family where procliticization is obligatory with finite verbs in Spanish, as in (4a), but optional or, more correctly, context-dependent in such languages as Galician, as in (4b): (4) a. b.

me= 1SG quéres you-love

quieres? you-love -me? 1SG

The examples just quoted are of course minimal sentences. It turns out that, depending on context, the clitic pronouns are found to procliticize to finite verbs as well, in both Cappadocian and Galician. This is all the more intriguing in view of the fact that the clitic pronouns behave more or less the same in the historical antecedents of Cappadocian and Galician. The examples just quoted are in fact taken from the fourth gospel which exhibits the same order in the original Greek text (5 a) and in the Latin translation of the Vulgate (5b): (5) a. b.

phileîs amas you-love

=me? =mē? 1SG

"do you love me?" (Jo 21.17)

In the remainder of this paper, I will take a closer look at the causes which have led to the grammaticalization of the preverbal position of the clitic pronouns in Standard Modern Greek and the majority of the Romance languages on the one hand, and the variability of their position in the easternmost modern Greek dialects and the westernmost modern Romance languages on the other. I will start with a brief review of the situation in some of the earliest attested docu­ ments of Ancient Greek and Latin.

1. Ancient Greek and Latin Ever since Wackernagel's seminal article (1892), which put his name to what has been called "[o]ne of the few generally accepted syntactic statements about Indo-European" (Watkins 1964: 1036), it is a generally accepted fact that Ancient Greek and Latin are prototypical Wackernagel-type languages (MeilletVendryes 1948: 580). Wackernagel's Law is often, but wrongly, interpreted as

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

233

having to do with sentence-second position.3 Wackernagel himself used the ambiguous term "Satz", but it is clear that he was talking about clause-second, not sentence-second position.4 Consider, for instance, the consistently clausesecond position of the clitic pronouns mēd and tēd in the archaic Latin Duenosinscription (CIL I2 4, cf. Eichner 1988-90: 210; Janse 1994a: 110): (6) a.

b.

c.

d.

e.

f.

iouesāt deiuōs qoi =mēd mitāt s/he-swears gods REL 1SG s/he-sends "to the gods swears he who sends me" nei -tēd endō cosmis uircō siēd SUB 2SG at smiling virgin s/he-should-be "should the girl not smile at you" as(t) -tēd noisi op(p)etoit co 2SG SUB s/he-should-come-towards "and should she not approach you" esiai pāca riuois her appease with-smells "appease her with delicious smells" duenos =mēd fēced en good 1SG s/he-has-made ready "Good One has made me" nē =mēd malo(s) statōd NEG 1SG envious s/he-should-steal "don't let Envious steal me"

Wackernagel (1892: 406) suggested that "die Stelle unmittelbar hinter dem ersten Wort des Satzes mit Tonschwäche verbunden sei", a suggestion Dunn (1989: 7) has treated as "confirmation, on the front of the sentence, of a sentence stress". 5 Elsewhere, I have elaborated the prosodic basis of Wackernagel's Law and shown that the Wackernagel position is best taken as second within the intonation unit rather than second within the sentence or clause, even though the former and the latter are often coextensive.6 Consider, for instance, the follow-

3

Cf., e.g., Watkins (1964: 1036), Collinge (1985: 217), Klavans (1985: 96), Anderson (1993: 69). 4 Cf., e.g., Spencer (1991: 355), Zwicky (1994: xiv). 5 Cf. Jakobson (1935: 384 1 ), Wanner (1996: 538). 6 On earlier occasions (Janse 1990b; 1993a), I have used the term "sentence segment" which was intended to be coterminous with "intonation unit". Compare also the similar claims by Hock (1996: 264).

234

MARK JANSE

ing verses from Homer's Odyssey where the clitics me, t(e), d ,7 min and hoi are placed second within their intonation unit:8 (7) a.

b.

(8) a.

b.

allá =me \ sós PRT 1SG your phaídim' Odusseû noble Odysseus

-te cc

póthos longing

|

sá your

=te cc

m dea counsels

\

"but me, longing for you and your counsels, noble Odysseus" (11.202) s =t' aganophrosúnē | meliēdéa thumòn apēúra your cc kindliness honey-sweet spirit s/he-reft-away "and your kindliness, they reft my honey-sweet spirit away" (11.203) hoì =d -min | pen k ri | theòn h s \ tim santo DEM CP 3SG very in-heart god like they-honored "and they honored him, very heartily, like a god" (19.280) kaí -hoi pollà dósan | pémpein -té -min \ co 3SG many they-gave send CC 3SG thelon autoì they-wished themselves "and gave him many things, and wished themselves to send him" (19.281)

If "second position" has to do with Tonschwäche, as Wackernagel suggested, it stands to reason to assume that "first position" has to do with Tonstärke.9 This is tantamount to saying that the inherently unaccented clitics are attached to words or phrases which, for discourse pragmatic reasons, are stressed and hence placed first within their intonation unit. From this point of view, the syntax of the clitic pronoun min would be "ordinary" in (8b), since it is attached to the infinitive pémpein on which it depends syntactically. In a way, this might be considered the "natural" order, since the phonological dependence of the clitic here coincides with its syntactic dependence (Janse 1993a: 19). 10 A fortiori, the position of min in (8a) would be "special", since its phonological dependence is inconsistent with its syntactic dependence. Moreover, it is separated by four other words from the finite verb tim santo on which it depends. In the examples just quoted, the clitics are variously attached to function words which are obligatorily placed first in their clause and to content words which are placed first in their intonation unit for discourse pragmatic reasons. If first position has to do with Tonstärke, as suggested earlier, then the question 7

On the status of d as a clitic see Wackernagel (1892: 378), Ruijgh (1990: 222). On the principles of intonational phrasing in Homer see Janse (1998d: 135ff.). 9 Cf. Weil (1879: 96), Jakobson (1935: 3841), Dunn (1989: 7), Janse (1990: 2648), Wanner (1996: 538). 10 Cf. Wackernagel (1892: 362), Jakobson (1935: 384), Janse (1990b: 2648; 1993b: 86). 8

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

235

should be asked why it is that certain function words should obligatorily come first in their clause. It is common knowledge that Greek and Latin are languages with a relatively "free" word order. If such languages have fixed positions for certain words or word classes, it stands to reason to assume that the position of these words must have been grammaticalized in the course of time. Given a broad definition of grammaticalization as the "organization of grammatical, especially morphosyntactic material" (Hopper-Traugott 1993: 50), the most common mechanism of language change leading to grammaticalization is the "fixing of discourse strategies" (Traugott 1994: 1483). The coordinator allá (literally "otherwise") in (7a) is a case in point. Etymologically, allá is the plural accusative of the pronoun állos "other" used prosthothetically, i.e. as a left-dislocated parenthetical. It developed into a first position coordinator through reanalysis involving boundary loss (Schwyzer 1950: 578). Its value as a function word is defined by Schwyzer (ibid.) as "scharf gegensätzlich" which accounts for both its position and its being stressed. The same holds for the coordinator kaí (8b), which was originally used as a focusing particle with the value "auch, sogar", "und besonders (auch)" (Schwyzer 1950: 567). In comparison with the clitic coordinator te (7a-b), kaí is characterized) as "lautstärker, gewichtiger" and hence "regelmäβig voranstehend" (ibid.). This accounts for the use of the clitic pronoun me in (7a) instead of the orthotonic emé. In later Greek, kaí gradually took over from te and developed into a proclitic coordinator no longer capable of being combined with clitics. The Alexandrian grammarian Apollonius Dyscolus explicitly contrasts the following minimal pair (Uhlig 1910: 170.8ff.): (9) a.

b.

etímësé =se kal Dionēsion s/he-honoured 2SG CO Dionysius "he honoured you, and Dionysius (as well)" Dionúsion tim i kal emé Dionysius s/he-honours co 1SG (NON-CL) "he honours Dionysius, and me (too)"

Whereas coordinators are either first (Greek kaí, Latin et) or second position (Greek te, Latin -que) words, subordinators are canonical first position words. Many subordinators are pronouns or have a pronominal etymology (Janse 1997a: 113). The archaic Latin relative pronoun qoi (6a), for instance, is formally identical to the interrogative pronoun (Leumann 1977:472). Both are first position words for discourse pragmatic reasons, the former functioning as topic, the latter as focus. The grammaticalization of their position is again a case of "fixing of discourse strategies" through reanalysis involving boundary loss:

236

MARK JANSE

(6)

a'.

[[iouesāt][deiuōs]]

|

[[qoi][mēdmitāt]]

[iouesāt [[deiuōs][[qoi][mēd mitāt]]]

The originally non-subordinating status of the relative pronoun is still particularly evident in Homeric Greek, where the relative pronoun hoí (8a) is formally identical with the "démonstratif en fonction de relatif" (Chantraine 1958: 277), the interpretation of which is ambiguous between "they" and "who". So far we have seen that canonical first position words ("Anfangswörter", as Wackernagel (1892: 410) called them) are the result of fixed discourse strategies. However, the same discourse strategies underlie the placement of noncanonical first position words such as the possessive adjective sós (sá, s) (7a-b) or the adjective duenos (6e). The various forms of sós are all focused, as is duenos, which at the same time serves as the discourse topic of intonation units (6e-f). We may hypothesize that first position words are placed first within their intonation unit for discourse pragmatic reasons, that primary stress is a corollary of first position and, finally, that it is the Tonstärke of first position words which is responsible for attracting clitics in second position (Janse 1990b: 2648; 1993a: 22). One of the corollaries of the fixing of discourse strategies in this particular case is the reanalysis of "collocations" (Janse 1994a: 112) consisting of canonical first position words and second position clitics, especially if the former belong to specific word classes. One such case is the above mentioned collocation of relative pronouns with clitics, as in (6a) and (8a), a fact already observed by Wackernagel (1892: 409f.) with regard to Latin. 11 It has already been remarked that the Latin relative pronoun is etymologically related to the interrogative pronoun, which is also a canonical first position word in both Greek and Latin. 12 As with relative pronouns, clitics are regularly attracted to interrogative pronouns. 13 A particularly telling example is the following, where the collocation quid me is uttered as a separate intonation unit (Janse 1994a: 117f.):14 (10)

quid INTER

=mē | 1SG

inquit s/he-says

\

rogās? you-ask

"why do you ask me? he says" (Petronius, Satyricon 52.5)

11

Cf. Adams (1994a: 44ff.; 1994b: 134f., 144ff.), Janse (1994a: 116; 1997a: 112f.), Salvi (1996: 13). Adams (1994b: 122) justly refers to the relative pronoun as the "favoured host for clitics". 12 Cf. Dover (1960: 20), Marshall (1987: 1, 12), Ruijgh (1990: 226). 13 Cf. Wackernagel (1892: 409, 428f.), Krisch (1990: 68ff.), Janse (1993a: 20; 1993b: 89f.; 1994a: 116ff.; 1997a: 113), Adams (1994b: 116, 148f.). 14 Whether (10) comprises two or three separate intonation units is moot question, but there can be no doubt that quid mē (inquit) and rogās are separated by a potential pause.

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

237

The fact that the clitic pronoun me has been separated from the verb on which it depends is, of course, very remarkable in a sentence which is, by all standards, minimal. The same applies to the following example from Ancient Greek, where it is the clitic particle án which is separated from the verb it modifies: (11)

tí =oûn =άι | éphēn \ eíē ho= INTER CP CP I-said s/he-would-be ART-CL "what then, I said, would Eros be?" (Plato, Symposium 202d)

érōs? Eros

2. Biblical Greek and Latin (and beyond) There are many more collocations of first position words with second position clitics. I will illustrate these from New Testament Greek, because it oc­ cupies a pivotal position in the development of the Greek clitic pronouns. Before going into details, it will be instructive to take another look at the position of min in relation to pémpein in (8b). Since pémpein comes first in its intonation unit, we may conclude that it owes its position to its being focused, which is perfectly consistent with what can be gathered from the linguistic context. Consequently, the attachment of te and min to pémpein fits in with the attachment of d and min to hoí in (8a), and of hoi to kaí in (8b). As has already been noted, however, min has "ordinary" syntax in (8b), but "special" syntax in (8a). It has also been remarked that the position of min in (8b) is in some sense more natural than that of min in (8a). In fact, in minimal sentences such as (5a) "ordinary" syntax is the only option available, given the enclitic nature of the Greek (and Latin) clitic pronouns. It should come as no surprise, then, that "ordinary" syntax, i.e. postverbal placement, came to be a major factor in the development of the Greek (and Latin) clitic pronouns. As Marshall (1987: 15f.) puts it: "there are two preferred positions (which may coincide) for q [= clitics— MJ] in a sentence, (i) peninitial in sentence, clause or colon [= intonation unit— MJ], (ii) directly postverbal". In New Testament Greek, postverbal placement is by far the preferred order (Janse 1993b: 86, 106). This preference for "ordinary" syntax is a general tendency in the Greek of the Hellenistic period, as Wifstrand (1949: 178ff.) emphasizes. It stands to reason to assume that in the case of New Testament Greek, this general tendency was reinforced by the particular contact situation in 1st century Palestine in which Aramaic and Hebrew, either directly or via the Greek translation of the Hebrew Old Testament, played a significant part (Janse 1998e; forthcoming). Like any other Semitic language, both Aramaic and Hebrew have pronominal suffixes instead of clitic pronouns. This may ex-

238

MARK JANSE

plain, at least partly, the extreme variation in the many manuscripts which have come down to us, but it will be important to remember that postverbal placement was but one of the two "preferred" positions. Consider, for instance, the variation in the Greek translation of Jesus' cry from the cross: (12) a. b.

eis= ti =me to INTER 1SG eis= tí egkatélipés to INTER you-have-forsaken

egkatélipés? you-have-forsaken =me? 1SG

(A C Θ f1-13 157

)

( Β Ψ 059 pc)

"why have you forsaken me?" (Mk 15.34)

Jesus' last saying is a quotation from the Old Testament (Psalm 22.2): (12) c. to

lā -mâ ăzavtā INTER you-have-forsaken

-ni? 1SG

It is the Greek version (12b) which renders the order of the Hebrew text (12c) exactly, the -nî in 'ăzavtānî being the first person singular pronominal suffix corresponding to the Greek clitic pronoun me. The fact that the manuscript tradition offers an alternative to the Greek "calque" version (12b) is significant, since postverbal placement was not just one of the two "preferred" positions in Marshall's sense, but it would have been the preferred position according to Wifstrand's observations. The same variation occurs in the following quotation from the Old Testament book of Habakkuk, where the Greek "calque" version (13b) again copies the order of the Hebrew original (13c): (13) a.

hina tí =moi to INTER 1SG b. hína tí édeiksás c. lā -mâ θar'ē -nî 'āwεn to INTER you-have-shown 1SG troubles

édeiksas you-have-shown mo i

kópous troubles kópous

(A) (B S)

"why have you shown me troubles?" (Habakkuk 1.3)

Here again, the -nî in θar'ēnî is the first person singular pronominal suffix corresponding to the moi in the Greek translations. The very existence of the variant readings (12b) and (13b) testifies to the collocational nature of combinations of interrogative pronouns with clitics. The postverbal position of the clitic pronouns in (12a) and (13a) is in every sense their "ordinary" syntax in that it reflects the unmarked order in New Testament Greek. It should come as no surprise that for practically every instance of "special" syntax a variant reading exhibiting "ordinary" syntax is available, but not the other way around.

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

239

To round off the discussion of "special" syntax in the presence of an interrogative pronoun, consider the following parallel texts from the synoptic gospels with their Latin translations (Janse 1993a: 20): (14) a.

tis

hoART

INTER

b.

(15) a. b.

hapsámenós having-touched

quis est quī mē INTER s/he-is-CL REL 1SG "who is it that touched me?" (Lk 8.45) hëpsató =moú =tis tetigit -mē =aliquis s/he-touched 1SG someone-CL

-mou? 1SG

tetigit? s/he-has-touched

"someone touched me" (Lk 8.46)

These examples are uttered in the same context. Examples (15a) and (15b) exhibit the "ordinary" syntax which is the only option available, given the en­ clitic nature of the clitic pronouns (including tis). The position of mou in (14a) is in accordance with the following "rule" formulated by Marshall (1987: 21): "Belonging to an articular verbal phrase (participial or infinitival), a postpositive [= clitic—MJ] does not precede the article or stand later than directly after the last mobile [= full word—MJ] in the phrase". 15 It should be noted that the same order is obligatory in Standard Modern Greek (Janse 1993b: 111). It should also be noted that a relative clause is used in the Latin translation, with the clitic pro­ noun attached to the relative pronoun, as in (6a). Now compare Mark's version of (14a): (16) a. tís -mou hpsato? b. quis -mē tetigit? INTER

1SG

s/he-touched

"who touched me?" (Mk 5.31)

As a result of Mark's using a finite verb instead of a participial phrase, mou has been attracted to the interrogative pronoun, this time without variant readings exhibiting "ordinary" syntax. I now turn to other collocations of first position words with clitic pronouns (Janse 1993b: 88-107). The Latin evidence comes again from the translation of the Vulgate, but the same collocations recur in "non-translation" Latin (Janse

15

To the few expceptions quoted or referred to by Marshall (ibid.) the following may be added: tines eisí me hoi dzëtoüntes? who they-are me the looking "who are the ones who are looking for me?" (Aristophanes, Birds 95).

240

MARK JANSE

1994a: 113ff.). 16 As already remarked, Wackernagel's Law is regularly trig­ gered off by the presence of subordinators (including relative pronouns). The first example is taken from Mark's account of the death of John the Baptist (6.25). Note again the variation in the written tradition: (17) a. b.

thélō hína dis I-want SUB you-give thélō hína =moi I-want SUB 1SG "I want you to give me right

=moi eksauts 1SG immediately dis eksauts you-give immediately now" (Mk 6.25)

(A C 3 Θ 544 579

)

(Ν Ο Σ 080)

The Latin translation is based on yet another variant reading and therefore of little importance for our purpose. The following example is more instructive: (18) a. b.

ouNEG nōnNEG

khreían need opus need

ékheis you-have =est s/he-is-CL

hína SUB -tibi ut 2SG SUB

-tís -se someone-CL 2SG -quis -tē someone-CL 2SG

erōtâi s/he-asks interroget s/he-asks

"there is no need for you to have someone ask you questions" (Jo 16.30)

In both cases that subordinator hína/ut has attracted both tis/quis and se/te. Note that there is no variation in the written tradition, which again testifies to the collocational nature of combinations of subordinators with clitic pronouns.17 A canonical first position word which can be used as a subordinator in Greek as well is the negative m, corresponding to the Latin në in (6f). Moorhouse (1959:147) remarks that the "primitive position for the nexal [= sentence —MJ] negative was at the start of the sentence". However, he also notes (ibid.) that "the association of negative and verb is greater in the post-Homeric language". "In the New Testament," he maintains (1959: 148), "the negative regularly precedes the verb". This is obviously correct in the case of the negative ou which, like its Latin counterpart nōn, has been grammaticalized to the point where it practically procliticizes to the verb (but see below). M, on the other hand, like its Latin counterpart nē, is a negative "with more independence", not "closely attached to a single word", to quote Moorhouse (1959: 28) once again. "We would on such grounds expect it to be accented", he continues (ibid.), "The testimony of antiquity is unanimous in always giving m an accent, and there is no good reason why we should disregard it". Again we may conclude that it is the Tonstärke of m (and nē) which attracts clitics, as in the following example: 16 17

For similar evidence from Classical Latin see Adams (1994b) and Salvi (1996). On the order of the clitics est and tibi in (18b) see Janse (1997a: 107f.).

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

(19) a. b.

241

m

-me basanízēis nē -mē torqueās NEG 1SG you-torture "don't torture me!" (Lk 8.28)

The combination ou m, which is the "schärfste Form der Verneinung" in the words of Radermacher (1925: 168), has the same attraction on clitic pro­ nouns. In the Latin translation of the following example, the negative nōn is used to render ou m, but the fact that it has attracted the clitic tē indicates its dis­ course pragmatic status as focus (the idea of denial is given from the preceding context): (20) a. ou- m -se aparnsomai b. nōn- =tē negābō NEG NEG 2SG I-will-disown "I will never disown you!" (Mt 26.35 = Mk 14.31)

Wackernagel's Law is regularly triggered off by subject pronouns as well. It will be obvious that subject pronouns share the discourse pragmatic properties of focused constituents: "die Nominative der Personalpronomina (oder die Nominative dafür eintretender Demonstrativa) wurden nur gesetzt, wenn ein besonderer Nachdruck auf sie gelegt werden sollte" (Schwyzer 1950: 187; cf. Leumann 1977: 173). It should come as no surprise, then, that subject pronouns are what Dover (1960: 20) calls "preferential" words, i.e. words which are "disproportionately common at the beginning of a clause". As in the case of canonical first position words, the written tradition often vacillates between "special" and "ordinary" syntax, as in the following variant readings (significantly, the Vulgate (21c) has "special" syntax as well): (21) a. b. c.

eg epitássō =soi I I-command 2SG eg =soi epitássō ego -tibi praecipiō I 2SG I-command "I command you!" (Mk 9.25)

( B C L Δ 33 pc) (A D X θ Π f1.13

)

Another particularly telling example can be found in the following excerpt from the Apocalypse, where the manuscripts agree in both the Greek (22a) and the Latin (22b) tradition. As matter of fact, the text is another quote from the Old Testament (22c), but note that, surprisingly, the Septuagint (22d) has "special" syntax:

242

MARK JANSE

(22) a. eg ēgápēsá =se b. ego dilexī =tē c. 'ărî 'ăhavtî -xā I I-have-loved 2SG d. eg -se ēgápēsa I 2SG I-have-loved "I have loved you" (Apocalypse 3.9 = Isaiah 43.4))

In (22c), the -xā in 'ăhavtîxā is the Hebrew second person singular pro­ nominal suffix corresponding to the se in the Greek texts (22a) and (22b). It will be clear, however, that only the "ordinary" syntax of the former matches the order of the Hebrew text. Other examples of subject pronouns attracting clitic pronouns include the following: (23) a.

hoekeinos

poisas having-made =moi

DEM

1SG

quī ille

-mē =mihi

ART

b.

DEM

lSG

who

me

=me 1SG

eîpen s/he-said fēcit dīxīt s/he-said s/he-has-made

hugi well

sānum

well

"the one who made me well, he said to me" (Jo 5.11)

What we have here is a left-dislocated participial phrase, serving as the topic of the sentence, which is referred to in the second intonation unit by the "resumptive" demonstrative pronoun ekeinos. Like its Latin counterpart ille, ekeinos has attracted the clitic pronoun for very much the same reasons as the relative pronouns qoi in (6a) and hoi in (8a). Note, too, the position of me within the Greek participial phrase in (23 a), which is in accordance with Marshall's rule discussed apropos of (14a), and the position of me in the Latin relative clause in (23b). The attraction of clitic pronouns to "resumptive" demon­ strative pronouns is not limited to subject pronouns. In fact, it seems safe to assume that it is the "preferential" status of demonstratives which accounts for the frequent attraction of clitic pronouns, as can be gathered from the following example:18 (24) a. b.

hò quod REL

=dè -autem CP

ékhō | habeō I-have

|

toûtó hoc DEM

=soi -tibi 2SG

dídōmi dö I-give

"but what I have I give you" (Acts 3.6)

On the "preferential" status of demonstratives see Dover (1960: 21f.), Ruijgh (1990: 226).

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

243

As has already been remarked on several occasions, canonical first position words such as relative or interrogative pronouns and, to a lesser degree, "prefer­ ential" words such as subject pronouns or demonstratives owe their preference for first position to the fixing of discourse strategies. However, the same dis­ course strategies account for the attraction of clitic pronouns in the case of "nonpreferential" words. Consider, for instance, the following example: (25) a. b.

ho= ART

kósmos mundus world

-se -tē 2SG

ouk= nōn= NEG

égnō cognōuit s/he-knew

"the world did not know you" (26) a. b.

eg ego I

=dé =autem PRT-CL

-se =tē 2SG

égnōn cognōuī I-knew

"but I knew you" (Jo 17.25)

In (25a), se has been attracted to the noun phrase ho kósmos which is clearly a focalized constituent on account of its contrast with ego. The same ap­ plies, mutatis mutandis, to the attraction of tē to mundus in (25b). Note that the subject pronouns have attracted the clitic pronouns, together with the clitic particles, in (26a-b), as might be expected. Another example is the following, where the Greek manuscripts again vacillate between "special" and "ordinary" syntax. I begin with the reading of the bilingual Codex Bezae Cantabrigiensis (D) (27) a. kaì ho= krits paradsei =se ti- práktori b. et iüdex trädat -të exäctöri co ART judge s/he-turns-over

2SG

ART

to-officer

"and the judge turns you over to the officer" (28) a. kaì ho= práktōr baleî =se eis= phulakn b. et exāctor mittat =tē in= carcerem co ART officer s/he-throws

2SG

in

prison

"and the officer throws you in prison" (Lk 12.58 D)

In both the Greek version and its Latin translation, the clitic pronouns are all postverbal. The majority of witnesses, however, have "special" syntax: (29) a. kaì ho- krits =se paradsei ti= práktori b. et idex -tē trādat exāctōri CO ART judge 2SG s/he-turns-over "and the judge turns you over to the officer"

ARTT

to-officer

244

MARK JANSE

(30) a. kaì ho- práktōr =se baleî eis- phulakn b. et exāctor =tē mittat in= carcerem CO ART officer 2SG s/he-throws in prison "and the officer throws you in prison" (Lk 12.58 A B T ƒ 1 3 157 167 346 αl)

What has happened here is that the clitic pronouns are attached to the contrastive subjects of the respective clauses. In (29a), the subject switches to ho krits, which is contrasted with both ho antídikos, the subject of the preceding clause, and ti práktori, the new subject of (30a). As a result of the double subject switch, Wackernagel's Law has been triggered off. The last example comes from Jesus' pardon of the sinful woman, while he was a guest of Simon, a Pharisee. In the following excerpt, two aspects of the woman's behaviour are contrasted with two expressions of hospitality Simon had not shown to Jesus: (31) a.

húdōr water

=moi ISG

epìon

pódas feet

ouk= NEG

édōkas you-have-given

"water for my feet you did not give me" (Lk 7.44) b.

phílēmá =moi ouk= kiss ISG NEG "a kiss you did not give me" (Lk 7.45)

édōkas you-have given

It will be obvious that the clitic pronouns have again been attracted to words which are the focus of the sentence. What is remarkable about (31a), however, is the separation of moi from the verb phrase on which it depends. This was not at all uncommon in older Greek, as can be gathered from the position of me in (7a), min in (8a), or hoi in (8b), but in New Testament Greek this separation is exceptional. From this perspective, the reading of D is particularly intriguing: (32) húdōr epì= pódas =moi ouk= édōkas (D) water on feet ISG NEG you-have-given "water for my feet you did not give me" (Lk 7.44)

One possible explanation is to consider húdōr epì pódas as a complex constituent, a possibility which cannot be ruled out altogether since it could be taken as a single expression of hospitality, comparable to the phüëma in (31b). It cannot be precluded, however, that the separation of moi from the verb phrase on which it depends was no longer possible in the Vth century, which is when D was written down. In Medieval Greek, clitic pronouns are placed either "immediately before or immediately after the verb", according to Mackridge (1993: 327). He correctly argues (1993: 339) that "the clitic object pronoun ceased to be a freely moving part of the clause and instead became part of the verb phrase".

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

245

Rollo (1989: 136) contends that preverbal clitics are ipso facto proclitic in Medieval Greek. However, "ordinary" syntax is obligatory "when the verb phrase stands at the beginning of a clause" (Mackridge 1993: 340). Put differently, in minimal sentences comprising just the verb and a clitic pronoun, the latter is necessarily postverbal, as in Ancient Greek. Whether or not the Medieval Greek clitics were accentually proclitic is therefore a moot question (but see below). "Special" syntax is attested in Medieval Greek under exactly the same conditions as in Ancient Greek, as can be gleaned from the "summary of rules" given by Mackridge (1993: 340-341). It should be noted, however, that these rules are labelled "more or less obligatory" (1993:340), which is tantamount to saying that there is still room for variation. In any event, it turns out that the clitic pronouns are always preverbal in the presence of canonical first position words such as interrogative and relative pronouns, subordinators and negatives. In the presence of "preferential" words, the position of the clitic pronouns vacillates between "ordinary" and "special" syntax. Mackridge (1993:329f.) observes: "In Medieval Greek the basic order of constituents in the clause (i.e. where no particular item is semantically emphasized) requires that adverbs [...], object noun phrases (direct and indirect), and subject and object complements [...] should follow the verb. Thus when one of these elements precedes the verb, it usually receives special semantic emphasis and attracts the pronoun to the preverbal position". It should be obvious that Medieval Greek has retained the same discourse strategies as Ancient Greek and that it is again the Tonstärke of the preverbal words or constituents which is responsible for the attraction of the clitic pronouns. Mackridge (1993: 329) maintains that the "special" syntax of clitic pronouns in the presence of canonical first position words "can be explained in purely syntactical terms". Put differently, combinations of canonical first position words with clitic pronouns in Medieval Greek are no longer the result of a discourse strategy, but have been reanalyzed as a syntactic rule involving specific words and word classes. The attraction of clitic pronouns to "preferential" words and focussed words or constituents remains an optional discourse strategy, but as Joseph (1990: 129) remarks, "the tendency toward placement on the left [of the verb—MJ] became stronger and stronger, eventually becoming the only possibility in Standard Modern Greek". The situation in medieval Romance is basically the same: "i clitici seguono il verbo quando questo è il pnmo elemento della frase, e lo precedono se il verbo non è il primo elemento della frase" (Salvi 1991: 443). This well-known observation has come to be known as the Tobler-Mussafia Law which, as will have become clear, can be considered a variant of Wackernagel's Law as defined here (Salvi 1991: 443; Wanner 1996: 541). Since the Tobler-Mussafia Law has been

246

MARK JANSE

discussed at great length in the literature on clitic placement in medieval Romance, I will concentrate on a crucial phase in the development of the clitic pronouns, viz. the reanalysis of preverbal position involving boundary loss. I am referring to the change in the "direction of clitization [...], and with it the analysis of the phenomenon in question", represented as follows by Wanner (1996: 538):

(33) a. [S[X=clitic]P1[V]P2Z S] b. [S[X]Pl[clitic=V]P2Z S]

In both (33a) and (33b), the clitic appears in second position (P2), but whereas it is enclitic on the first position (P1) word X in (33a), it is reanalyzed as proclitic on the verb in (33b). The diachronic development is sketched by Wanner (1996: 555f.): "The linear contact between unstressed object pronouns [...] and V creates a large zone of ambiguity between P2-F [second position— MJ] and VB [verb base—MJ], which eventually leads to reanalysis and establishment of VB as a relevant hypothesis". Elsewhere, I have argued that the historical shift from enclisis to proclisis in clitic pronouns depending on a finite verb in Greek can be explained only if clitics in Ancient Greek were "inherently unaccented" (Janse 1995-96: 165), i.e. inherently "clitic", not exclusively enclitic. In this connection, Devine and Stephens (1994: 365ff.) provide intriguing evidence of post-caesural instances of enclitic personal pronouns in the iambic trimeter, particularly when "its governing word or phrase follows the enclitic, so that it coheres syntactically to its right" (1994: 366). They further specify that "rhythmic cohesion to the right requires more syntactic support than rhythmic cohesion to the left, the unmarked condition" (1994: 367). They conclude that "[t]he strong association of postcaesural enclitic pronouns with rightward syntactic cohesion indicates that they were both rhythmically prepositive and accentually proclitic" {ibid.). The proviso to be made has to do with the criteria for determining the position(s) of caesurae (Janse 1998d: 135ff.). Whether or not we accept the possibility of proclisis in Ancient Greek, it must be emphasized that the reanalysis of preverbal clitics and the establishment of a continous verb phrase must have proceeded along these lines in Medieval Greek (and Romance).

3.

Modern Greek and Romance In Standard Modern Greek and in such modern Romance languages as French and Spanish, the reanalysis of preverbal clitics and the establishment of a

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

247

continuous verb phrase has led to a generalization of the "special" syntax with finite verbs, as exemplified in (2b), (3a) and (4a). That the erstwhile enclitic pronouns have indeed been subject to reanalysis is proven by the fact that on account of their having become proclitic, they may now be placed in first position. With non-finite verbs, however, the "ordinary" syntax has been generalized, as in the following Greek examples where the clitic pronouns are enclitic on the imperative, (34a) being the original text, (34b) its translation in Standard Modern Greek: (34) a. phéreté -moi =autous b. férte =moú =ta bring 1SG

3PL

"bring me them" (Mt 14.18)

The Latin translation of the Vulgate (34c) has reversed the order of the clitic pronouns, as in the French (34d), but not the Spanish (34e) version: (34) c. d e.

offerte apporte bring traéd bring

=illōs =mihi =les=moi 3PL 1SG =me=los 1SG 3PL

What we have here is again an illustration of the "fixing of discourse strategies" in that the imperative is another "preferential" word placed in first position on account of its discourse pragmatic status (Janse 1993b: 94, 108f.; 1994a: 120f.). In Standard Modern Greek, the clitic pronouns are obligatorily postverbal with participles as well, which may have to do with Marshall's rule discussed apropos of (14a) and (23a). In the modern Greek dialects and Ro­ mance languages which have generalized the "ordinary" syntax, the clitic pro­ nouns are a fortiori enclitic on the imperative as in the Cappadocian (34f), Portu­ guese (34g) and Galician (34h) equivalents of (34a-c):19 (34) f. fér =me g. traga =m h. traéde =m bring 1SG

=ta =os =os 3PL

In the remainder of this section, I will be concerned with finite verb phrases only. As has already been exemplified in the introduction to the present paper, the clitic pronouns are obligatorily postverbal in minimal sentences in the Cappadocian examples are quoted after D[awkins] (1916).

248

MARK JANSE

easternmost modern Greek dialects and the westernmost modern Romance languages, but they may be preverbal if the resulting verb phrase itself is not initial, in accordance with the Tobler-Mussafia Law referred to above. I will start with a discussion of the situation in Cappadocian. As in Ancient and Medieval Greek, the Cappadocian clitic pronouns have "special" syntax in the presence of canonical first position words or words which have developed out of erstwhile first position words, such as the proclitic modal particle ηά which developed historically from the Ancient Greek sub­ ordinate" hína: (35)

m -to fajís | nά =to kimís | CP 3SG you-feed CP 3SG you-lodge "feed her, lodge her, wash her" (D404)

nά CP

=to 3SG

níps you-wash

Another example is the modal particle ás, which is a contracted form of either ease or áphes (Thumb 1910: 121), both imperatives and hence preferential words. It developed into a first position subordinator through reanalysis involving boundary loss, and subsequently into a proclitic modal particle, like nά: (36)

ás -to pjáso | ás -to kópso | ás CP 3SG I-will-take CP 3SG I-will-kill CP "I will catch it, I will kill it, I will eat it up" (D366)

=to 3SG

fóyo I-will-eat

Other canonical first position words include such subordinators as án (from Ancient Greek än < eán < ei án) (37), negatives like én (dén) (from Ancient Greek oudén) (38), relative (39), and interrogative pronouns (40): (37) án =to šikóis ke án =to sékis eká SUB 3SG you-lift-up CO SUB 3SG you-put there "if you lift it up and if you put it over there" (D354) (38) dén =to xálasa ke deré dén -to xaláno NEG 3SG I-have-crossed co now NEG 3SG I-cross "I haven't crossed them, and I won't cross them now" (D322) (39) avuca tó me píken ekučís ne thus REL lSG s/he-has-done ox-driver s/he-is "he who treated me thus is the ox-driver" (D422) (40)

tí

=to

INTER 3SG

pikes? you-have-done

"what have you done to him?" (D436)

It will be obvious that the Cappadocian system is a continuation of the medieval situation, where the ancient discourse strategies were fixed and subse-

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

249

quently reanalyzed as a syntactic rule involving specific word classes, viz. canonical first position words and clitic pronouns. On the other hand, the "ordinary" syntax of the Cappadocian clitic pronouns is not just limited to minimal sentences of the (3b) type, but has developed into the unmarked order, as in the following example: (41) piren -cin ke lúsen =čin ke éplinén =čin s/he-took 3SG co s/he-washed 3SG CO s/he-cleansed 3SG "he took her and washed her and cleansed her" (D324)

Interestingly, Cappadocian has retained the possibility of exploiting the ancient discourse strategy optionally with focused words or constituents, par­ ticularly in the case of what I have called "echo structures" (Janse 1994b: 439; 1998a: 262f.): (42) a.

b.

(43) a.

čís =s' épke INTER 2SG s/he-has-made "who made you black?" etá =m' épke DEM 1SG s/he-has-made "she made me black" (D334) čís -t' álakse?

mávro? black mávro? black

INTER 3PL s/he-has-changed

"who has changed them?" b. ογό =t' álaksa I 3PL I-have-changed "I changed them" (D356)

In the (b) examples of (42-43), the clitic pronouns have been attracted into preverbal position by "preferential" subject pronouns, which are obviously focused since they carry new, i.c. requested, information. The order in the (b) sentences "echoes" that of the (a) sentences, where the clitic pronouns have been attracted to interrogative pronouns, as in (40). The following example illustrates the same phenomenon, but here the clitic pronoun has been attracted to a focused constituent rather than a "preferential" word: (44) a.

pjós -se píken na- jelášis? INTER 2SG s/he-has-made CP you-laugh "who made you laugh?" b. éna maimún =me jélasen ART monkey 1SG s/he-has-made-laugh "a monkey made me laugh" (D452)

250

MARK JANSE

Jakobson (1935: 384) once remarked that "ordinary" syntax, if generalized, would inevitably lead to the "univerbation" and morphologization of the erstwhile clitic pronouns. As a matter of fact, there is evidence of progressive morphologization of the clitic pronouns in Cappadocian (Janse 1998c). In Cappadocian, as in Ancient, Medieval and Standard Modern Greek, the attachment of a (postverbal) clitic pronoun results in the addition of a so-called "enclitic" accent on the verb if the latter is a proparoxytone as in (41). There are, however, numerous cases where the attachment of a pronominal clitic has not resulted in the addition of an enclitic accent, even though the host is a proparoxytone. In many cases, contrasting pairs are available from one and the same text: (45) a. galidzepsén =ta b. galídepsen =ta s/he-mounted 3PL "he mounted it" (D496) (46) a. éfayán =ta b. éfayan =ta they-ate 3PL "they ate it" (D520)

In the case of multiple clitic pronouns, the enclitic accent is systematically omitted in Cappadocian. Compare, for instance, example (47a) with its Standard Modern Greek equivalent (47b): (47) a. íkse =me =to b. íkse =mú =to show 1SG

3SG

"show me it" (D422)

However, an entirely different accentual pattern is attested as well. In this particular pattern, no enclitic accent is added on the last syllable of proparoxytone verbs, but instead the accent is shifted to the penultimate syllable as if the clitic were a suffix. Consider, for instance, the following contrasting pairs: (48) a. sérepsán =ta b. serépsan =ta they-gathered

3PL

"they gathered them up" (D398) (49) a. éširén =to b. eširen =to s/he-dragged 3SG "he dragged it" (D402)

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

251

The accentual behaviour of the verbs in the (b) examples can only be accounted for if the clitic pronoun is taken as a suffix, i.e. as an extra syllable added to the verb. Compare, for instance, the shifted accent in the first person plural as opposed to the first person singular in the following example: (50) /éeka/ => /éeka+me/ => [e'ekame] I-gave I-gave+PL we-gave

What we have here is in fact evidence of extreme variation. Apart from "true" cliticization, exemplified in (45-46a) and (48-49a), we find both agglutination, exemplified in (45-46b) and (47a), and "true" affixation, exemplified in (48-49b). But what of the status of preverbal clitic pronouns? The accentual behaviour of the modal particles is of crucial importance here. As in Standard Modern Greek, the modal particles are preverbal and proclitic. However, as soon as a preverbal clitic pronoun is added, they are accented, as in the following examples: (51) a. na= péyo nά= =to féro CP I-go CP 3SG I-bring "I will go and bring her" (D310)

Since there is hardly any reason to question the phonetic accuracy of Dawkins' transcriptions, as Mackridge (1990: 206-207) correctly points out, the above examples could theoretically be represented phonetically as follows: (51) b. [na' peγo ' nato ' fero]

However, it is equally well possible that the accent on na is in fact a secon­ dary accent (Janse 1995-96: 165), in which case the examples would have to be represented as follows: (51) c.

[na' peγo , nato' fero]

In the Standard Modern Greek equivalents of (51), both the modal particle and the clitic pronoun would be considered proclitic on the verb, as in (51c), even though an interpretation along the lines of (51b) cannot be precluded (Janse 1995-96: 165). The available evidence is unfortunately inconclusive as far as Cappadocian is concerned. Turning now to the westernmost Romance dialects, we find that the situation is very much like the one observed in Cappadocian. The general rule for Portuguese is formulated as follows by Vázquez Cuesta and Mendes da Luz

252

MARK JANSE

(1980: 493): "opronomepessoal [...] vão sempre emposição enclítica. [...] Nas orações afirmativas o pronome segue normalmente o verbo". The same general rule holds for Galician (Ålvarez-Monteagudo-Regueira 1989: 183) and Asturian (Sánchez Vicente-Rubiera Tuya 1985: 72). The Galician exemple (4b) = (52b), a translation of the Latin original (5b) = (52a), has already been quoted as a case in point. They are repeated here together with their Portuguese (52c) and Asturian (52d) equivalents: (52) a. amas b. quéres c. amas d. quies you-love

=me? =me? =me? =me? 1SG

"do you love me?" (Jo 21.17)

Whenever the finite verb is preceded by a canonical first position word, Wackernagel's Law is triggered off. Consider, for instance, the following examples involving a subordinator:20 (53) a. ei agapâté me b. sī dīligitis =mē SUB you-love 1SG c. se =me amades d. se =me amas e. se =me queréis SUB 1SG you-love "if you love me" (Jo 14.15)

It should be noted that the Vulgate has calqued the order of the original Greek, where the postverbal position of the clitic pronoun is possibly related to the focusing of the verb. Preverbal placement is also found in the presence of a negative:21 (54) a. nē =mē torqueās b. non =me atormentes c. não =me atormentes d. nun =m' atormentes NEG 1SG you-torture "don't torture me" (Lk 8.28)

20

Cf. Vázquez Cuesta-Mendes da Luz (1980: 495), Sánchez Vicente-Rubiera Tuya (1985: 73), Álvarez-Monteagudo-Regueira (1989: 194ff.), Salvi (1990: 181). 21 Cf. Vázquez Cuesta-Mendes da Luz (1980: 494), Sánchez Vicente-Rubiera Tuya (1985: 72), Ålvarez-Monteagudo-Regueira (1989: 185-186), Salvi (1990: 180).

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

253

The last class of canonical first words to be exemplified here is that of the interrogatives, which also systematically attract clitic pronouns in these lan­ guages:22 (55) a. quis =mē cōnstituit b. quen =me nomeu c. quem -me nomeu d. quién -me punxo como xuez? INTER 1SG s/he-appointed as judge

idicem? xuíz? juiz?

"who appointed me a judge or an arbiter between you?" (Lk 12.14)

Interestingly, preferential words such as subject pronouns and other focused constituents also attract clitic pronouns in preverbal position, as in the following "echo structures":23 (56) a. nēmo =tē condemnāuit? b. ninguén =te condenou? c. ninguém =te condenou? no-one 2SG s/he-has-condemned d. nun =te condergó dengún? NEG 2SG s/he-has-condemned anyone "has no one condemned you?" (John 8.10) (57) a. nec ego b. nin yo =te condergó NEG 2SG I-condemn

=tē

condemnō

c. tampouco eu non =te condeno PRT I NEG 2SG I-condemn d. tambén eu =te não condeno PRT I 2SG NEG I-condemn "then neither d o I c o n d e m n y o u " (Jo 8.11)

Note that the order of the subject pronoun and the negative varies in the westernmost Romance languages: whereas the clitic pronoun is attracted to the subject pronoun in the Portuguese (57d) and Asturian (57b) translations, it is placed between the negative and the finite verb in the Galician version (57c).

22

Cf. Sanchez Vicente-Rubiera Tuya (1985: 78), Álvarez-Monteagudo-Regueira (1989: 185), Salvi (1990: 180). 23 Cf. Vázquez Cuesta-Mendes da Luz (1980: 494), Sánchez Vicente-Rubiera Tuya (1985: 76ff.), Salvi (1990: 180).

254

4.

MARK JANSE

Conclusion

The picture emerging from the preceding sections is one of a competition between opposing principles. The most important of these is the gradual evolution towards a verb-centred system with the clitic pronouns floating around the verb, i.e. either immediately pre- or immediately postverbal. As a corollary, the clitic pronouns either encliticize to the preceding, or procliticize to the following verb. The latter cannot be easily proven in every single case as long as the clitic pronouns are not allowed to come first in minimal sentences such as (1) to (5). Another important principle is the tendency to grammaticalize the position of the clitic pronouns. In Ancient Greek and Latin, the placement of clitic pronouns was essentially a discourse pragmatic phenomenon, their position being dependent on canonical first position, preferential and focused words or constituents in general. A number of the resulting combinations must have gained the status of collocations in the course of time. The most important of these were collocations of clitic pronouns with subordinators, negatives, and relative and interrogative pronouns. Grammaticalization is also at work in the tendency to fix the position of the clitic pronouns vis-å-vis the verb. In the easternmost Greek dialects and the westernmost Romance languages postverbal placement has developed into the unmarked position, possibly under the influence of imperative-cum-clitic and infinitive/participle-cum-clitic sequences. The ancient collocations, however, still had a part to play in that preverbal placement was possible and even obligatory in a number of syntactic contexts. Preverbal placement has survived as an optional discourse strategy to the present day. In Standard Modern Greek and most of the other Romance languages a divergent evolution has taken place in that Wackernagel's Law has been generalized. In these languages, preverbal position is the obligatory order with finite verbs to the point where the preverbal clitics are also phonologically proclitic on the following verb. Only the ancient collocations of (postverbal) clitic pronouns with imperatives and infinitives/participles has survived. The evolution of the Greek and Latin clitic pronouns thus provides us a very intricate picture of variation, fixing of discourse strategies, grammaticalization and reanalysis. The evolution of the Greek and Latin clitic pronouns is an exemplary illustration of the fact that linguistic change is not necessarily teleological. The erstwhile discourse strategies have been fixed to a large degree, but at the same time there are convergent and divergent developments, there are obligatory syntactic rules and optional discourse strategies. I would like to conclude with the following words from Traugott (1994: 1481): "What distinguishes work on

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

255

grammaticalization from much other work on morphosyntax is the perspective taken: focus on the interdependence of langue and parole, of the categorial and the less categorial, of the fixed and the less fixed in language, of the motivated and the arbitrary. The study of grammaticalization therefore highlights the tension between relatively unconstrained lexical expressions and more constrained morphosyntactic coding, and points to relative indeterminacy in language and to the basic non-discreteness of categories. In this sense, much work on grammaticalization challenges some of the fundamental constructs of linguistics since Saussure, including categoriality, homogeneity, arbitrariness, and (more recently) the discreteness and autonomy of syntax".

REFERENCES Adams, J. N. 1994a. Wackernagel's Law and the placement of the copula esse in Classical Latin. Cambridge: Cambridge Philological Society. . 1994b. "Wackernagel's Law and the position of unstressed personal pronouns in Classical Latin". Transactions of the Philological Society 92.103178. , Mark Janse & Simon Swain eds. Fortcoming. Bilingualism in ancient society. Language contact and the written text. Oxford: Oxford University Press. Álvarez, Rosario, Henrique Monteagudo & Xosé Luís Regueira. 1980. Gramática galega. Vigo: Galaxia. Anderson, Stephen R. 1993. "Wackernagel's revenge". Language 69.68-98. Asher, R. E. ed. 1994. The encyclopedia of language and linguistics. Oxford: Pergamon Press. Bahner, Werner, Joachim Schildt & Dieter Viehweger eds. 1990. Proceedings of the 14th international congress of linguists. Berlin: Akademie-Verlag. Brixhe, Claude ed. 1993. La Koiné grecque antique. I. Nancy: Presses Universitaires de Nancy. . 1998. La Koiné grecque antique. III. Paris: de Boccard. Chantraine, Pierre. 1958. Grammaire homérique. I. 3rd ed. Paris: Klincksieck. Collinge, N. E. 1985. The laws of Indo-European. Amsterdam: Benjamins. Comrie, Bernard. 1989. Language universals and linguistic typology. 2nd ed. Oxford: Blackwell. Crochetière, André, Jean-Claude Boulanger & Conrad Ouellon eds. 1993. Actes du 15e Congrès international des linguistes. Sainte-Foy: Presses de l'Université Laval. Dawkins, R. M. 1916. Modern Greek in Asia Minor. Cambridge: Cambridge University Press. Devine, A. M. & Laurence D. Stephens. 1994. The prosody of Greek speech. Oxford: Oxford University Press.

256

MARK JANSE

Dover, Kenneth J. 1960. Greek word order. Cambridge: Cambridge University Press. Drachman, Gaberell, Angeliki Malikouti-Drachman, Sila Klidi & Jannis Fykias eds. 1997. Greek linguistics '95. Proceedings of the 2nd international conference on Greek linguistics. Graz: Neugebauer Dunn, Graham. 1989. "Enclitic pronoun movement and the Ancient Greek sentence accent". Glotta 67.1-19. Eichner, Heiner. 1988-90. "Reklameiamben aus Roms Königszeit (Erster Teil)". Die Sprache 34.207-238. & Helmut Rix eds. 1990. Sprachwissenschaft und Philologie. Jacob Wackernagel und die Indo germanistik heute. Wiesbaden: Reichert. Halpern, Aaron L. & Arnold M. Zwicky eds. 1996. Approaching second. Second position clitics and related phenomena. Stanford: CSLI Publications Hock, Hans Henrich. 1996. "Who's on first? Toward a prosodic account of P2 clitics". Halpern-Zwicky 1996.199-270. Hopper, Paul J. & Elizabeth Closs Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press. Jakobson, Roman. 1935. "Les enclitiques slaves". Migliorini-Pisani 1935.384390. Janse, Mark. 1990a. "A note on the Fibula Praenestina". Indogermanische Forschungen 95.101-103. . 1990b. "Wackernagel's Law". Bahner et al. 1990.2645-2649. . 1993a. "The prosodic basis of Wackernagel's Law". Crochetière et al. 1993.4.19-22. . 1993b. "La position des pronoms personnels enclitiques en grec néotestamentaire à la lumière des dialectes néo-helléniques". Brixhe 1993.83-121. . 1994a. "De la Loi de Wackernagel et ses extensions en latin. A propos de la collocation pronominale chez Pétrone". TEMA 1.109-146. . 1994b. "Son of Wackernagel. The distribution of object clitic pronouns in Cappadocian". Philippaki-Warburton et al. 1994.435-442. . 1995-96. "Phonological aspects of clisis in Ancient and Modern Greek". Glotta 73, 155-167. . 1997a. Review of Adams 1994a. Kratylos 42.105-115. . 1997b. "Synenclisis, metenclisis, and dienclisis in Cappadocian". Drachman et al. 1997.695-706. . 1998a. "Cappadocian clitics and the syntax-morphology interface". Joseph et al. 1998.257-281. ed. 1998b. Productivity and creativity. Studies in general and descriptive linguistics in honor of EM. Uhlenbeck. Berlin: Mouton de Gruyter. . 1998c. "Grammaticalization and typological change. The clitic cline in Inner Asia Minor Greek". Janse 1998b.521-547. . 1998d. "Homerische metriek. Orale poëzie in de praktijk". Didactica Classic a G andens ia 38.125-151. . 1998e. "La koiné au contact des langues sémitiques, de la Septante au Nouveau Testament". Brixhe 1998.99-111.

THE DEVELOPMENT OF GREEK AND LATIN CLITIC PRONOUNS

257

. Fortcoming. "Language contact, bilingualism and interference in the history of the Greek language". Adams et al. fortcoming. Joseph, Brian D. 1990. Morphology and universals in syntactic change. New York: Garland Publishing. , Geoffrey Horrocks & Irene Philippaki-Warburton eds. 1998. Themes in Greek linguistic. II. Amsterdam: Benjamins Klavans, Judith L. 1985. "The independence of syntax and phonology in cliticization". Language 61.95-120. Krisch, Thomas. 1990. "Das Wackernagelsche Gesetz aus heutiger Sicht". Eichner-Rix 1990.64-81. Leumann, Manu. 1977. Lateinische Laut- und Formenlehre. 2nd ed. München: Beck. Lunt, Horace G. ed. 1964. Proceedings of the 9th international congress of linguists. The Hague: Mouton Mackridge, Peter H. 1990. " 'Some pamphlets on dead Greek dialects'. R. M. Dawkins and Modern Greek dialectology". Annual of the British School at Athens 85.201-212. . 1993. "An editorial problem in Medieval Greek texts. The position of the object clitic pronoun in the Escorial Digenes Akrites". Panayotakis 1993. 325-342. Marshall, M. H. B. 1987. Verbs, nouns, and postpositives in Attic prose. Edinburgh: Scottish Academic Press. Meillet, Antoine & Joseph Vendryes. 1948. Traité de grammaire comparée des longues classiques. 2nd ed. Paris: Champion. Migliorini, Bruno & Vittore Pisani eds. Atti del III congresso internazionale dei linguisti. Firenze: Le Monnier. Moorhouse, A. C. 1959. Studies in the Greek negatives. Cardiff: University of Wales Press. Nevis, Joel Α., Brian D. Joseph, Dieter Wanner & Arnold M. Zwicky eds. 1994. Clitics. A comprehensive bibliography 1892-1991. Amsterdam: Benjamins. Panayotakis, N. ed. 1993. Origini della letteratura neogreca. Atti del Secondo Congresso Internazionale "Neograeca Medii Aevi". I. Venezia: Istituto Ellenico di Studi Bizantini e Postbizantini di Venezia. Philippaki-Warburton, Irene, Katerina Nicolaidis & Maria Sifianou eds. 1994. Themes in Greek linguistics. I. Amsterdam: Benjamins Radermacher, Ludwig. 1925. Neutestamentliche Grammatik. 2nd ed. Tübingen: Mohr. Rollo, Antonio. 1989. "L'uso dell'enclisi nel greco volgare dal XII al XVII secolo e la legge Tobler-Mussafia". 'Ίταλοελληνικά 2.135-146. Ruijgh, C. J. 1990. "La place des enclitiques dans l'ordre des mots chez Homère d' après la loi de Wackernagel". Eichner-Rix 1990.213-233. Salvi, Giampaolo. 1991. "Difesa e illustrazione della legge di Wackernagel applicata alle lingue romanze antiche. La posizione delle forme pronominale clitiche". Vanelli-Zamboni 1991.439-462. Padova: Unipress.

258

MARK JANSE

. 1996. From Latin weak pronouns to Romance clitics. Budapest: Institution Linguisticum Academiae Scientiarum Hungaricae. Sánchez Vicente, Xuan Xosé & Carlos Rubiera Tuya. 1985. "Enclisis, proclisis y pronomes átonos n'asturianu". Lletres Asturianes 18.71-81. Schwyzer, Eduard. 1950. Griechische Grammatik. Vol. 2. München: Beck. Spencer, Andrew. 1991. Morphological theory. Oxford: Blackwell. Thumb, Albert. 1910. Handbuch der neugriechischen Volkssprache. 2nd ed. Strassburg: Trübner. Traugott, Elizabeth Closs. 1994. "Grammaticalization and lexicalization". Asher 1994.1481-1486. Uhlig, Gustav ed. 1910. Apollonii Dyscoli quae supersunt. II. Leipzig: Teubner. Vanelli, Laura & Alberto Zamboni eds. 1991. Per Giovan Battista Pellegrini. Scritti degli allievi padovani. Padova: Unipress. Vázquez Cuesta, Pilar & Maria Albertina Mendes da Luz. 1980. Gramática da lingua portuguesa. Lisboa: Edições 70. Wackernagel, Jacob. 1892. "Über ein Gesetz der indogermanischen Wortstellung". Indogermanische Forschungen 1.333-446. Wanner, Dieter. 1996. "Second position clitics in Medieval Romance". In: Halpern-Zwicky 1996.537-578. Watkins, Calvert. 1964. "Preliminaries to the reconstruction of Indo-European sentence structure". Lunt 1964.1035-1045. Weil, Henri. 1879. De l'ordre des mots dans les langues anciennes comparées aux langues modernes. 3rd ed. Paris: Vieweg. Wifstrand, Albert. 1949. "A problem concerning the word order in the New Testament". Studia Theologica 3.172-184. Zwicky, Arnold M. 1977. On clitics. Bloomington: Indiana University Linguistics Club. . 1994. "What is a clitic?" Nevis et al. 1994.xii-xx.

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH: STABILITY, VARIATION AND CHANGE WILLEM KOOPMAN & WIM VAN DER WURFF University of Amsterdam University of Leiden 0. Introduction In this paper, we examine the history of two word order patterns in English, and try to determine to what extent they show gradualness or abruptness in their development. First, we consider the ordering of direct object (DO) and indirect object (IO), and secondly we look at the order of the direct object vis-à-vis the verb (V). In both cases, the historical development has gone from more or less variable ordering in the Old English period to a fairly strict order in Modern English. Thus, in Old English texts one can find sentences in which the indirect object precedes the direct object, as in (la), but also sentences like (lb), in which the order is the reverse. Similarly, the direct object precedes the verb in (2a) but follows it in (2b). (1) a.

b.

(2) a.

b.

gif se sacerd ne mæg ðam læwedum mannum (ΙΟ) if the priest not can the lay people larspel (DO) secgan homily say "if the priest cannot say a homily to the lay people" (ÆCHom II, 41.306.66) Öonne when ða that "when set"

he nyle ða bisne (DO) oðrum (ΙΟ) eowian he not-wants the example others show he mid ryhte eowian sceal he properly show must he does not want to set the example to others that he properly ought to (CP 59.449.29)

pa he pæt folc (DO) costigan (V) when he the people tempt "when he allowed the people to be tempted"

let let

pæt heo scolde geberan (V) Godes sunu (DO) (WHom 7.38) "that she must bear God's son" (Or 6.3.136.8)

260

WILLEM KOOPMAN & W I M V A N DER WURFF

Modern English in both cases allows only one of the orders (IO-DO and V - D O , or V O to use the standard label). In themselves, these facts might suggest that the development of the word orders was of a gradual nature, since it is tempting to interpret the Old English situation as a halfway station between a prehistoric state of the language when the order was categorically X Y (i.e. DO-IO and OV) and a later stage when there was categorical Y X (i.e. IO-DO and V O ) . However, our ideas about this prehistoric stage, and certainly about the categoricality or variability of its word orders, can only be based on conjecture. And even if Old English was a halfway station, this does not help us to decide whether the subsequent development, i.e. the loss of some of the ordering options, was gradual or not. Of course, the notions of graduality and abruptness are relative, and there can be few linguists who would deny that, one way or another, nearly every change is gradual. In discussions of specific changes in word order, one usually finds statements to the effect that the older order is still attested for a while after the new order has become dominant, but one seldom finds any serious attempt at analysis or even description of these late attestations of the older pattern. Instead, researchers typically focus on the nature and causes of the change in dominance. This is also true for the two changes in English mentioned above: they have received more than their fair share of attention in the literature, but it is only in recent work that the full extent of the graduality of both developments has become clear. In what follows, we describe the empirical evidence for the two changes, based on recent work that we and others have carried out, and we present some reflections on their implications for one specific influential approach to syntactic change. The basic data w i l l be presented in § 1 (for the ordering of direct and indirect objects) and § 2 (for direct object and verb). In § 3, we use the data to reflect on the viability of the approach to word order change which employs the notion of grammar competition (Kroch 1994). This type of competition at first sight seems well-suited to serve as a model for gradual change, but we shall voice some doubts about the explanatory value of the overall approach. Furthermore, we w i l l argue that the precise nature and locus of the competing grammars needs to be specified more carefully and that facts of specific changes, such as those described in §§ 1-2, may raise problems for the timescales involved.

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

261

1.

The order of direct and indirect object in English: a diachronic survey English has undergone massive changes in word order when one compares Old English with Middle English and later stages of the language, but close inspection reveals that many of these are the accumulated results of rather slow processes. One such change involves constructions with verbs taking two objects (traditionally called ditransitive verbs); as the following survey w i l l make clear, several of these constructions stay stable and the constructions that do change do so in a gradual rather than abrupt way. In the present-day language the following orders are possible for the objects of ditransitive verbs: (3) a.

NOUN-NOUN:

John gave Mary (IO) a book (DO) b.

PRONOUN-NOUN:

John gave her (IO) a book (DO) C.

PRONOUN-PRONOUN:

John gave it (DO) her (I0) d.

' T O ' PHRASE:

John gave a book/it (DO) to Mary/her

The patterns illustrated in (3) can all be found in Old English too: (4) a.

b.

C.

d.

NOUN-NOUN:

and noldon Iuliuse (IO) nænne weorpscipe (DO) don and not-wanted Iulius no worship do "and did not want to worship Iulius" (Or 5.10.124.9) PRONOUN-NOUN:

He "He

gehet him (IO) pa Godes yrre (DO) promised him then God's anger" (ÆL,S(Martin)803)

PRONOUN-PRONOUN:

ðæt hie hit (DO) him (IO) da io ondredon (CP 32.213.22) that they it themselves then long feared "that they had feared it then for themselves a long time" TO' PHRASE:

And ic "And I

hæbbe asend have sent

pis gewrit (DO) to eow this letter to you" (HomU 36(Nap45)116)

In addition, the direct object (whether nominal or pronominal) could precede the indirect object in Old English:

262

WILLEM KOOPMAN & W I M V A N DER WURFF

(5) a.

b.

NOUN-NOUN:

ðonne he nyle ða bisne (DO) oðrum (IO) eowian da he mid ryhte when he not-wants the example others show that he properly eowian sceal show must "when he does not want to set the example to others that he properly ought to set" (CP 59449.29) PRONOUN-NOUN:

hum he hine (DO) Gode (IO) bebeode (HomS14(BlHom4)145) indeed he him God commend "indeed, he should commend himself to God" (HomS14(BlHom4)145)

The 'to'-phrase is not yet common in Old English. It is found regularly only after verbs of saying (see Cassidy 1938; Ogura 1980:60), where the direct object often takes the form of a reported clause ("He said to John he wanted to go home"). Pronouns show less positional freedom in Old English than nouns and they generally occur before nominal objects (Mitchell 1985; Kohonen 1978; Koopman 1997). Most of these pronouns are in the dative, presumably because the dative indirect object usually refers to a person and is more likely to be 'given' information (Kuno 1987). We can summarize the frequency of constructions with two objects in Old English as follows (for detailed figures, see Koopman 1990:133-223): (6) a. b. c.

d.

NOUN-NOUN: objects found in either order (with about equal frequency) PRONOUN-NOUN: pronominal object before nominal object (more than 90%) PRONOUN-PRONOUN: DO before IO

'TO' PHRASE: very few examples

(95%)

Some of these patterns remain in the Middle English period. Pronoun objects still precede nominal objects, though the occasional example of the other order can be found (as in Old English), especially in early Middle English texts: (7)

(8)

leafdi do me are Lady do me mercy "Lady, give me mercy"

(Ancr. 10b.27)

pe walde pis bead him makien who wanted this bath him make "Who wanted to make this bath for him"

(Ancr. 107a.4)

DO-IO remains the favoured pattern when both objects are pronouns: (9)

for ha ne seide hit (DO) me (IO) naht she not said it me not for "for she did not say it to me" (Sward 75)

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

263

However, certain new developments also take place: the 'to'-phrase becomes a great deal more common and can now be used with any verb, while DO-IO patterns (other than with two pronouns) decrease in frequency, and are rare at the end of the period. We can summarize the frequency of constructions with two objects in Middle English as follows: (10) a. NOUN-NOUN: initially both orders are still found, later only IO-DO b. PRONOUN-NOUN: almost always in this order (IO-DO) c. PRONOUN-PRONOUN: predominantly DO-IO d.

'TO' PHRASE: common

Nominal objects are still found in both orders in early Middle English texts: (11)

ne schaw pu nan non (IO) pi wlite (DO) not show you no man your face "do not show your face to anyone" (Ancr. 26a.8)

(12)

deð hearm (DO) moni ancre (IO) does harm many anchoress "does harm to many an anchoress" (Ancr. 31b.21)

In the course of the Middle English period, DO-IO patterns with two nouns became gradually less and less frequent and by the end of the X l V t h century, they are no longer found (Allen 1995:419 says: " I have found no convincing examples with two nominals in texts written after 1340"). The DO-IO pattern survived for a while when DO was a pronoun and IO a noun, but by the end of the X V t h century, this option had also become obsolete (Allen 1995:420ff). The development is illustrated in Table 1 with data from some Old and Middle English texts, partly based on Koopman (1990:187) and Allen (1995:418).

noun-noun ÆCHom (ca. 1000) Mor.Hom. (late XlIthC) Ancr. (ca. 1230) Rob.Glo (early XlVthC) MKempe (1438) (pp. 1-60)

IO-DO 115(47%) 13 (72%) 19 (63%) 33 (79%) 13 (100%) Table 1

DO-IO 128 (53%) 5 (28%) 11 (37%) 9(21%)

Tot. 243 18 30 42 13

264

WILLEM KOOPMAN & W I M V A N DER WURFF

It is interesting to note that, in absolute terms, Ælfric has considerably more instances of this construction than the Middle English texts. This is partly due to lexical loss. Many Old English verbs taking two objects are lost in Middle English and this loss is only partially offset by borrowings from French such as grant. As Table 1 shows, the loss of DO-IO patterns spanned most of the Middle English period, though we have not much data for the early Middle English period. The loss could be connected with the general decline and loss of case endings in the late Old English/early Middle English period, combined with the increase in the use of prepositions where in Old English nominal inflections were used. During the Old English period there was comparatively rich nominal inflexion (for details, see Campbell 1959; Hogg 1992) and the freedom of objects to occur in either order in Old English has often been linked to this. In support of this view we can point to the fact that objects of double object verbs often occur in either order (IO-DO and DO-IO) in German, a language which also has a relatively full set of nominal inflexions. Modern Dutch, which is like English in having lost practically all nominal inflexions, does not exhibit this freedom: (13)

Jan wil Piet een boek geven *Jan wil een boek Piet geven "John wants to give Peter a book"

However, as Allen (1995) has argued convincingly, the loss of case endings by itself cannot explain why the order DO-IO became obsolete, since DO-IO patterns survived long after almost all nominal inflexions had disappeared. Although it cannot be doubted that the loss of morphology contributed to the fixing of word order, there must be some other factor that played a decisive role in the demise of DO-IO patterns. Although it is not the case that the use of prepositions showed any overall increase towards the end of the Old English period (Mitchell 1985: § 1225), the material collected by Cassidy (1938) does show that there was a slow spread in the use of the 'to'-phrase. Initially it was restricted to only certain verbs, but in late Old English texts, incidental examples are found with other verbs as well. By the time we have substantial Middle English texts, the 'to'-phrase is common with a great number of verbs. It can thus be seen as competing with patterns which depended on inflexional marking. As the syntax becomes increasingly V O (see next section), the preferred position for PPs is after the object. In this way the 'to'-phrase comes to compete with DO-IO patterns.

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

265

Data on the frequency of the various options are given in Table 2, which repeats the figures for some of the Middle English texts in Table 1, but adds figures for the 'to'-phrase (i.e. the DO (noun)-to+noun pattern): noun-noun Mor.Hom.(late 12thC) Ancr. (ca. 1230) MKempe (1438)

IO-DO 13 (65%) 19 (49%) 13 (59%)

DO-IO 5 (25%) 11(28%) Table 2

to 2 (10%) 9(23%) 9(41%)

Tot. 20 39 22

Although the figures are rather low, and it is therefore dangerous to draw firm conclusions from them, it seems that IO-DO occurs at roughly the same frequency as DO-IO+ 'to'-phrase. Figures extracted from the Helsinki Corpus (Kytö 1991) show the same picture: noun-noun (1150-1250) (1250-1350)

IO-DO DO-IO 19 (52%) 9 (24%) 11 (44%) Table 3: Helsinki Corpus

to 9 (24%) 14 (56%)

Tot. 37 25

The evidence we have presented here shows that the demise of DO-IO patterns took several centuries to come to completion and that a new pattern (the 'to'-phrase) played a crucial role in this. It would therefore be quite wrong to think that there was an abrupt change from DO-IO to IO-DO order at any point. Moreover, the DO-IO pattern remains stable in cases where both objects are pronouns. The alternative construction with the 'to'-phrase arises as an option here too, but it does not oust DO-IO. The order DO-IO remained fully productive, as can be seen from its use in Middle English with originally French verbs, such as grant in (14): (14)

Thou shalt have it seyde the kynge I graunte hit the {Morte Arthur, Vol i:297) "you will have it, said the king, I grant it you"

When both objects are pronouns, DO-IO is still the dominant present-day order in southern British English. Some northern varieties have the order IODO here, as do American and Australian English (though the 'to'-phrase is probably more common). This IO-DO order may have developed as a new variant at some point (Kirk 1985), though there is some evidence that it is a continuation of a (rare) option in Old English. A few examples with two pronouns in IO-DO order have been found in Old English (Koopman 1990:173-175), but they are not all equally convincing. The surviving Old

266

WILLEM KOOPMAN & W I M V A N DER WURFF

English texts clearly show DO-IO as the preferred order and this continues during the Middle English period. A search through the Helsinki Corpus and the University of Virginia electronic Middle English Texts revealed that the order IO-DO was used in about 5% of all cases with two pronouns, which is only marginally higher than what was found for Old English. A Middle English example is: (15)

pæt tu 3etti me (IO) ham (DO) that you grant me them "that you grant them to me" (Ancr. 7a. 10)

Throughout the history of English, IO-DO with two pronouns has always been infrequent in texts. In Shakespeare, for instance, there are many examples of DO-IO, as in (16): (16)

who gave it thee? (Two Gentlemen of Verona 1.2.37)

On the other hand there is only one possible instance of IO-DO, given in (17). However, this example is somewhat doubtful since them all can be interpreted as forming one phrase, which would make (17) similar to the usual pattern with a pronominal indirect object and a nominal direct object: (17)

Ile

tell you them all by their names (Troilus and Cressida 1.2.182)

A quick search through the large collection of electronic texts at the University of Virginia for later periods likewise resulted in very few tokens of the order IO-DO when both objects are pronouns. 2.

The order of direct object and verb in English: a diachronic survey It has been said that Old English is closer to modern Dutch and German than to modern English, as far as the order of the direct object and the verb is concerned. In his careful and judicious summary of work done on word order change in English, Denison (1993) describes the situation as follows: (18) Ignoring a large body of exceptions and some evidence of non-homogeneity within Old English, we might claim that Old English was a mixed V-2/V-F language like Dutch or German, with V-2 predominant in main clause declaratives and V-F predominant in subordinate clauses. Modern English, on the other hand, is consistently V-3 or SVO. How and why did word order change? This I take to be the main problem to be investigated. (Denison 1993:29-30)

Here, the label V - F means verb-final, i.e. O V in clauses with a direct object; V-2 refers to the property of many main clauses in Old English of having the finite verb in second position. Since this property is in principle

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

267

independent of that of the object and the verb (see, for example, Koopman 1985), we shall disregard it here, using Old English examples with non-finite lexical verbs to illustrate relevant points. While Denison quite rightly begins his review by asking how and why things changed, most of the actual work that he reviews pays little attention to the question how the change took place, and instead focuses almost exclusively on why it happened. The reason for this is no doubt that the path followed by the change looks so extremely easy to identify. Thus, consider the Old English sentence in (19). (19)

pæt he hwelc gerisenlic wundor wyrcean that he some suitable miracle perform "that he is able to perform some suitable miracle"

mæge can (CP 17.119.10)

Here, the direct object, hwelc gerisenlic wundor, precedes the lexical verb, wyrcean, an order that appears to be impossible in present-day English. Speas (1997:172), for example, has the following to say about V O and OV orders in present-day English: (20) a. Pat ate an apple b. *Pat an apple ate English speakers recognize that (a) is a fully well-formed sentence of English, but that (b) is not. (Speas 1997:172)

What has happened seems crystal-clear: English has undergone a change from OV (in Old English) to V O (in Modern English). On the basis of rather limited empirical work, it was decided in the early 1980s that the modern situation had basically been reached by ca. 1200, with just a few instances of the older order surviving for some centuries. In subsequent work, the only remaining question was usually taken to be why there had been this word order change, and Denison (1993:27-58) makes clear that a very wide variety of answers to this question has been proposed. However, as we shall show in what follows, a closer look at the empirical data reveals that there is still much more to be said about how and when the change took place. A good starting point for exploring this change may be a footnote in Speas (1997). In it, she adds the following qualification to her statement given in (20): (21)

This is not to say that [b] cannot be used under certain circumstances, such as in a poem. (Speas 1997:172)

268

WILLEM KOOPMAN & W I M V A N DER WURFF

What Speas is telling us here is that the order OV still exists, which must come as a surprise to anyone believing that this order was lost in the X l l t h century. Of course, the status in present-day English of the order in (20b) is rather different from that of (20a), but ignoring data like this unavoidably results in an impoverished picture of the change. Specifically, it can create the erroneous impression that OV order was lost abruptly. A closer look at the Old English data also makes clear that the idea of an abrupt change from OV to V O around 1200 is a distortion of the facts, since VO order is found even in the earliest texts. In (22), we give an example from the poem Beowulf, which is usually regarded as representing an early stage of Old English. (22)

Scildinga Pa of wealle geseah I weard then from cliff saw guard of-Scyldings se pe holm-clifu I healdan scolde, who that sea-cliff guard should beran ofer bolca I beorhte randas, carry over gangway bright shields fyrd-searu fuslicu armour ready "Then, from the promontory, the watchman of the Scyldings, who had the duty of guarding the sea-cliff, saw [them] carry bright shields and ready armour down the gangway." (Beo 229-232)

Here the object beorhte randas, fyrd-searu fuslicu follows the verb beran, resulting in V O order. This is not to say that, at this stage of the language, the use of V O was completely unrestricted. On the basis of metrical evidence, Pintzuk & Kroch (1989) argue that the normal word order in Beowulf is OV and that all clauses with V O feature an object which is in some way heavy and which has undergone rightward movement from preverbal position. Although there may be problems in basing our ideas about early Old English word order on one single poem whose date of composition has been fiercely debated (see the contributions in Chase 1981), the limitations of the textual evidence make it difficult to do much better. We shall therefore accept that early Old English had basic OV order, with V O being a permissible variant for heavy objects, which would be marked by an intonational break or a pause. In present-day English, as we just saw, the basic order is V O , but OV is permissible under certain restricted circumstances. The question that naturally arises is: what happened in between? In the major Old English prose works, which were written in the late IXth century and the Xth century, OV order is still predominant, but V O order occurs with heavy objects, not-so-heavy objects, and even (though very rarely)

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

269

pronouns (see Koopman 1997; Pintzuk 1991, 1996). In (23), we give an example of a postverbal object that appears to be quite short and not especially informative, and that cannot plausibly be preceded by an intonational break. In (24), there is an example with a postverbal pronoun that does not appear to have special prominence either. (23)

pæt he nolde niman mancynn neadunga of ðam deofle that he not-would take mankind forcibly from the devil "that he would not have taken mankind forcibly from the devil" (ÆCHom 1,14.1.216.5)

(24)

He nolde genyman us neadunge of deofles anwealde he not-wanted take us forcibly of devil's power "He would not forcibly take us from the devil's power" (ÆCHom 1,1.26.29)

On the basis of these facts, Pintzuk (1991) suggests that there was a qualitative change in word order some time in the VIIIth or IXth century, which made V O word order also available for rather 'light' objects. However, it is true that examples as in (24) remained rare, and OV order was by no means lost: it continued as a very productive pattern throughout the Old English period, as in (25) and (26). (25)

pæt hi pone heofonlican æðeling manfullice acwellan woldon that they the heavenly prince evilly kill wanted "that they evilly wanted to kill the heavenly prince" (ÆCHom 1,28.402.6)

(26)

swa hi nu heora swuran to his geðylde nellað gebigan as they now their necks to his patience not-want bow "as they now will not bow their necks to his patience" (ÆCHom 1,40.610.32)

Data presented in Kroch & Taylor (1994, 1998) show that in the XIIIth century, OV order still occurs in 30% or more of all cases, with south-eastern texts in particular still having very high percentages of this order (especially with short and most particularly with pronominal objects, which also in other dialects still show frequent OV order). Some examples are given in (27-29). (27)

(28)

pæt ic nule pe that I will-not you "that I will not forsake you"

forsaken forsake (St. Juliana (Bod) 278)

ðat we moten mid 3eure helpe and mid his hale grace that we may with your help and with his holy grace swa pis scorte lif her laden thus this short life here lead "that we may lead thus this short life here with your help and his holy grace." (Vices&V 21.23)

WILLEM KOOPMAN & W I M V A N DER WURFF

270

(29)

pat he deap scolde that he death should "that he should suffer death."

poli3en suffer (Vices&V 113.21)

Note that in both (28) and (29), the object is separated from the lexical verb: there is an intervening adverbial, her, in (28) and an auxiliary, scolde, in (29). A comparison with examples (25) and (26) shows that these patterns form a continuation of Old English word order options. In the course of the XIVth century, the use of OV order undergoes a decline from ca. 20% to 5% of all cases in prose texts (for full data and references to earlier work, see Foster & van der Wurff 1995). In spite of this quantitative decrease, there is no apparent qualitative change, since OV continues to be found with both pronouns and full NPs (though usually short ones, predominantly constituting old information; see Foster & van der Wurff 1997), and with or without elements intervening between the verb and the preverbal object. Examples of the various patterns are given in (30)-(34). (30)

bote yef he hit yelde per ha ssel but if he it yield where he must "unless he gives it where he must' {Ayenb. 10.1)

(31)

me ssel hine loky and urepie zo holyliche one must it observe and honour so religiously "one must observe and honour it so religiously" (Ayenb .122)

(32)

Ne he ne may habbe skele: pet he him mo3e excusi nor he not may have reason that he him may excuse "Nor may he have a reason to excuse himself' {Ayenb. 7.1)

(33)

alle that him feith berith and obeieth "all that have faith in him and obey him"

(34)

I may my persone and myn hous so kepen and deffendea " I can keep and defend myself and my house in such a way." Melibee 1334)

(Chaucer Astrolabe 58) (Chaucer

In verse, the use of OV order remains somewhat more common than in prose. In XlVth-century poetry, it is still found in ca. 20% of all possible cases, and XVth century poetry has it in ca. 10%. Qualitatively too, OV is holding its own in verse from these two centuries: the preverbal object can be a pronoun or a full NP and there may be an adverbial or auxiliary intervening between the object and its lexical verb (for a structural analysis of late Middle English OV order making crucial use of these distributional facts, see van der Wurff 1997a). There are some X l V t h and XVth century examples such as (35), in which the intervening element is actually another lexical verb.

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

(35)

271

Fortune his howve entended bet to glaze Fortune his hood intended better to glaze "Fortune intended to glaze his hood better (i.e. to fool him more completely)" (Chaucer Troilus and Criseyde 5.469)

Here the NP his howve functions as object of the verb glaze (in fact, the collocation to glaze his howve/hood was an idiomatic expression at the time), but the matrix verb entended intervenes. Examples of such a construction also exist in Dutch and German (where this pattern goes by the name of 'the third construction" or "remnant VP extraposition"; see Broekhuis et al 1995), and some instances have been found in Old English texts (see Los 2000:272). The continuation of this specific OV pattern reinforces the point that the shift to V O order was by no means complete even in late Middle English. In XVth-century prose, the frequency of OV order has become quite low, since in many texts it occurs in less than 1 % of all possible cases. This may look like a negligible amount and certainly in historical work, data found at such a low frequency are often silently ignored. A n explicit attempt at justification of this practice is made by Santorini (1992:609, n 17), who states that "[å]ccording to detailed quantitative work of my own and others, wellestablished generalizations in a language are violated in naturally-occurring usage at a low, relatively constant rate of about 1 % . " She goes on to formulate generalizations about the history of Yiddish which would exclude some patterns with postverbal particles and unstressed pronouns that are attested, but at a frequency below 1 % . Quite apart from other objections that can be raised against this equation of low textual frequency with marginality or even ungrammaticality (for some general discussion, see Koopman 1990:11-12), there is no doubt that infrequent patterns too may very well show significant regularity and systematicity, and these w i l l not be perceived if the relevant data are simply ignored. In the case of OV order in the X V t h century, it indeed turns out that the data, though infrequent, are easily robust enough to show that there was an important qualitative change in the use of OV order in prose texts at this time. In a study of a large number of texts, van der Wurff (1997b) found that, among the 270 OV tokens identified, there were three patterns that became increasingly prominent in the course of the X V t h century: OV in a clause with an auxiliary and a negated object, as in (36), OV in a clause with an auxiliary and a quantified object, as in (37), and OV in a clause with an empty subject, as in the coordinated clause in (38) and the relative clause in (39). (36)

pei schuld no meyhir haue "They were not allowed to have a mayor"

(Capgrave Cronicles 199.6)

W I L L E M K O O P M A N & W I M V A N DER WURFF

272 (37)

he hap on vs mercy, for he may al pynge do "he has mercy on us, for he can do everything"

(Barlam

2740)

(38)

summe he exiled, summe he put out here rite eyne, and namely pe worthi man, Athanasius, whech mad Quicuncque vult, pur sewed so feruently pat he durst not appere openly vii 3ere "Some he exiled, o f some he put out their right eye, and especially the honourable man Athanasius, who composed the Quicumque v u l t , he pursued so fiercely that he did not dare to show himself in public for seven years." (Capgrav Cronicles 62.24)

(39)

alle pat his writinge reden or heere "all that w i l l read or hear this w r i t i n g "

(Thorpe Sermon 2250)

In the Paston Letters, a large XVth-century collection of letters written by members of the Norfolk Paston family, these three patterns account for nearly 90% of all OV tokens, and the comparable (but smaller) collection of Cely Letters shows a similar picture (for details, see Moerenhout & van der Wurff, forthcoming). The more literary prose texts examined in van der Wurff (1997b) still feature a fair number of other OV patterns, but each of these is represented by just a handful of tokens, and it is possible that these were becoming more and more stylistically marked. On the basis of these data, it can be concluded that OV underwent not only a decline in frequency but also a narrowing of structural possibilities in XVth-century prose writings (on the reasons why this narrowing took the form it did, see van der Wurff 1999). In verse, the use of OV survives more tenaciously, both quantatively and qualitatively. By 1550, OV has virtually ceased to be used as a productive pattern in prose writings. The rare examples from the first half of the X V I t h century that we have found so far are similar to the ones found in XVth-century prose texts in having either a negative or quantified object following an auxiliary, as in (40) and (41), or a preverbal object in a subjectless clause, as in (42). 1 (40)

any great faute and intollerable sholde they none finde "they w o u l d f i n d no great and intolerable error" (More Apology

(41)

ƒ [ . . . ] wist not where I myghte any such fynde " I did not know where I might find any like that"

(42)

(More Apology

2.12) 44.13)

there was no man that any medling had wyth theym, into whose handes they were more lothe to come "there was no man that had any dealings w i t h them into whose hands they were more sorry to f a l l " (More Apology 54.12)

Note that in (40) the object is as it were split, w i t h part {any great faute and intolerable) being in clause-initial position and part (none) preceding the lexical verb and thus instantiating O V order; on the analysis o f this pattern, see Beukema & van der W u r f f (1999).

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

273

In poetry, however, OV remains available as a standard option until at least 1900. In (43)-(46), we give one example each century from the X V I t h to the XlXth. (43)

When that churl death my bones with dust shall cover (Shakespeare Sonnet 32.2)

(44)

But while your daies on publick thoughts are bent Past ills to heal, and future to prevent; Some vacant houres allow to your delight, Mirth is the pleasing buisness of the night (Dryden, 1681, The Epilogue Spoken to the King at the opening of the PlayHouse at Oxford, 23-26)

(45)

Comfort came down the trembling wretch to raise, And his last faltering accents whispered praise (Goldsmith, 1770, The Deserted Village 177-178)

(46)

To Learning's second seats we now proceed, Where humming students gilded primers read (Crabbe, 1810, The Borough IV.77-78)

Leech (1969:13-19) discusses several features of "standard archaic usage" in poetry in the period 1600-1900. They include certain lexical items, such as betimes and wight, grammatical features like do-less questions and negatives, and also what he calls "jumbled clause structures", of which the use of OV order may be considered a specific instance. As Leech points out, these usages petered out after 1900, and are now felt to be artificial. We have in fact found one piece of explicit stylistic commentary on the status of OV order by a major late XlXth-century poet, who evidently regarded it as a pattern too obsolete to be freely usable in poetry. It is found in a letter written on 14 April 1879 by Gerald Manley Hopkins to Robert Bridges; we give the relevant passage in (47). (47) By the by, inversions ~ As you say, I do avoid them, because they weaken and because they destroy the earnestness and in-earnestness of the utterance. Nevertheless in prose [...] they have great advantages of another sort. Now these advantages they should have in verse too, but they must not seem to be due to the verse: that is what is so enfeebling (for instance the finest of your sonnets to my mind has a line enfeebled by inversion plainly due to the verse, "Tis joy the falling of her fold to view' ~ but how it should be mended I do not see). [...] For it seems to me that the poetical language of an age should be the current language heightened, to any degree heightened and unlike itself, but not [...] an obsolete one.

These strictures did not prevent Hopkins from sometimes using OV order in his own poetry. A n example can be seen in (48), though it can be argued that the pattern is used here not just for the sake of the verse.

274

WILLEM KOOPMAN & W I M V A N DER WURFF

(48)

Thou art indeed just, Lord, if I contend With thee; but, sir, so what I plead is just, Why do sinners' ways prosper? and why must Disappointment all I endeavour end? (Gerard Manley Hopkins, 1889, 'Thou art indeed just, Lord' 3-4)

In XVIth-century poetry, van der Wurff & Foster (1997) found a preponderance of informationally light preverbal objects, including pronouns. So far, there has been no systematic study of OV usage in poetry in the period 1600-1900. but informationally light objects in preverbal position seem to become less usual as time goes by. Separation of the preverbal object from its lexical verb appears to remain possible; in (49), the intervening element is another lexical verb, just as in late Middle English (35), though in the absence of further data it is not clear to what extent this pattern has become limited to creative language play. (49)

Those two old Bachelors without loss of time The nearly purpledicular crags at once began to climb (Edward Lear, 1877, 'The Two Old Bachelors' 25-26)

As we saw above, Speas (1997) points out that in present-day English too, OV order is sometimes used in poetry. A n example can be seen in the fifth line of the following passage from a seasonal poem by L. Horn and G. Nunberg, which appeared in Natural Language and Linguistic Theory 14. (50)

God rest you merry, S J . Keyser; We just adored the last LI, sir. To you who labor on Synthese, We dip our derby, flip our fez; And heartily our topper doff To Lg. czar Mark Aronoff.

The use of OV order in present-day English is not entirely limited to humorous verse, as illustrated by (51), with preverbal this in the first line. (51)

And this remembering now I mark that what Light was leaving some of them at least then. Forty-two years ago, will never arrive In time for me to catch it (Louis MacNeice, 1963, 'Star Gazer' 10-13)

In addition, instances of OV order, often together with other obsolete usages, can also still be found in several fixed sayings and traditional rhymes, as in (52-55).

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

(52)

He that mischief hatches, mischief catches,

(53)

With this ring I thee wed.

(54)

One swallow doth not a summer make.

(55)

Now I lay me down to sleep, I pray the Lord my soul to keep And should I die before I wake, I pray the Lord my soul to take.

275

These well-known pieces of language are sometimes echoed or imitated, resulting in novel uses of O V order, as in (56) and (57). (56)

WITH THIS RING I THEE SHED (headline in Daily Mirror 29/8/96)

(57)

"Dedication doth not a Good Teacher Make" (title of student's essay; Barnet & Stubbs 1995:22)

Another occasion when speakers of modern English may produce, and be exposed to, clauses featuring OV order is when older texts are recited, sung, listened to, or read silently. This would include examples like (43-46), (48), (49) and also (58), which is from a hymn. (58)

God moves in a mysterious way His wonders to perform; He plants his footsteps in the sea, And rides upon the storm (Cowper, 1779, Olney Hymns 35)

The first sentence in (58) in fact seems to have attained a status as a more or less independent expression in English, as can be seen from its use in cases like (59). (59)

Lottery regulator Oflot moves in a mysterious way its wonders to perform {Private Eye 21/8/98)

A final context in present-day English where OV can sometimes be found is in translations of Old or Middle English texts. For example, Brinton (1990:51) quotes the three lines from the Old English poem Andreas given in (60) and provides the translation given in (61), no doubt counting on her readership's familiarity with OV order to enable them to understand this usage in the translation. (60)

Hwæt, ic hwile nu haliges lare, leoðgiddinga, lof fræs pe worhte wordum wemde (Andreas 1478-1480)

(61)

"What, I for a while now the story of the holy one, a song, praise of that which he wrought, announced with words."

276

WILLEM KOOPMAN & W I M V A N DER WURFF

The general conclusion that must be drawn from this lengthy diachronic survey is this: the change in the order of object and verb in English has been a long and extremely gradual one. A rough-and-ready summary of the process would run as follows: in early Old English, OV order is the norm, with V O being a special order, which is restricted to heavy NPs and had prosodic effects. In later Old English, V O order becomes more widely available for light objects (including pronouns) as well, and in the period until ca. 1400 there is competition between OV and V O order, with V O slowly becoming the predominant order in terms of frequency and in terms of information-status. In prose text from the X V t h century, OV order is quite rare and it becomes limited to just three syntactic contexts, before petering out in the course of the X V I t h century. In verse, OV stays alive much longer: it continues in a wide range of contexts and is regularly found until ca. 1900, when it comes to be regarded as an obsolete pattern, use of which betrays a certain lack of craftsmanship. In modern English, OV order can be encountered in a number of fixed sayings, proverbs and verse passages from earlier days, and these sometimes lead to the creation of novel examples of OV order by a process of echoing or allusion. 3.

Word order change and competing grammars In several generative studies of syntactic change published over the last ten years, use is made of the concept of grammar competition as an explanatory device. Here, we shall present some reflections on the viability of this approach, on the basis of the data described in the two preceding sections. The essential idea of the competition approach is that, where a language exhibits grammatical variation taking the form of apparent optionality, the two variants actually derive from different grammars. In a discussion of alternations such as verb-second vs. verb-third and OV vs. V O , Lightfoot (1999) explains the notion as follows: (62) Where a language has such an alternation, we say that this manifests diglossia, and that speakers have access to two grammars. Certain speakers have access only to one grammar; others have access to the other grammar; and others have access to both grammars in an internalized diglossia (Lightfoot 1999:94)

The hypothesis of grammar competition, or the 'double base hypothesis', derives from work dating from the late 1980s and early 1990s, carried out and inspired by Anthony Kroch, but recently it has found wider favour, as reflected in the enthusiastic endorsal by David Lightfoot in the book just quoted from and in other work.

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

277

Among the studies invoking the notion of grammar competition there are also some that deal with the shift from OV to V O order, in the history of Greek (by Ann Taylor) and in English (by Susan Pintzuk, and by Kroch & Taylor). These authors provide the following types of characterization of the competi­ tion that they see in their data: (63) [I]ncremental linguistic change is accomplished by the gradual replacement of one grammatical system (or subsystem) by another. While the change is in progress, speakers can make use of forms from either grammar, but as time passes, a higher and higher percentage of the output is derived from the innovative grammar at the expense of the one it is replacing. (Taylor 1994:2) From these results it is clear that Herodotus cannot be derived solely from either the verb-final grammar of Homer or the verb-medial grammar of the Koiné. On the other hand, a model in which a little over 60% of the data is from a verb-final base and the rest from a verb-medial one gives a fairly good fit. (Taylor 1994:33) (64) Instead of a uniform grammatical system during the Old English period [...] there are two competing grammars, one underlyingly OV, the other underlyingly VO. The VO grammar emerges during the Old English period and competes with the old OV grammar through at least part of the Middle English period, until the old system dies out. (Pintzuk 1998:27)

Some further general discussion of this approach to variation can be found in Kroch (1994); there, grammar competition is linked to morphological variation, via the idea that syntactic variation resides in functional heads, which are comparable to morphological formatives. Santorini (1992), Kroch (1994) and Kroch & Taylor (1994) also address a number of criticisms that have been levelled at the notion of grammar competition, concerning a possible lack of restrictiveness and problems of learnability; further arguments in favour of grammar competition are provided in Lightfoot (1999:92-101). We regard the studies cited above, and several others using the double base hypothesis, as excellent pieces of scholarship, whose authors combine admirable inventiveness with meticulous attention to empirical detail. Nevertheless, we have a gut feeling that the grammar competition hypothesis is wrong. One reason for this feeling is that, at a certain level, the entire approach smacks of brute force. Where others have devised the most subtle and ingenious adjustments to existing analyses in order to reconcile apparent optionality in grammars with the assumption that grammatical theory should not allow optional operations (see, for example, several of the papers in Abraham et al 1996), proponents of grammar competition simply postulate two entire grammatical systems operating next to each other. Moreover, one may wonder about the other properties of the systems, in particular their

278

WILLEM KOOPMAN & W I M V A N DER WURFF

morphological ones. As pointed out in § 1, there may be a unidirectional link between the overt expression of case distinctions and freedom of positioning of direct and indirect objects. How could this link ever be expressed in a grammar competition model? Similarly, the decline of OV order in late Middle English prose, as described in § 2, has been linked to the decline of morphological marking on verbs (see van der Wurff 1997a) and nouns (see Roberts 1997). Still, OV order remained a productive option in verse until 1900, while the morphological marking did not. What did the OV grammar licensing these poetic OV orders look like? Another general feature of the grammar competition hypothesis that we find less than appealing is the way it can discourage the search for explana­ tions. Kroch (1994) explicitly attributes the eventual disappearance of the conservative grammar, as shown by the demise of the older variant, to the blocking effect familiar from the study of morphology. This would mean, for example, that the order DO-IO with nominal objects, or the order OV, was ousted just because the coexistence of two orders was less than optimal. No specific other causative factors need have been present. If this is correct, the analyst's job is to determine the existence and nature of grammar variation, but not to advance explanations for the eventual loss of one of the grammars. This strikes us as an impoverished job description for a historical syntactician, and we would urge for inclusion of the duty to provide explanations for why certain variants are lost at the time that they are. At a more specific level, we think that there are two issues in competition theory that are in urgent need of further explication: the nature and locus of the competition between the two grammatical systems, and the length of the timespans that can be involved. That the first issue has so far remained somewhat underdeveloped is also pointed out by Lightfoot (1999:109, n 6). However, his own specific interpretation of the competition as involving diglossia (compare the quotation from his work given in (62)) cannot be regarded as a real solution to the problem. The standard meaning attached to the term 'diglossia' is that of two languages or language varieties being systematically distinguished according to the perceived formality of the situation in which they are used. In a survey article on diglossia, Schiffman (1997:205) speaks of "[d]omains of linguistic behaviour parceled out in a kind of complementary distribution", where the formal or 'high' situations, calling for the use of the H-variety, would include public speaking, religious texts and practice, education, and other prestigious kinds of usage, while the 'low' situations, calling for the L-variety, would include informal conversation,

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

279

jokes, the street and the market, the telephone and any other domains (letter writing, cinema, television) not reserved for the H-norm. This widely accepted meaning of the term diglossia, as involving entire varieties that systematically differ in level of formality, is not easily applicable to the sort of case that has been or could be described using the notion of grammar competition. In the two changes discussed in §§ 1-2, for example, there is clearly variation in the ordering of objects and verbs, but it does not appear to be the case that the variant word orders are parcelled out in anything resembling complementary distribution: the variation is found in many individual Old and Middle English texts, without there being any sign of a stylistic break separating the relevant places of occurrence. It is to be noted that other linguists adopting the idea of grammar competition do not use the term diglossia to describe the phenomenon. Santorini (1992:619), for example, points out that phenomena like multilingualism, diglossia and code-mixing provide evidence for speakers' ability to operate with two different grammati­ cal systems, but she does not argue that grammar competition is in fact identical to any of these. This then leaves us with the question: what is the competition between the two grammars like and what is it determined by? The evidence presented in §§ 1-2 suggests that the answer could depend on different factors at different historical stages. Thus, in the change affecting the order of the object relative to the verb, factors having to do with prosodic weight, with syntactic categorisation, with dialect, with information-status, with style, with register and even with specific collocations (e.g. of the verb wed and the object thee in present-day English) have all played a role. Describing the variation between OV and V O orders in terms of grammar competition would commit us to recognising the existence of grammar competition in myriads of cases. In the area of register variation alone, there are countless frequency differences in the use of grammatical options between different registers of English (see Biber 1988 and subsequent work along these lines) but there appears little to be gained by describing such facts in terms of grammar competition. In Kroch (1994) and Lightfoot (1999:92-101), much is made of the idea that the sort of alternations analysed in terms of grammar competition involve the coexistence of features that usually do not coexist and that are diachronically unstable. This would mean that many instances of grammatical register variation may not actually be instances of grammar competition. However, the relevant distinction is far from easy to make. The data in §§ 1-2, for example, involve word ordering phenomena that in present-day English seem to admit of no variation, making them suitable candidates for a grammar

280

WILLEM KOOPMAN & W I M V A N DER WURFF

competition analysis. Nevertheless, these two changes are so slow and gradual that the phenomena cannot plausibly be called unstable. Alternation between DO-IO and IO-DO orders with nominal objects is found during a period of some six centuries, and the coexistence of OV and V O orders is actually attested throughout the recorded history of English. The timespan over which the existence of grammar competition has to be postulated is therefore much longer than its proponents might care to allow. As both Kroch (1994) and Lightfoot (1999) point out, modern generative theory has explored the idea that there can be no true optionality and variability in grammars. There have been attempts to neutralize the not so hard-to-find evidence of seemingly optional phenomena in modern languages, but close inspection of the historical record of just two changes in English suggests that optionality and variability may be too widespread to be considered unstable, transient, and marginal.

REFERENCES P R I M A R Y SOURCES Citations of Old English texts are taken from the Dictionary of Old English Corpus (apart from ÆCHom I where Thorpe's text is cited) with the line references and short titles employed by the Dictionary of Old English (Healey & Venezky 1980). Ancr. = J. R. R. Tolkien, ed. 1962. The English Text of the Ancrene Riwle, Ancrene Wisse, CCCC MS 402 (EETS 249) By folio and line. Ayenb. = R. Morris and P. Gradon, eds. 1965-79. Dan Michel's Ayenbite of lnwit, corr. and repr. (EETS 23, 278) Barlam = J. C Hirsh, ed. 1986. Barlam and losaphat (EETS 290). Capgrave Chronicles = P. J. Lucas, ed. 1983. John Capgrave's Abbreuiacion of Chronicles (EETS 285). Chaucer = L. D. Benson, ed. 1988. The Riverside Chaucer 3rd edn. Oxford: Oxford University Press. MKempe = S. B. Meech, ed. 1940. The Book of Margery Kempe (EETS 212). More= qouted from The Complete Works of St. Thomas More. New Haven: Yale University Press. Mor.Hom= Richard Morris, ed. 1868. Old English Homilies and Homiletic Treatises of the Twelfth and Thirteenth Centuries (EETS 34). Morte Arthur = E. Vinaver, ed. 1947. The Works of Sir Thomas Malory, vol. I, Oxford: Clarendon Press. Rob.Glo=William Aldis Wright, ed. 1887. The Metrical Chronicle of Robert of Gloucester. 2 vols. Rolls Series 86.

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

281

Thorpe Sermon - Testimony of Thorpe in A . Hudson, ed. 1993. Two Wycliffite Texts: Sermon of Taylor, Testimony of Thorpe (EETS 301). St, Juliana (Bod) = M S . Bodley 34 in S. R. T. O. d'Ardenne, ed. 1961. þe Liflade ant te Passiun of Seinte Iuliene (EETS 248). Sward = Sawles Warde in J. A . W . Bennett and G. V . Smithers, eds. 1968. Early Middle English Verse and Prose 2nd edn, Oxford: Clarendon Press. Vices&V = F. Holthausen, ed. 1888-1921. Vices and Virtues (EETS 89,159).

SECONDARY SOURCES Abraham, Werner, Samuel David, Epstein, Höskuldur Thráinsson & C. Jan-Wouter Zwart, eds. 1996. Minimal Ideas: Syntactic Studies in the Minimalist Framework. Amsterdam & Philadelphia: John Benjamins. Allen, Cynthia L. 1995. Case Marking and Reanalysis. Oxford: Clarendon Press. Barnet, Sylvan & Marcia Stubbs. 1995. Barnet and Stubbs's Practical Guide to Writing, with Readings (7th ed.). New York: HarperCollins. Beukema, Frits & W i m van der Wurff. 1999. "Modals, Objects, and Negation i n Late Middle English." M S , University of Leiden. Biber, Douglas. 1988. Variation Across Speech and Writing. Cambridge: Cambridge University Press. Brinton, Laurel J. 1990. "The Development of Discourse Markers in English." Historical Linguistics and Philology ed. by Jacek Fisiak, 45-71. Berlin & New York: Mouton de Gruyter. Broekhuis, Hans, Hans den Besten, Kees Hoekstra & Jean Rutten. 1995. "Infinitival Complementation i n Dutch: on Remnant Extraposition." The Linguistic Review 12.93-122. Campbell, Alistair. 1959. Old English Grammar. Oxford: Clarendon Press. Cassidy, Frederic G. 1938. The Backgrounds in Old English of the Modern English Substitutes for the Dative-Object in the Group Verb + Dative-Object + Accusative Object. PhD dissertation, University of Michigan. Chase, Colin, ed. 1981. The Dating of Beowulf. Toronto: University of Toronto Press. Denison, David. 1993. English Historical Syntax: Verbal Constructions. London: Longman. Foster, Tony & W i m van der Wurff. 1995. "The Survival of Object-Verb Order in Middle English: Some Data."Neophilologus 79.309-327. .1997. "From Syntax to Discourse: The Function of Object-Verb Order i n late Middle English." Studies in Middle English Linguistics ed. by Jacek Fisiak, 135156. Berlin & New York: Mouton de Gruyter. Healey, Antonette diPaolo & Richard L. Venezky. 1980. A Microfiche Concordance to Old English. Toronto: Pontifical Institute of Medieval Studies. Hogg, Richard M . 1992. "Phonology and Morphology." The Cambridge History of English, volume I, The Beginnings to 1066 ed. by Richard M . Hogg, 67-167. Cambridge: Cambridge University Press. Kirk, John M . 1985. "Linguistic Atlases and Grammar: The Investigation and Description of Regional Variation in English Syntax". Studies in Linguistic

282

WILLEM KOOPMAN & W I M V A N DER WURFF

Geography ed. by John M . Kirk, Stewart Sanderson & J. D. A . Widdowson, 130-156. London: Croom Helm. Kohonen, V i l j o . 1978. On the Development of English Word Order in Religious Prose around 1000 and 1200 A.D. Åbo: Åbo Akedemi. Koopman, W i l l e m F. 1985. "The Syntax of Verb and Particle Combinations in Old English." Linguistics in the Netherlands 1985 ed. by Hans Bennis & Frits Beukema, 91-99. Dordrecht: Foris. .1990. Word Order in Old English, with Special Reference to the Verb Phrase, PhD dissertation, University of Amsterdam. .1997. "Another Look at Clitics in Old English". Transactions of the Philological Society 95.73-93. Kroch, Anthony. 1994. "Morphosyntactic Variation." Chicago Linguistic Society 30/2.180-201. .& A n n Taylor. 1994. "Remarks on the X V / V X Alternation in Early Middle English." M S , University of Pennsylvania. .1998. "Dialect Variation in the Transition from X V to V X i n Early Middle English". Paper read at 10th I C E H L , Manchester, September 1998. Kuno, Susumu. 1987'. Functional Syntax. Anaphora, Discourse and Empathy. Chicago & London: The University of Chicago Press. Kytö, Merja. 1991. Manual to the Diachronic Part of the Helsinki Corpus of English Texts. Helsinki: Department of English. Leech, Geoffrey. 1969. A Linguistic Guide to English Poetry. London: Longman. Lightfoot, David. 1999. The Development of Language. Acquisition, Change and Evolution. Oxford: Blackwell. Los, Bettelou. 2000. Infinitival Complementation in Old and Middle English. PhD dissertation, Vrije Universiteit Amsterdam. Mitchell, Bruce. 1985. Old English Syntax. 2 vols. Oxford: Clarendon Press. Moerenhout, Mike & W i m van der Wurff. forthcoming. "Remnants of the Old Order: OV i n the Paston Letters". To appear in English Studies. Ogura, Michiko. 1980. The Syntactic and Semantic Rivalry of Quoth, Say and Tell in Medieval English. Japan: Kufs Publication. Pintzuk, Susan. 1991. Phrase Structures in Competition: Variation and Change in Old English Word Order. PhD dissertation, University of Pennsylvania. .1996. "Old English Verb-Complement Word Order and the Change from O V to V O " . York Papers in Linguistics 17.241-264. .1998. "From O V to V O in the History of English". M S , University of Y o r k . .& Anthony S. Kroch. 1989. "The Rightward Movement of Complements and Adjuncts in the Old English of Beowulf". Language Variation and Change 1.115-143. Roberts, Ian. 1997. "Directionality and Word Order Change in the History of English". Parameters of Morphosyntactic Change ed. by Ans van Kemenade and Nigel Vincent, 397-426. Cambridge: Cambridge University Press. Santorini, Beatrice. 1992. "Variation and Change in Yiddish Subordinate Clause Word Order". Natural Language and Linguistic Theory 10.595-640. Schiffman, Harold F. 1997. "Diglossia as a Sociolinguistic Situation". The Handbook of Sociolinguistics ed. by Florian Coulmas, 205-216. Oxford: Blackwell.

TWO WORD ORDER PATTERNS IN THE HISTORY OF ENGLISH

283

Speas, Margaret. 1997. "Optimality Theory and Syntax: N u l l Pronouns and Control". Optimality Theory: An Overview ed. by Dana Archangeli & D. Terence Langendoen, 171-199. Oxford: Blackwell. Taylor, A n n . 1994. "The Change from SOV to SVO i n Ancient Greek". Language Variation and Change 6.1-37. Wurff, W i m van der. 1997a. "Deriving Object-Verb Order in Late Middle English". Journal of Linguistics 33.485-509. .1997b. "OV-volgorde in vijftiende-eeuws Engels proza. Taal in Tijd en Ruimte" ed. by A . van Santen & M . van der W a l , 73-84. Leiden: Stichting Neerlandistiek Leiden. .1999. "Objects and Verbs in Modern Icelandic and Fifteenth-Century English: A Word Order Parallel and its Causes". Lingua 109.237-265. .& Tony Foster. 1997. "Object-Verb Word Order in 16th Century English: a Study of its Frequency and Status". Language History and Linguistic Modelling ed. by Raymond Hickey & Stanislaw Puppel, v o l . I, 439-453. Berlin & New York: Mouton de Gruyter.

G E N I T I V E CONSTRUCTIONS I N E A R L Y M O D E R N E N G L I S H NEW EVIDENCE FROM A CORPUS ANALYSIS

ANETTE ROSENBACH & L E T I Z I A VEZZOSI Heinrich-Heine- Universität Düsseldorf Università di Firenze

0.

Introduction Traditionally, the development of the English genitive constructions1 has been described as the gradual replacement of the inflected genitive (henceforth referred to as 's-genitive') by the of-prepositional phrase (henceforth referred to as 'of-genitive'), a process that is widely assumed to have stopped by the end of the X l V t h century, leaving the s-genitive in certain circumscribed contexts in the English language. The present paper, however, suggests another interpretation. First, the development did not end in the X l V t h century, but continued up to the X V t h century. Second, from the X V I t h century to the early X V I I t h century a significant increase of the s-genitive can be observed. We analysed a corpus based on texts ranging from 1400 to 1630 and investigated the internal factors determining the occurrence of the s-genitive vs. the of-genitive. In particular, we looked at the distribution of the two forms according to the function and the semantics of what we call "genitive N P " (see below) 2 .

* The introduction and the conclusion were written by A. Rosenbach & L.Vezzosi. Chapters 13.1 were written by Rosenbach and chapters 3.2-6 by Vezzosi. Given the traditional genitive functions (cf. § 3), we apply the term 'genitive construction' to the morphosyntactic devices that can express them. If not indicated otherwise, the examples used in this paper are taken from our corpus.

286

ANETTE ROSENBACH & LETIZIA VEZZOSI

Gods the constable's

s-genitive word son

genitive NP

head

of-genitive the word an image

of God

head

genitive NP

of a man

Our analysis shows how the alternation between the two competing constructions reveals both aspects of change and stability. 1.

Historical development In Old English, an inflectional language, genitive case-marking was characterised by various suffixes according to noun class and gender with agreement between the constituents of a NP, i.e. they all had to be case-marked (see examples (2) and (3), below). Genitive suffixes Noun Adjective

masc. sg. -(e)s -(e)s

S t r o n g Declension f e m . sg. neut. sg. pl. -(e)s -e -a l-ra -re -(e)s -ra

W e a k Declension sg. pl. -an -(e)na -e(n) -al-e

The examples (1-3) given below show that, in general, there were no restrictions on word order and function. The genitive NP could be both prenominal (1) and postnominal (2-3) and could express all genitive functions 3 . Genitive case-endings could still differ according to number (singular in example (1) and (2), plural in (3)), gender (masculine in (1) and (3), feminine in (2)) and noun class (a-stem in (1), o-stem in (2) and athematic stem in (3)). Although, generally speaking, there were already pragmatic factors for the linear arrangement of genitive NP and head within the inflected genitive (e.g. topics already tended to occur prenominally) the position of the genitive NP had not yet become fixed to prenominal position, as in Modern English. Moreover, the genitive function did not yet play a role in selecting the morphosyntactic form. (1)

pæs cyninges pegnas the-GEN king-GEN thanes "the king's thanes" {Chron A 755)

(2)

se forma dæg ðyssere worulde the first day this-GEN world-GEN "the first day of this world" (ÆCHom i.100.5)

(3)

heafod ealra haligra manna head all-GEN holy-GEN men-GEN "the head (leader) of all saints" (ÆCHom i i . 14.23)

For specification of the genitive functions, see § 3.1.

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH

In the course of Middle English, the -(e)s suffix was generalized to all nominal declensions and was the only genitive marker. The nominal paradigm was then reduced to two forms: one for the genitive, ending with the suffix -es for all singular nouns, and the other for all the other cases. Case agreement within the whole NP was no longer compulsory. For example, in (5) the possessive adjective still agrees with its head and is casemarked, but it does not in (4). Moreover, the genitive NP became restricted to prenominal position only. (4)

ure lordes beringe our lord-Sg-Gen birth "our lord's birth" (Lawman B 18426)

(5)

for mines drihtenes luve for my-Sg-M-Gen lord-Gen love "for my Lord's love" (Lawman A 19728)

In this same period, the s-genitive was largely replaced by analytic patterns, mainly the of-genitive, a process which is in accordance with the overall drift of English from a synthetic to an analytic language. Mustanoja (1960:75) presents the following distribution of the two genitive constructions throughout the Middle English period.

■ -

(e)s

-•—of

Figure 1: Occurrences of the s-genitive and the oi-genitive according to Mustanoja (1960:75)4(in %)

The figures presented in Mustanoja (1960:75) rely on Fries (1940:74), who again quotes an unpublished doctoral dissertation by his student Russell Thomas (University of Michigan, 1931).

288

ANETTE ROSENBACH & LETIZIA VEZZOSI

Apart from Mustanoja's often quoted figures (taken from Thomas's statistical analysis), which stops at the end of the X l V t h century, there are no other studies which provide any exact figures about the distribution of the different genitive constructions for the time after the X l V t h century. Stahl (1927) gives a detailed account of the distribution of the 5-genitive and the of-genitive in Middle English and early Modern English, starting at the end of the Old English period. For the whole X V t h century, however, his figures are based on two authors (Caxton and Malory, late X V t h century); for the X V I t h century his corpus consists solely of Shakespeare plays (late X V I t h century/early X V I I t h century), which cannot be taken as representative for the language of the whole century. Den Breejen (1937) provides a very elaborate discussion of the 5-genitive and the of-genitive in the second half of the X V I t h century, but he does not quantify his data. Altenberg's extensive and sophisticated corpus analysis (Altenberg 1982) of the distribution of the sgenitive and the of-genitive covers only the X V I I t h century. Although he mentions the fact that the 5-genitive was almost dying out in late Middle English and Early Modern English and got consolidated again in later centuries (cf. Altenberg 1982:13), this claim is, firstly, based on observation only and not supported by any quantitative data, and, secondly, it refers to the replacement of the 5-genitive by s-less forms (e.g. my father house) in some dialects, but not to its relative distribution together with the of-genitive. Recently, the history of English genitive constructions has experienced a sort of renaissance among linguists. These studies, however, either focus on a discussion about the status of the 5-genitive as an inflection or a clitic (cf. Allen 1997; Seppänen 1997a), or on the s-less genitive (cf. Allen 1998; Klemola 1997), but neither of them aims to provide any empirical data for the relative distribution of the 5-genitive and the of-genitive. Appropriate data for the X V t h century onwards are therefore lacking. This might be the reason why there seems to be a certain consensus among handbooks that the whole process of replacing the 5-genitive had been completed by the end of the X l V t h century. 2.

Corpus analysis: 1400-1630 To fill the gap, we put together a corpus consisting of texts mainly taken from the Helsinki Corpus, covering the period between 1400 to 1630. The advantage of using the Helsinki Corpus lies in the fact that it contains text excerpts covering a large variety of text types and authors. In selecting appropriate texts we tried to ensure to have a heterogeneous text sample ranging from formal prose (e.g. religious texts) to informal prose (e.g. correspondence, diaries, travelogues). The results of this study are therefore not biased towards certain text types or the stylistic preferences of single authors.

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH

289

Our corpus comprises almost 10,000 tokens which include not only the sgenitive and the of-genitive, but all nominal constructions that can encode genitive functions (cf. § 2.1). In contrast to the Helsinki Corpus, which takes as synchronic stages time intervals of 70 years, our corpus is divided into four periods: 1400-1449 (I), 1450-1499 (II), 1500-1559 (III), and 1560-1630 (IV). This study is based on quantification of the data, i.e. the frequencies of the tokens were recorded. 2.1 Distribution of genitive constructions In the period investigated there are several nominal constructions that can encode the genitive functions 5 . Apart from the s-genitive we find the following constructions with a preposed genitive NP: the his-genitive, the zero-genitive and the s-less genitive. In the his-genitive (Monsieur Boissy his army), also referred to as "possessive dative" (cf. e.g. Mustanoja 1960:159-160) or as "separated genitive" (Allen 1997:113ff.), a possessive pronoun (mainly his) is used to link the genitive NP (mainly a proper name) with the head in a posses­ sive relationship (cf. Mustanoja 1960:159-160). In early Modern English the his-genitive was not yet as stigmatized as in present-day English 6 and could therefore still occur in written texts (cf. den Breejen 1937:10). It was, however, as Altenberg (1982:45) notes, "never ... popular enough to threaten the regular form", i.e. the s-genitive. Furthermore, there are genitive constructions with prenominal genitive NPs which lack a phonologically realized genitive marker. Here we further distinguish between zero-genitives and s-less genitives. We regard as zero-genitives all those constructions in which there is a plural genitive NP (certin howers tasks) and where a genitive marking sometimes is orthographically indicated by an apostrophe (boys' game). Our category of sless genitives covers all the other cases (e.g. the butcher wyff, mr perkins newe booke, S. Thomas Evin). In the of-genitive and the to-genitive the genitive NP follows the head. The analysis of the frequency of genitive constructions reveals the following distribution.

For a discussion and specification of the genitive functions, see § 3. In present-day English the his-genitive can still be found in colloquial speech and regional dialects (cf. Seppänen 1997b:160-161).

290 Forms

of

(e)s 's total s his to zero s-less

ANETTE ROSENBACH & LETIZIA VEZZOSI 1560-1630 (IV) 1500-1559 (III) 1400-1449 (I) 1450-1499 (II) n n n n % % % % 2059 82.6 77.5 1341 90.7 82.8 1826 2257 369 14.8 8 118 7.9 184 6.8 188 0.1 2 0.1 5 261 11.1 1 136 119 14.9 11.8 8 371 320 449 19.1 1 0.4 0.1 39 1.6 0.4 11 10 3 0.2 1 1.2 41 1.5 29 6 0.4 8 0.4 0.3 12 0.5 10 9 14 0.6 3.1 0.6 84 31 1.3 Table 1: Distribution of genitive constructions according to time intervals

In each interval the of-genitive clearly prevails over the other genitive constructions by a wide margin. The only other form that occurs with any significance is the s-genitive, which comprises both the (e)s ending and the apostrophe 's7'. Thus, the of-genitive and the s-genitive turn out to be the major competing forms. Therefore, our further quantificational analysis is focussed on them only. What is much more interesting is that - unnoticed by previous studies change within the English genitive after the end of the X l V t h century continued. In particular, a further decrease of the s-genitive can be observed in the X V t h century. This is not only a decrease in quantity, but also a change in quality. In fact, in the first part of the X V t h century (period (I)) the s-genitive predominantly occurs with certain lexical items (i.e. God, king, queen and man), which indicates that the s-genitive in this period is no longer fully productive. The default form is represented by the of-genitive. However, later, as can be seen in Figure 2, the s-genitive increases again in occurrence at the expense of the o/-genitive.

Note that the old inflectional (e)s suffix was gradually replaced by the apostrophe 's during the XVIth and XVIIth centuries. Accordingly, these two forms vary in our corpus.

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH

291

Ms-genitive mof-genitive

13thc (first half) Mustanoja

14thc Mustanoja

I5thc (first half) our data

around 1600 our data

Figure 2: Distribution of the s-genitive vs. the of-genitive: Mustanoja (1960) and our data (in %)

This development runs counter to predictions of typology and language change theories. From a typological point of view, it is extremely unusual for a language to have the genitive case as the only inflectional form in the nominal paradigm (cf. Janda 1980:245; Silverstein 1976:162; Taylor 1996:118). I f any­ thing, English genitive case marking should have been lost rather than revived 8 , since in the X V t h century it had almost fallen out of use. Therefore, it is hard to explain why it has become productive again. Our analysis aims at identifying the linguistic context of this change, focusing on the semantic-pragmatic factors that could possibly influence the occurrence of the s-genitive in favour of the of-genitive. In particular, the following factors are analysed: genitive functions, animacy and topicality /definiteness.

We refer to the concept of natural morphology (cf. Dressler et al. 1987), which predicts change from a marked (i.e. unnatural) situation to an unmarked (i.e. natural) one.

292

ANETTE ROSENBACH & LETIZIA VEZZOSI

3.

Genitive functions For the classification of the genitive functions, the traditional categories, such as possessive, subjective, objective and partitive are adopted, which are based on Latin grammar and which are partly inconsistent. 3.1 Classification of genitive functions The main inconsistency is the distinction between the possessive, subjective and objective function, and the source of such inconsistency is twofold. First, the possessive function is a semantic/conceptual category, while the subjective and objective functions are grammatical. Second, the notions of subject and object themselves are highly controversial. Accordingly, in several cases they cannot be unambiguously categorized. The different nature of the possessive vs. the subjective/objective function leads to problems of classification concerning both the semantic and syntactic level. On the semantic level, ambiguity arises from the degree of 'verbiness' of the head noun. Given the two conceptually identical phrases John's error and John's fault (our examples), John's error would have to be classified as subjective because it can be paraphrased as "John errs", but a possessive interpretation is to be preferred for "John's fault", because there is no verbal counterpart to the noun fault (*"John faults"). In general, if there is a corresponding verb to the head noun, then the bias is towards a subjective/ objective interpretation. Otherwise rather the possessive function seems to be appropriate although this may not be conceptually motivated (cf. John's error vs. John's fault). Furthermore, if there is a verbal counterpart to the head noun the classification also seems to depend on the aspectual reading of this nominalized verb. In the writings of Plinius the head noun "writings" refers to the texts Plinius wrote. However, due to the underlying verb to write the action of "writing" can be foregrounded, in contrast to the term 'books', which simply refers to the object itself. I f the telic reading of "writing" (as in Plinius' books) is preferred, the writings of Plinius has to be classified as possessive. The dy­ namic reading of "writing", however, would suggest a subjective interpretation. On the syntactic level, genitive constructions expressing properties (e.g. the fertility of the soil) cause a general problem. I f the fertility of the soil is paraphrased with a copula construction (i.e. "the soil is fertile"), then, grammatically speaking, these cases have to be classified as subjective. Semantically, however, these constructions are rather cases of loose possession.

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH

As far as the controversial status of the concept of 'subject' and 'object' is concerned, the question, in general, is whether to proceed from their syntactic properties or from their semantic value. For example, on a purely syntactic level the increase of worship can be regarded as subjective ("worship increases"). Looking at the semantics of the verb, however, we note that the subject is not identical with the agent, but rather with the patient. Therefore, this example cannot be classed as subjective, but rather as an intermediate category between subjective and objective. Moreover, it appears that we cannot do without the notion of argument structure in our classification. For example, the lack of maintenance could be classified as both objective ("someone lacks maintenance") and as subjective ("maintenance is lacking"). To sum up, the traditional categories possessive, subjective and objective cannot always be discretely distinguished. However, the overlaps (as indicated in Figure 3) that arise are not so dramatic that this classification system has to be changed.

Figure 3: Genitive functions: methodological problems

We w i l l therefore, for the sake of comparison and compatibility, follow the taxonomic categories traditionally used in the description of the historical development of the English genitive (e.g. den Breejen 1937; Mustanoja 1960; Fischer 1992). These taxonomies vary from author to author. The operational definitions for the categorization used in the present study can be seen in Table 2.

294 function Possessive

ANETTE ROSENBACH & LETIZIA VEZZOSI operational definition and examples a) prototypical: Master Slender's purse, Slender's servant; Page's wife; Caesars letter b) non-prototypical: the housewives cleanlinesse; the fertility of the soil

Subjective

GEN-NP in S-V relation to head noun: Peter's going to the market his wife's love my lady's presence

Objective

GEN-NP in O-V relation to head noun: the milking of a cow Goddis pleasuer

Descriptive

1

additional information that qualifies/specifies the head noun: a daughter of noble fame chains of gold

Partitive Appositive Lexical Lexicalised Time Location

other

part/whole relationship: one of his servants, the rest of his life both head NP and GEN-NP are referentially identical: the city of Rome, the month of May, the book of Exodus certain head nouns require one, and only one, form of the genitive: some manner of difference, any kind of diseases a) fossilized prepositional forms: on top of, in the sted of be cause of b) fixed phrases (mainly titles): the kynges highnes, my ladys grace time indication: the first day of August, the 7th year of king Edward additional locative information: Hamlet of Denmark, M. Brown of Christchurch, the old woman of Brentford any other occurrences of the genitive that do not fit into the above mentioned categories

Table 2: Operational definitions of the genitive functions

The possessive function does not only, as the name suggests, refer to possession proper, but rather covers a wide range of relationships. Taylor (1989:202ff.) points out that the concept of possession is a difficult one and that it cannot be regarded as a semantic primitive, but rather as a category with more or less prototypical members, and he provides a list of prototypical properties of possession. For our analysis it was not feasible to isolate all the different degrees of possessive relationships, and we had to limit the distinction to possessive vs. non-possessive functions. Therefore, we w i l l not specify different types of possession, but we w i l l subsume them all under possessive function. 3.2 Analysis and results In our functional analysis, the realization of the various functions by the two alternative genitive constructions is investigated.

(e)s

1400-1449(1)

Others %

Time %

4

- -

- -

-

- -

922 51 662 33.8 299 2 0.7 99.3 26 2 82.9 7.1 35 12 74.5 -25.5 96 100

- 26

11731

totals of 224 738 23.3 76.7 338 71 17.3 82J6 - mm 355

(e)s 107 123 16 11.1 4 13 21 132 5 13 3 6.7

81.3

27

7.6

14 88 1 0.3 3 6.7

35 6 22 1.6

124 78 366 98.4 39 86.7 44 55.7 217 100 137 100

iïïilÉfi 212 1 092 18.7 W3ÊÊ 973

totals of 201 672 77 23 117

's 94 10.7 11

1500-1559(III)

12 16.4 16 6.1 3 5.1

1

5.1

- -

- -

- -

- -3.6

-

82 100

29 100 100

165 100 76

94.9 27 96.4

's totals of 186 605 202 314 653 151 48 34.1 68 21 6 71 9 mm 341 32 25 ' l i f e 963 1 13 60 1.4 82.2 31 231 15 17.8 5.7 11.8 88.2 3 56

1560-1630 (IV)

(e)s 128 139 23 10.4 4 1.1

2 183 0.7 97.8 2 37 7.1 97.4 6 1 37 12 27 8 13.4 36 97.4 25.5 34.1 35 10.1 69 - 44.2 100 i 42 53 - - 1 100 1 1.9 98.1 1 16 1.9 25 249 - 3.8 100 96.2 1 100 17 1 40 65 - 3.8 1 - 100 2.4... 100 2.4 97.6 Table 3: Function Form correlation

1 2.6

2.6

- -

-

- 2.1 1

-

■

-

Descr. 4 2.1 % 1 Lexicalized 2.6 % 1 Lexicalized 2.6 % Part. % Loc. % -

Poss. %

's 1 0.1 1 02

1450-1499 ( I I )

's totals of (e)s 1 85 420 223 84 0.2 16.8 83.2232 70 20 16.6 - 20 193 Subj 9.4 9.4 90.6 17.1 6 fioit 268 30 % % t i l l 973 73 22 2 58 26 - 2 App %33 33 96.7 333

Function

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH 295

296

ANETTE ROSENBACH & LETIZIA VEZZOSI

The figures in Table 3 reveal two things. First, each function can be codified by the of-genitive, while the s-genitive is mainly restricted to possessive, subjective and appositive functions and some lexicalized phrases (e.g. the Kyngis Grace, the Kyngis Highnes). Moreover, in each function the of-genitive is always the more frequent variant. The s-genitive thus turns out to be the marked option as far as the realization of the functions and the frequency of its occurrence within these functions are concerned. Second, in a diachronic perspective, the preferences for the s-genitive in the possessive and the subjective functions increase at the expense of the of-genitive throughout the four intervals as can be seen in Figure 5.

Figure 4: Realization ofpossessive and subjective functions by s-genitive (in %)

The relatively high proportions of s-genitives in the appositive function (cf. Table 3) in the second to the last intervals are due to a certain type of appositives in which the genitive NP is a saint's name and the head refers to an institution (e.g. a church, parish, hospital), such as S. Cuthbertes chirch, saynt Dunstonys parryche, S. John's hospitale, or designates a certain day (Seynt Margaretes Day). Although these cases are certainly not prototypical instances of the appositive function, there is evidence that the genitive NP and the head seem to be coreferential. Sometimes the head is omitted and the genitive NP alone refers to the referent of the genitive construction (e.g. at saynt Antonys); these elliptic constructions have also been called the 'absolute genitive' (van der Gaaf 1932) or 'genitive of place' (Mustanoja 1960:83). In contrast to the possessive and subjective functions, however, the frequency of the s-genitive with this type of appositive function clearly decreases over time (cf. Table 3). So far, the 5-genitive has been shown to be marked with regard to function. The use of a marked construction, however, is usually characterized by various factors. Therefore, we looked for further factors that could possibly favour the occurrence of the 5-genitive in contrast to the unmarked of-genitive.

297

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH

4.

Animacy One good candidate for the preference for the 5-genitive is the factor of animacy. In Old English, genitive functions were primarily expressed by inflection (see Figure 1; cf. Mustanoja 1960:75) irrespective of the animacy of the genitive NP (cf. examples (1) and (2)). Only in Middle English, when the two genitive constructions became competitive, linguistic factors make sense to be investigated. In fact, most descriptions of the English genitive, both synchronic (e.g. Quirk et al. 1972) and diachronic (e.g. den Breejen 1937; Mustanoja 1960; Stahl 1927), point to the preference of the 5-genitive for animate NPs 9 . However, this statement has never been empirically tested through any quantitative analysis for the X V t h century onwards. The only exception is the study of Stahl (1927), whose selection of texts, however, cannot be considered representative for the X V t h and X V I t h centuries (see above; for a similar argumentation; cf. also Altenberg 1982:19). In this study, the frequency of occurrence of the two genitive constructions is analysed according to the feature [± animate] of the genitive NP. This feature includes [+ human10] and [+ animal]. As emerges from Table 4 and Figure 5a, the s-genitive hardly ever occurs with inanimate genitive NPs. So the only possible chance for the 5-genitive to be realized is the [+ animate] context (cf. Table 4 and Figure 5b).

Form Animate Genitive-NP

Inanimate Genitive NP

of (e)s 's total s

of

(e)s 's

toal s

8

1400-1449 n % 398 78.4

1450-1499 n % 735 68.1

110

21.6

344

31.9

110 732

21.6 98.9

31.9 97.7

8

1.1

344 108 2 25

1.1

2.5

2.3

2.3

1500-1559 n % 512 63.2

171

127 298 161 4 11 6

17

1560-1630 n % 398 49.7

21.1 15.7 36.8 98.9

183 219 402 1287

22.9 27.4 50.3 96.8

0.7 0.4

4 38

0.3 2.9

1.1

42

3.2

Table 4: Distribution of the s-genitive and of-genitive according to [± animate]

Note that they do not refer to animate NPs, but rather to human beings, preferably proper names. Including metaphysical entities, such as God, angels and saints (e.g. Saynt James, saynt Marten).

298

ANETTE ROSENBACH & LETIZIA VEZZOSI

■ of-genitive ■ s-genitive

Figure 5a: Distribution of the s-genitive and of-genitive according to [-animate] genitive NP (in %)

■ of-genitive ■ s-genitive

Figure 5b: Distribution of s-genitive and of-genitive according to [+animate] genitive NP (in %)

Diachronically, within the animate context the s-genitive increases over time until, finally in the last period (IV), it is as frequent as the of-genitive (see Figure 5b). Thus, variation of the two genitive constructions is restricted to the [+ animate] context, only here is there a true option for the s-genitive. Therefore, only in this context can we apply a statistical test (chi2) to show that the increase of the s-genitive is not due to random variation but statistically significant.

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH

299

The increase of frequency of the s-genitive with [+ animate] genitive NPs at the expense of the of-genitive from the first interval (1400-49) to the last interval (1560-1630) in our corpus turned out to be significant (chi2, p < 0.05, using Yates' correction). 5.

Topicality Idefiniteness Given that the s-genitive is restricted to the feature [+ animate], our analysis focuses on this context, this time distinguishing between [± human]. Since the frequency of the s-genitive with [+ animate] [- human] is not relevant (i.e. there are hardly any [- human] s-genitives in the corpus; cf. Table 5), exclusively the feature [+ human] is taken into account when looking at the factors of topicality and definiteness. Defïniteness/ Topicality Human Non-referential (generic) Referential Proper name/ unique ref. common noun/ DET ref common noun/ INDET ref. animals

1400-1449 (I)

1450-1499(II) 1500-1559 (III) 1560-1630 (IV) 's total s (e)s 's total s (e)s 's total s

(e)s 's total s (e)s 25

1 26/

42

-

23.6% 32 164

-

32/ 9.3 2 166/

6

11

17/ 5.7% 55 53 108/ 36.2% 108 62 170/ 57.1% 2 - 2/ 0.7% 1 1/ 0.3%

19 99

42/ 48.2% 38.2% 137 38 52 - 38/ - 137/ 34.5% 39.8% 6 9 3 - 6/ 3/ 1 - 2.7 3 4 - 1.7% 3/ 1/ % 0.9% Topicality and 0.9% Table 5: S-genitive: definiteness in [+ animate]

32

51/ 12.7% 96 195/ 48.5% 74 126/ 31.3% 13 22/ 5.5 4 8/ 2

'Definiteness' is used in the sense of 'referential definiteness' (Givón 1989). This notion includes all those terms that can be referentially identified both extralinguistically (i.e. proper names, such as John, and unique reference entities, such as God) and contextually (i.e. common nouns with determinate reference, such as the father, the king). Accordingly, 'indefiniteness' is used in the sense of 'referential indefiniteness', referring to all those terms that cannot be referentially identified (e.g. a poor man11). In contrast, 'non-referential' corresponds to 'generic' (e.g. 0 lions, a/the lion). Figure 6 shows how the sgenitive is distributed according to referential definiteness [(+ref.] [+def.]) and indefiniteness ([+ref.] [-def.]).

11 In a non-generic reading.

ANETTE ROSENBACH & LETIZIA VEZZOSI

300

Figure 6: Distribution of the s-genitive according to referential definiteness ([+ref.] [+def]) and indefiniteness ([+ref] [-def]) of the genitive NP (in %)

Throughout all the intervals, the s-genitive hardly ever occurs with indefinite genitive NPs ([+ref.] [-def.]), but has a strong preference for definite genitive NPs ([+ref.] [+def.]). [+ s-genitive]

[+definite] [+referential]

[- s- genitive]

[-definite] [+referential]

the kinges children mon. Morret's brother Gods word strength of other princes the endeuors of a priuate man

How can this neat distribution be interpreted? In our opinion, this result is in conformity with the discourse pragmatic function of the s-genitive. The features [+definite] [+referential] correspond to given information, and thus are part of the topic (Gundel 1986). According to theories of information structure (Firbas 1964; Gundel 1986), given information precedes new information in a sentence. Assuming that this principle does not only work at the sentence level but also at the phrase level, then the topical information expressed in the genitive NP facilitates the referential identification of the head noun. For example, it is easy to identify the aduersairs "adversaries" and house in the phrases the kynges Aduersairs "the king's adversaries" and mr. Browne's house, if the kyng and mr. Browne are known. So, it is not surprising that the s-genitive almost never occurs with referential indefinites. If the genitive NP is topical (referentially definite), only in the s-genitive can the pragmatically more efficient structure be realized.

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH

syntactic word order

GEN-NP

s-genitive HEAD- NP

pragmatic word order

Topic/Theme

Comment/Rheme

301

6.

Conclusion According to our data the process of replacing the s-genitive by the ofgenitive has not been completed by the end of the XlVth century, as handbooks claim. The of-genitive continues to extend its domain until the XVth century. During the XVIth and XVIIth centuries, the s-genitive increases again, although it is both functionally and semantically marked. First, the s-genitive preferably expresses possessive and subjective functions. Second, the s-genitive is almost exclusively restricted to genitive NPs with the semantic features [+animate] [+human] [+definite] [+referential]. Within these contexts the s-genitive stays in the English language and increases in occurrence. The English morphological system shows a formal continuity in that the formal element for encoding the genitive case has been maintained (-(e)s I 's) from Old English to Modern English. However, the s-genitive has changed functionally. First, it becomes restricted to possessive and subjective functions mainly, and it is in these two functions that we can observe the increase of the s-genitive, while it is constantly decreasing with the appositive function. Second, and more importantly, what motivates the occurrence of the s-genitive becomes pragmatic in nature. The s-genitive is preserved only in prenominal position, and in this position it occurs only with topical genitive NPs, thereby acquiring an anchoring function. This new function might explain why the s-genitive stays in English against the odds and even increases in frequency. Thus, the development of the s-genitive is not only a case of change, but also exhibits aspects of stability. What at first sight seems to be a contradiction, namely that the development of the s-genitive is both a matter of stability and change, can be explained when distinguishing its formal and functional aspects. Form Function

Stability + (e)s/'s -

Change + restriction to possessive and subjective function mainly new anchoring function

302

ANETTE ROSENBACH & LETIZIA VEZZOSI

Continuity (i.e. stability) is a property of the morphological level. I n fact, a formal innovation must have a strong motivation and external conditions w h i c h can favour i t , because morphology is an inventory o f a more or less fixed set o f elements. I n contrast, functional innovations are not so constrained, since they do not belong to the ' v i s i b l e ' part of grammar and therefore are not much affected by linguistic norms. A n d thus they are not so resistant to linguistic change.

REFERENCES P R I M A R Y SOURCES 1400-1449 portions read from Helsinki Corpus: An Anthology of Chancery English. 1984. ed. by J. H. Fisher, M . Richardson & J. L. Fisher. Knoxville: The University of Tennessee Press. A Book of London English 1384-1425. 1967. [1931]. ed. by R. W. Chambers and M . Daunt. Oxford: Clarendon Press. Early Middle English Texts. 1956. [1951]. ed. by B. Dickins and R. M . Wilson. London: Bowes & Bowes. Gaytrydge, Dan Jon. 1969. [1914]. Dan Jon Gaytrydge's Sermon. Religious Pieces in Prose and Verse. EETS O.S. 26 ed. by G. G. Perry. London: Oxford University Press. Kempe, Margery. 1940. The Book of Margery Kempe. V o l . I. EETS 212 ed. by S. B. Meech and H. E. Allen. London: Oxford University Press. Mandeville. 1919. Mandeville's Travels Translated from the French of Jean D'Outremeuse. V o l . I. EETS O.S. 153 ed. by P. Hamelius. London: Oxford University Press. Paston Letters and Papers of the Fifteenth Century. 1971. Part I ed. by N. Davis. Oxford: Clarendon Press. Rolle, Richard. 1952. The Bee and the Stork. A Handbook of Middle English ed. by F. Mosse. Translated by J. A . Walker. Baltimore: The Johns Hopkins Press. . 1921. [1866]. English Prose Treatises of Richard Rolle of Hampule. EETS O.S. 20 ed. by G. Perry. London: Oxford University Press. The Book of Vices and Virtues. A Fourteenth Century English Translation of The Somme Le Roi of Lorens D'Orleans. 1942. EETS 217 ed. by W. N. Francis. London: Oxford University Press. 1450-1499: portions read from Helsinki Corpus: Capgrave, John. 1983. John Capgrave's Abbreviation of Chronicles. EETS 285 ed. by P. J. Lucas. Oxford: Oxford University Press. . 1971. [1910]. John Capgrave's lives of St. Augustine and St. Gilbert of Sempringham and a Sermon. EETS O.S. 140 ed. by J. J. Munro. London: Oxford University Press.

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH

303

Caxton, W i l l i a m . 1956. [1928]. The Prologues and Epilogues. EETS 176 ed. by W . J. B. Crotch. London: Oxford University Press. . 1970. The History of Reynard the Fox. Translated from the Dutch Original by William Caxton. EETS 263 ed. by N. F. Blake. London: Oxford University Press. Cely, George. 1975. The Cely Letters 1472-1488. EETS 273 ed. by A . Hanham. London: Oxford University Press. The Early South-English Legendary or Lives of Saints. [The life of St. Edmund]. 1887. EETS O.S. 87 ed. by C. Horstmann. London: N. Trïbner & co. Fitzjames, Richard. 1907. Sermo die lune in Ebdomada Pasche. Westminster, Wynkyn de worde (1495?) ed. by F. Jenkinson. Cambridge: Cambridge University Press. Gregory, W i l l i a m . 1876. The historical collections of a citizen of London in the fifteenth century. Camden Society, N.S. X V I I ed. by J. Gairdner. Westminster. Hilton, Walter. 1967. Walter Hilton's Eight Chapters on Perfection ed. by F. Kuriyagawa. Tokyo: The Keyo Institute of Cultural and Linguistic Studies. Julian of Norwich. 1978. Julian of Norwich's Revelations of Divine Love. The Shorter Version ed. from B. L. A D D . MS 37790. Middle English Texts ed. by F. Beer. Heidelberg: Carl Winter. Malory, Thomas. 1954. The works of Sir Thomas Malory ed. by E. Vinaver. London: Oxford University Press. Methan, John. 1916. The Works of John Methan including the Romance of Amonyus and Cleopes. EETS O.S. 132 ed. by H. Craig. London: Kegan Paul, Trench Trïbner &co. Middle English Sermones edited from British Museum MS. Royal 18 B XXIII. 1940. EETS 209 ed. by W . O. Ross. London: Oxford University Press. Mirk, John. 1905. Mirk's Festial: a Collection of Homilies by Johannes Mirkus (John Mirk). Part I. EETS E.S. 96 ed. by T. Erbe. London: Kegan Paul, Trench Trïbner & co. Paston Letters and Papers of the Fifteenth Century. 1971. Part I ed. by N. Davis. Oxford: Clarendon Press. Reynes, Robert. 1980. The Commonplace Book of Robert Reynes of Acle. An Edition of Tanner MS 407. Garland Medieval Texts 1 ed. by C. Louis. New York and London: Garland. Two Sermons Preached by the Boy Bishop, at St. Paul's Temp. Henry VII, and at Gloucester, Temp. Mary. 1875. Camden Society Miscellany V I I . Camden Society N.S. X I V ed. by J. G. Nichols. London. Shillingford, John. 1965. [1871]. Letters and Papers of John Shillingford, Mayor of Exter 1447-1450. Camden Society N.S. I I ed. by S. A . Moore. New York. The Statutes of the Realm. Printed by command of his Majesty king George the third in pursuance of an address of the house of Commons of Great Britain. 1963. [1816]. V o l . I I . London: Dawson of Pall Mall. Stonor, Elizabeth. 1919. Stonor Letters and Papers, 1290-1483. V o l l . I-II. Camden Society Third Series X X I X - X X X ed. by C. L. Kingsford. London. Other sources: Paston Letters and Papers of the Fifteenth Oxford: Clarendon Press.

Century. 1971. Ed. by N. Davis, 1-20.

304

ANETTE ROSENBACH & LETIZIA VEZZOSI

1500-1559 portions read from Helsinki Corpus: Cromwell, Gregory. 1846. Original Letters, illustrative of English history; including numerous royal letters. Third Series, V o l . I ed. by H. Ellis. London: Richard Bentley. Cumberland, Eleanor. 1962. Clifford Letters of the Sixteenth Century. Surtees Society, C L X X I I ed. by A . G. Dickens. Durham and London. Edward V I . 1963. [1857]. The Diary of Edward VI. Literary remains of King Edward the Sixth, V o l . I I [Burt Franklin Research & Source Works Series, 51] ed. by J. G. Nichols. New York. Elyot, Thomas. 1907. [1531]. The Boke named the Gouernour. Everyman's Library ed. by E. Rhys. With an Introduction by F. Watson. London & New York: J. M . Dent & Co. and E. P. Dutton & Co. Leland, John. 1964. The Itenary of John Leland in or about the Years 1535-1543. Vol. I, Parts I to III ed. by L. T. Smith. London: Centaur Press Ltd. More, Thomas. 1846. Original Letters, illustrative of English history; including numerous royal letters. Third Series, V o l . I ed. by H. Ellis. London: Richard Bentley. Mowntayne, Thomas. 1859. The Autobiography of Thomas Mowntayne. Narratives of the Days of the Reformation, chiefly from the Manuscripts of John Foxe the Martyrologist. Camden Society, L X X V I I ed. by J. G. Nichols. London. Roper, Margaret. 1947. The Correspondence of Sir Thomas More ed. by E. F. Rogers. Princeton: Princeton University Press. Roper, W i l l i a m . 1958. [1935]. The Lyfe of Sir Thomas Moore, Knighte, written by William Roper, Esquire, whiche maried Margreat, Daughter of the sayed Thomas Moore. EETS, 197 ed. by E. V . Hitchcock. London: Oxford University Press. Scrope, Katherine. 1962. Clifford Letters of the Sixteenth Century. Surtees Society, C L X X I I ed. by A . G. Dickens. Durham and London. Torkington, Richard. 1884. Ye oldest Diarie of Englysshe Traveil: Being the hitherto unpublished narrative of the pilgrimage of Sir Richard Torkington to Jerusalem in 1517. The Vellum-Parchment Shilling Series of Miscellaneous Literature, V I ed. by W . J. Loftie. London: Field & Tuer, Ye Leadenhalle Presse, E.C., ETC. Other sources: Machyn, Henry. 1968 (repr.). The Diary of Henry Machyn ed. by John Gough Nichols. London/New York: A M S Press. 1-40. Medwall, Henry. Fulgens and Lucrece. I n : Five Pre-Shakespearean Comedies ed. by Frederick S. Boas, 1-72. Oxford: Oxford University Press. 1560-1630: portions read from Helsinki Corpus: Ascham, Roger. 1870. [1570, 1571]. The Scholemaster. Written between 1563-8. Posthumously published. 1st ed., 1570; collated with the 2nd ed., 1571. English reprints ed. by E. Arber. London.

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH

305

Bacon, Francis. 1970. [1605]. The Twoo Bookes of the Proficience and Advancement of Learning (1605). English Experience, 218. Amsterdam: Theatrvm Orbis Terrarvm Ltd. and New York: Da Capo Press. Facsimile. Barrington Family Letters, 1628-1632. 1983. Camden Fourth Series, 28. ed. by A . Searle. London. The Correspondence of Lady Katherine Paston, 1603-1627. 1941. Norfolk Record Society, X I V . ed. by R. Hughey. Norwich: Norfolk Record Society. Forman, Simon. 1849. The Autobiography and Personal Diary of Dr. Simon Forman, the celebrated Astrologer, from A.D. 1552, to A.D. 1602 ed. by J. O. Halliwell. London: Privately printed. Gifford, George. 1931. [1593]. A Handbook on Witches and Witchcraft. A Dialogue concerning Witches and Witchcraftes. 1593. Shakespeare Association Facsimiles, 1. With an Introduction by B. White. London: Humphrey Milford and Oxford University Press. Harman, Thomas. 1937. [1869, 1898]. A Caveat of Warning for Commen Cursetors Vulgarely called Vagabones. Collated with the 2 nd edition of 1567 in the Bodleian Library, Oxford, and with the reprint of the 4 th edition of 1573. EETS, E.S. 9 ed. by E. Viles & F. Funrivall. London. Hoby, Margaret. 1930. Diary of Lady Margaret Hoby, 1599-1605 ed. by D. M . Meads. London: George Routledge & Sons, Ltd. Letters of Philip Gawdy of West Harling, Norfolk, and of London to various Members of his Family, 1579-1616.1906 ed. by I. H. Jeayes. London: J. B. Nichols and Sons. Madox, Richard. 1976. An Elizabethan in 1582: The Diary of Richard Madox. Fellow of All Souls ed. by E. S. Donno. London: Hakluyt Society. Markham, Gervase. 1973. Countrey Contentments, 1615. The English Experience, 613. Amsterdam: Theatrvm Orbis Terrarvm Ltd. and New York: Da Capo Press Inc. Facsimile. Plumpton Correspondence. A Series of Letters, chiefly domestick, written in the reigns of Edward TV. Richard III. Henry VII. and Henry VIII. 1839. Camden Society, I V ed. by T. Stapleton. London. Stow, John. 1580. The Chronicles of England from Brute unto this present Year e of Christ. London: printed by Ralphe Newberie. Other sources: Shakespeare, W i l l i a m . 1989. The Merry Wives of Windsor ed. by T. W . Craik. Oxford: Clarendon Press. . 1917. The tragedy of Hamlet, Prince of Denmark ed. by Tucker Brook & Jack Randal Crawford. New Haven: Yale University Press and London: Geoffrey Cumberlege and Oxford University Press.

306

ANETTE ROSENBACH & LETIZIA VEZZOSI

SECONDARY SOURCES Allen, Cynthia. 1997. "The Origins of the 'Group Genitive' in English". Transactions of the Philological Society 95.111-131. . 1998. "Genitives and the Creolization Question". English Language and Linguistics 2.129-134. Altenberg, Bengt. 1982. The Genitive vs. the of-construction. Malmö: Gleerup. Breejen, Bastiaan den. 1937. The Genitive Case and its 'Of'-Equivalent in the Latter Half of the 16th Century. Amsterdam dissertation. Dressier, Wolfgang U., W i l l i Mayerthaler, Oswald Panagl & Wolfgang U. Wurzel, eds. 1987.. Leitmotif s in Natural Morphology. Amsterdam: John Benjamins. Firbas, Jan. 1964. "On Defining the Theme and Functional Sentence Analysis". Travaux linguistique de Prague 1.267-280. Fischer, Olga. 1992. "Syntax". The Cambridge History of the English Language. Volume II: 1066-1476 ed. by N. Blake, 207-408. Cambridge: Cambridge University Press. Fries, Charles C. 1940. American English Grammar. New York: Appleton-CenturyCroft. Gaaf, W . van der. 1932. "The Absolute Genitive". English Studies 14.49-65. Givón, Talmy. 1989. Syntax. Amsterdam: John Benjamins. Gundel, Jeanette K. 1986. "Universals of Topic-Comment Structure". Studies in Syntactic Typology ed. by M . Hammond et al, 209-239. Amsterdam & Philadelphia: John Benjamins. Janda, Richard D. 1980. "On the Decline of Declensional Systems: the Overall Loss of OE Nominal Case Inflections and the M E Reanalysis of -es as his"'. Papers from the 4th International Conference on Historical Linguistics ed. by Elizabeth Closs Traugott, Rebecca Labrum & Susan Shepherd, 243-252. Amsterdam: John Benjamins. Klemola, Juhani. 1997. "Dialect Evidence for the Loss of Genitive Inflection in English". English Language and Linguistics 1:2.350-353. Mustanoja, Tauno. 1960. A Middle English Syntax. Helsinki: Société Néophilologique. Quirk, Randolph, Sidney Greenbaum, Jeoffrey Leech & Jan Svartvik. 1972. A Grammar of Contemporary English. London: Longman. Seppänen, A i m o . 1997a. "The Genitive and the Category of Case in the History of English". Language History and Linguistic Modelling. A Festschrift for Jacek Fisiak on his 60th Birthday ed. by Raymond Hickey & Stanislaw Puppel, v o l . I, 193-214. Berlin: Mouton de Grayter. . 1997b. "The Genitives of the Relative Pronouns in Present-Day English". Taming the Vernacular: from dialect to written standard language ed. by Jenny Cheshire & Dieter Stein, 152-169. Cambridge: Cambridge University Press.. Silverstein, Michael. 1976. "Hierarchy of Features and Ergativity". Grammatical Categories in Australian Languages ed. by R. M . W . Dixon, 112-171. Canberra: Australian Institute of Aboriginal Studies. Stahl, Leon. 1927. "Der adnominale Genitiv und sein Ersatz im Mittelenglischen und Frühneuenglischen". Giessener Beitrage 3.1-35.

GENITIVE CONSTRUCTIONS IN EARLY MODERN ENGLISH

307

Taylor, John. 1989. [1995]. Linguistic Categorization: Prototypes in Linguistic Theory. 2 nd edition. Oxford: Oxford University Press. . 1996. Possessives in English: An Exploration in Cognitive Grammar. Oxford: Clarendon Press. Thomas, Russell. 1931. Syntactical Processes Involved in the Development of the Adnominal Periphrastic Genitive in the English Language. Ann Arbor: University of Michigan. Unpublished Ph.D. dissertation.

INDEX OF NAMES

A. Abesa3e, Nia; 159 Abraham, Werner; 277; 281 Adams, J. N.; 112-114; 236; 240; 255; 256 Aissen, Judith; 154; 159 Aitchison, Jean; 152; 159 Alam, Yukiko Sasaki; 165; 169; 170; 176; 1 8 1 ; 183 Allen, Cynthia L.; 263; 264; 281; 288; 289;305 Altenberg, Bengt.; 288; 289; 297; 305 Alvarez, Rosario; 252; 253; 255 Andersen, Paul Kent; 134; 145; 149; 160 Anderson, Stephen; 187; 188; 201; 233; 255 Appleton, George; 172; 183 Apridoni3e, Sukia; 141; 160 Aristar, Anthony Rodrigues; 63; 69; 133; 134; 160 Asatiani, Irine; 148; 160 Asher, R. E.; 255; 258 Axvlediani, G.; 160; 201 B. Bahner, Werner; 255; 256 Barbato, M.; 104; 107; 110; 115 Barlow, Micheal; 220; 228 Barnet, Sylvan; 275; 281 Basilaia,N.; 142; 143; 160 Behlmer, Heike; 24-30; 33-38 Benoist, J.-P.; 9; 18; 21 Benveniste, Emile; 180; 183 Bergel'son, M. B.;21 Bergh, Lars; 108; 114 Bernini, G.; 91; 98-100; 115 Besten, Hans den; 281 Beukema, Frits; 272; 281; 282 Biber, Douglas; 279; 281 Blake, Barry J.; 44; 133 Blake, Robert P.; 135; 159; 160; 200 Boeder, Winfried; 191; 192; 201 Bonnot, C ; 18; 21 Borsley, Robert D.; 73; 98 Boulanger, Jean-Claude; 255

Braunmüller, Kurt.; 122; 131 Breejen, Bastiaan den.; 288; 289; 293; 297; 306 Brière, Maurice; 135; 159; 160 Brinton, Laurel J.; 275; 281 Brixhe, Claude; 255; 256 Broekhuis, Hans; 271; 281 Brunot, Fernand; 106; 108; 114 Buridant, Claude; 106; 114 C. C'ereteli, G.; 160 Cagareli, Aleksandre; 189; 201 Cameron, Deborah; 218; 228 Campbell, Alistair; 264; 281 Campbell, Lyle; 59; 69; 120; 134; 139; 1 4 1 ; 1 4 8 ; 154 Cassidy, Frederic G.; 262; 264; 281 Cesarano, M.; 104; 107; 110; 115 Chafe, Wallace L.; 31; 38; 46; 49 Champollion Ie Jeune; 71 Chantraine, Pierre; 236; 255 Chase, Colin; 268; 281 Cherchi, Marcello; 190; 191; 193; 201 Chincharauli, A l ; 191; 201 Chomsky, Noam; 120; 131; 204; 228 Cikobava, Arnold; 143; 148; 159; 160 Christoffersen, Marit; 124; 131 Collinge,N. E.; 233; 255 Comrie, Bernard; 231; 255 Cooreman, Anne; 219; 228 Coseriu, Eugenio; 225; 228 Crochetière, André; 255; 256 Croft, William; 24-27; 34; 38; 306 Crum, Walter E.; 36; 38; 39 Currie, Oliver; 43; 49; 217; 224; 228 Cxubianisvili, D.; 144; 160

D. Dahl, O.; 21 Danes, F.; 10; 11; 13; 21 Davies, Sioned; 48; 49; 207; 216; 221; 226;227

INDEX OF NAMES

310

Davydov, 1.1.; 8; 21 Dawkins, R. M.; 251; 255; 257 De Boer, Charles; 74; 98 Deeters, Gerhard; 188; 189; 195; 196; 198; 201 Denison, David; 266; 267; 281 Depuydt, Leo; 32; 38 Devine, A. M ; 246; 255 Dik, Simon C ; 47; 49 Dirr, Adolph; 157; 160 Dover, Kenneth J.; 236; 241; 242; 256 Drachman, Gaberell; 256 Dressier, Wolfgang U.; 291; 306 Dryer, Matthew S.; 57; 67; 69; 133; 146; 147; 160 Dunn, Graham; 233; 234; 256 Duszak, Anna; 45; 49 Dzidziguri, Shota; 201; 202 E. Ebeling, C. L.;21;69 Eckert, Penelope; 218; 228 Edel, Doris; 48; 49 Eichner, Heiner; 233; 256; 257 Enukise, L.; 144; 160 Epstein, Richard; 90; 98 Epstein, Samuel David; 281 Erckert, Roderich von; 189; 201 Erckert, Roderich von.; 189 Erdmann, Peter; 47; 49 Erman,Adolf;73;98 Evans, D. Simon; 49; 212; 224; 226; 229 F. Faarlund, Jan Terje; 122; 132; 165; 167; 170; 183 Fähnrich, Heinz; 141; 160; 201 Feydit, Frederic; 153; 160 Fife, James; 43; 45; 49; 50; 229 Fillmore, Charles J.; 204; 229 Firbas, Jan; 11; 12; 21; 300; 306 Fischer, Olga; 293; 306 Fleischman, Suzanne; 75; 92; 98 Fortunatov, F. F.; 11; 21 Foster, Tony; 226; 227; 270; 274; 281; 283 Fougeron, I.; 18; 21 Foulet, Lucien; 105; 106; 108; 114

Friedrich, Johannes; 165; 183 Friedrich, Paul; 152; 160 Fries, Charles C ; 287; 306 Fykias, Jannis; 256 G. Gaaf, W. van der; 296; 306 Gardiner, Alan Henderson; 73; 98 Givón, Talmy; 26; 27; 32; 38; 133; 160; 211; 229; 299; 306 Glinert, Lewis; 28; 38 Goebel, Hans; 97; 98 Goetze, Albrecht; 171; 183 Gonda,Jan; 113; 114 Grech,N. I.; 8; 21 Greenbaum, Sidney; 183; 306 Greenberg, Joseph H.; 120; 132-134; 161 Griffith, Frank Lloyd; 39; 73; 98; 207; 215; 216; 221; 227; 228 Giildenstädt, Johann Anton; 189; 201 Gundel, Jeanette K.; 300; 306 Gunn, Battiscombe; 73; 98 H. Haegeman, Liliane; 120; 132 Haider, Hubert; 134; 161 Halpern, Aaron L.; 256; 258 Harris, Alice C ; 59; 69; 120; 132; 134; 139; 141; 144; 148; 153; 154; 1 6 1 ; 1 9 4 ; 195;201 Haugan,Jens; 124; 128; 132 Hawkins, John A.; 63; 64; 67; 69; 74; 98; 1 2 0 ; 1 3 2 ; 1 3 3 ; 161 Healey, Antonette diPaolo; 280; 281 Heine, Bernd; 133; 134; 161; 181; 183; 187; 285 Herman, József; 107; 108; 113; 114 Herzog, Marvin L; 203; 229 Hewitt, B.G.; 196; 201 Hickey, Raymond; 98; 100; 283; 306 Hilty, Gerold; 104; 114 Hintze, Fritz; 24; 38 Hock, Hans Heinrich; 26; 35; 38; 233; 256 Hoekstra, Kees; 281 Hofmann, J. B.; 132 Hogg, Richard M.; 264; 281 Holes, Clive; 26; 38

311

INDEX OF NAMES

Holisky, Dee Ann; 148; 161 Hook, Peter Edwin; 181; 183 Hopper, Paul J.; 24; 26; 38; 79; 90; 99; 1 2 2 ; 1 3 2 ; 1 6 6 ; 1 7 4 ; 183;235;256 Horrocks, Geoffrey; 257 Householder, Fred W., Jr.; 155; 161 Hyman, Larry M.; 122; 132 I. Imnaisvili, I.; 135; 136; 139; 146; 161 Isaac, Graham H.; 43-45; 49; 80; 99 Isachenko, A.; 21 Ishiguro, Teruhiro; 168; 183 J. Jakobson, Roman; 12; 21; 233; 234; 250; 256 Jamieson, Stephanie W.; 80; 99 Janda, Richard D.; 291; 306 Janse, Mark; 233-237; 239; 240; 246; 247; 249; 250; 251; 256 Jikia, Sergi; 152; 161; 162 Johnstone, Barbara; 218; 229 Joseph, Brian D.; 245; 257 Junge, Friedrich; 23; 38; 72; 99 Justus, Carol F.; 172-174; 182; 183 K. K'art'ozia, Guram; 144; 148; 162 Kaldani, Maxime; 145; 162; 200 Kamrnerzell, Frank; 30; 31; 38 Karamzin, N. M.; 8-10; 13; 20; 21 Kasser, Rodolphe; 23; 38; 100 Kay, Paul; 228 Keenan, Edward L.; 65; 69 Kemmer, Suzanne; 220; 228 Kibrik,A. E.;21 Kindt, Walther; 76; 99 King Gareth; 45; 46; 49; 87; 158; 159; 1 7 3 ; 174;273;304 Kiparsky; 187 Kipshidze, I.; 193; 194; 195; 202 Kirk, John M.; 265; 281; 282 Klaproth, Jules; 189; 201; 202 Klavans, Judith L.; 233; 257 Kleiber, Georges; 45; 50

Klemola, Juhani; 288; 306 Klidi, Sila; 256 Koch, Roland; 24; 25; 38 Kogian, Fr. S.L.; 153; 156; 162 Kohonen, Viijo; 262; 282 Koopman, Willem F.; 262; 263; 265; 267; 269;271; 282 Kortmann, Bernd; 79; 99 Kovtunova, I.1.; 9; 11; 17; 18; 21 Krisch, Thomas; 236; 257 Kristoffersen, Kristian E.; 124; 132 Kroch, Anthony; 260; 268; 269; 276; 277279;280;282 Krushel'nickaja, K. G.; 11; 21 Kuno, Susumu; 147; 162; 168; 169; 183; 262; 282 Kytö, Merja;265;282 L. Labov, William; 203; 228; 229 Langacker, Ronald W.; 204; 229 Lapesa, Rafael; 109; 114 Lausberg, Heinrich; 191; 202 Lavandera, Beatriz; 203; 229 Layton, Bentley; 73; 99 Lebrun, Réné; 170; 171; 183 Leech Geoffrey; 183; 273; 282; 306 Lehmann, Winfred P.; 99; 122; 132; 165; 182; 183 Leumann, Manu; 122; 132; 235; 241; 257 Lieber, Rochelle; 188; 202 Lightfoot, David W.; 120; 132; 182; 183; 276-280; 282 Linde, P.; 112; 114 Lomonosov, M. V.; 4; 5; 8; 13; 20; 21 Lomouri, Tamar; 160 Loprieno, Antonio; 23; 25; 34; 39; 71; 73; 99; 111 Lord, Carol; 134; Los, Bettelou;271;282 Lotfi, Mansour; 155; 161 Lunt, Horace G.; 115; 163; 257; 258 M. Malikouti-Drachman, Angeliki; 256 Mallinson, Graham; 44; 50; 133; 162 Marchello-Nizia, Christiane; 106; 114

INDEX OF NAMES

312

Marouzeau, Jules; 107; 114 Marshall, M. H. B.; 236-239; 242; 247; 257 Mart'irosovi, Ar.; 146; 147; 162 Martin, Robert; 114 Matesius,V.; 11; 12; 21 Mauss, Marcel; 180; 183 Mayerthaler, Willi; 306 Meillet, Antoine; 232; 257 Mendes da Luz, Maria Albertina; 251-253; 258 Migliorini, Bruno; 256; 257 Milroy, James.; 229 Milroy, Lesley; 229 Misteli, Franz; 72; 100 Mitchell, Bruce; 262; 264; 282 Mithun, Marianne; 79; 99 Moerenhout, Mike; 272; 282 Molinelli, P.; 100 Monteagudo, Henrique; 255 Moorhouse, A.C.; 240; 257 Morenz, Siegfried; 29; 32; 35; 39 Mouraviev, Serge N.; 153; 162 Murav'ev, Serge N.; 162 Mussafia; 245; 248; 257 Mustanoja, Tauno; 287-289; 291; 293; 296; 297; 306 Myhill, John; 219; 229 N. Nadareisvili, L.; 146; 162 Nagel, Peter; 30; 31; 39; 100 Nakau, Minoru; 168; 183 Nevis, Joel A.; 257; 258 Newman, John; 180; 181; 183 Nicolaidis, Katerina; 257 Nikolaeva, T. M.; 18; 19; 21 O. O'Connor, Catherine; 228 Ogura, Michiko; 262; 282 Ouellon, Conrad; 255 P. Panagl, Oswald; 306 Panayotakis, N.; 257

Panhuis, Dirk G.; 112; 114 Pancvi3e, V l ; 157; 162 Parkinson, Richard, B.; 23; 24; 39 Patsch, Gertrud; 135; 158; 159; 162 Philippaki-Warburton, Irene; 256; 257 Pinkster, Harm; 112; 114 Pintzuk, Susan; 268; 269; 277; 282 Pisani, Vittore; 256; 257 Plank, Frans; 137; 162 Polotsky, Hans Jacob; 23-25; 28; 31; 33; 34; 37; 39; 71; 73; 74; 83; 93; 99; 100 Poppe, Erich; 43; 45; 47-50; 111; 187; 206;215; 222; 223; 229 Počxua,B.; 141; 162 Price, Glanville; 105; 108; 114 Puolato, D.; 104; 107; 110; 115 Puppel, Stanislaw; 98; 100; 283; 306

Q. Quakenbusch, Stephen.; 219; 229 Quirk, Randolph; 165; 166; 174; 183; 297; 306 R. Radermacher, Ludwig; 241; 257 Ramat,P.;99; 100; 181 Regueira, Xosé Luis; 252; 253; 255 Reh, Mechtild; 133; 134; 161 Reich, Nathaniel; 39 Reinhart, Tanya; 43; 44; 50 Reintges, Chris; 23; 30; 39 Richards, Melville; 42; 49; 50 Richter, Elise; 113; 114 Rix, Helmut; 256; 257 Roberts, Brynley F.; 50; 225 Roberts, Ian; 278; 282 Rollo, Antonio; 245; 257 Romaine, Suzanne; 218; 229 Rosen, Georg; 189; 202 Rubiera Tuya, Carlos; 252; 253; 258 Ruelle, Pierre.; 105; 110; 115 Ruijgh, C. J.; 234; 236; 242; 257 Rutten, Jean; 281 S. Sadock, Jerrold M.; 131; 132; 188; 202

313

INDEX OF NAMES

Sáenz-Badillos, Angel; 30; 39 Salvi, Giampaolo; 104; 115; 236; 240; 245;252;253;257 Sánchez Vicente, Xuan Xosé; 252; 253; 258 Šani3e, Ak'ak'i; 141; 162 Santorini, Beatrice; 271; 277; 279; 282 Saussure Ferdinand de; 78; 255 Scaglione, Aldo; 104; 115 Schenkel, Wolfgang; 100 Schiffman, Harold F.; 278; 282 Schildt, Joachim; 255 Schmidt, Annette; 225; 229 Schmidt, Karl Horst; 156; 163; 202 Schwyzer, Eduard; 235; 241; 258 Seppänen, Aimo; 288; 289; 306 Šerozia, Revaz; 143; 163 Sethe, Kurt; 28; 39; 73; 94; 100 Sgall,P.;ll;21 Shisha-Halevy, Ariel; 23; 26; 28; 33; 34; 36; 39; 45; 50; 73; 76; 80; 81; 84; 88; 89; 91; 93; 94; 97; 100; 111; 187 Siewierska, Anna; 204; 229 Sifianou, Maria; 257 Silverstein, Michael; 291; 306 Singh, Rajendra; 188; 202 Sirotinina, O. B.; 11; 22 Slanskij,V.; 11 ; 22 Slotkin, Edgar M.; 48; 50 Soden, Wolfram von; 58; 69 Sornicola, Rosanna; 41; 45; 47; 50; 101; 103; 104; 107; 1 1 0 ; 1 1 1 ; 1 1 5 ; 1 8 7 Speas, Margaret; 267; 268; 274; 283 Spencer, Andrew; 187; 188; 202; 233; 258 Spiegelberg, Wilhelm; 73; 100 Stahl, Leon; 288; 297; 306 Stassen, Leon; 149; 163 Steinthal, Haymann; 8; 72; 100 Stephens, Laurence D.; 246; 255 Stern, Ludwig; 73; 100 Stubbs,Marcia;275;281 Svartvik, Jan; 183; 306 Swain, Simon; 255 Szantyr, Anton; 132

T. Taylor, Ann; 228; 269; 277; 281-283; 291; 294;306 Thomas, Peter Wynn; 229 Thomas, Russell; 288; 306 Thráinsson, Höskuldur; 281 Thumb, Albert; 248; 258 Till, Walter C ; 23; 34; 39 Tischler, Johann; 175; 184 Tobler; 245; 257 Tomlin, Russell S.; 99; 100 Topuria,V.; 160; 200 Traugott, Elizabeth Closs; 24; 26; 38; 166; 1 7 4 ; 1 8 3 ; 235; 254; 256; 258; 306 Trubetzkoy, Nikolaj; 188; 202 Tschenkéli, Kita; 141; 155; 163 Tuite, Kevin; 190; 196-198; 200; 202 U. Uhlig, Gustav; 235; 258 Underhill, Robert; 154; 155; 163 V. Vanelli, Laura; 257; 258 Vázquez Cuesta, Pilar; 251; 252; 253; 258 Vendryes, Joseph; 232; 257 Venezky, Richard L.; 280; 281 Vennemann, Theo; 73; 100; 120; 132 Veyrenc, J.; 22 Viehweger, Dieter; 255 Vinogradov, V. V.; 5; 22 Vittmann, Günther; 23; 39 Vogt, Hans; 144; 163 W. Wackernagel, Jacob; 97; 193; 201; 232234;236;240; 241; 244; 245; 252; 254258 Walther, G.; 22; 99 Wanner, Dieter; 233; 234; 245; 246; 257; 258 Watkins, Calvert; 113; 115; 152; 163; 232; 258

314

INDEX OF NAMES

Watkins, T. Arwyn; 41; 42; 43; 50; 51; 152; 233 Weil, Henri; 8; 234; 258 Weinreich, Uriel; 203; 230 Weiss, Hans-Friedric; 23; 39 Westendorf Wolfhart; 28; 39 Wifstrand, Albert; 237; 238; 258 Williams, Ifor; 51 Willis, David W. E.; 42; 46; 51; 219; 222; 224; 230 Wilmet, Marc; 114

Winand,Jean.;91; 100 Wurff, Wim van der; 270-272; 274; 278; 281-283 Wurzel, Wolfgang U.; 306 Z. Zakszewska, Ewa D.; 39 Zamboni, Alberto; 257; 258 Zwanziger, Ronald; 134; 161 Zwart, Jan-Wouter; 281 Zwicky, Arnold M.; 231; 233; 256-258

I N D E X O F SUBJECTS

A. adjective; 137; 141; 149-151 adposition; 62; 134; 135; 136; 141; 149; 162 adverb adverb-rheme nexus; 82; 86; 91 adverbial adverbial clauses; 26; 63; 66 adverbial phrase; 33; 41; 206; 214 agentive hearer; 179 ambiposition; 135; 141 analytic features; 24 analytic pattern; 23; 287 autonomy; 76; 255 auxiliary; 24; 45; 80; 85; 86; 90-92; 133; 134; 1 4 0 ; 145-149; 166; 168; 1 6 9 ; 1 7 2 176; 178-182; 217; 223; 270-272 B. binding relations; 131 Book of Aneirin; 43; 49; 99 borrowing; 23; 56; 63; 122; 152; 154; 156; 180 C. case case marking; 124; 127; 129; 137; 138; 187; 291 case system; 55; 58-60; 62; 68 case agreement; 287 cataphoric; 31; 109; 178 change abruptness of change; 259-260 circumposition; 136 circumstantial; 25; 71; 84; 85 cleft-construction; 45; 47 cleft-sentence; 42 clitic; 24; 121; 193; 231; 232-235; 237258; 288 enclitic; 42; 91; 97; 146; 178; 237; 239; 246; 247; 250 proclitic; 123; 235; 245-248; 251; 254 collocation; 236; 256; 271 comment; 9; 10; 33; 43; 45; 119

communicative salience; 32 comparative construction; 133; 139; 149; 1 5 0 ; 1 5 1 ; 1 5 4 ; 155; 157 complementizer; 67; 140; 159 complex predicates; 167; 169; 175; 176 Construction Grammar; 204 construction reanalysis; 133; 134; 157 contrastive stress; 110 copula; 95-97; 125; 128; 255; 292 corpus analysis; 288 correlative pronoun; 144 D. Dafydd ap Gwilym; 43; 207; 225 dative speaker; 176-178 deep structure; 120; 123 demonstrative anaphoric demonstrative; 105 dependants; 62 diachrony; 75; 77; 85; 91; 94; 104 dialects; 26; 55; 72; 75; 81; 97; 126; 152; 190; 231; 232; 247; 248; 251; 254; 257; 269; 288; 289 diglossia; 276- 279 directive; 165; 166; 170; 173; 174; 177; 180-182 deictic directives; 182 discourse function; 99; 119; 121 ditransitive verbs; 261 double base hypothesis; 276; 277 dummy adverb; 27 dummy subject; 208; 211-221; 224 dyad; 133-135; 146; 151-153; 157 dynamic models; 103 E. echo structures; 249; 253 embedding; 25; 67; 148; 151 emphasis; 32; 42; 68; 104; 113; 245 ergative verb; 125 expletive subject; 126; 128; 129 external factors; 111; 112 extraposition; 23; 28; 30; 32; 33-37; 124; 271 extraposed subject; 31; 36; 126; 127

INDEX OF SUBJECTS

316 F.

finite sentential complements; 57; 66 first person imperative; 165; 166; 170; 171; 172; 174; 180; 181 flexibility of structure; 26, 37, 56, 59, 61, 65. focalization; 25; 105; 109 focus; 33; 34; 36; 37; 45-48; 81; 84; 85; 93; 106; 1 0 7 ; 108-111; 151; 203; 212; 219; 220; 235; 241; 244; 255; 260; 288 Functional Grammar; 49; 204 future tense; 148 G. garden path; 64 genitive genitive case-marking; 286 genitive construction; 139; 285; 287290;292; 294; 296 genitive functions; 285; 286; 289; 291; 292;294;297 of-genitive; 285-291; 296; 297-299; 301 s-genitive; 285-291; 296-301 s-less genitive; 288; 289 to-genitive; 289 zero-genitive; 289 GIVEN; 44; 300 graduality; 260 grammar competition; 260; 276-279 grammatical categories; 91; 92; 182; 200 grammatical function; 168 grammaticalization; 20; 24; 28; 34; 38; 63; 95; 161; 165; 166; 181; 183; 205; 223; 232; 235; 254-256; 258 grammaticization; 188; 198; 200 gwneuthur-periphrasis;45 H. harmony of order; 62; 154; 194 Helsinki Corpus; 265; 266; 282; 288; 289; 302;303;304;305 head; 27, 31-33, 38, 62, 64-65, 101-102, 135-139, 143-146, 154, 158, 168, 177, 197, 277, 286-287, 289, 292, 294, 296, 300.

hierarchy of animacy; 27 hierarchy of topicality; 31 highlighting; 86; 87; 90; 135; 136; 139; 1 4 0 ; 1 4 1 ; 1 4 4 ; 150;212; 255 historical models; 102 I. identity-definition; 74 idioms; 217 imperative imperative particle; 166 imperative suffix; 172; 174 person-differentiated imperatives; 166; 170 indefinite articles; 90 indefinite NP; 126 infinitive; 24; 92; 95; 138; 143; 160; 175; 176-179; 234; 254 information structure; 31; 80; 119; 125; 300 informational weight; 35; 38 intransitive verb; 107; 126; 127 interference linguistic interference; 30 inversion of S; 104, 108, 111 L. language acquisition; 131 M. microscopic approach; 102 minimalism; 204 Mischsprache; 73; 74 modality; 75; 177; 178; 180; 181; 193 modifier modifying constituent; 101 modularity; 119; 131 morphological case; 127 movement rule; 124 multilingualism; 279 multistratal theory; 120 N. nature and locus of competition between two grammatical systems; 278 nci-construction; 29; 30; 31; 32; 34; 36

INDEX OF SUBJECTS

negation; 87; 90; 91; 130; 189 NEW; 44, 105, 119, 121, 125,300. nexus; 72; 79-86; 91; 96 nominalization; 26; 37 nominative ease marking; 127 nominative NP; 124 non-specific subject; 27; 31; 37 noun phrase; 89; 98; 175; 243. see also NP nuclearity of the determinators; 89 O. Object; 28; 36; 37; 41; 42; 44-45; 56-57; 59-62; 65; 112; 119; 120-121; 123; 125-126; 128-131; 133-134; 143; 147148; 154; 168-169; 176; 178; 206; 223; 244-246; 256-257; 259-262; 264; 266272; 274; 276; 279; 292-293 object pronoun suffixes; 57 oblique; 159 operators; 62 optionality in grammar; 277 order abnormal order; 42 basic order; 43; 103; 112; 120; 134; 245; 268 dominant order; 103-104; 111-112; 152; 215 free word order; 60; 124; 152 linear order; 119; 125; 189 order of the direct object vis-à-vis the verb; 259 ordering of direct object (DO) and indirect object (IO); 259 OSV; 104 OVS; 104; 107; 110 SOV; 26; 59; 61; 106; 107; 110; 112113; 132; 141; 146-148; 152-153; 205; 283 SV; 26; 31-33; 37-38; 78-80; 84; 86; 91; 99; 196; 197; 214; 219-221; 225 SVO; 23-24; 26; 28; 30; 37; 59; 79; 104 ; 107; 110; 112; 132; 141; 147; 152-153; 170; 205; 266; 283 VO; 26; 55-58; 62-63; 120-123; 133; 135; 141; 146; 147-148; 151-153; 157; 260; 264; 267-269; 271; 276277; 279; 280; 282

317

VOS; 107-108 VSO; 23-24; 26-28; 30; 42-43; 49; 56; 74; 79; 107-110; 152; 205-206; 222 OV language; 62; 120; 123 P. parallel subject; 32 participial relative; 144 particle imperfectivizing particle; 193 tmetic partiele; 189, 191-192; 198-199 passive verb; 125; 128 pattern; 28; 31; 36-37; 41-42; 45; 48; 61; 72; 74; 78-79; 81; 84; 86; 93; 95; 9798; 1 0 5 ; 1 0 7 ; 110-111; 120; 1 2 2 ; 1 2 7 ; 182; 206; 207; 209; 211; 213-214; 215; 219-220; 224; 250; 260; 262-263; 265266; 269; 271-274; 276 periphrastic SVO-pattems; 24 poetic texts; 56 possession; 27; 37; 88; 90; 97; 137; 292; 294 postposition; 7; 12; 133-134; 136; 141; 151; 154-156 postverbal NP; 128; 129; 130; 131 postverbal subject; 128 pragmatic factors pragmatic function; 33; 47; 203; 300 prayer request / petition; 172-175; 179 preposition; 37; 82-83; 88; 133; 136 prepositional phrase; 27; 109; 285 present participle; 138 presupposition; 37; 78-79 preverb; 189 Principles and Parameters; 204-205; 222 processing; 62-68 prosodie length; 35 prosodically long subject; 29

Q. quantitative data. See also corpus analysis quasi-auxiliary; 172; 175

INDEX OF SUBJECTS

318 R.

reanalysis; 63; 119-120; 123-124; 127131; 134; 147; 151; 205; 235-236; 246; 247-248; 254 reciprocity; 169 reflexive verb; 125 relative clause; 63-65; 67; 133-134; 139; 144; 146; 154; 239; 242; 271 relative pronoun; 139; 144-145; 231; 235236; 239 rhematization; 32; 36 rheme; 25; 30; 32; 34-36; 38; 43; 80; 8183; 86; 91; 93; 95-97; 119 rightward movement; 268 S. second tense; 25; 33; 36-37 semantic controller; 167 semantic path of grammaticalization; 181 semantic role; 24; 128; 131 sentence conjugation; 24 sentential phrase; 63 sociolinguistic(s); 203, 228-229; 282 speaker beneficiary; 165; 168; 170; 174 spoken registers; 110; 112 stylistic variation; 113, 203-204, 218, 221, 224-225 subject pronoun; 28; 241-243; 249; 253 subordinative; 64-65; 68 suffix conjugation; 27-28; 37 surface manifestation; 120 synchronic variation; 68; 188 Syntactic; 51; 69; 132; 161; 202; 229; 281; 282;306 syntax ordinary syntax; 231 special syntax; 231; 237 'syntax-to-morphology' agglutination process; 92 synthetic; 23; 85; 91; 212; 287 T. teleology; 131 Tempuslehre; 76; 84 tense marker; 24; 26; 31-33 text-linguistics; 75

that-form; 25; 92 thematic continuity; 32; 108; 110; 211 thematic element; 82; 105 thematic function; 32 theme; 25; 31-32; 35; 38; 43; 77; 80-81; 83; 85-86; 91; 9 5 - 9 7 ; 1 1 9 tmesis; 191; 192; 193; 198; 199. see also particle. Tobler-Mussafia Law; 245; 248 top-down; 204-205; 221-222 topic topic continuity; 44; 203; 211; 219-220 topic position; 27; 45-46 topic shift; 219-220 topicalization; 25; 33; 61-62; 65; 68; 8586; 111; 129 transitive verb; 83-84; 101; 104; 108-109; 111-112; 121; 125; 127-129; 154; 166; 211; 219 transitive verb; 168 typological correlations; 102-103 typological inconsistency; 55-56; 58; 6263 typological models; 102 U. underlying order; 124 underlying structure; 46; 120 Universal Grammar; 69; 163; 222 universals; 38; 69; 98; 132; 160-161; 306 unmarked order; 24; 121; 135-137; 142; 144; 145-146; 150; 157; 238; 249 V. V-2 (languages); 107; 266 variation; 12; 47-48; 66; 77; 101-103; 112113; 139-140; 182; 187-188; 190; 193; 195; 198; 203-207; 209-211; 213-215; 217-221; 223-224; 238; 240; 245; 251; 254; 276-279; 299 verb agreement; 124; 131; 212 verb of existence; 27 verb second language; 124 verbal agreement; 60; 68 verbal nexus; 79; 81; 84; 91 Verbal nexus; 83 verbal phrase; 31; 33; 37; 45; 239

INDEX OF SUBJECTS

verb-final; 56-59; 63; 67; 170; 174; 181182; 266; 277 verb-initial order; 46; 206; 211; 222

319 W.

Wackernagel's Law; 97; 2

INDEX OF LANGUAGES

A. Akkadian; 34; 55; 56; 63; 68; see also Babylonian; Assyrian Albanian; 153; 157 Amharic; 73 Ancient Nordic; 122 Arabic; 23; 26; 34; 38; 72; 73; 98; 109 Aramaic; 55; 58; 237 Armenian; 133; 138; 152-154; 156; 162; Old Armenian; 152 Assyrian; 55; 69 Asturian; 252-254

283; 285-291; 293; 297; 301-305-307; Early Modern English; 288; Middle English; 261-264; 266; 271; 275; 277; 281; 287; 288; 297; Modern English; 165; 166; 168; 231; 259; 266; 267; 281; 288; 301; Old English; 182; 259-262; 264-268; 275-277; 280; 288; 301; Present-day English, see English F. French; 20; 104; 106-108; 110; 111; 121; 123; 264; 265; 302; Modern French; 231; 246; 247; Old French; 107

B.

G.

Babylonian; 55-69; Neo-BabyIonian; 56; 59; 69; Old Babylonian; 56; 59; 66; 67 Bohairic; 28; 30; 31; 39; 75; 81; 82; 89; 91; 95; 100

Galician: Modern Galician; 232; 247; 252; 253 Georgian; 135-141; 143; 146; 147; 149157; 159-163; 187-193; 195; 196; 198201; Modern Georgian; 134; 141; 143147; 152-154; 157; 159; 160; 195; Old Georgian; 134; 135; 143; 145; 146; 154; 155; 157; 158-162; 189; 200 German; 11; 26; 63; 64; 76; 121; 141; 169; 264; 266; 271 Germanic; 26; 75; 89; 132; 181; Modern High German; 46; North Germanic; 119; 122; Proto-Germanic; 26; 132; see also Ancient Nordic; German; English; Norwegian; Old Icelandic; Old Norse; Scandinavian

C. Cappadocian; 231; 232; 247-251; 256; see also Kartvelian, Caucasian; 153; 154; 156; see also Kartvelian; Lezgian; North East Caucasian Celtic; 49; 50; 51; 73; 78; 80; 93; 111; 228; 229; Brythonic Celtic; 80; see also Welsh; Irish Coptic; 23-39; 71; 73; 74-76; 80-82; 8486; 89; 91-93; 95; 97; 99; 100 D. Demotic; 23; 71-73; 82; 83; 86; 90; 97; 98 E. Egyptian; 23-28; 30; 37; 38; 39; 71; 72; 73- 77; 80-83; 85; 89-94; 97; 98-100; Earlier Egyptian; 23; Later Egyptian; 23; 24; see also Coptic; Demotic English; 27; 37; 76; 97; 99; 169; 170; 172174; 181-183; 192; 216; 259-271; 274-

Greek: Standard Greek; 250 Greek; 23; 30-32; 35; 38; 72; 78; 85; 86; 93; 97; 138; 152; 154; 156; 173-175; 177; 178; 180-182; 238; 241; 242; 243; 247; 252; 254-257; 277; Ancient Greek; 232- 237; 239; 240; 245; 254; 283; Ancient Greek; 232; Biblical Greek; 237-246; Homeric Greek; 172; 174; 175; 178; 179; Medieval Greek; 232; 244- 246; 248; 250; Standard Greek; 231; 232; 239; 245-247; 250- 251; 254; see also Cappadocian

INDEX OF LANGUAGES H. Hamito-Semitic; 73; 111; see also Egyptian Hebrew; 237; Biblical Hebrew; 237; 238; 242 Hittite; 165; 166; 170; 172-175; 180; 181 Hopi; 67

322 N.

Norse: Old Norse; 122; 124; 125; 127; 128; 131; 132 Norwegian; 126-128; 165-169; 172; 174; 180-183; Modern Norwegian; 124; 125; 128; 129; Standard Norwegian; 126

I.

P.

Icelandic; 124; Modern Icelandic; 283; Old Icelandic; 46 Indo-European; 73; 74; 78; 79; 100-113; 115; 122; 135; 156; 160; 163; 165; 170; 1 7 2 ; 179-183; 191; 193; 232;258; Proto-Indo-European; 26; 115; 122; 132; 160; 183; 232; see also Armenian; Celtic; Latin; Albanian; Germanic; Romance; Greek; Hittite; Persian Irish: Modern Irish; 82 Italian; 50; 104; 107; 108; 110; 115

Persian; 67; 161 Portuguese: Modern European Portuguese; 247; 251-254

J. Japanese; 147; 148; 162; 165-169; 174; 176; 179-183 K. Kartvelian; 134; 148; 155; 159; 161; 163; 188; 189; 191; 195; 199-201; see also Georgian; Laz; Mingrelian; Svan L. Latin: Ancient Latin; 232-237; 240; 254; Biblical Latin; 109; 237-247; Classical Latin; 77; 172; 174; 240; 255; Medieval Latin; 104; Vulgar Latin; 77; 114 Laz; 134; 148; 149; 152; 154; 155; 160; 161; 162; 188; 190 Lezgian; 153; 156; see also Tabassaran; Udi M. Mingrelian; 154; 155; 160; 161; 188-190; 193; 1 9 5 ; 198-202

R. Romance; 75; 77; 82; 89; 98; 101; 104; 106; 107; 110-115; 123; 181; 188; 232; 245-248; 251; 253; 254; 258; 303; see also Italian; French; Spanish; Portuguese Russian; 3-5; 8-17; 20; 21; 22; 193 S. Sahidic; 28; 30; 31; 75; 81; 89; 93; 99 Scandinavian; 125; 132; Modern Scandinavian; 119; 124; 127; 128 Semitic; 26; 34; 55-57; 63; 73; 74; 95; 237; see also Akkadian, Amharic, Arabic, Aramhaic, Assyrian, Hebrew, Babylonian Slavic; 161; 188; 201 Spanish; 104; 108-110; 232; 246; 247; Old Spanish; 108 Sumerian; 56; 58 Svan; 145; 162; 188; 190; 1 9 6 ;198-202 T. Tabassaran; 156 Turkic; 152; 154; 155; see also Turkish Turkish; 84; 133; 152; 154-156; 161-163 U. Udi; 142; 153; 157; 161; 162

INDEX OF LANGUAGES

w. Wappo; 67 Welsh: Early Modern Welsh; 204-207; 212; 215; 217-219; 221-225; 228; Middle Welsh; 41-50; 77; 80; 84; 100; 206; 207; 212; 219; 222-225; 228; Modern Welsh; 42; 45; 49; 50; 77; 206;

323

212; Old Welsh; 42; 50; 80 West European Sprachbund; 75 X. Xinalug; 156 Y. Yaqui; 67